2018-09-19 Richard Biener <rguenther@suse.de>
[official-gcc.git] / gcc / ChangeLog
blob68d8be20e71dcb78e679251906acadbf279f749d
1 2018-09-19  Richard Biener  <rguenther@suse.de>
3         PR tree-optimization/87349
4         PR tree-optimization/87342
5         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
7 2018-09-18  Marek Polacek  <polacek@redhat.com>
9         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
10         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
12 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
14         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
16 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
18         PR rtl-optimization/86882
19         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
21 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
23         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
24         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
26 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
28         PR other/87353
29         * doc/invoke.texi (Link Options): Fix formatting and grammar.
31 2018-09-18  Richard Biener  <rguenther@suse.de>
33         PR middle-end/63155
34         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
35         (compute_samebase_partition_bases): Likewise.
36         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
37         (gimple_can_coalesce_p): Simplify.
39 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
41         Handle a library implementation of ffs calling __builtin_ffs.
42         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
43         (mmix_init_libfuncs): New function: make __builtin_ffs expand
44         to __ffsdi2.
46 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
48         * diagnostic-show-locus.c (class layout_range): Add field
49         "m_original_idx".
50         (layout_range::layout_range): Add "original_idx" param and use it
51         to initialize new field.
52         (make_range): Use 0 for original_idx.
53         (layout::layout): Pass in index to calls to
54         maybe_add_location_range.
55         (layout::maybe_add_location_range): Add param "original_idx" and
56         pass it on to layout_range.
57         (layout::print_any_labels): Pass on range->m_original_idx to
58         get_text call.
59         (gcc_rich_location::add_location_if_nearby): Use 0 for
60         original_idx.
61         * gcc-rich-location.h (text_range_label::get_text): Update for new
62         param.
63         (range_label_for_type_mismatch::get_text): Likewise.
65 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
67         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
69 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
71         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
72         format_string_diagnostic_t.
73         (fmtwarn_n): Likewise.
74         * substring-locations.c
75         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
76         (format_warning_n_va): Convert to...
77         (format_string_diagnostic_t::emit_warning_n_va): ...this.
78         (format_warning_va): Convert to...
79         (format_string_diagnostic_t::emit_warning_va): ...this.
80         (format_warning_at_substring): Convert to...
81         (format_string_diagnostic_t::emit_warning): ...this.
82         (format_warning_at_substring_n): Convert to...
83         (format_string_diagnostic_t::emit_warning_n): ...this.
84         * substring-locations.h (class format_string_diagnostic_t): New
85         class.
86         (format_warning_va): Convert to
87         format_string_diagnostic_t::emit_warning_va.
88         (format_warning_n_va): Convert to
89         format_string_diagnostic_t::emit_warning_n_va.
90         (format_warning_at_substring): Convert to
91         format_string_diagnostic_t::emit_warning.
92         (format_warning_at_substring_n): Convert to
93         format_string_diagnostic_t::emit_warning_n.
95 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
96             Bernd Schmidt <bernds_cb1@t-online.de>
98         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
99         SImode args.
101 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
103         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
104         operand 0 predicate to nonimmediate operand.
105         (rint<mode>2_frndint): Remove insn pattern.
106         (rint<mode>2): Change operand 1 predicate to general_operand.
107         Extend operand 1 to XFmode and generate rintxf2 insn.
108         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
109         Do not use X87MODEF mode macro.
110         (frndintxf2_<rounding>_i387): Rename from
111         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
112         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
113         to XFmode and generate significandxf3 insn.
115 2018-09-17  Richard Biener  <rguenther@suse.de>
117         PR tree-optimization/87328
118         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
119         visiting unexecutable backedges when not iterating.
120         (do_rpo_vn): Mark all edges not executable even when not
121         iterating.
123 2018-09-17  Martin Jambor  <mjambor@suse.cz>
125         PR c/63886
126         * doc/invoke.texi (Warning Options): Likewise.
128 2018-09-17  Richard Biener  <rguenther@suse.de>
130         PR tree-optimization/87301
131         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
132         clean EH info from leftover copy assignments.
134 2018-09-17  Martin Liska  <mliska@suse.cz>
136         PR gcov-profile/85871
137         * gcov.c (output_intermediate_file): Fix out of bounds
138         access.
140 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
142         * config/arc/arc.c: Object attributes for core4 not reflected
143         correctly.
144         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
145         core3).
147 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
149         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
151 2018-09-17  Martin Liska  <mliska@suse.cz>
153         * doc/gcov.texi: Document new option --use-hotness-colors.
154         * gcov.c (struct source_info): Declare new field.
155         (source_info::source_info): Set default for maximum_count.
156         (print_usage): Add new -q option.
157         (process_args): Process it.
158         (accumulate_line_info): Save src->maximum_count.
159         (output_line_beginning): Make color line number if
160         flag_use_hotness_colors is set.
161         (output_line_details): Pass default argument value.
162         (output_lines): Pass src->maximum_count.
164 2018-09-17  Martin Liska  <mliska@suse.cz>
166         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
167         Use processor_names table.
168         * config/i386/i386.c (ix86_default_align): Use
169         processor_cost_table for alignment values.
170         (ix86_option_override_internal): Use processor_names.
171         (ix86_function_specific_print): Likewise.
172         * config/i386/i386.h (struct processor_costs):
173         Add alignment values.
174         (struct ptt): Remove and replace with const char *.
175         * config/i386/x86-tune-costs.h (struct processor_costs):
176         Declare default alignments for all costs.
178 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
180         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
181         symbolics or VR_VARYING ranges for ABS_EXPR.
182         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
183         when range will wrap.
185 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
187         PR middle-end/86864
188         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
189         before and after a JUMP_TABLE_DATA.
191 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
193         PR middle-end/87188
194         * dojump.c (do_compare_and_jump): Canonicalize function pointers
195         when one operand is a function pointer.  Use POINTER_TYPE_P and
196         FUNC_OR_METHOD_TYPE_P.
197         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
198         * fold-const.c (build_range_check): Likewise.
199         * match.pd (simple_comparison): Likewise.
201 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
203         PR c/82967
204         * spellcheck.c (get_edit_distance_cutoff): New function.
205         (selftest::test_edit_distance_unit_test_oneway): Rename to...
206         (selftest::test_get_edit_distance_one_way): ...this.
207         (selftest::test_get_edit_distance_unit): Rename to...
208         (selftest::test_get_edit_distance_both_ways): ...this.
209         (selftest::test_edit_distances): Move tests to this new function,
210         and test some more pairs of strings.  Update for above renaming.
211         (selftest::get_old_cutoff): New function.
212         (selftest::test_get_edit_distance_cutoff): New function.
213         (selftest::assert_suggested_for): New function.
214         (ASSERT_SUGGESTED_FOR): New macro.
215         (selftest::assert_not_suggested_for): New function.
216         (ASSERT_NOT_SUGGESTED_FOR): New macro.
217         (selftest::test_suggestions): New function.
218         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
219         tests to selftest::test_edit_distances and call it.  Add calls to
220         selftest::test_get_edit_distance_cutoff and
221         selftest::test_suggestions.
222         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
223         (best_match::consider): Replace hard-coded cutoff calculation with
224         a call to...
225         (best_match::get_cutoff): New declaration.
226         (best_match::get_best_meaningful_candidate): Likewise.
228 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
230         * builtins.c (fold_builtin_strlen): Remove TODO comment.
232 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
234         revert:
235         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
237         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
238         terminated string literal.
240 2018-09-14  Martin Sebor  <msebor@redhat.com>
242         * builtins.c (unterminated_array): Handle ARRAY_REF.
243         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
244         * builtins.h (unterminated_array): Declare extern.
245         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
246         arrays.
247         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
248         calls.
250 2018-09-14  Martin Sebor  <msebor@redhat.com>
251             Jeff Law  <law@redhat.com>
253         * builtins.c (unterminated_array): New.
254         (expand_builtin_strcpy): Adjust.
255         (expand_builtin_strcpy_args): Detect unterminated arrays.
256         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
257         unterminated arrays.
258         * gimple-fold.h (get_maxval_strlen): Add argument.
259         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
261         * gimple-fold.c (get_range_strlen): Add argument.
262         (get_maxval_strlen): Adjust.
263         * gimple-fold.h (get_range_strlen): Add argument.
265 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
267         * config/i386/movdirintrin.h: Fix copyright year.
269 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
271         * reg-stack.c: Include regs.h.
272         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
273         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
274         FIRST_STACK_REG, not DFmode.
275         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
276         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
277         (change stack): Default register mode to the reg_raw_mode of
278         FIRST_STACK_REG, not DFmode.
279         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
280         (*swapxf): Rename from swapxf.
282 2018-09-14  Carl Love  <cel@us.ibm.com>
284         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
285         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
287 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
289         PR target/87224
290         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
291         alternatives.
293 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
295         PR target/85628
296         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
298 2018-09-14  Jason Merrill  <jason@redhat.com>
300         Fix --enable-gather-detailed-mem-stats.
301         * hash-table.c (hash_table_usage): Change from variable to function.
302         * hash-table.h: Adjust.
303         * Makefile.in: Add missing dependencies on hash-table.h.
305 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
307         PR tree-optimization/87259
308         PR lto/87283
309         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
310         execute_cse_reciprocals_1 has tried transforming.
312 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
314         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
315         VR_VARYING for PLUS/MINUS_EXPR.
317 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
319         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
320         formatting.
322 2018-09-14  Richard Biener  <rguenther@suse.de>
324         PR middle-end/63155
325         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
326         bits for the merged partition.
328 2018-09-13  Martin Sebor  <msebor@redhat.com>
329             Bernd Edlinger  <bernd.edlinger@hotmail.de>
331         * builtins.h (c_srlen): Add argument.
332         * builtins.c (warn_string_no_nul): New function.
333         (c_strlen): Add argument and use it.  Update recursive calls.
334         Pass DECL argument to string_constant to get info on non
335         terminated strings.  Update *NONSTR as needed.
336         (fold_builtin_strlen): Add argument to calls to c_strlen.
337         Warn for unterminated arrays.
338         (warn_string_no_null): Add prototype.
339         * expr.c (string_constant): Update arguments.  Update recursive
340         calls appropriately.  Detect missing NUL terminator and outermost
341         declaration its missing in.
342         Improve checks for arrays with nonzero lower bound or elements
343         that are not a single byte.  Simplify offset computation.
344         Simplify checks for non-NUL terminated strings.
345         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
346         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
348 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
350         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
351         correctly.
352         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
353         argument.  Fix range checks.
354         * fold-const.h (c_getstr): Adjust protoype.
355         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
356         string is constant but contains no NUL byte.
358         * expr.c (string_constant): Adjust function comment.
359         Remove bogus check for zero termination.
361         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
363         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
364         (get_constant_size): Don't make STRING_CSTs larger than they are.
365         (check_string_literal): New check function for STRING_CSTs.
366         (output_constant): Use it.
368 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
370         PR target/86812
371         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
373 2018-09-13  Richard Biener  <rguenther@suse.de>
375         PR tree-optimization/87263
376         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
377         (struct unwind_state): Add max_rpo field.
378         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
379         Compute max_rpo, the max RPO number a block can be backwards reached
380         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
381         separating it from the iterating mode.
383 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
385         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
386         (rfs_decision): New scheduling decision.
388 2018-09-13  Richard Biener  <rguenther@suse.de>
390         PR bootstrap/87134
391         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
392         (vn_nary_op_insert_pieces_predicated): Do not write useless
393         valid_dominated_by_p entry outside of the allocated storage.
395 2018-09-13  Omar Sandoval  <osandov@osandov.com>
396             Tom de Vries  <tdevries@suse.de>
398         PR debug/86985
399         * dwarf2out.c (is_c): New function.
400         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
402 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
404         PR target/85628
405         * config/aarch64/aarch64.md (*aarch64_bfxil):
406         Define.
407         * config/aarch64/constraints.md (Ulc): Define.
408         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
409         Define.
410         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
411         New function.
413 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
415         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
416         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
417         aarch64_layout_frame call. 
418         (aarch64_expand_epilogue): Likewise.
419         (aarch64_initial_elimination_offset): Likewise.
420         (aarch64_get_separate_components): Likewise.
421         (aarch64_use_return_insn_p): Likewise.
422         (aarch64_layout_frame): Remove unneeded check.
424 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
426         * configure.ac: Only append
427         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
428         gcc_config_arguments if it was never reconfigured or last reconfigure
429         was with different arguments.
430         * configure: Regenerated.
432 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
433             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
435         PR middle-end/87290
436         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
437         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
439 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
441         PR tree-optimization/87287
442         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
443         X % C == 0 to X % (unsigned) C == 0 optimization to ...
444         * match.pd (X % C == 0): ... here.  New optimization.
446 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
448         PR middle-end/82853
449         * expr.h (maybe_optimize_mod_cmp): Declare.
450         * expr.c (mod_inv): New function.
451         (maybe_optimize_mod_cmp): New function.
452         (do_store_flag): Use it.
453         * cfgexpand.c (expand_gimple_cond): Likewise.
455 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
456             Julian Brown  <julian@codesourcery.com>
458         PR middle-end/86336
459         * gimplify.c (gimplify_scan_omp_clauses): Set
460         target_firstprivatize_array_bases in OpenACC parallel and kernels
461         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
462         OpenACC data regions.
464 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
466         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
467         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
469 2018-09-12  Richard Biener  <rguenther@suse.de>
471         PR tree-optimization/87280
472         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
473         edge but unreachable target.
474         (do_rpo_vn): For conservatively handling a PHI only mark
475         the backedge executable but not the block reachable.
477 2018-09-12  Richard Biener  <rguenther@suse.de>
479         PR tree-optimization/87266
480         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
481         visited blocks.
483 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
485         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
486         constants.
487         ("trunc<BFP:mode><DFP_ALL:mode>2")
488         ("trunc<DFP_ALL:mode><BFP:mode>2")
489         ("extend<BFP:mode><DFP_ALL:mode>2")
490         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
491         according to the target operand type.
493 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
494             Andreas Krebbel  <krebbel@linux.ibm.com>
496         PR tree-optimization/86844
497         * gimple-ssa-store-merging.c
498         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
499         there are any overlapping stores in between them, make sure they are
500         also coalesced or we give up completely.
502 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
504         PR middle-end/87248
505         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
506         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
508 2018-09-12  Tom de Vries  <tdevries@suse.de>
510         * common.opt (gdescribe-dies): Add option.
511         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
512         attribute for artifical and nameless decls.
513         (dwarf2out_register_external_die): Add description attribute to
514         external reference die.
515         (add_desc_attribute): New functions.
516         (gen_subprogram_die): Add description attribute to
517         DW_TAG_call_site_parameter.
518         * tree-pretty-print.c (print_generic_expr_to_str): New function.
519         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
520         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
521         -gno-describe-dies.
522         (@item -gdescribe-dies): Add.
524 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
526         * tree-vrp.c (vrp_shift_undefined_p): Remove.
527         (extract_range_from_binary_expr_1: Call
528         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
529         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
530         depend on sign.
532 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
534         * gimple-ssa-warn-alloca.c
535         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
536         field for ALLOCA_BOUND_*_LARGE.
538 2018-09-11  Nathan Sidwell  <nathan@acm.org>
540         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
542 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
544         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
545         for clobbers.  Remove obsolete comment.
547 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
549         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
550         mpxchk, mpxld and mpxst types.
551         (define_attr length_immediate): Remove all processing of mpx types.
552         (define_attr prefix_0f): Ditto.
553         (define_attr memory): Ditto.
555 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
557         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
558         (log<mode>2): Change operand 1 predicate to general_operand.
559         Extend operand 1 to XFmode and generate logxf3 insn.
560         (log10<mode>2): Change operand 1 predicate to general_operand.
561         Extend operand 1 to XFmode and generate log10xf3 insn.
562         (log2<mode>2): Change operand 1 predicate to general_operand.
563         Extend operand 1 to XFmode and generate log2xf3 insn.
564         (fyl2xp1_extend<mode>xf3_i387): Remove.
565         (log1p<mode>2): Change operand 1 predicate to general_operand.
566         Extend operand 1 to XFmode and generate log1pxf3 insn.
567         (fxtract_extend<mode>xf3_i387): Remove.
568         (logb<mode>2): Change operand 1 predicate to general_operand.
569         Extend operand 1 to XFmode and generate logbxf3 insn.
570         (ilogb<mode>2): Change operand 1 predicate to general_operand.
571         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
572         (significand<mode>2): Change operand 1 predicate to general_operand.
573         Extend operand 1 to XFmode and generate significandxf3 insn.
575 2018-09-11  Nathan Sidwell  <nathan@acm.org>
577         * gcc.c (perror_with_name, pfatal_with_name): Delete.
578         (load_specs): Use fatal_error.
579         (DELETE_IF_ORDINARY, process_command): Use error.
580         (execute, run_attempt): Use fatal_error.
582 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
584         * diagnostic-core.h (sorry_at): New prototype.
585         * diagnostic.c (sorry_at): New function.
587 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
589         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
590         by zero as VR_UNDEFINED.
592 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
594         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
595         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
596         (<sincos>mode2): New expander.
597         (sincos_extend<mode>xf3_i387): Remove insn pattern.
598         (sincos -> sin, cos splitters): Remove splitter patterns.
599         (sincos<mode>3): Change operand 2 predicate to general_operand.
600         Extend operand 2 to XFmode and generate sincosxf3 insn.
601         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
602         Change operand 3 predicate to const1_operand.
603         (fptan_extend<mode>xf4_i387): Remove insn pattern.
604         (tanxf2): Update operands in the call to fptanxf4_i387.
605         (tan<mode>2): Change operand 1 predicate to general_operand.
606         Extend operand 1 to XFmode and generate tanxf3 insn.
607         (atan2xf3): Rename from *fpatanxf3_i387.
608         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
609         (atan2xf3): Remove expander.
610         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
611         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
612         (atan<mode>2): Change operand 1 predicate to general_operand.
613         Extend operand 1 to XFmode and generate atanxf3 insn.
615 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
617         * config/i386/i386.md (x87/SSE constant load splitter): Use
618         memory_operand instead of nonimmediate_operand for input operand
619         predicate.
621 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
623         * config/i386/i386.md (float partial SSE register stall splitter): Move
624         splitter near its instruction pattern.
625         (float_extend partial SSE register stall splitter): Ditto.
626         (float_truncate partial SSE register stall splitter): Ditto.
628 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
630         PR target/86794
631         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
632         to speculation_safe_value_not_needed.
634         PR target/85666
635         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
636         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
637         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
638         leaf_function_p, instead use has_hard_reg_initial_val.
640 2018-09-09  Nathan Sidwell  <nathan@acm.org>
642         * gcc.h (pfatal_with_name): Don't declare here.
643         * gcc.c (pfatal_with_name): Make static.
645 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
647         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
648         earlyclobber.
650 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
652         PR rtl-optimization/85458
653         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
654         priority hook to reduce the priority of EXPR.
656 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
658         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
659         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
660         with X87_ENABLE_FLOAT.  Remove preparation code.
661         (*float<SWI48:mode><MODEF:mode>2): Rename from
662         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
663         math using "enabled" attribute.
664         (*floatdi<MODEF:mode>2_i387): Rename from
665         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
666         enable for 32bit targets only.
667         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
668         splitter.
669         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
670         as operand 1 predicate.  Rewrite as define_insn_and_split.
671         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
673 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
675         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
676         to fallthru to FLOAT case.
678 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
680         PR target/86731
681         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
682         around folding of vec_sl to handle out of range shift values.
684 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
686         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
687         Update callers to gen_fix_trunc<mode>_i387_fisttp
688         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
689         nonimmediate_operand.
690         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
691         and corresponding splitters.
692         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
693         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
694         (fix_truncdi_i387_with_temp): Remove insn pattern
695         and corresponding splitters.
696         (fix_trunc<mode>_i387): Change operand 0 predicate to
697         nonimmediate_operand.
698         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
699         and corresponding splitters.
700         (*fistdi2_1): Remove.
701         (fistdi2): Ditto.
702         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
703         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
704         (*fist<mode>2_1): Remove.
705         (fist<mode>2): Ditto.
706         (fist<mode>2_with_temp): Remove insn pattern and corresponding
707         splitters.
708         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
709         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
710         (fistdi2_<rounding>): Change operand 0 predicate to
711         nonimmediate_operand.
712         (fistdi2_<rounding>_with_temp): Remove insn pattern
713         and corresponding splitters.
714         (fist<mode>2_<rounding>): Change operand 0 predicate to
715         nonimmediate_operand.
716         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
717         and corresponding splitters.
719         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
721 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
723         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
724         the init value.
726 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
728         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
729         early gimple folding of vec_splat().
730         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
731         * gimple-fold.h: Add an extern define for tree_vec_extract().
733 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
735         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
736         wrappers around TREE_TYPE comparisons.
738 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
740         PR target/80080
741         * config/s390/predicates.md: Add nonsym_memory_operand.
742         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
743         contains a SYMBOL_REF, load it into an intermediate pseudo.
744         (s390_emit_compare_and_swap): Legitimize operand.
745         * config/s390/s390.md: Use the new nonsym_memory_operand
746         with UNSPECV_CAS patterns.
748 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
750         PR target/80080
751         * config/s390/s390-passes.def: New file.
752         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
753         declaration.
754         (make_pass_s390_early_mach): Add declaration.
755         * config/s390/s390.c (make_pass_s390_early_mach):
756         (s390_option_override): Remove dynamic registration.
757         * config/s390/t-s390: Add s390-passes.def.
759 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
761         * config/s390/s390.c (s390_decompose_constant_pool_ref):
762         Remove UNSPEC_LTREL_BASE check.
763         (annotate_constant_pool_refs): Likewise.
764         (find_constant_pool_ref): Likewise.
765         (find_ltrel_base): Removed.
766         (replace_ltrel_base): Removed.
767         (s390_mainpool_finish): Remove replace_ltrel_base call.
768         (s390_chunkify_start): Remove pending LTREL_BASE logic.
769         (s390_chunkify_finish): Remove replace_ltrel_base call.
770         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
772 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
774         PR target/86779
775         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
776         to speculation_safe_value_not_needed.
778 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
779             Bernd Schmidt  <bernds_cb1@t-online.de>
781         * config/nvptx/nvptx-opts.h: New file.
782         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
783         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
784         (ASM_SPEC): Define.
785         (TARGET_SM35): New macro.
786         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
787         correct predicate.
788         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
789         values.
790         (misa=): New option.
791         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
793 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
795         * config/i386/i386.md (truncdfsf2): Remove expander.
796         (truncdfsf2_with_temp): Ditto.
797         (truncxf<mode>2): Ditto.
798         (*truncdfsf_fast_mixed): Remove insn pattern.
799         (*truncdfsf_fast_i387): Ditto.
800         (*truncdfsf_mixed): Ditto.
801         (*truncdfsf_i387): Ditto.
802         (*truncdfsf2_i387_1): Ditto.
803         (*truncxfsf2_mixed): Ditto.
804         (*truncxfdf2_mixed): Ditto.
805         (*truncxf<mode>2_i387_noop): Ditto. Update callers
806         to call gen_truncxf<mode>2 instead.
807         (*truncxf<mode>2_i387): Remove.
808         (reg->reg splitters): Remove splitter pattern.
809         (reg->mem splitters): Ditto.
811         (truncdfsf2): New insn pattern.
812         (truncxf<mode>2): Ditto.
814 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
816         * tree-ssa-math-opts.c (is_mult_by): New function.
817         (is_square_of): Use the above.
818         (optimize_recip_sqrt): New function.
819         (pass_cse_reciprocals::execute): Use the above.
821 2018-09-05  Richard Biener  <rguenther@suse.de>
823         PR bootstrap/87134
824         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
825         to zero-init the emplaced vec.
827 2018-09-05  Martin Liska  <mliska@suse.cz>
829         PR tree-optimization/87205
830         * tree-switch-conversion.c (pass_lower_switch::execute):
831         Group cases for switch statements.
833 2018-09-05  Richard Biener  <rguenther@suse.de>
835         PR tree-optimization/87217
836         * tree-ssa-sccvn.c (vuse_valueize): New.
837         (vn_reference_lookup_pieces): Use it.
838         (vn_reference_lookup): Likewise.
840 2018-09-05  Nathan Sidwell  <nathan@acm.org>
842         PR c++/87137
843         * stor-layout.c (place_field): Scan forwards to check last
844         bitfield when ms_bitfield_placement is in effect.
846 2018-09-05  Richard Biener  <rguenther@suse.de>
848         PR bootstrap/87225
849         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
850         return.
852 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
853             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
855         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
856         * config.gcc (extra_objs): Build it.
857         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
858         Likewise.
859         * config/aarch64/aarch64-passes.def
860         (pass_tag_collision_avoidance): New pass.
861         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
862         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
863         (aarch64_classify_address): Remove static qualifier.
864         (aarch64_address_info, aarch64_address_type): Move to...
865         * config/aarch64/aarch64-protos.h: ... here.
866         (make_pass_tag_collision_avoidance): New function.
867         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
868         New tuning flag.
870 2018-09-05  Martin Liska  <mliska@suse.cz>
872         * doc/gcov.texi: Update documentation of humar
873         readable mode.
874         * gcov.c (format_count): Print one decimal place, it provides
875         more fine number of situations like '1G' vs. '1.4G'.
877 2018-09-05  Martin Liska  <mliska@suse.cz>
879         PR target/87164
880         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
881         * optc-gen.awk: Allow 'Var' for Deprecated options in order
882         to generate a MASK value.
884 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
886         PR debug/86593
887         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
888         if frame pointer isn't used.
889         (compute_frame_pointer_to_fb_displacement): Likewise.
891 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
893         PR target/87198
894         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
895         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
896         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
897         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
898         and OPTION_MASK_ISA_XSAVEC_UNSET.
900 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
902         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
903         XOR operations in NAND case.
905 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
907         * wide-int-range.cc (wide_int_range_convert): New.
908         * wide-int-range.h (wide_int_range_convert): New.
909         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
910         code into wide_int_range_convert.
911         (extract_range_into_wide_ints): Do not munge anti range constants
912         into the entire domain.  Just return the range back.
914 2018-09-04  Martin Liska  <mliska@suse.cz>
916         * genmatch.c (output_line_directive): Add new argument
917         fnargs.
918         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
920 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
922         * doc/invoke.texi (Option Summary): Add whitespace.
924         * doc/invoke.texi (Option Summary): Add -Waligned-new.
926 2018-09-04  Richard Biener  <rguenther@suse.de>
928         PR tree-optimization/87211
929         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
930         backedge value we're supposed to treat as VARYING also number
931         the PHI to VARYING in case it got a different value-number already.
933 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
935         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
936         (extract_range_from_binary_expr_1): Do not call
937         vrp_can_optimize_bit_op.
938         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
939         static.
940         (wide_int_range_get_mask_and_bounds): New.
941         (wide_int_range_optimize_bit_op): New.
942         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
943         (wide_int_range_bit_and): Same.
944         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
945         (wide_int_range_optimize_bit_op): New.
946         (wide_int_range_get_mask_and_bounds): New.
948 2018-09-04  Richard Biener  <rguenther@suse.de>
950         PR tree-optimization/87176
951         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
952         variable.  When value-numbering a virtual PHI node make sure
953         to not value-number to the backedge value.
955 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
957         * doc/extend.texi (Long Long, Hex Floats): Document support for
958         long long and hex floats in more recent versions of ISO C++.
960 2018-09-03  Richard Biener  <rguenther@suse.de>
962         PR tree-optimization/87177
963         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
964         cleanup.
966 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
968         * bb-reorder.c (edge_order): Convert to C-qsort-style
969         tri-state comparator.
970         (reorder_basic_blocks_simple): Change std::stable_sort to gcc_stablesort.
972 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
974         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
975         tri-state comparator.
976         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
978 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
980         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
981         (mergesort): ... here as maximum count for using netsort.
982         (gcc_qsort): Set nlim to 3 if stable sort is requested.
983         (gcc_stablesort): New.
984         * system.h (gcc_stablesort): Declare.
986 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
988         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
989         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
990         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
992 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
994         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
995         lxsdx and stxsdx alternatives.
996         (*mov<mode>_hardfloat64): Ditto.
997         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
999 2018-09-03  Richard Biener  <rguenther@suse.de>
1001         PR tree-optimization/87200
1002         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
1003         simplify result.
1005 2018-09-03  Martin Liska  <mliska@suse.cz>
1007         PR tree-optimization/87201
1008         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1009         Fix parenthesis in an expression.
1011 2018-09-03  Richard Biener  <rguenther@suse.de>
1013         PR tree-optimization/87197
1014         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
1015         visited.  CSE the VN_INFO hashtable lookup.
1017         PR tree-optimization/87169
1018         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
1019         iterating make sure there's no extra backedges from irreducible
1020         regions feeding the header.  Mark the destination block
1021         executable.
1023 2018-09-03  Martin Liska  <mliska@suse.cz>
1025         PR driver/83193
1026         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1027         * common/common-targhooks.c (default_get_valid_option_values):
1028         New function.
1029         * common/common-targhooks.h (default_get_valid_option_values):
1030         Likewise.
1031         * common/config/i386/i386-common.c: Move processor_target_table
1032         from i386.c.
1033         (ix86_get_valid_option_values): New function.
1034         (TARGET_GET_VALID_OPTION_VALUES): New macro.
1035         * config/i386/i386.c (struct ptt): Move to i386-common.c.
1036         (PTA_*): Move all defined masks into i386-common.c.
1037         (ix86_function_specific_restore): Use new processor_cost_table.
1038         * config/i386/i386.h (struct ptt): Moved from i386.c.
1039         (struct pta): Likewise.
1040         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
1041         * doc/tm.texi.in: Likewise.
1042         * opt-suggestions.c (option_proposer::suggest_option):
1043         Pass prefix to build_option_suggestions.
1044         (option_proposer::get_completions): Likewise.
1045         (option_proposer::build_option_suggestions): Use the new target
1046         hook.
1047         * opts.c (struct option_help_tuple): New struct.
1048         (print_filtered_help): Use the new target hook.
1050 2018-09-03  Martin Liska  <mliska@suse.cz>
1052         PR middle-end/59521
1053         * predict.c (set_even_probabilities): Add likely_edges
1054         argument and handle cases where we have precisely one
1055         likely edge.
1056         (combine_predictions_for_bb): Catch also likely_edges.
1057         (tree_predict_by_opcode): Handle gswitch statements.
1058         * tree-cfg.h (find_case_label_for_value): New declaration.
1059         (find_taken_edge_switch_expr): Likewise.
1060         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
1061         Find pivot in decision tree based on probabily, not by number of
1062         nodes.
1064 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
1066         * doc/standards.texi (Standards): Update Objective-C reference.
1068 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1070         * doc/install.texi (Prerequisites): Update link for MPC.
1072 2018-09-01  Michael Matz  <matz@suse.de>
1074         PR tree-optimization/87074
1075         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
1076         PHIs for outer-loop uses.
1078 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1080         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
1081         * doc/invoke.texi (C Dialect Options): Ditto.
1083 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
1085         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
1087 2018-08-31  Richard Biener  <rguenther@suse.de>
1089         PR tree-optimization/87168
1090         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
1091         (rpo_elim::eliminate_avail): When OP was not visited it must
1092         be available.
1094 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
1096         * tree-vrp.c (copy_value_range): Convert param "from" from
1097         "value_range *" to "const value_range *".
1098         (range_is_null): Likewise for param "vr".
1099         (range_int_cst_p): Likewise.
1100         (range_int_cst_singleton_p): Likewise.
1101         (symbolic_range_p): Likewise.
1102         (value_ranges_intersect_p): Likewise for both params.
1103         (value_range_nonnegative_p): Likewise for param "vr".
1104         (value_range_constant_singleton): Likewise.
1105         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
1106         (extract_range_into_wide_ints): Likewise for param "vr".
1107         (extract_range_from_multiplicative_op): Likewise for params "vr0"
1108         and "vr1".
1109         (vrp_can_optimize_bit_op): Likewise.
1110         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
1111         "vr1_".
1112         (extract_range_from_unary_expr): Likewise.
1113         (debug_value_range): Likewise for param "vr".
1114         (value_range::dump): Add "const" qualifier.
1115         (vrp_prop::check_array_ref): Convert local "vr" from
1116         "value_range *" to "const value_range *".
1117         (vrp_prop::check_mem_ref): Likewise.
1118         (vrp_prop::visit_stmt): Likewise for local "old_vr".
1119         (vrp_intersect_ranges_1): Likewise for param "vr_1".
1120         (vrp_intersect_ranges): Likewise.
1121         (simplify_stmt_for_jump_threading): Likewise for local "vr".
1122         (vrp_prop::vrp_finalize): Likewise.
1123         * tree-vrp.h (value_range::dump): Add "const" qualifier.
1124         (vrp_intersect_ranges): Add "const" qualifier to params as above.
1125         (extract_range_from_unary_expr): Likewise.
1126         (value_range_constant_singleton): Likewise.
1127         (symbolic_range_p): Likewise.
1128         (copy_value_range): Likewise.
1129         (extract_range_from_binary_expr_1): Likewise.
1130         (range_int_cst_p): Likewise.
1131         (vrp_set_zero_nonzero_bits): Likewise.
1132         (range_int_cst_singleton_p): Likewise.
1134 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
1136         * config/aarch64/arm_neon.h (vabsd_s64): New.
1137         (vnegd_s64): Likewise.
1139 2018-08-31  Martin Jambor  <mjambor@suse.cz>
1141         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
1143 2018-08-31  Martin Liska  <mliska@suse.cz>
1145         * ipa-icf.c (sem_item::add_type): Use
1146         sem_item::m_type_hash_cache.
1147         * ipa-icf.h: Move the cache from sem_item_optimizer
1148         to sem_item.
1150 2018-08-31  Nathan Sidwell  <nathan@acm.org>
1152         * doc/extend.texi (Backwards Compatibility): Remove implicit
1153         extern C leeway of () being (...).
1155 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1157         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
1159 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
1161         PR target/86684
1162         PR target/87149
1163         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1165 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
1167         PR middle-end/87138
1168         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
1169         gen_int_mode.  Formatting fixes.
1171 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
1173         * target.def (custom_function_descriptors): Improve documentation.
1174         * doc/tm.texi.in (Trampolines): Expand discussion of function
1175         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
1176         beginning of the section.
1177         * doc/tm.texi: Regenerated.
1179 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
1181         * cfg.h (class auto_edge_flag): Spell out the template-id of the
1182         base class in the initializer list.  This is a workaround for
1183         building with older GCC.
1184         (class auto_bb_flag): Likewise.
1186 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
1188         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
1189         (altivec_vcmpequ<VI_char>_p): Remove star.
1190         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
1191         vector load modes.
1192         (expand_strncmp_vec_sequence): New function.
1193         (emit_final_str_compare_vec): New function.
1194         (expand_strn_compare): Add support for vector strncmp.
1195         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
1196         length specification to bytes.
1197         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
1198         (vcmpnezb_p): New pattern.
1199         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
1200         for option -mstring-compare-inline-limit.
1202 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
1204         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
1205         (PTA_SKYLAKE): Add PTA_AES.
1206         (PTA_GOLDMONT): Likewise.
1208 2018-08-29  Jan Hubicka  <jh@suse.cz>
1210         PR lto/86517
1211         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
1212         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
1214 2018-08-29  Jan Hubicka  <jh@suse.cz>
1216         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
1217         TYPE_STUB_DECL.
1218         (hash_tree): Do not visit TYPE_STUB_DECL.
1219         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
1220         stream TYPE_STUB_DECL.
1221         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
1222         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
1223         after free_lang_data.
1224         (type_in_anonymous_namespace_p): Likewise.
1226 2018-08-29  Jan Hubicka  <jh@suse.cz>
1228         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1229         comment that it has to be even number.
1230         (class sreal): Change m_sig type to int32_t.
1231         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1232         int64_t for temporary calculations.
1233         (sreal_verify_basics): Drop one bit from minimum and maximum.
1235 2018-08-30  Richard Biener  <rguenther@suse.de>
1237         PR tree-optimization/87147
1238         * tree-ssa-sccvn.c (SSA_VISITED): New function.
1239         (visit_phi): When the degenerate result is from the backedge and
1240         we didn't visit its definition yet drop to VARYING.
1241         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
1243 2018-08-29  Jan Hubicka  <jh@suse.cz>
1245         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
1246         DECL_VINDEX.
1247         (hash_tree): Likewise.
1249 2018-08-29  Jan Hubicka  <jh@suse.cz>
1251         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
1252         and TYPE_NEXT_REF_TO.
1254 2018-08-29  Jan Hubicka  <jh@suse.cz>
1256         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
1257         comment that it has to be even number.
1258         (class sreal): Change m_sig type to int32_t.
1259         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
1260         int64_t for temporary calculations.
1261         (sreal_verify_basics): Drop one bit from minimum and maximum.
1263 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
1265         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
1267 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
1269         PR middle-end/86995
1270         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
1271         if to_add is negative.
1273 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1275         PR middle-end/87053
1276         * builtins.c (c_strlen): Improve range checks.
1278 2018-08-29  Martin Sebor  <msebor@redhat.com>
1279             Jeff Law  <law@redhat.com>
1281         PR tree-optimization/86714
1282         PR tree-optimization/86711
1283         * builtins.c (c_strlen): Add arguments to call to string_constant.
1284         * expr.c (string_constant): Add argument.  Detect missing nul
1285         terminator and outermost declaration it's missing in.
1286         * expr.h (string_constant): Add argument.
1287         * fold-const.c (read_from_constant_string): Add arguments to call to
1288         string_constant.
1289         (c_getstr): Likewise.
1290         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1291         to string_constant.
1292         * tree-ssa-strlen.c (get_stridx): Likewise.
1294 2018-08-29  Jan Hubicka  <jh@suse.cz>
1296         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
1297         Do not stream DECL_VINDEX.
1298         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
1299         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
1300         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
1302 2018-08-29  Richard Biener  <rguenther@suse.de>
1304         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
1305         virtual operands that are not default defs to honor region
1306         boundaries.
1307         (rpo_vn_valueize): Remove ineffective code here.
1309 2018-08-29  Richard Biener  <rguenther@suse.de>
1311         PR tree-optimization/87132
1312         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
1313         when skipping defs reachable over backedges.
1315 2018-08-29  Richard Biener  <rguenther@suse.de>
1317         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
1318         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
1319         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
1320         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
1321         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
1322         (vn_reference_lookup_3): Remove use of const_parms.
1323         (free_rpo_vn): Do not free const_parms.
1324         (do_rpo_vn): Do not call init_const_parms.
1325         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
1326         SSA_NAME_POINTS_TO_READONLY_MEMORY.
1327         (call_may_clobber_ref_p_1): Likewise.
1329 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
1331         PR other/86726
1332         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
1333         (-O): Ditto.
1334         (-ftree-scev-cprop): Document.
1336 2018-08-29  Jan Hubicka  <jh@suse.cz>
1338         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
1339         parameters.
1340         (sreal constructor): Update.
1341         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
1342         sreal:operator/): Update.
1344 2018-08-29  Martin Liska  <mliska@suse.cz>
1346         * tree-switch-conversion.c (switch_conversion::expand):
1347         Strenghten assumption about gswitch statements.
1349 2018-08-29  Richard Biener  <rguenther@suse.de>
1351         PR tree-optimization/87117
1352         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
1353         re-value-number released SSA VDEFs.
1355 2018-08-29  Richard Biener  <rguenther@suse.de>
1357         PR tree-optimization/87126
1358         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
1360 2018-08-28  Jim Wilson  <jimw@sifive.com>
1362         * config/riscv/pic.md: Rewrite.
1363         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
1364         invalid address.
1365         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
1366         (SOFTF, default_load, softload, softstore): New.
1368 2018-08-28  Jeff Law  <law@redhat.com>
1370         * fold-const.c (fold_binary_loc): Remove recently added assert.
1372 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
1374         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
1375         to OP parmeter of generated function.
1377 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
1379         PR tree-optimization/87009
1380         * match.pd: Add boolean optimizations.
1382 2018-08-28  Martin Sebor  <msebor@redhat.com>
1384         PR middle-end/86631
1385         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
1386         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
1387         (pass_walloca::gate): Use it.
1388         (alloca_call_type): Same.
1389         (pass_walloca::execute): Same.
1390         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
1392 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
1394         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
1395         GCC_VERSION for usage of "__gcc_dump_printf__" format from
1396         >= 3005 to >= 9000.
1398 2018-08-28  Richard Biener  <rguenther@suse.de>
1400         PR tree-optimization/87124
1401         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
1402         constants before looking up avail.
1404 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
1406         PR middle-end/87099
1407         * calls.c (maybe_warn_nonstring_arg): Punt early if
1408         warn_stringop_overflow is zero.  Don't call get_range_strlen
1409         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
1410         Swap comparison operands to have constants on rhs.  Only use
1411         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
1412         increment lenrng[0].
1414 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
1416         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
1417         use of tree_to_shwi.  Remove duplicated test for the size being
1418         a whole number of bytes.
1420 2018-08-28  Richard Biener  <rguenther@suse.de>
1422         PR tree-optimization/87117
1423         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
1424         Handle removed stmt without LHS (GIMPLE_NOP).
1426 2018-08-28  Richard Biener  <rguenther@suse.de>
1428         PR tree-optimization/87117
1429         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
1430         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
1432 2018-08-28  Richard Biener  <rguenther@suse.de>
1434         PR tree-optimization/87117
1435         * tree-ssa-pre.c (compute_avail): Do not make expressions
1436         with predicated values available.
1437         (get_expr_value_id): Assert we do not run into predicated value
1438         expressions.
1440 2018-08-28  Richard Biener  <rguenther@suse.de>
1442         PR tree-optimization/87117
1443         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
1444         get virtual operands.
1445         (get_expr_operands): Handle STRING_CST like other decls.
1447 2018-08-28  Martin Liska  <mliska@suse.cz>
1449         * tree.h: Update documentation of fndecl_built_in_p
1450         functions.
1453 2018-08-27  Jeff Law  <law@redhat.com>
1454         PR tree-optimization/87110
1455         * tree-ssa-dse.c (compute_trims): Handle non-constant
1456         TYPE_SIZE_UNIT.
1458 2018-08-27  Martin Sebor  <msebor@redhat.com>
1460         PR tree-optimization/86914
1461         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
1463 2018-08-27  Martin Sebor  <msebor@redhat.com>
1465         PR tree-optimization/87112
1466         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
1467         the type of the bound argument.
1469 2018-08-27  Jeff Law  <law@redhat.com>
1471         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
1472         type does not have a TYPE_SIZE_UNIT.
1474 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
1476         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
1477         with include of backend.h.
1479 2018-08-27  Richard Biener  <rguenther@suse.de>
1481         PR tree-optimization/86927
1482         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
1483         use const cond reduction code.
1485 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
1487         PR tree-optimization/85758
1488         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
1490 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
1492         PR c++/87091
1493         * diagnostic-show-locus.c (class layout_range): Update for
1494         conversion of show_caret_p to a tri-state.
1495         (layout_range::layout_range): Likewise.
1496         (make_range): Likewise.
1497         (layout::maybe_add_location_range): Likewise.
1498         (layout::should_print_annotation_line_p): Don't show annotation
1499         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
1500         (layout::get_state_at_point): Update for conversion of
1501         show_caret_p to a tri-state.  Bail out early for
1502         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
1503         underlining or source colorization.
1504         (gcc_rich_location::add_location_if_nearby): Update for conversion
1505         of show_caret_p to a tri-state.
1506         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
1507         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
1508         Likewise.
1509         (selftest::test_one_liner_labels): Likewise.
1510         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
1511         conversion of show_caret_p to a tri-state.
1512         * pretty-print.c (text_info::set_location): Likewise.
1513         * pretty-print.h (text_info::set_location): Likewise.
1514         * substring-locations.c (format_warning_n_va): Likewise.
1515         * tree-diagnostic.c (default_tree_printer): Likewise.
1516         * tree-pretty-print.c (newline_and_indent): Likewise.
1518 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
1520         PR c++/87091
1521         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
1522         line above for line-insertion fix-it hints.
1523         (selftest::test_fixit_insert_containing_newline): Update the
1524         expected results, and add a test with line-numbering enabled.
1526 2018-08-27  Martin Liska  <mliska@suse.cz>
1528         PR sanitizer/86962
1529         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
1530         params with DECL_HAS_VALUE_EXPR_P.
1532 2018-08-27  Martin Liska  <mliska@suse.cz>
1534         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
1535         selected expansion strategy.
1537 2018-08-27  Martin Liska  <mliska@suse.cz>
1539         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
1540         * builtins.c (is_builtin_fn): Likewise.
1541         * attribs.c (diag_attr_exclusions): Use new function
1542         fndecl_built_in_p and remove check for FUNCTION_DECL if
1543         possible.
1544         (builtin_mathfn_code): Likewise.
1545         (fold_builtin_expect): Likewise.
1546         (fold_call_expr): Likewise.
1547         (fold_builtin_call_array): Likewise.
1548         (fold_call_stmt): Likewise.
1549         (set_builtin_user_assembler_name): Likewise.
1550         (is_simple_builtin): Likewise.
1551         * calls.c (gimple_alloca_call_p): Likewise.
1552         (maybe_warn_nonstring_arg): Likewise.
1553         * cfgexpand.c (expand_call_stmt): Likewise.
1554         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
1555         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
1556         (cgraph_node::verify_node): Likewise.
1557         * cgraphclones.c (build_function_decl_skip_args): Likewise.
1558         (cgraph_node::create_clone): Likewise.
1559         * config/arm/arm.c (arm_insert_attributes): Likewise.
1560         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
1561         * dse.c (scan_insn): Likewise.
1562         * expr.c (expand_expr_real_1): Likewise.
1563         * fold-const.c (operand_equal_p): Likewise.
1564         (fold_binary_loc): Likewise.
1565         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
1566         * gimple-low.c (lower_stmt): Likewise.
1567         * gimple-pretty-print.c (dump_gimple_call): Likewise.
1568         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Likewise.
1569         * gimple.c (gimple_build_call_from_tree): Likewise.
1570         (gimple_call_builtin_p): Likewise.
1571         (gimple_call_combined_fn): Likewise.
1572         * gimplify.c (gimplify_call_expr): Likewise.
1573         (gimple_boolify): Likewise.
1574         (gimplify_modify_expr): Likewise.
1575         (gimplify_addr_expr): Likewise.
1576         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
1577         * ipa-cp.c (determine_versionability): Likewise.
1578         * ipa-fnsummary.c (compute_fn_summary): Likewise.
1579         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
1580         * ipa-split.c (visit_bb): Likewise.
1581         (split_function): Likewise.
1582         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1583         * lto-cgraph.c (input_node): Likewise.
1584         * lto-streamer-out.c (write_symbol): Likewise.
1585         * omp-low.c (setjmp_or_longjmp_p): Likewise.
1586         (lower_omp_1): Likewise.
1587         * predict.c (strip_predict_hints): Likewise.
1588         * print-tree.c (print_node): Likewise.
1589         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
1590         * trans-mem.c (is_tm_irrevocable): Likewise.
1591         (is_tm_load): Likewise.
1592         (is_tm_simple_load): Likewise.
1593         (is_tm_store): Likewise.
1594         (is_tm_simple_store): Likewise.
1595         (is_tm_abort): Likewise.
1596         (tm_region_init_1): Likewise.
1597         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
1598         * tree-cfg.c (verify_gimple_call): Likewise.
1599         (move_stmt_r): Likewise.
1600         (stmt_can_terminate_bb_p): Likewise.
1601         * tree-eh.c (lower_eh_constructs_2): Likewise.
1602         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
1603         * tree-inline.c (remap_gimple_stmt): Likewise.
1604         (copy_bb): Likewise.
1605         (estimate_num_insns): Likewise.
1606         (fold_marked_statements): Likewise.
1607         * tree-sra.c (scan_function): Likewise.
1608         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
1609         (optimize_stack_restore): Likewise.
1610         (pass_fold_builtins::execute): Likewise.
1611         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
1612         (mark_all_reaching_defs_necessary_1): Likewise.
1613         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
1614         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
1615         (pass_forwprop::execute): Likewise.
1616         * tree-ssa-loop-im.c (stmt_cost): Likewise.
1617         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
1618         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
1619         * tree-ssa-strlen.c (get_string_length): Likewise.
1620         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
1621         (find_func_aliases_for_call): Likewise.
1622         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
1623         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
1624         * tree-tailcall.c (find_tail_calls): Likewise.
1625         * tree.c (need_assembler_name_p): Likewise.
1626         (free_lang_data_in_decl): Likewise.
1627         (get_call_combined_fn): Likewise.
1628         * ubsan.c (is_ubsan_builtin_p): Likewise.
1629         * varasm.c (incorporeal_function_p): Likewise.
1630         * tree.h (DECL_BUILT_IN): Remove and replace with
1631         fndecl_built_in_p.
1632         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
1633         (fndecl_built_in_p): New.
1635 2018-08-27  Martin Liska  <mliska@suse.cz>
1637         PR tree-optimization/86847
1638         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
1639         Dump also subtree probability.
1640         (switch_decision_tree::do_jump_if_equal): New function.
1641         (switch_decision_tree::emit_case_nodes): Handle special
1642         situations in balanced tree that can be emitted much simpler.
1643         Fix calculation of probabilities that happen in tree expansion.
1644         * tree-switch-conversion.h (struct cluster): Add
1645         is_single_value_p.
1646         (struct simple_cluster): Likewise.
1647         (struct case_tree_node): Add new function has_child.
1648         (do_jump_if_equal): New.
1650 2018-08-27  Martin Liska  <mliska@suse.cz>
1652         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
1653         Add new argument to bit_test_cluster constructor.
1654         (bit_test_cluster::emit): Set bits really number of values
1655         handlel by a test.
1656         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
1657         probability argument.
1658         * tree-switch-conversion.h (struct bit_test_cluster):
1659         Add m_handles_entire_switch.
1661 2018-08-27  Martin Liska  <mliska@suse.cz>
1663         PR tree-optimization/86702
1664         * tree-switch-conversion.c (jump_table_cluster::emit):
1665         Make probabilities even for values in jump table
1666         according to number of cases handled.
1667         (switch_decision_tree::compute_cases_per_edge): Pass
1668         argument to reset_out_edges_aux function.
1669         (switch_decision_tree::analyze_switch_statement): Likewise.
1670         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
1671         Make it static.
1673 2018-08-27  Martin Liska  <mliska@suse.cz>
1675         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
1676         cfun argument explicitly.
1677         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
1678         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
1679         function gimple_switch_default_bb.
1680         (convert_switch_statements):
1681         (expand_builtins):
1682         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
1683         * stmt.c (label_to_block_fn): Use label_to_block and pass
1684         cfun argument explicitly and use gimple_switch_label_bb.
1685         (expand_case): Likewise.
1686         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
1687         cfun argument explicitly. Likewise.
1688         (make_edges_bb): Likewise.
1689         (make_cond_expr_edges): Likewise.
1690         (get_cases_for_edge): Likewise.
1691         (make_gimple_switch_edges): Likewise.
1692         (label_to_block_fn): Likewise.
1693         (label_to_block): Likewise.
1694         (make_goto_expr_edges): Likewise.
1695         (make_gimple_asm_edges): Likewise.
1696         (main_block_label): Likewise.
1697         (group_case_labels_stmt): Likewise.
1698         (find_taken_edge_computed_goto): Likewise.
1699         (find_taken_edge_switch_expr): Likewise.
1700         (gimple_verify_flow_info): Likewise.
1701         (gimple_redirect_edge_and_branch): Likewise.
1702         (gimple_switch_label_bb): New function.
1703         (gimple_switch_default_bb): Likewise.
1704         (gimple_switch_edge): Likewise.
1705         (gimple_switch_default_edge): Likewise.
1706         * tree-cfg.h (label_to_block_fn): Remove and replace ...
1707         (label_to_block): ... with this.
1708         (gimple_switch_label_bb): New.
1709         (gimple_switch_default_bb): Likewise.
1710         (gimple_switch_edge): Likewise.
1711         (gimple_switch_default_edge): Likewise.
1712         * tree-cfgcleanup.c (convert_single_case_switch): Use
1713         new gimple functions and pass new argument to label_to_block.
1714         (cleanup_control_flow_bb):
1715         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
1716         cfun argument explicitly.
1717         (make_eh_edges): Likewise.
1718         (redirect_eh_dispatch_edge): Likewise.
1719         (lower_resx): Likewise.
1720         (lower_eh_dispatch): Likewise.
1721         (maybe_remove_unreachable_handlers): Likewise.
1722         (unsplit_eh): Likewise.
1723         (cleanup_empty_eh): Likewise.
1724         (verify_eh_edges): Likewise.
1725         (verify_eh_dispatch_edge): Likewise.
1726         * tree-ssa-dom.c (record_edge_info): Likewise.
1727         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
1728         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
1729         (thread_through_normal_block): Likewise.
1730         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
1731         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
1732         * tree-switch-conversion.c (switch_conversion::collect): Use new
1733         gimple functions.
1734         (switch_conversion::check_final_bb): Likewise.
1735         (switch_conversion::gather_default_values): Pass new argument
1736         to label_to_block.
1737         (switch_conversion::build_constructors): Likewise.
1738         (switch_decision_tree::compute_cases_per_edge): Use new
1739         gimple_switch_edge function.
1740         (switch_decision_tree::analyze_switch_statement): Pass new argument
1741         to label_to_block.
1742         (switch_decision_tree::try_switch_expansion): Use
1743         gimple_switch_default_edge.
1744         * tree-vrp.c (find_switch_asserts): Pass new argument
1745         to label_to_block.
1746         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
1747         (vr_values::simplify_switch_using_ranges): Likewise.
1749 2018-08-27  Richard Biener  <rguenther@suse.de>
1751         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
1752         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
1754         * tree-ssa-sccvn.h (struct vn_pval): New structure.
1755         (struct vn_nary_op_s): Add unwind_to member.  Add
1756         predicated_values flag and put result into a union together
1757         with a linked list of vn_pval.
1758         (struct vn_ssa_aux): Add name member to make maintaining
1759         a map of SSA name to vn_ssa_aux possible.  Remove no longer
1760         needed info, dfsnum, low, visited, on_sccstack, use_processed
1761         and range_info_anti_range_p members.
1762         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
1763         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
1764         New functions.
1765         (vn_valueize): New global.
1766         (vn_context_bb): Likewise.
1767         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
1768         VN_INFO_PTR_INFO): Remove.
1769         * tree-ssa-sccvn.c: ... (rewrite)
1770         (pass_fre::execute): For -O2+ initialize loops and run
1771         RPO VN in optimistic mode (iterating).  For -O1 and -Og
1772         run RPO VN in non-optimistic mode.
1773         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
1774         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
1775         * doc/invoke.texi (sccvn-max-scc-size): Remove.
1776         (rpo-vn-max-loop-depth): Document.
1777         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
1778         when valuezing the VUSE signals we walked out of the region.
1779         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
1780         (phi_translate): Set VN context block to use for availability
1781         lookup.
1782         (compute_avail): Likewise.
1783         (pre_valueize): New function.
1784         (pass_pre::execute): Adjust to the RPO VN API.
1786         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
1787         (propagate_constants_for_unrolling): Remove.
1788         (tree_unroll_loops_completely): Perform value-numbering
1789         on the unrolled bodies loop parent.
1791 2018-08-27  Richard Biener  <rguenther@suse.de>
1793         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
1794         for partial antic compute.
1796 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
1798         PR rtl-optimization/87065
1799         * combine.c (simplify_if_then_else): Formatting fix.
1800         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
1801         check.
1802         (known_cond): Don't return const_true_rtx for vector modes.  Use
1803         CONST0_RTX instead of const0_rtx.  Formatting fixes.
1805 2018-08-27  Martin Liska  <mliska@suse.cz>
1807         PR gcov-profile/87069
1808         * gcov.c (process_file): Record files already processed
1809         and warn about a file being processed multiple times.
1811 2018-08-27  Martin Liska  <mliska@suse.cz>
1813         PR driver/83193
1814         * config/aarch64/aarch64.c (aarch64_override_options_internal):
1815         Set default values for x_aarch64_*_string strings.
1816         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
1817         prefix.  For -mabi do not print '=ABI' in help and use
1818         <option_value> format for -msve-vector-bits and -moverride
1819         options.
1821 2018-08-26  Jeff Law  <law@redhat.com>
1823         * config/mips/frame-header-opt.c: Include "backend.h" rather than
1824         "cfg.h"
1826 2018-08-26  Marek Polacek  <polacek@redhat.com>
1828         PR c++/87029, Implement -Wredundant-move.
1829         * doc/invoke.texi: Document -Wredundant-move.
1831 2018-08-25  Martin Sebor  <msebor@redhat.com>
1833         PR tree-optimization/87059
1834         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
1835         to the same type as the other.
1836         * fold-const.c (fold_binary_loc): Assert expectation.
1838 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
1840         * config/darwin.c (machopic_legitimize_pic_address): Clean up
1841         extraneous parentheses, dead code section and formatting.
1843 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
1845         PR c++/87091
1846         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
1847         wide enough for jumps in the line-numbering to be visible.
1848         (layout::print_gap_in_line_numbering): New member function.
1849         (layout::calculate_line_spans): When using line numbering, merge
1850         line spans that are only 1 line apart.
1851         (diagnostic_show_locus): When printing line numbers, show gaps in
1852         line numbering directly, rather than printing headers.
1853         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
1854         line-numbering with multiple line spans.
1855         (selftest::test_fixit_insert_containing_newline_2): Add test of
1856         line-numbering, in which the spans are close enough to be merged.
1858 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
1860         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
1861         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
1862         * tree-vrp.c (range_is_nonnull): Remove.
1863         (range_includes_zero_p): Accept value_range instead of min/max.
1864         (extract_range_from_binary_expr_1): Do not early bail on
1865         POINTER_PLUS_EXPR.
1866         Use range_includes_zero_p instead of range_is_nonnull.
1867         (extract_range_from_unary_expr): Use range_includes_zero_p instead
1868         of range_is_nonnull.
1869         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
1870         special case VR_ANTI_RANGE.
1871         (vrp_finalize): Same.
1872         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
1873         instead of min/max.
1874         (range_is_nonnull): Remove.
1875         * vr-values.c (vrp_stmt_computes_nonzero): Use
1876         range_includes_zero_p instead of range_is_nonnull.
1877         (extract_range_basic): Pass value_range to range_includes_zero_p
1878         instead of range_is_nonnull.
1880 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
1882         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
1883         * emit-rtl.h (rtl_data): Remove return_bnd.
1884         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
1885         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
1886         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
1887         (POINTER_BOUNDS_MODE): Remove definition.
1888         (make_pointer_bounds_mode): Remove.
1889         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
1890         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
1891         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
1892         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
1893         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
1894         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
1895         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
1897         * config/i386/i386-modes.def (BND32, BND64): Remove.
1898         * config/i386/i386.c (dbx_register_map): Remove bound registers.
1899         (dbx64_register_map): Ditto.
1900         (svr4_dbx_register_map): Ditto.
1901         (indirect_thunk_bnd_needed): Remove.
1902         (indirect_thunks_bnd_used): Ditto.
1903         (indirect_return_bnd_needed): Ditto.
1904         (indirect_return_via_cx_bnd): Ditto.
1905         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
1906         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
1907         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
1908         (output_indirect_thunk_function): Remove handling of
1909         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
1910         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
1911         (ix86_save_reg): Remove handling of crtl->return_bnd.
1912         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
1913         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
1914         and UNSPEC_BNDLX_ADDR.
1915         (ix86_output_indirect_branch_via_reg): Remove handling of
1916         indirect_thunk_prefix_bnd.
1917         (ix86_output_indirect_branch_via_push): Ditto.
1918         (ix86_output_function_return): Ditto.
1919         (ix86_output_indirect_function_return): Ditto.
1920         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
1921         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
1922         (CALL_USED_REGISTERS): Ditto.
1923         (REG_ALLOC_ORDER): Update for removal of bound registers.
1924         (HI_REGISTER_NAMES): Ditto.
1925         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
1926         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
1927         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
1928         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
1929         (FIRST_PSEUDO_REG): Update.
1930         (BND): Remove mode iterator.
1931         * config/i386/predicates.md (bnd_mem_operator): Remove.
1933 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
1935         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
1936         vectors.
1938 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
1940         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
1941         the case in which the permute needs only a single element and
1942         repeats for every vector of the result.  Extend that case to
1943         handle variable-length vectors.
1944         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
1946 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
1948         PR debug/79342
1949         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
1950         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
1952 2018-08-24  Richard Biener  <rguenther@suse.de>
1954         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
1955         bb_flags_allocated members.
1956         (auto_flag): New RAII class for allocating flags.
1957         (auto_edge_flag): New RAII class for allocating edge flags.
1958         (auto_bb_flag): New RAII class for allocating bb flags.
1959         * cfgloop.c (verify_loop_structure): Allocate temporary edge
1960         flag dynamically.
1961         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
1962         in favor of temporarily allocated BB flag.
1963         * hsa-brig.c: Re-order includes.
1964         * hsa-dump.c: Likewise.
1965         * hsa-regalloc.c: Likewise.
1966         * print-rtl.c: Likewise.
1967         * profile-count.c: Likewise.
1969 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
1971         PR target/86989
1972         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1973         the TOC register.
1975 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
1977         PR 87073/bootstrap
1978         * wide-int-range.cc (wide_int_range_div): Do not ignore result
1979         from wide_int_range_multiplicative_op.
1981 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1983         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
1984         "permutaion".
1986 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
1988         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
1989         to 'expanded'.
1991 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
1993         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
1994         full GENERIC expression used for replacement.
1996 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
1998         * tree-vrp.c (abs_extent_range): Remove.
1999         (extract_range_into_wide_ints): Pass wide ints by reference.
2000         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
2001         Pass wide ints by reference in all calls to
2002         extract_range_into_wide_ints.
2003         * wide-int-range.cc (wide_int_range_div): New.
2004         * wide-int-range.h (wide_int_range_div): New.
2005         (wide_int_range_includes_zero_p): New.
2006         (wide_int_range_zero_p): New.
2008 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
2010         * config/aarch64/aarch64.md (arches): New enum.
2011         (arch): New enum attr.
2012         (arch_enabled): New attr.
2013         (enabled): Now uses arch_enabled only.
2014         (simd, sve, fp16): Removed attribute.
2015         (fp): Attr now defined in terms of 'arch'.
2016         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
2017         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
2018         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
2019         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
2020         attributes into 'arch'.
2021         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
2022         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
2023         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
2024         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
2025         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
2026         'simd' attribute into 'arch'.
2027         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
2028         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
2029         Convert use of 'fp' attribute to 'arch'.
2030         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
2031         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
2032         into 'arch'.
2033         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
2034         (different modes) Merge 'fp' and 'simd' into 'arch'.
2035         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
2036         'simd' into 'arch'.
2038 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
2040         PR rtl-optimization/87026
2041         * expmed.c (canonicalize_comparison): If we can no longer create
2042         pseudoregisters, don't.
2044 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
2046         PR target/86951
2047         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2048         prototype.
2049         * config/arm/arm.c (speculation_barrier_libfunc): New static
2050         variable.
2051         (arm_init_libfuncs): Initialize it.
2052         (arm_emit_speculation_barrier): New function.
2053         * config/arm/arm.md (speculation_barrier): Call
2054         arm_emit_speculation_barrier for architectures that do not have
2055         DSB or ISB.
2056         (speculation_barrier_insn): Only match on Armv7 or later.
2058 2018-08-23  Richard Biener  <rguenther@suse.de>
2060         PR middle-end/87024
2061         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
2062         calls.
2064 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
2066         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
2067         of single-vector TBLs.
2068         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
2069         one input is given.
2071 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
2073         PR target/85910
2074         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
2075         aarch64_evpc_tbl guard.
2077 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2079         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
2080         behaviour.
2082 2018-08-22  Martin Sebor  <msebor@redhat.com>
2084         PR middle-end/87052
2085         * tree-pretty-print.c (pretty_print_string): Add argument.
2086         (dump_generic_node): Call to pretty_print_string with string size.
2088 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
2090         PR rtl-optimization/86771
2091         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
2092         of two SETs into those two SETs, one to be placed at i2, if that SETs
2093         destination is modified between i2 and i3.
2095 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2097         PR tree-optimization/86725
2098         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
2099         function.
2100         (vect_analyze_scalar_cycles_1): Check it.
2102 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2104         PR tree-optimization/86725
2105         * tree-vect-loop.c (vect_is_simple_reduction): When treating
2106         an outer loop phi as a double reduction, make sure that the
2107         single user of the phi result is an inner loop phi.
2109 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2111         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
2112         grouped stores with gaps to a strided group.
2114 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
2116         * tree-vect-stmts.c (get_group_load_store_type)
2117         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
2118         first statement in a group.
2120 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2122         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
2123         the sequence used in gcc/gcc.c.
2125 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2127         PR other/704
2128         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
2129         building it.
2131 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2133         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
2134         Darwin10-specific unwinder-shim.
2135         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
2136         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC): 
2137         New to cater for Darwin10 Rosetta.
2139 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2141         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
2142         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
2144 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
2146         PR bootstrap/81033
2147         PR target/81733
2148         PR target/52795
2149         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
2150         (dwarf2out_switch_text_section): Generate a local label for the
2151         second function sub-section and apply it as the second FDE start
2152         label.
2153         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
2154         second sub-section start.
2156 2018-08-22  Richard Biener  <rguenther@suse.de>
2158         PR tree-optimization/86988
2159         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
2161 2018-08-22  Richard Biener  <rguenther@suse.de>
2163         PR tree-optimization/86945
2164         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
2166 2018-08-22  Alexandre Oliva <oliva@adacore.com>
2168         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
2169         a comment about how uses of r2 for .sdata2 come about.
2171 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
2173         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
2175 2018-08-21  Marek Polacek  <polacek@redhat.com>
2177         PR c++/86981, Implement -Wpessimizing-move.
2178         * doc/invoke.texi: Document -Wpessimizing-move.
2180 2018-08-21  Jan Hubicka  <jh@suse.cz>
2182         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
2183         * tree.h (is_redundant_typedef): Remove.
2184         * dwarf2out.c (is_redundant_typedef): Turn into static function.
2186 2018-08-21  Jan Hubicka  <jh@suse.cz>
2188         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
2189         when possible.
2191 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
2193         * expmed.c (extract_low_bits): Reject invalid subregs early.
2195 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2197         PR middle-end/86121
2198         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
2199         behaviour.
2201 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2203         * config/vxworks.h: Guard vxworks_asm_out_constructor and
2204         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
2205         * config/vxworks.c: Likewise.
2207 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2209         * config/vxworks.c: Set targetm.have_ctors_dtors
2210         if HAVE_INITFINI_ARRAY_SUPPORT.
2211         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
2212         if HAVE_INITFINI_ARRAY_SUPPORT.
2214 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
2216         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
2217         default search path for VxWorks < 7.
2219 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2221         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
2222         (get_format_string): Refer to c_getstr.
2224 2018-08-21  Tom de Vries  <tdevries@suse.de>
2226         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2227         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2228         (debuginfo_early_stop): Declare.
2229         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
2230         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
2231         (debuginfo_early_stop): New function.
2232         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
2233         and debuginfo_early_stop.
2234         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
2235         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
2236         (general_init): Call debuginfo_early_init.
2237         (finalize): Call debuginfo_fini.
2238         (do_compile): Call debuginfo_init.
2239         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
2240         -fdump-early-debug.
2241         (@item -fdump-debug, @item -fdump-earlydebug): Add.
2243 2018-08-21  Tom de Vries  <tdevries@suse.de>
2245         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
2246         flag_dump_noaddr and flag_dump_unnumbered.
2248 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
2250         * wide-int-range.cc (wide_int_range_abs): New.
2251         (wide_int_range_order_set): Rename from wide_int_range_min_max.
2252         * wide-int-range.h (wide_int_range_abs): New.
2253         (wide_int_range_min_max): New.
2254         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
2255         case to call wide_int_range_abs.
2256         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
2257         (extract_range_from_abs_expr): Delete.
2259 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
2261         PR target/87033
2262         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2263         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
2264         for indexed loads.
2266 2018-08-20  Nathan Sidwell  <nathan@acm.org>
2267             Jeff Law <law@redhat.com>
2269         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
2270         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
2272 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
2274         PR other/84889
2275         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
2276         (decl_attributes): Likewise.
2277         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
2278         instance.
2279         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2280         * diagnostic-core.h (class auto_diagnostic_group): New class.
2281         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
2282         (diagnostic_report_diagnostic): Handle the first diagnostics within
2283         a group.
2284         (emit_diagnostic): Add auto_diagnostic_group instance.
2285         (inform): Likewise.
2286         (inform_n): Likewise.
2287         (warning): Likewise.
2288         (warning_at): Likewise.
2289         (warning_n): Likewise.
2290         (pedwarn): Likewise.
2291         (permerror): Likewise.
2292         (error): Likewise.
2293         (error_n): Likewise.
2294         (error_at): Likewise.
2295         (sorry): Likewise.
2296         (fatal_error): Likewise.
2297         (internal_error): Likewise.
2298         (internal_error_no_backtrace): Likewise.
2299         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
2300         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
2301         * diagnostic.h (struct diagnostic_context): Add fields
2302         "diagnostic_group_nesting_depth",
2303         "diagnostic_group_emission_count", "begin_group_cb",
2304         "end_group_cb".
2305         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
2306         Add auto_diagnostic_group instance(s).
2307         (find_explicit_erroneous_behavior): Likewise.
2308         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
2309         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
2310         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
2311         (gimplify_va_arg_expr): Likewise.
2312         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
2313         (HSA_SORRY_AT): Likewise.
2314         * ipa-devirt.c (compare_virtual_tables): Likewise.
2315         (warn_odr): Likewise.
2316         * multiple_target.c (expand_target_clones): Likewise.
2317         * opts-common.c (cmdline_handle_error): Likewise.
2318         * reginfo.c (globalize_reg): Likewise.
2319         * substring-locations.c (format_warning_n_va): Likewise.
2320         * tree-inline.c (expand_call_inline): Likewise.
2321         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
2322         * tree-ssa-loop-niter.c
2323         (do_warn_aggressive_loop_optimizations): Likewise.
2324         * tree-ssa-uninit.c (warn_uninit): Likewise.
2325         * tree.c (warn_deprecated_use): Likewise.
2327 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
2329         PR target/87014
2330         * config/i386/i386.md (eh_return): Always update EH return
2331         address in word_mode.
2333 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
2335         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
2336         TARGET_SPLIT_COMPLEX_ARG is defined.
2338 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2340         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
2342 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2344         PR target/86984
2345         * expr.c (expand_assignment): Assert that bitpos is positive.
2346         (store_field): Likewise
2347         (expand_expr_real_1): Make sure that bitpos is positive.
2348         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
2349         integer overflow.
2351 2018-08-20  Nathan Sidwell  <nathan@acm.org>
2353         * Makefile.in (CPP_ID_DATA_H): Delete.
2354         (CPP_INTERNAL_H): Don't add it.
2355         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
2356         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
2358 2018-08-20  Richard Biener  <rguenther@suse.de>
2360         PR tree-optimization/78655
2361         * tree-vrp.c (extract_range_from_binary_expr_1): Make
2362         pointer + offset nonnull if either operand is nonnull work.
2364 2018-08-20  Tom de Vries  <tdevries@suse.de>
2366         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
2367         unless the referenced die describes the added property using
2368         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
2369         Otherwise, add a DW_AT_location to the referenced die.
2371 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
2373         PR target/86994
2374         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
2375         register_operand when calling ix86_set_reg_reg_cost.
2376         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
2377         Set *total to 0 for operands that satisfy x86_64_immediate_operand
2378         predicate and to 1 otherwise.
2380 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
2382         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
2383         emit a diagnostic that it is not supported and reset the option.
2384         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
2385         supported and consume the option.  (ASM_FINAL_SPEC): New.
2387 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
2389         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
2390         a sentence.
2392 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
2394         C-SKY port: Documentation
2396         * doc/extend.texi (C-SKY Function Attributes): New section.
2397         * doc/invoke.texi (Option Summary): Add C-SKY options.
2398         (C-SKY Options): New section.
2399         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
2401 2018-08-17  Jojo  <jijie_rong@c-sky.com>
2402             Huibin Wang  <huibin_wang@c-sky.com>
2403             Sandra Loosemore  <sandra@codesourcery.com>
2404             Chung-Lin Tang  <cltang@codesourcery.com>
2406         C-SKY port: Backend implementation
2408         * config/csky/*: New.
2409         * common/config/csky/*: New.
2411 2018-08-17  Jojo  <jijie_rong@c-sky.com>
2412             Huibin Wang  <huibin_wang@c-sky.com>
2413             Sandra Loosemore  <sandra@codesourcery.com>
2414             Chung-Lin Tang  <cltang@codesourcery.com>
2415             Andrew Jenner  <andrew@codesourcery.com>
2417         C-SKY port: Configury
2419         * config.gcc (csky-*-*): New.
2420         * configure.ac: Add csky to targets for dwarf2 debug_line support.
2421         * configure: Regenerated.
2423 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
2425         * dump-context.h: Include "dumpfile.h".
2426         (dump_context::dump_printf_va): Convert final param from va_list
2427         to va_list *.  Convert from ATTRIBUTE_PRINTF to
2428         ATTRIBUTE_GCC_DUMP_PRINTF.
2429         (dump_context::dump_printf_loc_va): Likewise.
2430         * dumpfile.c: Include "stringpool.h".
2431         (make_item_for_dump_printf_va): Delete.
2432         (make_item_for_dump_printf): Delete.
2433         (class dump_pretty_printer): New class.
2434         (dump_pretty_printer::dump_pretty_printer): New ctor.
2435         (dump_pretty_printer::emit_items): New member function.
2436         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
2437         function.
2438         (dump_pretty_printer::emit_item): New member function.
2439         (dump_pretty_printer::stash_item): New member function.
2440         (dump_pretty_printer::format_decoder_cb): New member function.
2441         (dump_pretty_printer::decode_format): New member function.
2442         (dump_context::dump_printf_va): Reimplement in terms of
2443         dump_pretty_printer.
2444         (dump_context::dump_printf_loc_va): Convert final param from va_list
2445         to va_list *.
2446         (dump_context::begin_scope): Reimplement call to
2447         make_item_for_dump_printf.
2448         (dump_printf): Update for change to dump_printf_va.
2449         (dump_printf_loc): Likewise.
2450         (selftest::test_capture_of_dump_calls): Convert "stmt" from
2451         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
2452         with %T, %E, and %G.
2453         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
2454         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
2455         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
2456         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
2457         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
2458         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
2459         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
2460         within a dump_printf_loc call to "%wu".
2461         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
2462         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
2463         missing space after "=".
2464         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
2465         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
2466         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
2467         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
2468         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
2469         duplicate "vectorized" from message.
2471 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
2473         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
2474         polyNxK_t element's TYPE_STRING_FLAG.
2476 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
2478         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
2479         (they were unnamed before).  Fix comments.
2481 2018-08-17  Nathan Sidwell  <nathan@acm.org>
2483         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
2485 2018-08-17  Richard Biener  <rguenther@suse.de>
2487         PR tree-optimization/86841
2488         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
2490 2018-08-17  Martin Liska  <mliska@suse.cz>
2492         * common.opt: Remove Warn, Init and Report for options with
2493         Ignore/Deprecated flag. Warning is done automatically for
2494         Deprecated flags.
2495         * config/i386/i386.opt: Likewise.
2496         * config/ia64/ia64.opt: Likewise.
2497         * config/rs6000/rs6000.opt: Likewise.
2498         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
2499         Remove usage of flag_check_pointer_bounds.
2500         * lto-wrapper.c (merge_and_complain): Do not handle
2501         OPT_fcheck_pointer_bounds.
2502         (append_compiler_options): Likewise.
2503         * opt-functions.awk: Do not handle Deprecated.
2504         * optc-gen.awk: Check that Var, Report and Init are not
2505         used for an option with Ignore/Deprecated flag.
2506         * opts-common.c (decode_cmdline_option): Do not report
2507         CL_ERR_DEPRECATED.
2508         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
2509         options.
2510         * opts.h (struct cl_option): Remove cl_deprecated flag.
2511         (CL_ERR_DEPRECATED): Remove error enum value.
2513 2018-08-17  Richard Biener  <rguenther@suse.de>
2515         PR middle-end/86505
2516         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
2517         across a va-arg-pack using call adjust its return value accordingly.
2519 2018-08-16  Martin Sebor  <msebor@redhat.com>
2521         PR tree-optimization/86853
2522         * gimple-ssa-sprintf.c (struct format_result): Rename member.
2523         (struct fmtresult): Add member and initialize it in ctors.
2524         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
2525         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
2526         (format_directive): Set POSUNDER4K when MAYFAIL is set.
2527         (parse_directive): Handle %C same as %c.
2528         (sprintf_dom_walker::compute_format_length): Adjust.
2529         (is_call_safe): Adjust.
2531 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2533         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
2534         for determining how to count the elements.
2535         * builtins.h (c_strlen): Adjust prototype.
2536         * expr.c (string_constant): Add new parameter mem_size.
2537         Set *mem_size appropriately.
2538         * expr.h (string_constant): Adjust protoype.
2539         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
2540         * gimple-fold.h (get_range_strlen): Adjust prototype.
2541         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
2542         (format_string): Call get_string_length with eltsize.
2544 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
2546         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
2547         to emit the span, rather than setting it as the prefix.
2549 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
2551         * diagnostic-show-locus.c (layout::start_annotation_line): Add
2552         "margin_char" parameter, defaulting to space.  Use it in place
2553         of pp_space for the initial part of the margin.
2554         (layout::print_leading_fixits): Use '+' when filling the margin
2555         of line-insertion fix-it hints.
2557 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2559         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
2560         Delete.
2562 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2564         * config/rs6000/altivec.md: Don't set length attribute to the default
2565         value.
2566         * config/rs6000/darwin.md: Ditto.
2567         * config/rs6000/dfp.md: Ditto.
2568         * config/rs6000/htm.md: Ditto.
2569         * config/rs6000/rs6000.md: Ditto.
2570         * config/rs6000/sync.md: Ditto.
2571         * config/rs6000/vsx.md: Ditto.
2573 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2575         * config/rs6000/altivec.md: Don't set length attribute to the default
2576         value, for branch instructions.
2577         * config/rs6000/darwin.md: Ditto.
2578         * config/rs6000/rs6000.md: Ditto.
2580 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
2582         * config/rs6000/rs6000.md (length): Always define as const_int 4.
2583         (unnamed conditional branch define_insn): Set length to 4 or 8
2584         depending on offset.
2585         (<bd>_<mode>): Similar, for alternative 0.
2586         (<bd>tf_<mode>): Ditto.
2588 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
2590         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
2592 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
2594         * doc/rtl.texi: Replace old RTX class names with new names.
2597 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
2599         * expmed.h (canonicalize_comparison): New declaration.
2600         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
2601         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
2602         * optabs.c (prepare_cmp_insn): Likewise.
2603         * rtl.h (unsigned_condition_p): New function which checks if a
2604         comparison operator is unsigned.
2606 2018-08-16  Nathan Sidwell  <nathan@acm.org>
2608         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
2609         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
2611 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
2613         PR target/84711
2614         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
2615         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
2616         (mov<mov>): ..this and enable unconditionally.
2618 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
2620         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
2622 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
2624         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
2625         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
2626         "Common" with "Target".
2628 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
2630         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
2631         * doc/invoke.texi (mmitigate-rop): Remove.
2632         * config/i386/i386.c: Do not include "regrename.h".
2633         (ix86_rop_should_change_byte_p, reg_encoded_number)
2634         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
2635         Remove.
2636         (ix86_reorg): Remove call to ix86_mitigate_rop.
2637         * config/i386/i386.md (attr "modrm_class"): Remove.
2638         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
2639         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
2640         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
2642 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
2644         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
2645         allow folding of mergeh() and mergel() for the float and double types.
2646         (fold_mergehl_helper): Rework to handle building a permute tree
2647         for float vectors.
2649 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
2651         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
2652         for TARGET_SSE.
2654 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
2656         * common.opt (fdiagnostics-show-labels): New option.
2657         * diagnostic-show-locus.c (class layout_range): Add field
2658         "m_label".
2659         (class layout): Add field "m_show_labels_p".
2660         (layout_range::layout_range): Add param "label" and use it to
2661         initialize m_label.
2662         (make_range): Pass in NULL for new "label" param of layout_range's
2663         ctor.
2664         (layout::layout): Initialize m_show_labels_p.
2665         (layout::maybe_add_location_range): Pass in loc_range->m_label
2666         when constructing layout_range instances.
2667         (struct line_label): New struct.
2668         (layout::print_any_labels): New member function.
2669         (layout::print_line): Call it if label-printing is enabled.
2670         (selftest::test_one_liner_labels): New test.
2671         (selftest::test_diagnostic_show_locus_one_liner): Call it.
2672         * diagnostic.c (diagnostic_initialize): Initialize
2673         context->show_labels_p.
2674         * diagnostic.h (struct diagnostic_context): Add field
2675         "show_labels_p".
2676         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2677         -fno-diagnostics-show-labels.
2678         * dwarf2out.c (gen_producer_string): Add
2679         OPT_fdiagnostics_show_labels to the ignored options.
2680         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
2681         param.
2682         (gcc_rich_location::maybe_add_expr): Likewise.
2683         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
2684         label" param, defaulting to NULL.
2685         (gcc_rich_location::add_expr): Add "label" param.
2686         (gcc_rich_location::maybe_add_expr): Likewise.
2687         (class text_range_label): New class.
2688         (class range_label_for_type_mismatch): New class.
2689         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
2690         of format_warning_va.
2691         (fmtwarn_n): Likewise for new params of format_warning_n_va.
2692         * lto-wrapper.c (merge_and_complain): Add
2693         OPT_fdiagnostics_show_labels to the "pick one setting" options.
2694         (append_compiler_options): Likewise to the dropped options.
2695         (append_diag_options): Likewise to the passed-on options.
2696         * opts.c (common_handle_option): Handle the new option.
2697         * selftest-diagnostic.c
2698         (test_diagnostic_context::test_diagnostic_context): Enable
2699         show_labels_p.
2700         * substring-locations.c: Include "gcc-rich-location.h".
2701         (format_warning_n_va): Add "fmt_label" and "param_label" params
2702         and use them as appropriate.
2703         (format_warning_va): Add "fmt_label" and "param_label" params,
2704         passing them on to format_warning_n_va.
2705         (format_warning_at_substring): Likewise.
2706         (format_warning_at_substring_n): Likewise.
2707         * substring-locations.h (format_warning_va): Add "fmt_label" and
2708         "param_label" params.
2709         (format_warning_n_va): Likewise.
2710         (format_warning_at_substring): Likewise.
2711         (format_warning_at_substring_n): Likewise.
2712         * toplev.c (general_init): Initialize global_dc->show_labels_p.
2714 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
2716         PR testsuite/86519
2717         * builtins.c (expand_builtin_memcmp): Do not expand the call
2718         when overflow is detected.
2720 2018-08-15  Martin Sebor  <msebor@redhat.com>
2722         PR tree-optimization/71625
2723         * config/aarch64/aarch64-builtins.c
2724         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
2726 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
2728         * config/s390/s390.c (s390_reorg): Remove loop.
2730 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
2732         * config/darwin.c
2733          (darwin_function_switched_text_sections): Delete.
2734         * gcc/config/darwin.h
2735          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
2737 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
2739         PR target/81685
2740         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
2741         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
2742         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
2744 2018-08-15  Martin Liska  <mliska@suse.cz>
2746         PR tree-optimization/86925
2747         * predict.c (expr_expected_value_1): When taking
2748         later predictor, assign also probability.
2749         Use fold_build2_initializer_loc in order to fold
2750         the expression in -frounding-math.
2752 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
2754         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
2755         patterns.
2756         (expand_vec_perm_1): Try the new method.
2758 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
2760         PR target/86547
2761         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
2762         Check whether lra_live_max_point is 0 before dividing.
2764 2018-08-14  Martin Sebor  <msebor@redhat.com>
2766         PR tree-optimization/86650
2767         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
2768         (vrp_prop::check_mem_ref): Same.
2770 2018-08-13  Liu Hao <lh_mouse@126.com>
2772         * pretty-print.c (eat_esc_sequence): Swap the foreground and
2773         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
2774         and clear it thereafter, as it only works for DBCS.
2776 2018-08-13  Liu Hao <lh_mouse@126.com>
2778         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
2779         handle returned by _get_osf_handle().
2781 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
2783         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
2784         for folding vec_perm.
2786 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
2788         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
2789         Add support for gimple-folding of vec_pack() and vec_unpack()
2790         intrinsics.
2792 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
2794         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
2795         vec_xst variants to the list.
2796         (rs6000_gimple_fold_builtin): Add support for folding unaligned
2797         vector loads and stores.
2799 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
2801         * config.gcc (rs6000-ibm-aix4.x): Delete.
2802         (rs6000-ibm-aix5.1): Delete.
2803         (rs6000-ibm-aix5.2): Delete.
2804         (rs6000-ibm-aix5.3): Delete.
2805         * config/rs6000/aix43.h: Delete.
2806         * config/rs6000/aix51.h: Delete.
2807         * config/rs6000/aix52.h: Delete.
2808         * config/rs6000/t-aix43: Delete.
2810 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
2812         * config/s390/s390.c (s390_decompose_constant_pool_ref):
2813         New function.
2814         (s390_decompose_address): Factor out constant pool ref
2815         decomposition.
2817 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
2819         * config/nds32/nds32-predicates.c
2820         (nds32_can_use_bclr_p): Change return type as bool.
2821         (nds32_can_use_bset_p): Ditto.
2822         (nds32_can_use_btgl_p): Ditto.
2823         (nds32_can_use_bitci_p): Ditto.
2824         * config/nds32/nds32-protos.h
2825         (nds32_can_use_bclr_p): Change declaration.
2826         (nds32_can_use_bset_p): Ditto.
2827         (nds32_can_use_btgl_p): Ditto.
2828         (nds32_can_use_bitci_p): Ditto.
2830 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
2832         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
2833         Support -msched-prolog-epilog option.
2834         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
2836 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
2838         * common/config/nds32/nds32-common.c
2839         (nds32_option_optimization_table): Enalbe -malways-align.
2841 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
2843         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
2844         extra_headers.
2845         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
2846         OPT_misr_secure_ case.
2847         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
2848         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
2849         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
2850         secure attribute.
2851         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
2852         (nds32_isr_info): New field security_level.
2853         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
2854         * config/nds32/nds32.md (return_internal): Consider critical attribute.
2855         * config/nds32/nds32.opt (misr-secure): New option.
2856         * config/nds32/nds32_init.inc: New file.
2857         * config/nds32/nds32_isr.h: New file.
2859 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
2861         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
2862         Update comment for atomic instructions.
2863         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
2864         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
2865         Remove.
2866         (atomic_loaddi): Revise fence expansion to only emit fence prior to
2867         load for __ATOMIC_SEQ_CST model.
2868         (atomic_loaddi_1): Remove float register target.
2869         (atomic_storedi): Handle CONST_INT values.
2870         (atomic_storedi_1): Remove float register source.  Add special case
2871         for zero value.
2872         (memory_barrier): New expander and insn.
2874 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
2876         PR tree-optimization/86835
2877         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
2878         new_stmt after def_gsi, make sure to insert new_square_stmt after
2879         that stmt, not 2 stmts before it.
2881 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
2883         PR target/82418
2884         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2885         instead of SWI48.
2887 2018-08-10  Martin Liska  <mliska@suse.cz>
2889         PR target/83610
2890         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2891         function type.
2892         * builtins.c (expand_builtin_expect_with_probability):
2893         New function.
2894         (expand_builtin_expect_with_probability): New function.
2895         (build_builtin_expect_predicate): Add new argumnet probability
2896         for BUILT_IN_EXPECT_WITH_PROBABILITY.
2897         (fold_builtin_expect):
2898         (fold_builtin_2):
2899         (fold_builtin_3):
2900         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
2901         * builtins.h (fold_builtin_expect): Set new argument.
2902         * doc/extend.texi: Document __builtin_expect_with_probability.
2903         * doc/invoke.texi: Likewise.
2904         * gimple-fold.c (gimple_fold_call): Pass new argument.
2905         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
2906         also BUILT_IN_EXPECT_WITH_PROBABILITY.
2907         * predict.c (get_predictor_value): New function.
2908         (expr_expected_value): Add new argument probability. Assume
2909         that predictor and probability are always non-null.
2910         (expr_expected_value_1): Likewise.  For __builtin_expect and
2911         __builtin_expect_with_probability set probability.  Handle
2912         combination in binary expressions.
2913         (tree_predict_by_opcode): Simplify code by simply calling
2914         get_predictor_value.
2915         (pass_strip_predict_hints::execute): Add handling of
2916         BUILT_IN_EXPECT_WITH_PROBABILITY.
2917         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
2918         new predictor.
2919         * tree.h (DECL_BUILT_IN_P): New function.
2921 2018-08-10  Martin Liska  <mliska@suse.cz>
2923         PR tree-optimization/85799
2924         * passes.def: Add argument for pass_strip_predict_hints.
2925         * predict.c (class pass_strip_predict_hints): Add new argument
2926         early_p.
2927         (strip_predictor_early): New function.
2928         (pass_strip_predict_hints::execute): Call the function to
2929         strip predictors.
2930         (strip_predict_hints): New function.
2931         * predict.def: Fix comment.
2933 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2935         * Makefile.in: Clarify which tm.texi to copy over to assert the
2936         right to grant a GFDL license for all.
2938 2018-08-09  Jeff Law  <law@redhat.com>
2940         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
2941         unused variable.
2943 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
2945         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
2946         prototype.
2948 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
2950         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
2951         reductions for variable-length vectors.
2953 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
2955         PR other/84889
2956         * common.opt (fdiagnostics-show-line-numbers): New option.
2957         * diagnostic-show-locus.c (class layout): Add fields
2958         "m_show_line_numbers_p" and "m_linenum_width";
2959         (num_digits): New function.
2960         (test_num_digits): New function.
2961         (layout::layout): Initialize new fields.  Update m_x_offset
2962         logic to handle any left margin.
2963         (layout::print_source_line): Print line number when requested.
2964         (layout::start_annotation_line): New member function.
2965         (layout::print_annotation_line): Call it.
2966         (layout::print_leading_fixits): Likewise.
2967         (layout::print_trailing_fixits): Likewise.  Update calls to
2968         move_to_column for new parameter.
2969         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
2970         it to potentially call start_annotation_line.
2971         (layout::show_ruler): Call start_annotation_line.
2972         (selftest::test_line_numbers_multiline_range): New selftest.
2973         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
2974         and selftest::test_line_numbers_multiline_range.
2975         * diagnostic.c (diagnostic_initialize): Initialize
2976         show_line_numbers_p.
2977         * diagnostic.h (struct diagnostic_context): Add field
2978         "show_line_numbers_p".
2979         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
2980         -fno-diagnostics-show-line-numbers.
2981         * dwarf2out.c (gen_producer_string): Add
2982         OPT_fdiagnostics_show_line_numbers to the ignored options.
2983         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
2984         one setting" options.
2985         (append_compiler_options): Likewise to the dropped options.
2986         (append_diag_options): Likewise to the passed-on options.
2987         * opts.c (common_handle_option): Handle the new option.
2988         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
2990 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
2992         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
2993         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
2994         third argument of type "const signed char" to descriptions of
2995         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
2996         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
2997         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
2998         __builtin_bcdsub_ov functions.
3000 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3002         PR tree-optimization/86858
3003         * tree-vect-loop.c (vect_is_simple_reduction): Restore
3004         flow_bb_inside_loop_p calls.
3006 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
3008         PR tree-optimization/86871
3009         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
3010         instead of gimple_assign_lhs.
3012 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
3014         PR target/86887
3015         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
3016         register constraint to operand 0.
3017         (add<mode>3_carryinC): Likewise.
3018         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
3020 2018-08-09  Martin Liska  <mliska@suse.cz>
3022         PR c/86895
3023         * common.opt: Remove extra line.
3025 2018-08-09  Martin Liska  <mliska@suse.cz>
3027         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
3028         at the end of a line, make first letter capital and end up
3029         a sentence with a dot.
3030         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
3031         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
3032         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
3033         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
3034         (PARAM_MAX_ISL_OPERATIONS): Likewise.
3035         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
3036         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
3037         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
3038         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
3039         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
3040         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
3041         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
3042         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
3043         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
3044         (PARAM_TREE_REASSOC_WIDTH): Likewise.
3045         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
3046         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
3047         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
3049 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
3051         PR target/84332
3052         * config/s390/s390.c (s390_option_override_internal): Reduce the
3053         stack-clash-protection-probe-interval param if it would be too big
3054         for z900.
3056 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
3058         PR target/46179
3059         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
3060         * config/m68k/m68k.c (handle_move_double): Don't call
3061         m68k_final_prescan_insn.
3062         (m68k_adjust_decorated_operand): Renamed from
3063         m68k_final_prescan_insn, remove first and third operand and
3064         simplify.
3065         (print_operand): Call it.
3066         (print_operand_address): Call it.
3068 2018-08-08  Nathan Sidwell  <nathan@acm.org>
3070         * diagnostic.c (diagnostic_report_current_module): Use
3071         linemap_included_from & linemap_included_from_linemap.
3073 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
3075         * config/aarch64/aarch64-cores.def: Add phecda core.
3076         * config/aarch64/aarch64-tune.md: Regenerate.
3077         * doc/invoke.texi: Add phecda core.
3079 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
3081         PR target/85295
3082         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
3083         definitions.
3084         * config/s390/s390.md ("movti"): Add more alternatives for
3085         constant to GPR copies.
3087 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
3089         * config/s390/s390.c: Fix whitespace damage throughout the file.
3090         * config/s390/s390.h: Likewise.
3091         * config/s390/tpf.h: Likewise.
3093 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
3095         * config/s390/s390.c (s390_loadrelative_operand_p):
3096         Remove TARGET_CPU_ZARCH usages.
3097         (s390_rtx_costs): Likewise.
3098         (s390_legitimate_constant_p): Likewise.
3099         (s390_cannot_force_const_mem): Likewise.
3100         (legitimate_reload_constant_p): Likewise.
3101         (s390_preferred_reload_class): Likewise.
3102         (legitimize_pic_address): Likewise.
3103         (legitimize_tls_address): Likewise.
3104         (s390_split_branches): Removed.
3105         (s390_add_execute): Removed.
3106         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
3107         (s390_mainpool_start): Likewise.
3108         (s390_mainpool_finish): Likewise.
3109         (s390_mainpool_cancel): Removed.
3110         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
3111         (s390_chunkify_cancel): Likewise.
3112         (s390_return_addr_rtx): Likewise.
3113         (s390_register_info): Remove split_branches_pending_p uages.
3114         (s390_optimize_register_info): Likewise.
3115         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
3116         split_branches_pending_p usages.
3117         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
3118         (s390_load_got): Likewise.
3119         (s390_expand_split_stack_prologue): Likewise.
3120         (output_asm_nops): Likewise.
3121         (s390_function_profiler): Likewise.
3122         (s390_emit_call): Likewise.
3123         (s390_conditional_register_usage): Likewise.
3124         (s390_optimize_prologue): Likewise.
3125         (s390_reorg): Remove TARGET_CPU_ZARCH and
3126         split_branches_pending_p usages.
3127         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
3128         usages.
3129         (s390_output_indirect_thunk_function): Likewise.
3130         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
3131         (TARGET_CPU_ZARCH_P): Removed.
3132         (struct machine_function): Remove split_branches_pending_p.
3133         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
3135 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
3137         * common/config/s390/s390-common.c (processor_flags_table):
3138         Remove flags.
3139         * config.gcc: Remove with_arch/with_tune support.
3140         * config/s390/2064.md: Remove cpu attribute comparisons.
3141         * config/s390/driver-native.c (s390_host_detect_local_cpu):
3142         Remove MTN.
3143         * config/s390/linux.h (ASM_SPEC):
3144         Remove -march support.
3145         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
3146         Use a table to get an arch level.
3147         * config/s390/s390-opts.h (enum processor_type):
3148         Remove enum values.
3149         * config/s390/s390.c
3150         (processor_table): Remove entries, add arch_level values.
3151         (s390_issue_rate): Remove cases.
3152         (s390_option_override): Adjust
3153         s390_option_override_internal() call.
3154         (s390_option_override_internal): Remove deprecation warning.
3155         (s390_valid_target_attribute_tree): Adjust
3156         s390_option_override_internal() call.
3157         * config/s390/s390.h (struct s390_processor):
3158         Share with s390-c.c, add arch_level field.
3159         * config/s390/s390.md:
3160         Remove occurrences in cpu attribute.
3161         * config/s390/s390.opt: Remove -march/-mtune support.
3162         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
3163         * doc/invoke.texi: Remove deprecation warning.
3165 2018-08-08  Luis Machado  <luis.machado@linaro.org>
3167         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
3168         global.
3169         (qdf24xx_tunings): Set vector cost structure to
3170         qdf24xx_vector_cost.
3172         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
3173         <register_sextend>: Set to 3.
3175 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
3177         PR target/86838
3178         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
3179         * config/aarch64/aarch64-simd.md
3180         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
3181         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
3182         (aarch64_frecpx<mode>): ...this new pattern.
3183         * config/aarch64/aarch64-simd-builtins.def: Remove comment
3184         about aarch64_frecp<FRECP:frecp_suffix><mode>.
3186 2018-08-07  Martin Liska  <mliska@suse.cz>
3188         PR middle-end/83023
3189         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
3190         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
3191         * predict.def (PRED_MALLOC_NONNULL): New predictor.
3192         * doc/extend.texi: Document that malloc attribute adds
3193         hit to compiler.
3195 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
3197         PR target/86785
3198         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3199         Define to speculation_safe_value_not_needed.
3201 2018-08-06  Jeff Law  <law@redhat.com>
3203         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
3204         the vr_values instance to cprop_into_stmt.
3205         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
3206         (cprop_operand): Also query EVRP to determine if OP is a constant.
3208 2018-08-06  Nathan Sidwell  <nathan@acm.org>
3210         * diagnostic.c (diagnostic_report_current_module): Reroll
3211         included-at loop.  Translate text.
3213 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
3215         * function-tests.c (selftest::test_expansion_to_rtl): Call
3216         free_after_compilation.
3218 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3220         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
3222 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
3224         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
3225         loops with memory block operations from getting unrolled.
3227 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
3229         PR target/86807
3230         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3231         Define to speculation_safe_value_not_needed.
3233 2018-08-06  Jeff Law  <law@redhat.com>
3235         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
3236         assert.
3238 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3240         PR target/86662
3241         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
3242         with all enabled __intN types.
3244         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
3246 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3248         * alias.c (record_set): Check for clobber high.
3249         * cfgexpand.c (expand_gimple_stmt): Likewise.
3250         * combine-stack-adj.c (single_set_for_csa): Likewise.
3251         * combine.c (find_single_use_1): Likewise.
3252         (set_nonzero_bits_and_sign_copies): Likewise.
3253         (get_combine_src_dest): Likewise.
3254         (is_parallel_of_n_reg_sets): Likewise.
3255         (try_combine): Likewise.
3256         (record_dead_and_set_regs_1): Likewise.
3257         (reg_dead_at_p_1): Likewise.
3258         (reg_dead_at_p): Likewise.
3259         * dce.c (deletable_insn_p): Likewise.
3260         (mark_nonreg_stores_1): Likewise.
3261         (mark_nonreg_stores_2): Likewise.
3262         * df-scan.c (df_find_hard_reg_defs): Likewise.
3263         (df_uses_record): Likewise.
3264         (df_get_call_refs): Likewise.
3265         * dwarf2out.c (mem_loc_descriptor): Likewise.
3266         * haifa-sched.c (haifa_classify_rtx): Likewise.
3267         * ira-build.c (create_insn_allocnos): Likewise.
3268         * ira-costs.c (scan_one_insn): Likewise.
3269         * ira.c (equiv_init_movable_p): Likewise.
3270         (rtx_moveable_p): Likewise.
3271         (interesting_dest_for_shprep): Likewise.
3272         * jump.c (mark_jump_label_1): Likewise.
3273         * postreload-gcse.c (record_opr_changes): Likewise.
3274         * postreload.c (reload_cse_simplify): Likewise.
3275         (struct reg_use): Add source expr.
3276         (reload_combine): Check for clobber high.
3277         (reload_combine_note_use): Likewise.
3278         (reload_cse_move2add): Likewise.
3279         (move2add_note_store): Likewise.
3280         * print-rtl.c (print_pattern): Likewise.
3281         * recog.c (decode_asm_operands): Likewise.
3282         (store_data_bypass_p): Likewise.
3283         (if_test_bypass_p): Likewise.
3284         * regcprop.c (kill_clobbered_value): Likewise.
3285         (kill_set_value): Likewise.
3286         * reginfo.c (reg_scan_mark_refs): Likewise.
3287         * reload1.c (maybe_fix_stack_asms): Likewise.
3288         (eliminate_regs_1): Likewise.
3289         (elimination_effects): Likewise.
3290         (mark_not_eliminable): Likewise.
3291         (scan_paradoxical_subregs): Likewise.
3292         (forget_old_reloads_1): Likewise.
3293         * reorg.c (find_end_label): Likewise.
3294         (try_merge_delay_insns): Likewise.
3295         (redundant_insn): Likewise.
3296         (own_thread_p): Likewise.
3297         (fill_simple_delay_slots): Likewise.
3298         (fill_slots_from_thread): Likewise.
3299         (dbr_schedule): Likewise.
3300         * resource.c (update_live_status): Likewise.
3301         (mark_referenced_resources): Likewise.
3302         (mark_set_resources): Likewise.
3303         * rtl.c (copy_rtx): Likewise.
3304         * rtlanal.c (reg_referenced_p): Likewise.
3305         (single_set_2): Likewise.
3306         (noop_move_p): Likewise.
3307         (note_stores): Likewise.
3308         * sched-deps.c (sched_analyze_reg): Likewise.
3309         (sched_analyze_insn): Likewise.
3311 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3313         * cse.c (invalidate_reg): New function extracted from...
3314         (invalidate): ...here.
3315         (canonicalize_insn): Check for clobber high.
3316         (invalidate_from_clobbers): invalidate clobber highs.
3317         (invalidate_from_sets_and_clobbers): Likewise.
3318         (count_reg_usage): Check for clobber high.
3319         (insn_live_p): Likewise.
3320         * cselib.c (cselib_expand_value_rtx_1):Likewise.
3321         (cselib_invalidate_regno): Check for clobber in setter.
3322         (cselib_invalidate_rtx): Pass through setter.
3323         (cselib_invalidate_rtx_note_stores):
3324         (cselib_process_insn): Check for clobber high.
3325         * cselib.h (cselib_invalidate_rtx): Add operand.
3327 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3329         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
3330         (mark_not_eliminable): Likewise.
3331         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
3332         * lra-lives.c (process_bb_lives): Check for clobber high.
3333         * lra.c (new_insn_reg): Remember clobber highs.
3334         (collect_non_operand_hard_regs): Check for clobber high.
3335         (lra_set_insn_recog_data): Likewise.
3336         (add_regs_to_insn_regno_info): Likewise.
3337         (lra_update_insn_regno_info): Likewise.
3339 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3341         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
3342         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
3344 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3346         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
3347         (copy_insn_1): Likewise.
3348         (gen_hard_reg_clobber_high): New gen function.
3349         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
3350         * genemit.c (gen_exp): Likewise.
3351         (gen_emit_seq): Pass through info.
3352         (gen_insn): Check for CLOBBER_HIGH.
3353         (gen_expand): Pass through info.
3354         (gen_split): Likewise.
3355         (output_add_clobbers): Likewise.
3356         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
3357         (remove_clobbers): Likewise.
3358         * rtl.h (gen_hard_reg_clobber_high): New declaration.
3360 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
3362         * doc/rtl.texi (clobber_high): Add.
3363         (parallel): Add in clobber high
3364         * rtl.c (rtl_check_failed_code3): Add function.
3365         * rtl.def (CLOBBER_HIGH): Add expression.
3366         * rtl.h (RTL_CHECKC3): Add macro.
3367         (rtl_check_failed_code3): Add declaration.
3368         (XC3EXP): Add macro.
3370 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
3372         PR target/86386
3373         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
3374         cfun->machine->max_used_stack_alignment if needed.
3376 2018-08-04  Martin Sebor  <msebor@redhat.com>
3378         PR tree-optimization/86571
3379         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
3380         NaN output to 4.
3382 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
3384         PR target/86799
3385         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3386         Define.
3388 2018-08-03  Jeff Law  <law@redhat.com>
3390         PR target/86795
3391         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3392         Define to speculation_safe_value_not_needed.
3394 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
3396         * doc/gcov.texi (-x): Remove duplicate "to".
3397         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
3398         (-Wif-not-aligned): Remove duplicate "is".
3399         (-flto): Remove duplicate "the".
3400         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
3401         duplicate "v5.00.b".
3402         (MSP430 Options): Remove duplicate "and" from the description
3403         of "-mgprel-sec=regexp".
3404         (x86 Options): Remove duplicate copies of "vmldLog102" and
3405         vmlsLog104 from description of "-mveclibabi=type".
3407 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
3409         * internal-fn.h (first_commutative_argument): Declare.
3410         * internal-fn.c (first_commutative_argument): New function.
3411         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
3412         restrictions for pattern statements.  Use first_commutative_argument
3413         to look for commutative operands in calls to internal functions.
3415 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
3417         * Makefile.in (wide-int-range.o): New.
3418         * tree-vrp.c: Move all the wide_int_* functions to...
3419         * wide-int-range.cc: ...here.
3420         * tree-vrp.h: Move all the wide_int_* prototypes to...
3421         * wide-int-range.h: ...here.
3423 2018-08-03  Tom de Vries  <tdevries@suse.de>
3425         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
3426         UI_NONE.
3427         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
3428         * except.c (output_function_exception_table): Do early exit if
3429         targetm_common.except_unwind_info (&global_options) == UI_NONE.
3431 2018-08-03  Martin Liska  <mliska@suse.cz>
3433         * predict.c (dump_prediction): Change to 2 digits
3434         in fraction part.
3436 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
3438         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
3439         neon_dup_q to...
3440         (falkor_am_1_gtov_gtov): ... a new insn reservation.
3442 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
3444         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
3445         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
3446         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
3447         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
3448         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
3450 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
3452         * diagnostic-show-locus.c (diagnostic_show_locus): Use
3453         pp_take_prefix when saving the existing prefix.
3454         * diagnostic.c (diagnostic_append_note): Likewise.
3455         * langhooks.c (lhd_print_error_function): Likewise.
3456         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
3457         param's type.  Free the existing prefix.
3458         (pp_take_prefix): New function.
3459         (pretty_printer::pretty_printer): Drop the prefix parameter.
3460         Rename the length parameter to match the comment.
3461         (pretty_printer::~pretty_printer): Free the prefix.
3462         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
3463         parameter.
3464         (struct pretty_printer): Drop the "const" from "prefix" field's
3465         type and clarify memory management.
3466         (pp_set_prefix): Drop the "const" from the 2nd param.
3467         (pp_take_prefix): New decl.
3469 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
3471         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
3472         for word_mode_ok here instead of passing as argument.
3473         (expand_block_compare): Change select_block_compare_mode() call.
3474         (expand_strncmp_gpr_sequence): New function.
3475         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
3477 2018-08-02  Jeff Law  <law@redhat.com>
3479         PR target/86790
3480         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3481         Define to speculation_safe_value_not_needed.
3483         PR target/86784
3484         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3485         Define to speculation_safe_value_not_needed.
3487 2018-08-02  Tom de Vries  <tdevries@suse.de>
3489         PR target/86660
3490         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
3491         function.  Return UI_TARGET unconditionally.
3492         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
3493         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
3495 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3497         * genemit.c (print_overload_test): Fix typo.
3499 2018-08-02  Richard Biener  <rguenther@suse.de>
3501         PR tree-optimization/86816
3502         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
3503         which checks for value availability before querying it.
3504         (gvn_uses_equal): Use it.
3505         (same_succ_hash): Likewise.
3506         (gimple_equal_p): Likewise.
3508 2018-08-02  Nick Clifton  <nickc@redhat.com>
3510         PR target/86813
3511         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3512         Define to speculation_safe_value_not_needed.
3514         PR target/86810
3515         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3516         Define to speculation_safe_value_not_needed.
3518         PR target/86810
3519         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3520         Define to speculation_safe_value_not_needed.
3522         PR target/86803
3523         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3524         Define to speculation_safe_value_not_needed.
3526         PR target/86797
3527         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3528         Define to speculation_safe_value_not_needed.
3530         PR target/86791
3531         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3532         Define to speculation_safe_value_not_needed.
3534         PR target/86789
3535         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3536         Define to speculation_safe_value_not_needed.
3538         PR target/86787
3539         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3540         Define to speculation_safe_value_not_needed.
3542         PR target/86782
3543         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3544         speculation_safe_value_not_needed.
3546         PR target/86781
3547         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
3548         to speculation_safe_value_not_needed.
3550 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3552         * doc/md.texi: Expand the documentation of instruction names
3553         to mention port-local uses.  Document '@' in pattern names.
3554         * read-md.h (overloaded_instance, overloaded_name): New structs.
3555         (mapping): Declare.
3556         (md_reader::handle_overloaded_name): New member function.
3557         (md_reader::get_overloads): Likewise.
3558         (md_reader::m_first_overload): New member variable.
3559         (md_reader::m_next_overload_ptr): Likewise.
3560         (md_reader::m_overloads_htab): Likewise.
3561         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
3562         m_next_overload_ptr and m_overloads_htab.
3563         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
3564         (get_mode_token, get_code_token, get_int_token): New functions.
3565         (map_attr_string): Add an optional argument that passes back
3566         the associated iterator.
3567         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
3568         (md_reader::handle_overloaded_name, add_overload_instance): New
3569         functions.
3570         (apply_iterators): Handle '@' names.  Report an error if '@'
3571         is used without iterators.
3572         (initialize_iterators): Initialize the new iterator_group fields.
3573         * genopinit.c (handle_overloaded_code_for)
3574         (handle_overloaded_gen): New functions.
3575         (main): Use them to print declarations of maybe_code_for_* and
3576         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
3577         * genemit.c (print_overload_arguments, print_overload_test)
3578         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
3579         (main): Use it to print definitions of maybe_code_for_* and
3580         maybe_gen_* functions.
3581         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
3582         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
3583         instead of explicit mode checks.
3584         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
3585         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
3586         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
3587         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
3588         (aarch64_expand_compare_and_swap): Likewise
3589         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
3590         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
3591         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
3592         (aarch64_constant_pool_reload_icode): Delete.
3593         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
3594         instead of aarch64_constant_pool_reload_icode.  Use
3595         code_for_aarch64_reload_mov instead of explicit mode checks.
3596         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
3597         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
3598         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
3599         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
3600         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
3601         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
3602         (aarch64_atomic_load_op_code): Delete.
3603         (aarch64_emit_atomic_load_op): Likewise.
3604         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
3605         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
3606         instead of aarch64_emit_atomic_load_op.
3607         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
3608         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
3609         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
3610         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
3611         character before the pattern name.
3612         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
3613         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
3614         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
3615         (aarch64_frecps<mode>): Likewise.
3616         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
3617         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
3618         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
3619         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
3620         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
3622 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
3624         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
3625         Allow HFmode constants if TARGET_FP_F16INST.
3627 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
3629         PR target/86014
3630         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
3631         No longer check last store for clobber of address register.
3633 2018-08-02  Martin Liska  <mliska@suse.cz>
3635         PR gcov-profile/86817
3636         * gcov.c (process_all_functions): New function.
3637         (main): Call it.
3638         (process_file): Move functions processing to
3639         process_all_functions.
3641 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
3643         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
3644         "const" to the "gimple *" and "rtx_insn *" parameters.
3645         * dumpfile.h (dump_user_location_t::dump_user_location_t):
3646         Likewise.
3647         (dump_location_t::dump_location_t): Likewise.
3649 2018-08-01  Martin Sebor  <msebor@redhat.com>
3651         PR tree-optimization/86650
3652         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
3653         rather than a "gcall *".  Directly pass the data of interest
3654         to percent_K_format, rather than building a temporary CALL_EXPR
3655         to hold it.
3656         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
3657         (gimple_fold_builtin_strncat): Adjust.
3658         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
3659         gcall* argument with gimple*.
3660         * gimple-ssa-warn-restrict.c (check_call): Same.
3661         (wrestrict_dom_walker::before_dom_children): Same.
3662         (builtin_access::builtin_access): Same.
3663         (check_bounds_or_overlap): Same
3664         (maybe_diag_overlap): Same.
3665         (maybe_diag_offset_bounds): Same.
3666         * tree-diagnostic.c (default_tree_printer): Move usage of
3667         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
3668         to this callsite.
3669         * tree-pretty-print.c (percent_K_format): Add argument.
3670         * tree-pretty-print.h: Add argument.
3671         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
3672         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
3673         (maybe_diag_stxncpy_trunc): Same.
3674         (handle_builtin_stxncpy): Same.
3675         (handle_builtin_strcat): Same.
3677 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3679         * match.pd: Optimise pointer range checks.
3681 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3683         PR tree-optimization/86758
3684         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
3685         to remove pattern statements.
3687 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3689         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
3690         result of dfs_enumerate_from when constructing stmt_vec_infos,
3691         instead of additionally calling get_loop_body.
3693 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3695         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
3696         parameter.
3697         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
3698         When creating an iv, assert that the step is not known to be zero.
3699         (vect_setup_realignment): Update call accordingly.
3700         * tree-vect-stmts.c (vectorizable_store): Likewise.
3701         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
3703 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3705         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
3706         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
3707         (vectorizable_reduction): Likewise.
3708         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
3709         (vect_detect_hybrid_slp_stmts): Likewise.
3710         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
3712 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
3714         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
3715         (wide_int_set_zero_nonzero_bits): ...this.
3716         (zero_nonzero_bits_from_vr): Rename to...
3717         (vrp_set_zero_nonzero_bits): ...this.
3718         (extract_range_from_multiplicative_op_1): Abstract wide int
3719         code...
3720         (wide_int_range_multiplicative_op): ...here.
3721         (extract_range_from_binary_expr_1): Extract wide int binary
3722         operations into their own functions.
3723         (wide_int_range_lshift): New.
3724         (wide_int_range_can_optimize_bit_op): New.
3725         (wide_int_range_shift_undefined_p): New.
3726         (wide_int_range_bit_xor): New.
3727         (wide_int_range_bit_ior): New.
3728         (wide_int_range_bit_and): New.
3729         (wide_int_range_trunc_mod): New.
3730         (extract_range_into_wide_ints): New.
3731         (vrp_shift_undefined_p): New.
3732         (extract_range_from_multiplicative_op): New.
3733         (vrp_can_optimize_bit_op): New.
3734         * tree-vrp.h (value_range::dump): New.
3735         (wide_int_range_multiplicative_op): New.
3736         (wide_int_range_lshift):New.
3737         (wide_int_range_shift_undefined_p): New.
3738         (wide_int_range_bit_xor): New.
3739         (wide_int_range_bit_ior): New.
3740         (wide_int_range_bit_and): New.
3741         (wide_int_range_trunc_mod): New.
3742         (zero_nonzero_bits_from_bounds): Rename to...
3743         (wide_int_set_zero_nonzero_bits): ...this.
3744         (zero_nonzero_bits_from_vr): Rename to...
3745         (vrp_set_zero_nonzero_bits): ...this.
3746         (range_easy_mask_min_max): Rename to...
3747         (wide_int_range_can_optimize_bit_op): this.
3748         * vr-values.c (simplify_bit_ops_using_ranges): Rename
3749         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
3751 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3753         * tree-vectorizer.h (vect_orig_stmt): New function.
3754         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
3755         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
3756         (vect_create_epilog_for_reduction): Likewise.
3757         (vectorizable_live_operation): Likewise.
3758         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
3759         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
3760         * tree-vect-stmts.c (vectorizable_call): Likewise.
3761         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
3763 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3765         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
3766         argument.
3767         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
3768         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
3769         (vect_transform_loop): Likewise.
3770         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3772 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3774         * tree-vectorizer.h (vect_schedule_slp): Return void.
3775         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
3776         (vect_schedule_slp): Likewise.
3778 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3780         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
3781         argument.
3782         (vect_transform_loop): Update calls accordingly.  Schedule SLP
3783         instances before the main loop, if any exist.
3785 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
3787         PR tree-optimization/86749
3788         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
3789         If the lhs is used in a COND_EXPR, check that it is being used
3790         as the "then" or "else" value.
3792 2018-08-01  Tom de Vries  <tdevries@suse.de>
3794         PR target/86800
3795         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3796         speculation_safe_value_not_needed.
3798 2018-08-01  Richard Biener  <rguenther@suse.de>
3800         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
3801         as base and offset.
3803 2018-08-01  Martin Liska  <mliska@suse.cz>
3805         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
3806         format how successful transformation is dumped.
3807         (gimple_mod_pow2_value_transform): Likewise.
3808         (gimple_mod_subtract_transform): Likewise.
3809         (gimple_stringops_transform): Likewise.
3811 2018-08-01  Martin Liska  <mliska@suse.cz>
3813         PR value-prof/35543
3814         * value-prof.c (interesting_stringop_to_profile_p):
3815         Simplify the code and add BUILT_IN_MEMMOVE.
3816         (gimple_stringops_transform): Likewise.
3818 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
3820         * config/aarch64/aarch64-simd.md
3821         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
3822         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
3823         use GPI iterator instead of SI mode.
3825 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
3827         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
3828         rs6000_speculation_barrier.
3829         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
3830         new barrier pattern name.
3832 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
3834         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
3835         (speculation_barrier): New insn.
3837 2018-08-01  Richard Biener  <rguenther@suse.de>
3839         PR bootstrap/86724
3840         * graphite.h: Include isl/id.h and isl/space.h to allow build
3841         with ISL 0.20.
3843 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
3845         PR target/86651
3846         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
3847         mode for COFF targets.
3848         * defaults.h (TARGET_COFF): Define.
3849         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
3850         TARGET_COFF): Define.
3851         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
3852         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
3853         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
3855 2018-07-31  Alexandre Oliva <oliva@adacore.com>
3857         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
3858         * gimple-streamer-out.c (output_bb): Save it.
3859         * lto-streamer-in.c (input_struct_function_base): Restore
3860         instance discriminator if available.  Create map on demand.
3861         * lto-streamer-out.c (output_struct_function_base): Save it if
3862         available.
3863         * final.c (decl_to_instance_map): Document LTO strategy.
3865 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
3866             Olivier Hainque  <hainque@adacore.com>
3868         * debug.h (decl_to_instance_map_t): New type.
3869         (decl_to_instance_map): Declare.
3870         (maybe_create_decl_to_instance_map): New inline function.
3871         * final.c (bb_discriminator, last_bb_discriminator): New statics,
3872         to track basic block discriminators.
3873         (final_start_function_1): Initialize them.
3874         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
3875         bb_discriminator.
3876         (decl_to_instance_map): New variable.
3877         (map_decl_to_instance, maybe_set_discriminator): New functions.
3878         (notice_source_line): Set discriminator.
3880 2018-07-31  Ian Lance Taylor  <iant@golang.org>
3882         * targhooks.c (default_have_speculation_safe_value): Add
3883         ATTRIBUTE_UNUSED.
3885 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
3887         * dump-context.h: Include "pretty-print.h".
3888         (dump_context::refresh_dumps_are_enabled): New decl.
3889         (dump_context::emit_item): New decl.
3890         (class dump_context): Add fields "m_test_pp" and
3891         "m_test_pp_flags".
3892         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3893         (temp_dump_context::get_dumped_text): New decl.
3894         (class temp_dump_context): Add field "m_pp".
3895         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
3896         (dump_context::refresh_dumps_are_enabled): ...and add a test for
3897         m_test_pp.
3898         (set_dump_file): Update for above change.
3899         (set_alt_dump_file): Likewise.
3900         (dump_loc): New overload, taking a pretty_printer *.
3901         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
3902         to any test pretty-printer.
3903         (make_item_for_dump_gimple_stmt): New function, adapted from
3904         optinfo::add_gimple_stmt.
3905         (dump_context::dump_gimple_stmt): Call it, and use the result,
3906         eliminating the direct usage of dump_file and alt_dump_file in
3907         favor of indirectly using them via emit_item.
3908         (make_item_for_dump_gimple_expr): New function, adapted from
3909         optinfo::add_gimple_expr.
3910         (dump_context::dump_gimple_expr): Call it, and use the result,
3911         eliminating the direct usage of dump_file and alt_dump_file in
3912         favor of indirectly using them via emit_item.
3913         (make_item_for_dump_generic_expr): New function, adapted from
3914         optinfo::add_tree.
3915         (dump_context::dump_generic_expr): Call it, and use the result,
3916         eliminating the direct usage of dump_file and alt_dump_file in
3917         favor of indirectly using them via emit_item.
3918         (make_item_for_dump_printf_va): New function, adapted from
3919         optinfo::add_printf_va.
3920         (make_item_for_dump_printf): New function.
3921         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
3922         and use the result, eliminating the direct usage of dump_file and
3923         alt_dump_file in favor of indirectly using them via emit_item.
3924         (make_item_for_dump_dec): New function.
3925         (dump_context::dump_dec): Call it, and use the result,
3926         eliminating the direct usage of dump_file and alt_dump_file in
3927         favor of indirectly using them via emit_item.
3928         (make_item_for_dump_symtab_node): New function, adapted from
3929         optinfo::add_symtab_node.
3930         (dump_context::dump_symtab_node): Call it, and use the result,
3931         eliminating the direct usage of dump_file and alt_dump_file in
3932         favor of indirectly using them via emit_item.
3933         (dump_context::begin_scope): Reimplement, avoiding direct usage
3934         of dump_file and alt_dump_file in favor of indirectly using them
3935         via emit_item.
3936         (dump_context::emit_item): New member function.
3937         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
3938         Set up test pretty-printer on the underlying context.  Call
3939         refresh_dumps_are_enabled.
3940         (temp_dump_context::~temp_dump_context): Call
3941         refresh_dumps_are_enabled.
3942         (temp_dump_context::get_dumped_text): New member function.
3943         (selftest::verify_dumped_text): New function.
3944         (ASSERT_DUMPED_TEXT_EQ): New macro.
3945         (selftest::test_capture_of_dump_calls): Run all tests twice, with
3946         and then without optinfo enabled.  Add uses of
3947         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
3948         * dumpfile.h: Update comment for the dump_* API.
3949         * optinfo-emit-json.cc
3950         (selftest::test_building_json_from_dump_calls): Update for new
3951         param for temp_dump_context ctor.
3952         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
3953         and "m_owned" field.
3954         (optinfo_item::~optinfo_item): Likewise.
3955         (optinfo::add_item): New member function.
3956         (optinfo::emit): Update comment.
3957         (optinfo::add_string): Delete.
3958         (optinfo::add_printf): Delete.
3959         (optinfo::add_printf_va): Delete.
3960         (optinfo::add_gimple_stmt): Delete.
3961         (optinfo::add_gimple_expr): Delete.
3962         (optinfo::add_tree): Delete.
3963         (optinfo::add_symtab_node): Delete.
3964         (optinfo::add_dec): Delete.
3965         * optinfo.h (class dump_context): New forward decl.
3966         (optinfo::add_item): New decl.
3967         (optinfo::add_string): Delete.
3968         (optinfo::add_printf): Delete.
3969         (optinfo::add_printf_va): Delete.
3970         (optinfo::add_gimple_stmt): Delete.
3971         (optinfo::add_gimple_expr): Delete.
3972         (optinfo::add_tree): Delete.
3973         (optinfo::add_symtab_node): Delete.
3974         (optinfo::add_dec): Delete.
3975         (optinfo::add_poly_int): Delete.
3976         (optinfo_item::optinfo_item): Remove "owned" param.
3977         (class optinfo_item): Remove field "m_owned".
3979 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
3981         PR middle-end/86705
3982         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
3983         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
3984         requested variable alignment.
3985         (expand_one_ssa_partition): Likewise.
3986         (expand_one_var): Likewise.
3988 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
3990         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3991         to speculation_safe_value_not_needed.
3993 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
3995         * targhooks.h (speculation_safe_value_not_needed): New prototype.
3996         * targhooks.c (speculation_safe_value_not_needed): New function.
3997         * target.def (have_speculation_safe_value): Update documentation.
3998         * doc/tm.texi: Regenerated.
4000 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4002         * config/aarch64/iterators.md (ALLI_TI): New iterator.
4003         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
4004         expand.
4005         (despeculate_copy<ALLI:mode>_insn): New insn.
4006         (despeculate_copyti_insn): New insn.
4007         (despeculate_simple<ALLI:mode>): New insn
4008         (despeculate_simpleti): New insn.
4009         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
4010         function.
4011         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
4012         aarch64_speculation_safe_value.
4013         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
4015 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4017         * config/aarch64/aarch64-speculation.cc: New file.
4018         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
4019         before pass_reorder_blocks.
4020         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
4021         prototype.
4022         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
4023         X14 and X15 when tracking speculation.
4024         * config/aarch64/aarch64.md (register name constants): Add
4025         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
4026         (unspec): Add UNSPEC_SPECULATION_TRACKER.
4027         (speculation_barrier): New insn attribute.
4028         (cmp<mode>): Allow SP in comparisons.
4029         (speculation_tracker): New insn.
4030         (speculation_barrier): Add speculation_barrier attribute.
4031         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
4032         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
4033         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
4035 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4037         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
4038         aarch64_track_speculation is true.
4039         (tb<optab><mode>1): Likewise.
4040         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
4041         generate CB[N]Z when tracking speculation.
4042         (aarch64_split_compare_and_swap): Likewise.
4043         (aarch64_split_atomic_op): Likewise.
4045 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4047         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
4049 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4051         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
4052         (speculation_barrier): New insn.
4054 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4056         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
4057         * config/arm/arm.md (speculation_barrier): New expand.
4058         (speculation_barrier_insn): New pattern.
4060 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
4062         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
4063         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
4064         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
4065         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
4066         list.
4067         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
4068         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
4069         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
4070         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
4071         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
4072         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
4073         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
4074         * builtins.c (expand_speculation_safe_value): New function.
4075         (expand_builtin): Call it.
4076         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
4077         * doc/extend.texi: Document __builtin_speculation_safe_value.
4078         * doc/md.texi: Document "speculation_barrier" pattern.
4079         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
4080         TARGET_HAVE_SPECULATION_SAFE_VALUE.
4081         * doc/tm.texi: Regenerated.
4082         * target.def (have_speculation_safe_value, speculation_safe_value): New
4083         hooks.
4084         * targhooks.c (default_have_speculation_safe_value): New function.
4085         (default_speculation_safe_value): New function.
4086         * targhooks.h (default_have_speculation_safe_value): Add prototype.
4087         (default_speculation_safe_value): Add prototype.
4089 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
4091         * dump-context.h (dump_context::dump_loc): New decl.
4092         * dumpfile.c (dump_context::dump_loc): New member function.
4093         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
4094         and dump_gimple_stmt.
4095         (dump_context::dump_gimple_expr_loc): Likewise, using
4096         dump_gimple_expr.
4097         (dump_context::dump_generic_expr_loc): Likewise, using
4098         dump_generic_expr.
4099         (dump_context::dump_printf_loc_va): Likewise, using
4100         dump_printf_va.
4101         (dump_context::begin_scope): Explicitly using the global function
4102         "dump_loc", rather than the member function.
4104 2018-07-31  Martin Sebor  <msebor@redhat.com>
4106         PR tree-optimization/86741
4107         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
4109 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
4111         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
4112         depend on whether prefetch instructions will be emitted or not.
4113         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
4114         will be emitted or not.
4115         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
4116         (TARGET_SETMEM_PFD): New macros.
4118 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4120         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
4121         (NULL_STMT_VEC_INFO): Delete.
4122         (stmt_vec_info::operator*): Likewise.
4123         (stmt_vec_info::operator gimple *): Likewise.
4124         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
4125         of NULL_STMT_VEC_INFO.
4126         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4127         (vect_reassociating_reduction_p): Likewise.
4128         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4129         (vectorizable_store): Likewise.
4130         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
4131         (vec_info::free_stmt_vec_infos): Likewise.
4133 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4135         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
4136         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
4138 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4140         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
4141         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4142         (vec_info::free_stmt_vec_info): New private member functions.
4143         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
4144         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
4145         * tree-parloops.c (gather_scalar_reductions): Remove calls to
4146         set_stmt_vec_info_vec and free_stmt_vec_infos.
4147         * tree-vect-loop.c (_loop_vec_info): Remove call to
4148         set_stmt_vec_info_vec.
4149         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
4150         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
4151         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
4152         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
4153         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
4154         assignments in {vec_info::,}new_stmt_vec_info that are redundant
4155         with the clearing in the xcalloc.
4156         (stmt_vec_info_vec): Delete.
4157         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
4158         (vectorize_loops): Likewise.
4159         (vec_info::~vec_info): Remove argument from call to
4160         free_stmt_vec_infos.
4161         (vec_info::add_stmt): Remove vinfo argument from call to
4162         new_stmt_vec_info.
4164 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4166         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
4167         rather than a gimple stmt.
4168         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
4169         information for pattern statements when passed the original
4170         statement; instead wait to be passed the pattern statement itself.
4171         Don't call set_vinfo_for_stmt here.
4172         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
4173         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
4174         stmt_vec_infos here.
4175         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
4176         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
4177         stmt_vec_infos entry.
4179 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4181         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
4182         * tree-vectorizer.c (vec_info::replace_stmt): New function.
4183         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
4184         * tree-vect-stmts.c (vectorizable_call): Likewise.
4185         (vectorizable_simd_clone_call): Likewise.
4187 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4189         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
4190         * tree-vectorizer.c (vec_info::remove_stmt): New function.
4191         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
4192         * tree-vect-loop.c (vect_transform_loop): Likewise.
4193         * tree-vect-slp.c (vect_schedule_slp): Likewise.
4194         * tree-vect-stmts.c (vect_remove_stores): Likewise.
4196 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4198         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
4199         (vect_dr_stmt): Delete.
4200         * tree-vectorizer.c (vec_info::lookup_dr): New function.
4201         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
4202         of DR_VECT_AUX.
4203         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
4204         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
4205         (vect_verify_datarefs_alignment, vect_peeling_supportable)
4206         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4207         (vect_analyze_data_refs): Likewise.
4208         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
4209         argument.
4210         (vect_find_same_alignment_drs): Likewise.
4211         (vect_slp_analyze_node_dependences): Update calls accordingly.
4212         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
4213         instead of DR_VECT_AUX.
4214         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
4215         of a vector data references.  Use vec_info::lookup_dr instead of
4216         DR_VECT_AUX.
4217         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
4218         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
4219         instead of DR_VECT_AUX.
4221 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4223         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
4224         dr_vec_info.
4225         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4226         accordingly.
4227         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4228         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4229         (vect_gen_prolog_loop_niters): Likewise.
4231 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4233         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
4234         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
4235         (vect_known_alignment_in_bytes, vect_dr_behavior)
4236         (vect_get_scalar_dr_size): Take references as dr_vec_infos
4237         instead of data_references.  Update calls to other routines for
4238         which the same change has been made.
4239         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
4240         dr_vec_infos instead of stmt_vec_infos.
4241         (vect_analyze_data_ref_dependence): Update call accordingly.
4242         (vect_slp_analyze_data_ref_dependence)
4243         (vect_record_base_alignments): Use DR_VECT_AUX.
4244         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
4245         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
4246         (vector_alignment_reachable_p, vect_get_data_access_cost)
4247         (vect_peeling_supportable, vect_analyze_group_access_1)
4248         (vect_analyze_group_access, vect_analyze_data_ref_access)
4249         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
4250         (vect_compile_time_alias, vect_small_gap_p)
4251         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
4252         (vect_supportable_dr_alignment): Take references as dr_vec_infos
4253         instead of data_references.  Update calls to other routines for
4254         which the same change has been made.
4255         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
4256         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4257         (vect_slp_analyze_and_verify_node_alignment)
4258         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
4259         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4260         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
4261         above changes.
4262         (_vect_peel_info::dr): Replace with...
4263         (_vect_peel_info::dr_info): ...this new field.
4264         (vect_peeling_hash_get_most_frequent)
4265         (vect_peeling_hash_choose_best_peeling): Update accordingly.
4266         (vect_peeling_hash_get_lowest_cost):
4267         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
4268         routines for which the same change has been made.
4269         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
4270         data_reference.
4271         * tree-vect-loop-manip.c (get_misalign_in_elems)
4272         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
4273         above changes.
4274         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4275         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
4276         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
4277         (get_group_load_store_type, get_negative_load_store_type)
4278         (vect_get_data_ptr_increment, vectorizable_store)
4279         (vectorizable_load): Likewise.
4280         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
4281         Update calls to other routines for which the same change has been made.
4283 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4285         * tree-vectorizer.h (vec_info::move_dr): New member function.
4286         (dataref_aux): Rename to...
4287         (dr_vec_info): ...this and add "dr" and "stmt" fields.
4288         (_stmt_vec_info::dr_aux): Update accordingly.
4289         (_stmt_vec_info::data_ref_info): Delete.
4290         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
4291         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
4292         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
4293         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
4294         of data_ref.
4295         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
4296         (STMT_VINFO_DR_INFO): New macro.
4297         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
4298         (set_dr_misalignment): Update after rename of dataref_aux.
4299         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
4300         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
4301         initialization of STMT_VINFO_DATA_REF.
4302         * tree-vectorizer.c (vec_info::move_dr): New function.
4303         * tree-vect-patterns.c (vect_recog_bool_pattern)
4304         (vect_recog_mask_conversion_pattern)
4305         (vect_recog_gather_scatter_pattern): Use it.
4306         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
4307         the "dr" and "stmt" fields of dr_vec_info instead of
4308         STMT_VINFO_DATA_REF.
4310 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4312         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
4313         (is_pattern_stmt_p): Use it.
4314         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
4315         on pattern statements.
4317 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4319         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
4320         original stmt as a stmt_vec_info rather than a gimple stmt.
4321         (vect_pattern_recog_1): Take the statement directly as a
4322         stmt_vec_info, rather than via a gimple_stmt_iterator.
4323         Update call to vect_mark_pattern_stmts.
4324         (vect_pattern_recog): Update calls accordingly.
4326 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4328         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
4329         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
4330         a vect_def_type for the first argument.
4331         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
4332         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
4333         operand if it isn't defined by a vectorized statement.
4334         (vect_build_gather_load_calls): Remove the mask_dt argument and
4335         update calls to vect_get_vec_def_for_stmt_copy.
4336         (vectorizable_bswap): Likewise the dt argument.
4337         (vectorizable_call): Update calls to vectorizable_bswap and
4338         vect_get_vec_def_for_stmt_copy.
4339         (vectorizable_simd_clone_call, vectorizable_assignment)
4340         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
4341         (vectorizable_comparison): Update calls to
4342         vect_get_vec_def_for_stmt_copy.
4343         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
4344         vect_is_simple_use.
4345         (vect_get_loop_based_defs): Remove dt argument and update call
4346         to vect_get_vec_def_for_stmt_copy.
4347         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
4348         and vect_get_vec_def_for_stmt_copy.
4349         (vectorizable_load): Update calls to vect_build_gather_load_calls
4350         and vect_get_vec_def_for_stmt_copy.
4351         * tree-vect-loop.c (vect_create_epilog_for_reduction)
4352         (vectorizable_reduction, vectorizable_live_operation): Update calls
4353         to vect_get_vec_def_for_stmt_copy.
4355 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4357         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
4358         and gimple stmt arguments with a stmt_vec_info.
4359         (vect_record_base_alignments): Update calls accordingly.
4360         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
4361         and gimple stmt arguments with a stmt_vec_info.
4362         (vect_build_slp_tree_1): Remove vinfo argument and update call
4363         to vect_record_max_nunits.
4364         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
4365         and vect_record_max_nunits.
4367 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4369         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
4370         file and take a stmt_vec_info instead of a gimple stmt.
4371         (supportable_widening_operation, vect_finish_replace_stmt)
4372         (vect_finish_stmt_generation, vect_get_store_rhs)
4373         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
4374         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
4375         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
4376         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
4377         (vect_create_data_ref_ptr, bump_vector_ptr)
4378         (vect_permute_store_chain, vect_setup_realignment)
4379         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
4380         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
4381         (vectorizable_reduction, vectorizable_induction)
4382         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
4383         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
4384         than gimple stmts as arguments.
4385         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
4386         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
4387         (can_group_stmts_p, vect_check_gather_scatter)
4388         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
4389         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
4390         (vect_permute_load_chain, vect_shift_permute_load_chain)
4391         (vect_transform_grouped_load)
4392         (vect_record_grouped_load_vectors): Likewise.
4393         * tree-vect-loop.c (vect_fixup_reduc_chain)
4394         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
4395         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
4396         (vectorizable_reduction, vectorizable_induction)
4397         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
4398         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
4399         (vect_get_load_store_mask): Likewise.
4400         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
4401         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
4402         * tree-vect-stmts.c (vect_mark_relevant)
4403         (is_simple_and_all_uses_invariant)
4404         (exist_non_indexing_operands_for_use_p, process_use)
4405         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
4406         (vect_get_vec_def_for_operand, vect_get_vec_defs)
4407         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
4408         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
4409         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
4410         (get_negative_load_store_type, get_load_store_type)
4411         (vect_check_load_store_mask, vect_check_store_rhs)
4412         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
4413         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4414         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4415         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4416         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
4417         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
4418         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
4419         (supportable_widening_operation): Likewise.
4421 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4423         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
4424         a stmt_vec_info instead of a gcall.
4425         (vect_check_gather_scatter): Update call accordingly.
4426         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
4427         of a gphi.
4428         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
4429         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
4430         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
4431         instead of a gimple stmt.
4432         (vect_transform_loop): Update calls accordingly.
4433         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
4434         stmt_vec_infos instead of gimple stmts.
4435         (vect_analyze_slp_instance): Update use accordingly.
4436         * tree-vect-stmts.c (read_vector_array, write_vector_array)
4437         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
4438         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
4439         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
4440         (vect_gen_widened_results_half, vect_get_loop_based_defs)
4441         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
4442         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
4443         down to subroutines.
4445 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4447         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
4448         of the worklist from a vector of gimple stmts to a vector of
4449         stmt_vec_infos.
4450         * tree-vect-stmts.c (vect_mark_relevant, process_use)
4451         (vect_mark_stmts_to_be_vectorized): Likewise
4453 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4455         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
4456         statement before passing it to vect_analyze_stmt.
4457         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
4458         the chain of phi vector definitions.  Track the exit phi via its
4459         stmt_vec_info.
4460         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
4461         STMT_VINFO_REDUC_DEF.
4462         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
4463         stmt_vec_infos to handle the statement chains.
4464         (vect_get_slp_defs): Record the first statement in the node
4465         using a stmt_vec_info.
4466         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
4467         statements here and pass their stmt_vec_info down to subroutines.
4468         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
4469         down to vect_finish_stmt_generation.
4470         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
4471         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
4472         stmt_vec_infos to subroutines.
4473         (vect_remove_stores): Use stmt_vec_infos to handle the statement
4474         chains.
4476 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4478         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
4479         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
4480         (vect_permute_store_chain, vect_setup_realignment)
4481         (vect_permute_load_chain, vect_shift_permute_load_chain)
4482         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
4483         stmts internally, and when passing values to other vectorizer routines.
4484         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4485         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
4486         (vect_analyze_loop_operations, get_initial_def_for_reduction)
4487         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
4488         (vectorizable_reduction, vectorizable_induction)
4489         (vectorizable_live_operation, vect_transform_loop_stmt)
4490         (vect_transform_loop): Likewise.
4491         * tree-vect-patterns.c (vect_reassociating_reduction_p)
4492         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
4493         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
4494         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4495         (vect_slp_analyze_node_operations_1): Likewise.
4496         * tree-vect-stmts.c (vect_mark_relevant, process_use)
4497         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
4498         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
4499         (vect_finish_stmt_generation_1, get_group_load_store_type)
4500         (get_load_store_type, vect_build_gather_load_calls)
4501         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4502         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4503         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4504         (vectorizable_store, vectorizable_load, vectorizable_condition)
4505         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
4506         (supportable_widening_operation): Likewise.
4507         (vect_get_vector_types_for_stmt): Likewise.
4508         * tree-vectorizer.h (vect_dr_behavior): Likewise.
4510 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4512         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4513         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
4514         (vect_permute_store_chain, vect_permute_load_chain)
4515         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
4516         repeated stmt_vec_info lookups.
4517         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
4518         (vect_update_ivs_after_vectorizer): Likewise.
4519         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
4520         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4521         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
4522         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4523         (vect_bb_slp_scalar_cost): Likewise.
4524         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
4526 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4528         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
4529         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
4530         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
4531         to get gassigns and gcalls, rather than operating on generc gimple
4532         stmts.
4533         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
4534         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
4535         (vectorizable_load, vect_analyze_stmt): Likewise.
4536         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
4538 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4540         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
4541         return stmt_vec_infos rather than gimple stmts.  Do not accept
4542         null arguments.
4543         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
4544         of a gimple stmt.
4545         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
4546         Update use of get_later_stmt.
4547         (vect_get_constant_vectors): Update call accordingly.
4548         (vect_schedule_slp_instance): Likewise
4549         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
4550         (vect_slp_analyze_instance_dependence): Likewise.
4551         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
4553 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4555         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
4556         (stmt_info_for_cost::stmt_info): ...this new field.
4557         (add_stmt_costs): Update accordingly.
4558         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4559         (vect_get_known_peeling_cost): Likewise.
4560         (vect_estimate_min_profitable_iters): Likewise.
4561         * tree-vect-stmts.c (record_stmt_cost): Likewise.
4563 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4565         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
4566         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4567         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
4568         accordingly.
4569         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
4571 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4573         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
4574         a gimple stmt to a stmt_vec_info.
4575         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4577 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4579         * tree-vectorizer.h (vec_info::grouped_stores): Change from
4580         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4581         (_loop_vec_info::reduction_chains): Likewise.
4582         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
4583         accordingly.
4584         * tree-vect-slp.c (vect_analyze_slp): Likewise.
4586 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4588         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
4589         a gimple stmt to a stmt_vec_info.
4590         (_stmt_vec_info::next_element): Likewise.
4591         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
4592         (vect_slp_analyze_and_verify_node_alignment)
4593         (vect_analyze_group_access_1, vect_analyze_group_access)
4594         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
4595         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
4596         (vect_supportable_dr_alignment): Update accordingly.
4597         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
4598         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
4599         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
4600         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
4601         * tree-vect-slp.c (vect_build_slp_tree_1)
4602         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
4603         (vect_split_slp_store_group, vect_analyze_slp_instance)
4604         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
4605         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
4606         (get_group_load_store_type, get_load_store_type)
4607         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
4608         (vect_transform_stmt, vect_remove_stores): Likewise.
4610 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4612         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
4613         than a gimple stmt.
4614         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4615         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
4616         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
4617         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
4618         (vector_alignment_reachable_p, vect_get_data_access_cost)
4619         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
4620         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
4621         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
4622         (vect_analyze_group_access_1, vect_analyze_group_access)
4623         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
4624         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
4625         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
4626         result of vect_dr_stmt and use the stmt_vec_info instead of
4627         the associated gimple stmt.
4628         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
4629         (vect_gen_prolog_loop_niters): Likewise.
4630         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
4632 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4634         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
4635         to a vec<stmt_vec_info>.
4636         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
4637         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
4638         vec<stmt_vec_info>.
4639         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
4640         to a vec<stmt_vec_info>.
4641         (bst_traits::value_type, bst_traits::value_type): Likewise.
4642         (bst_traits::hash): Update accordingly.
4643         (vect_get_and_check_slp_defs): Change the stmts parameter from
4644         a vec<gimple *> to a vec<stmt_vec_info>.
4645         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
4646         (vect_build_slp_tree): Likewise.
4647         (vect_build_slp_tree_2): Likewise.  Update uses of
4648         SLP_TREE_SCALAR_STMTS.
4649         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
4650         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
4651         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
4652         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
4653         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
4654         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
4655         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
4656         (vect_get_constant_vectors, vect_get_slp_defs)
4657         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
4658         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
4659         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
4660         instead of gimple stmts.
4661         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
4662         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
4663         (vect_slp_analyze_instance_dependence): Update uses of
4664         SLP_TREE_SCALAR_STMTS.
4665         (vect_slp_analyze_and_verify_node_alignment): Likewise.
4666         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
4667         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
4668         (get_initial_defs_for_reduction): Likewise.
4669         (vect_create_epilog_for_reduction): Likewise.
4670         (vectorize_fold_left_reduction): Likewise.
4671         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
4672         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
4673         (can_vectorize_live_stmts): Likewise.
4675 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4677         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
4678         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
4679         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
4680         than gimple stmts.
4681         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
4682         of a gimple stmt.
4683         (gather_scalar_reductions): Update after above interface changes.
4684         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
4685         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
4686         than gimple stmts.
4687         (vect_force_simple_reduction): Likewise.
4688         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
4689         LOOP_VINFO_REDUCTIONS.
4690         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4692 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4694         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
4695         a gimple stmt to a stmt_vec_info.
4696         * tree-vect-loop.c (vect_active_double_reduction_p)
4697         (vect_force_simple_reduction, vectorizable_reduction): Update
4698         accordingly.
4700 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4702         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
4703         vec<gimple *> to a vec<stmt_vec_info>.
4704         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
4705         the reduction_phis argument from a vec<gimple *> to a
4706         vec<stmt_vec_info>.
4707         (vectorizable_reduction): Likewise the phis local variable that
4708         is passed to vect_create_epilog_for_reduction.  Update for new type
4709         of SLP_TREE_VEC_STMTS.
4710         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
4711         (vectorizable_live_operation): Likewise.
4712         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
4713         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
4715 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4717         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
4718         a gimple stmt to a stmt_vec_info.
4719         (vectorizable_condition, vectorizable_live_operation)
4720         (vectorizable_reduction, vectorizable_induction): Pass back the
4721         vectorized statement as a stmt_vec_info.
4722         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
4723         use of STMT_VINFO_VEC_STMT.
4724         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
4725         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
4726         as stmt_vec_infos rather than gimple stmts.
4727         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
4728         to a stmt_vec_info.
4729         (vectorizable_live_operation): Likewise.
4730         (vectorizable_reduction, vectorizable_induction): Likewise,
4731         updating use of STMT_VINFO_VEC_STMT.
4732         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
4733         of STMT_VINFO_VEC_STMT.
4734         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
4735         (vectorizable_simd_clone_call, vectorizable_conversion)
4736         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4737         (vectorizable_store, vectorizable_load, vectorizable_condition)
4738         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
4739         from a gimple stmt to a stmt_vec_info.
4740         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
4741         pointer to a stmt_vec_info to the vectorizable_* routines.
4743 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4745         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
4746         a gimple stmt to a stmt_vec_info.
4747         (is_pattern_stmt_p): Update accordingly.
4748         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
4749         (vect_record_grouped_load_vectors): Likewise.
4750         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
4751         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
4752         (vect_model_reduction_cost): Likewise.
4753         (vect_create_epilog_for_reduction): Likewise.
4754         (vectorizable_reduction, vectorizable_induction): Likewise.
4755         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4756         Return the stmt_vec_info for the pattern statement.
4757         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
4758         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
4759         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4760         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
4761         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
4762         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
4763         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
4764         (free_stmt_vec_info, vect_is_simple_use): Likewise.
4766 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4768         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
4769         (vect_finish_stmt_generation): Likewise.
4770         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4771         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
4772         (vect_build_gather_load_calls): Use the return value of the above
4773         functions instead of a separate call to vinfo_for_stmt.  Use narrow
4774         scopes for the input gimple stmt and wider scopes for the associated
4775         stmt_vec_info.  Use vec_info::lookup_def when setting these
4776         stmt_vec_infos from an SSA_NAME definition.
4777         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
4778         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
4779         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
4780         (vectorizable_store, vectorizable_load, vectorizable_condition)
4781         (vectorizable_comparison): Likewise.
4782         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
4783         (vectorizable_reduction): Likewise.
4785 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4787         * tree-vectorizer.h (vect_is_simple_use): Add an optional
4788         stmt_vec_info * parameter before the optional gimple **.
4789         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4790         (process_use, vect_get_vec_def_for_operand_1): Update callers.
4791         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
4792         * tree-vect-loop.c (vectorizable_reduction): Likewise.
4793         (vectorizable_live_operation): Likewise.
4794         * tree-vect-patterns.c (type_conversion_p): Likewise.
4795         (vect_look_through_possible_promotion): Likewise.
4796         (vect_recog_rotate_pattern): Likewise.
4797         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
4799 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4801         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
4802         a typedef to a wrapper class.
4803         (NULL_STMT_VEC_INFO): New macro.
4804         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
4805         (stmt_vec_info::operator*): New function.
4806         (stmt_vec_info::operator gimple *): Likewise.
4807         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
4808         (add_stmt_costs): Likewise.
4809         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
4810         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
4811         (vect_get_known_peeling_cost): Likewise.
4812         (vect_estimate_min_profitable_iters): Likewise.
4813         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4814         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
4815         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
4816         (vectorizable_store, free_stmt_vec_infos): Likewise.
4817         (new_stmt_vec_info): Change return type of xcalloc to
4818         _stmt_vec_info *.
4820 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4822         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
4823         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
4824         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
4825         a single_imm_use-based sequence.
4826         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
4828 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4830         * tree-vectorizer.h (vec_info::lookup_def): Declare.
4831         * tree-vectorizer.c (vec_info::lookup_def): New function.
4832         * tree-vect-patterns.c (vect_get_internal_def): Use it.
4833         (vect_widened_op_tree): Likewise.
4834         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4835         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
4836         (vectorizable_reduction): Likewise.
4837         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
4838         of a gimple *.
4839         (vect_is_slp_reduction): Update calls accordingly.  Use
4840         vec_info::lookup_def.
4841         (vect_is_simple_reduction): Likewise
4842         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
4844 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4846         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
4847         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
4848         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
4849         of vinfo_for_stmt.
4850         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
4851         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
4852         (vect_update_vf_for_slp, vect_analyze_loop_operations)
4853         (vect_is_slp_reduction, vectorizable_induction)
4854         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
4855         * tree-vect-patterns.c (vect_init_pattern_stmt):
4856         (vect_determine_min_output_precision_1, vect_determine_precisions)
4857         (vect_pattern_recog): Likewise.
4858         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
4859         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
4860         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
4861         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
4862         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
4863         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
4864         info field from a loop to a loop_vec_info.
4866 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4868         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
4869         (vec_info::add_stmt): Declare.
4870         * tree-vectorizer.c (vec_info::add_stmt): New function.
4871         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
4872         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
4873         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
4874         (vectorizable_induction): Likewise.
4875         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
4876         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
4877         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
4878         (vectorizable_load): Likewise.
4879         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
4880         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
4881         (vect_recog_gather_scatter_pattern): Likewise.
4882         (append_pattern_def_seq): Likewise.  Remove a check that is
4883         performed by add_stmt itself.
4885 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4887         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
4888         which make_ssa_name was called with new_stmt before new_stmt
4889         had been created.
4891 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4893         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
4894         split out from...
4895         (vect_is_slp_reduction): ...here...
4896         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
4897         that are already known to be false.
4899 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4901         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
4902         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
4903         STMT_VINFO_NUM_SLP_USES when it's true.
4904         (vect_free_slp_instance): Add a final_p parameter and pass it to
4905         vect_free_slp_tree.
4906         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
4907         (vect_analyze_slp_instance): Likewise.
4908         (vect_slp_analyze_operations): Likewise.
4909         (vect_slp_analyze_bb_1): Likewise.
4910         * tree-vectorizer.c (vec_info): Likewise.
4911         * tree-vect-loop.c (vect_transform_loop): Likewise.
4913 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4915         * tree-vect-loop.c (vectorizable_reduction): Assert that the
4916         function is not called for second and subsequent members of
4917         a reduction group.
4919 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
4921         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
4922         cases for nested loops from here to ...
4923         (vect_create_epilog_for_reduction): ...here.  Only call
4924         vect_is_simple_use for inner-loop reductions.
4926 2018-07-31  Martin Liska  <mliska@suse.cz>
4928         PR gcov-profile/85338
4929         PR gcov-profile/85350
4930         PR gcov-profile/85372
4931         * profile.c (struct location_triplet): New.
4932         (struct location_triplet_hash): Likewise.
4933         (output_location): Do not output a BB that
4934         is already recorded for a line.
4935         (branch_prob): Use streamed_locations.
4937 2018-07-31  Martin Liska  <mliska@suse.cz>
4939         PR gcov-profile/85370
4940         * coverage.c (coverage_begin_function): Do not mark target
4941         clones as artificial functions.
4943 2018-07-31  Martin Liska  <mliska@suse.cz>
4945         PR gcov-profile/83813
4946         PR gcov-profile/84758
4947         PR gcov-profile/85217
4948         PR gcov-profile/85332
4949         * profile.c (branch_prob): Do not record GOTO expressions
4950         for GIMPLE statements which locations are already streamed.
4952 2018-07-31  Olivier Hainque  <hainque@adacore.com>
4954         * gcc.c (handle_spec_function): Accept a soft_matched_part
4955         argument, as do_spec_1.  Pass it down to ...
4956         (eval_spec_function): Accept a soft_matched_part argument,
4957         and pass it down to ...
4958         (do_spec_2): Accept a soft_matched_part argument, and pass
4959         it down to do_spec_1.
4960         (do_spec_1): Pass soft_matched_part to handle_spec_function.
4961         (handle_braces): Update call to handle_spec_function.
4962         (driver::set_up_specs): Update calls to do_spec_2.
4963         (compare_debug_dump_opt_spec_function): Likewise.
4964         (compare_debug_self_opt_spec_function): Likewise.
4966 2018-07-31  Olivier Hainque  <hainque@adacore.com>
4968         * common.opt (nolibc): New option.
4969         * doc/invoke.texi (Link Options): Document it.
4970         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
4971         * config/alpha/linux.h: Likewise.
4972         * config/arc/elf.h: Likewise.
4973         * config/arm/uclinux-elf.h: Likewise.
4974         * config/arm/unknown-elf.h: Likewise.
4975         * config/avr/avrlibc.h: Likewise.
4976         * config/bfin/bfin.h: Likewise.
4977         * config/bfin/linux.h: Likewise.
4978         * config/bfin/uclinux.h: Likewise.
4979         * config/darwin.h: Likewise.
4980         * config/darwin10.h: Likewise.
4981         * config/darwin12.h: Likewise.
4982         * config/gnu-user.h: Likewise.
4983         * config/lm32/uclinux-elf.h: Likewise.
4984         * config/pa/pa-hpux11.h: Likewise.
4985         * config/pa/pa64-hpux.h: Likewise.
4986         * config/sparc/sparc.h: Likewise.
4988 2018-07-31  Olivier Hainque  <hainque@adacore.com>
4990         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
4991         undefined variables.
4993 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
4995         PR target/86640
4996         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
4997         instead of GEN_INT.
4999 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5001         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
5002         terminated string literal.
5004 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
5006         PR rtl-optimization/85160
5007         * combine.c (is_just_move): New function.
5008         (try_combine): Allow combining two instructions into two if neither of
5009         the original instructions was a move.
5011 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
5013         PR target/86673
5014         * doc/extend.texi (Global Register Variables): Discourage use of type
5015         qualifiers.
5016         (Local Register Variables): Likewise.
5018 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
5020         PR tree-optimization/86506
5021         * hwint.h (ceil_log2): Resync with hwint.c implementation.
5023 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
5025         PR target/86547
5026         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
5027         hard_regno, make sure no insn between `from` and `to` clobbers it.
5029 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
5030             Tom de Vries  <tdevries@suse.de>
5032         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
5033         (PTX_DEFAULT_RUNTIME_DIM): ... this.
5034         (nvptx_goacc_validate_dims): Set default worker and gang dims to
5035         PTX_DEFAULT_RUNTIME_DIM.
5036         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
5038 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
5040         * config/pa/pa.c (pa_output_addr_vec): Align address table.
5041         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
5042         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
5044 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
5046         * config/rs6000/constraints.md (wG constraint): Delete, no longer
5047         used.
5048         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
5049         predicate to reflect toc fusion has been deleted.
5050         (toc_fusion_mem_raw): Delete, no longer used.
5051         (toc_fusion_mem_wrapped): Likewise.
5052         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
5053         fusion mask bit.
5054         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
5055         Delete, no longer used.
5056         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
5057         meant to be used for toc fusion.
5058         (rs6000_debug_print_mode): Delete toc fusion debugging.
5059         (rs6000_debug_reg_global): Likewise.
5060         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
5061         fusion and secondary reload support that were never used.
5062         (rs6000_option_override_internal): Delete TOC fusion, that was only
5063         partially defined, and it did not work unless you also used the
5064         -mcmodel= switch.
5065         (rs6000_legitimate_address_p): Delete TOC fusion support.
5066         (rs6000_opt_masks): Likewise.
5067         (fusion_wrap_memory_address): Delete function, no longer used.
5068         (fusion_split_address); Delete TOC fusion support.
5069         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
5070         longer used with toc fusion being deleted.
5071         (TARGET_TOC_FUSION_FP): Likewise.
5072         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
5073         UNSPEC.
5074         (toc fusion spliter): Delete TOC fusion support.
5075         (toc_fusionload_<mode>): Likewise.
5076         (toc_fusionload_di): Likewise.
5077         (fusion_gpr_load_<mode>): Delete generator function, this insn no
5078         longer needs to be named.  Rename predicate to delete TOC fusion.
5079         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5080         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5081         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
5082         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
5083         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
5085 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5087         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
5088         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
5089         __int128 in built-in function prototypes.
5090         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
5091         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
5093 2018-07-27  Martin Sebor  <msebor@redhat.com>
5095         PR tree-optimization/86696
5096         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
5097         types, including enums.
5098         (handle_char_store): Be prepared for the above function to fail.
5100 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
5102         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
5103         when optimization level is lower than 2 or optimize for size.
5105 2018-07-26  Martin Sebor  <msebor@redhat.com>
5107         PR tree-optimization/86043
5108         PR tree-optimization/86042
5109         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
5110         (get_string_cst_length): Rename...
5111         (get_min_string_length): ...to this.  Add argument.
5112         (handle_char_store): Extend to handle multi-character stores by
5113         MEM_REF.
5114         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
5115         * tree.h (initializer_zerop): Add argument.
5117 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
5119         PR middle-end/86660
5120         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
5121         declare target to variables if they have always,{to,from,tofrom} map
5122         kinds.
5124 2018-07-26  Martin Liska  <mliska@suse.cz>
5126         PR lto/86548
5127         * lto-wrapper.c: Add linker_output as prefix
5128         for ltrans_output_file.
5130 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
5132         PR rtl-optimization/85805
5133         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
5134         value for hard registers if that was written in the same mode.
5136 2018-07-26  Martin Liska  <mliska@suse.cz>
5138         PR gcov-profile/86536
5139         * gcov.c (format_gcov): Use printf format %.*f directly
5140         and do not handle special values.
5142 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5144         * common/config/arc/arc-common.c (arc_option_optimization_table):
5145         Update default optimizations for size.
5147 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5149         * config/arc/arc.md (movsf_insn): Add short instruction selection.
5150         * config/arc/constraints.md (CfZ): New constraint.
5151         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
5152         (subsf3_fpu): Likewise.
5153         (cmpsf_fpu): Likewise.
5154         (cmpsf_fpu_uneq): Likewise.
5156 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5158         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
5159         accesses as well.
5160         (arc_is_uncached_mem_p): uncached applies to both the variable and
5161         the pointer.
5163 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
5165         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
5166         register names.
5168 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
5170         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
5171         field "m_scopes" from vec to auto_vec.
5173 2018-07-25  Martin Liska  <mliska@suse.cz>
5175         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
5176         return type.
5178 2018-07-25  Richard Biener  <rguenther@suse.de>
5180         PR debug/86654
5181         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
5182         special wrt context_die late.
5183         (gen_subprogram_die): Re-use DIEs in local scope.
5185 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
5187         PR tree-optimization/86644
5188         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
5190 2018-07-25  Martin Liska  <mliska@suse.cz>
5192         PR middle-end/86645
5193         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
5194         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
5196 2018-07-25  Martin Liska  <mliska@suse.cz>
5198         PR sanitizer/79635
5199         * params.def: Explain ASan abbreviation and provide
5200         a documentation link.
5202 2018-07-24  Martin Sebor  <msebor@redhat.com>
5204         PR tree-optimization/86622
5205         PR tree-optimization/86532
5206         * builtins.h (string_length): Declare.
5207         * builtins.c (c_strlen): Correct handling of non-constant offsets.
5208         (check_access): Be prepared for non-constant length ranges.
5209         (string_length): Make extern.
5210         * expr.c (string_constant): Only handle the minor non-constant
5211         array index.  Use string_constant to compute the length of
5212         a generic string constant.
5214 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
5216         PR tree-optimization/86618
5217         * tree-vect-stmts.c (vectorizable_call): Don't take the address
5218         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
5220 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
5222         PR tree-optimization/86636
5223         * json.cc (json::object::set): Fix comment.  Add assertions.
5224         (json::array::append): Move here from json.h.  Add comment and an
5225         assertion.
5226         (json::string::string): Likewise.
5227         * json.h (json::array::append): Move to json.cc.
5228         (json::string::string): Likewise.
5229         * optinfo-emit-json.cc
5230         (optrecord_json_writer::impl_location_to_json): Assert that we
5231         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
5232         wrapper around it.  Expand the location once, rather than three
5233         times.
5234         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5235         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
5236         wrappers.
5237         (optrecord_json_writer::optinfo_to_json): Likewise, in four
5238         places.  Fix some overlong lines.
5240 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
5242         * config/aarch64/aarch64-simd.md
5243         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
5244         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
5245         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
5246         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
5247         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
5248         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
5249         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
5250         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
5251         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
5253 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
5255         PR middle-end/86627
5256         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
5257         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
5258         and abs_d == d, do the power of two handling if profitable.
5260 2018-07-24  Richard Biener  <rguenther@suse.de>
5262         * match.pd: Add BIT_FIELD_REF canonicalizations.
5264 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5266         PR c/86617
5267         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
5269 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5271         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
5272         terminated STRING_CST object.
5274 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5276         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
5278 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
5280         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
5281         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
5282         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
5283         the elements into a register.
5284         (rs6000_split_v4si_init_di_reg): Delete.
5285         (rs6000_split_v4si_init): Delete.
5286         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
5287         (vsx_init_v4si): Rewrite as a define_expand.
5289 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
5291         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
5292         zero_extend argument from memory): New.
5294 2018-07-22  Martin Sebor  <msebor@redhat.com>
5296         PR bootstrap/86621
5297         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
5298         diagnosing calls with unknown arguments unless -Walloca-larger-than
5299         is restricted to less than PTRDIFF_MAX bytes.
5301 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
5303         * doc/gcov.texi (Invoking Gcov): Editorial changes.
5305 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
5307         * pretty-print.c (text_info::set_location): Remove redundant
5308         "line_table" parameter from call to rich_location::set_range.
5310 2018-07-20  Martin Sebor  <msebor@redhat.com>
5312         PR middle-end/82063
5313         * builtins.c (expand_builtin_alloca): Adjust.
5314         * calls.c (alloc_max_size): Simplify.
5315         * cgraphunit.c (cgraph_node::expand): Adjust.
5316         * common.opt (larger_than_size, warn_frame_larger_than): Remove
5317         variables.
5318         (frame_larger_than_size): Same.
5319         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
5320         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
5321         Initialize.
5322         * doc/invoke.texi (GCC Command Options): Document option arguments.
5323         Explain byte-size arguments and suffixes.
5324         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
5325         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
5326         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
5327         * doc/options.texi (UInteger): Expand.
5328         (Host_Wide_Int, ByteSize): Document new properties.
5329         * final.c (final_start_function_1): Include sizes in an error message.
5330         * function.c (frame_offset_overflow): Same.
5331         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
5332         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
5333         Diagnose unbounded alloca calls only for limits of less than
5334         PTRDIFF_MAX.
5335         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
5336         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
5337         for alloca(0).
5338         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
5339         only for limits of less than PTRDIFF_MAX.
5340         * langhooks-def.h (lhd_handle_option): Change function argument
5341         to HOST_WIDE_INT.
5342         * langhooks.c (lhd_handle_option): Same.
5343         * langhooks.h (handle_option): Same.
5344         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
5345         ByteSize flags.
5346         (var_type, var_type_struct): Same.
5347         (var_set): Handle ByteSize flag.
5348         * optc-gen.awk: Add comments to output to ease debugging.  Make
5349         use of HOST_WIDE_INT where appropriate.
5350         * opts-gen-save.awk:  Use %lx to format unsigned long.
5351         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
5352         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
5353         arguments.  Parse bytes-size suffixes.
5354         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
5355         (enum_value_to_arg): Same.
5356         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
5357         (handle_option): Adjust.
5358         (generate_option): Change function argument to HOST_WIDE_INT.
5359         (cmdline_handle_error): Adjust.
5360         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
5361         (set_option): Change function argument to HOST_WIDE_INT.
5362         (option_enabled): Handle cl_host_wide_int.
5363         (get_option_state): Handle CLVC_SIZE.
5364         (control_warning_option): Same.
5365         * opts.c (common_handle_option): Change function argument to
5366         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
5367         OPT_Wvla_larger_than_.
5368         * opts.h (enum cl_var_type): Add an enumerator.
5369         * stor-layout.c (layout_decl): Print a more meaningful warning.
5370         * toplev.c (output_stack_usage): Adjust.
5372 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
5374         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
5375         call to inline_expand_builtin_string_cmp.
5376         (expand_builtin_strcmp): Likewise.
5377         (expand_builtin_strncmp): Likewise.
5378         (inline_string_cmp): Delete the last parameter, change char_type_node
5379         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
5380         two operands.
5381         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
5382         the inlining expansion on target where the type of the call has same or
5383         narrower precision than unsigned char.
5385 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
5387         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
5388         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
5389         * common.opt (fsave-optimization-record): New option.
5390         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
5391         * doc/invoke.texi (-fsave-optimization-record): New option.
5392         * dumpfile.c: Include "optinfo-emit-json.h".
5393         (struct kv_pair): Move to coretypes.h.
5394         (optgroup_options): Make non-static.
5395         (dump_context::end_scope): Call
5396         optimization_records_maybe_pop_dump_scope.
5397         * dumpfile.h (optgroup_options): New decl.
5398         * json.cc: New file.
5399         * json.h: New file.
5400         * optinfo-emit-json.cc: New file.
5401         * optinfo-emit-json.h: New file.
5402         * optinfo.cc: Include "optinfo-emit-json.h".
5403         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
5404         (optinfo_enabled_p): Check optimization_records_enabled_p.
5405         (optinfo_wants_inlining_info_p): Likewise.
5406         * optinfo.h: Update comment.
5407         * profile-count.c (profile_quality_as_string): New function.
5408         * profile-count.h (profile_quality_as_string): New decl.
5409         (profile_count::quality): New accessor.
5410         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
5411         and optinfo_emit_json_cc_tests.
5412         * selftest.h (selftest::json_cc_tests): New decl.
5413         (selftest::optinfo_emit_json_cc_tests): New decl.
5414         * toplev.c: Include "optinfo-emit-json.h".
5415         (compile_file): Call optimization_records_finish.
5416         (do_compile): Call optimization_records_start.
5417         * tree-ssa-live.c: Include optinfo.h.
5418         (remove_unused_scope_block_p): Retain inlining information if
5419         optinfo_wants_inlining_info_p returns true.
5421 2018-07-20  Richard Biener  <rguenther@suse.de>
5423         PR debug/86585
5424         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
5425         to cover -flto-partition=none.
5427 2018-07-20  Martin Liska  <mliska@suse.cz>
5429         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
5430         (get_decl_source_range): Remove unused function.
5432 2018-07-20  Richard Biener  <rguenther@suse.de>
5434         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
5435         (struct vn_phi_s): Likewise.
5436         (struct vn_reference_s): Likewise.
5437         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
5438         for searching the slot of an entry known to be in the hash itself.
5439         (vn_phi_hasher::equal): Likewise.
5440         (vn_reference_hasher::equal): Likewise.
5441         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
5442         globals.
5443         (optimistic_info, current_info): Remove, keeping only valid_info.
5444         (vn_reference_lookup_1): Remove fallback lookup.
5445         (vn_reference_lookup_2): Likewise.
5446         (vn_nary_op_lookup_1): Likewise.
5447         (vn_phi_lookup): Likewise.
5448         (vn_nary_build_or_lookup_1): Make sure to not chain the built
5449         hash element.
5450         (vn_reference_insert): Adjust, chain the inserted hash element
5451         at last_inserted_ref.
5452         (vn_reference_insert_pieces): Likewise.
5453         (visit_reference_op_call): Likewise.
5454         (vn_nary_op_insert_into): Chain the inserted hash element at
5455         last_inserted_nary.
5456         (vn_nary_op_insert_pieces): Adjust.
5457         (vn_nary_op_insert): Likewise.
5458         (vn_nary_op_insert_stmt): Likewise.
5459         (vn_phi_insert): Adjust, chain the inserted hash element at
5460         last_inserted_phi.
5461         (process_scc): Remove clearing and copying the optimistic
5462         table.  Instead remove elements inserted during an optimistic
5463         iteration from the single table we maintain.
5464         (init_scc_vn): Adjust.
5465         (free_scc_vn): Likewise.
5466         (sccvn_dom_walker::record_cond): Likewise.
5467         (sccvn_dom_walker::after_dom_children): Likewise.
5469 2018-07-19  Martin Sebor  <msebor@redhat.com>
5471         PR tree-optimization/84047
5472         PR tree-optimization/83776
5473         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
5474         (check_array_bounds): Call it.
5476 2018-07-19  Martin Sebor  <msebor@redhat.com>
5478         * align.h (align_flags): Use member initialization.
5480 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
5482         * Makefile.in (OBJS): Add optinfo.o.
5483         * coretypes.h (class symtab_node): New forward decl.
5484         (struct cgraph_node): New forward decl.
5485         (class varpool_node): New forward decl.
5486         * dump-context.h: New file.
5487         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
5488         "tree-pass.h".
5489         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
5490         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
5491         (set_alt_dump_file): Likewise.
5492         (dump_context::~dump_context): New dtor.
5493         (dump_gimple_stmt): Move implementation to...
5494         (dump_context::dump_gimple_stmt): ...this new member function.
5495         Add the stmt to any pending optinfo, creating one if need be.
5496         (dump_gimple_stmt_loc): Move implementation to...
5497         (dump_context::dump_gimple_stmt_loc): ...this new member function.
5498         Start a new optinfo and add the stmt to it.
5499         (dump_gimple_expr): Move implementation to...
5500         (dump_context::dump_gimple_expr): ...this new member function.
5501         Add the stmt to any pending optinfo, creating one if need be.
5502         (dump_gimple_expr_loc): Move implementation to...
5503         (dump_context::dump_gimple_expr_loc): ...this new member function.
5504         Start a new optinfo and add the stmt to it.
5505         (dump_generic_expr): Move implementation to...
5506         (dump_context::dump_generic_expr): ...this new member function.
5507         Add the tree to any pending optinfo, creating one if need be.
5508         (dump_generic_expr_loc): Move implementation to...
5509         (dump_context::dump_generic_expr_loc): ...this new member
5510         function.  Add the tree to any pending optinfo, creating one if
5511         need be.
5512         (dump_printf): Move implementation to...
5513         (dump_context::dump_printf_va): ...this new member function.  Add
5514         the text to any pending optinfo, creating one if need be.
5515         (dump_printf_loc): Move implementation to...
5516         (dump_context::dump_printf_loc_va): ...this new member function.
5517         Start a new optinfo and add the stmt to it.
5518         (dump_dec): Move implementation to...
5519         (dump_context::dump_dec): ...this new member function.  Add the
5520         value to any pending optinfo, creating one if need be.
5521         (dump_context::dump_symtab_node): New member function.
5522         (dump_context::get_scope_depth): New member function.
5523         (dump_context::begin_scope): New member function.
5524         (dump_context::end_scope): New member function.
5525         (dump_context::ensure_pending_optinfo): New member function.
5526         (dump_context::begin_next_optinfo): New member function.
5527         (dump_context::end_any_optinfo): New member function.
5528         (dump_context::s_current): New global.
5529         (dump_context::s_default): New global.
5530         (dump_scope_depth): Delete global.
5531         (dumpfile_ensure_any_optinfo_are_flushed): New function.
5532         (dump_symtab_node): New function.
5533         (get_dump_scope_depth): Reimplement in terms of dump_context.
5534         (dump_begin_scope): Likewise.
5535         (dump_end_scope): Likewise.
5536         (selftest::temp_dump_context::temp_dump_context): New ctor.
5537         (selftest::temp_dump_context::~temp_dump_context): New dtor.
5538         (selftest::verify_item): New function.
5539         (ASSERT_IS_TEXT): New macro.
5540         (ASSERT_IS_TREE): New macro.
5541         (ASSERT_IS_GIMPLE): New macro.
5542         (selftest::test_capture_of_dump_calls): New test.
5543         (selftest::dumpfile_c_tests): Call it.
5544         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
5545         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
5546         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
5547         descriptive comment.
5548         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
5549         (dump_node, dump_bb): Move these unrelated decls.
5550         (class dump_manager): Add leading comment.
5551         * optinfo.cc: New file.
5552         * optinfo.h: New file.
5554 2018-07-19  Michael Collison  <michael.collison@arm.com>
5555             Richard Henderson <rth@redhat.com>
5557         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
5558         (subti): Handle op1 zero.
5559         (subvti4, usub4ti4): New.
5560         (*sub<GPI>3_compare1_imm): New.
5561         (sub<GPI>3_carryinCV): New.
5562         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
5563         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
5565 2018-07-19  Michael Collison  <michael.collison@arm.com>
5566             Richard Henderson <rth@redhat.com>
5568         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
5569         (addti3): Create simpler code if low part is already known to be 0.
5570         (addvti4, uaddvti4): New.
5571         (*add<GPI>3_compareC_cconly_imm): New.
5572         (*add<GPI>3_compareC_cconly): New.
5573         (*add<GPI>3_compareC_imm): New.
5574         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
5575         handle constants within this pattern..
5576         (*add<GPI>3_compareV_cconly_imm): New.
5577         (*add<GPI>3_compareV_cconly): New.
5578         (*add<GPI>3_compareV_imm): New.
5579         (add<GPI>3_compareV): New.
5580         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
5581         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
5582         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
5583         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
5584         with 'comparison' operator.
5585         (*add<GPI>3_compareV_cconly_imm): Ditto.
5586         (*add<GPI>3_compareV_cconly): Ditto.
5587         (*add<GPI>3_compareV_imm): Ditto.
5588         (add<GPI>3_compareV): Ditto.
5589         (add<mode>3_carryinC): Ditto.
5590         (*add<mode>3_carryinC_zero): Ditto.
5591         (*add<mode>3_carryinC): Ditto.
5592         (add<mode>3_carryinV): Ditto.
5593         (*add<mode>3_carryinV_zero): Ditto.
5594         (*add<mode>3_carryinV): Ditto.
5596 2018-07-19  Michael Collison  <michael.collison@arm.com>
5597             Richard Henderson <rth@redhat.com>
5599         * config/aarch64/aarch64-modes.def (CC_V): New.
5600         * config/aarch64/aarch64-protos.h
5601         (aarch64_addti_scratch_regs): Declare
5602         (aarch64_subvti_scratch_regs): Declare.
5603         (aarch64_expand_subvti): Declare.
5604         (aarch64_gen_unlikely_cbranch): Declare
5605         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
5606         for signed overflow using CC_Vmode.
5607         (aarch64_get_condition_code_1): Handle CC_Vmode.
5608         (aarch64_gen_unlikely_cbranch): New function.
5609         (aarch64_addti_scratch_regs): New function.
5610         (aarch64_subvti_scratch_regs): New function.
5611         (aarch64_expand_subvti): New function.
5613 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5615         * config/aarch64/aarch64-option-extensions.def: New entry for profile
5616         extension.
5617         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5618         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5619         extension.
5621 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5623         PR target/83009
5624         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5625         address check not strict.
5627 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
5629         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
5630         Umq with Umn.
5631         (store_pair_lanes<mode>): Likewise.
5632         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
5633         enum value 'ADDR_QUERY_LDP_STP_N'.
5634         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
5635         (aarch64_print_address_internal): Add declaration.
5636         (aarch64_print_ldpstp_address): Remove.
5637         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
5638         (aarch64_print_operand): Change printing of 'y'.
5639         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
5640         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
5641         'true' rather than '1'.
5642         * config/aarch64/constraints.md (Uml): Likewise.
5643         (Uml): Rename to Umn.
5644         (Umq): Remove.
5646 2018-07-19  Richard Biener  <rguenther@suse.de>
5648         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
5649         a trailing array.
5650         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
5651         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
5652         (vn_reference_hasher): Likewise.
5653         (struct vn_tables_s): Remove obstack and alloc-pool members.
5654         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
5655         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
5656         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
5657         (vn_reference_insert_pieces): Likewise.
5658         (alloc_vn_nary_op_noinit): Adjust.
5659         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
5660         (vn_phi_eq): Adjust.
5661         (shared_lookup_phiargs): Remove.
5662         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
5663         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
5664         (visit_reference_op_call): Likewise.
5665         (copy_nary, copy_phi, copy_reference): Remove.
5666         (process_scc): Rewind the obstack when iterating.  Do not
5667         copy the elements to valid_info but just move them from one
5668         hashtable to the other.
5669         (allocate_vn_table): Adjust.
5670         (free_vn_table): Likewise.
5671         (init_scc_vn): Likewise.
5672         (free_scc_vn): Likewise.
5674 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
5676         PR target/86560
5677         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
5678         indirect_return as function type attribute.
5679         (ix86_attribute_table): Change indirect_return to function
5680         type attribute.
5681         * doc/extend.texi: Update indirect_return attribute.
5683 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
5685         * wide-int.h (widest2_int): New.
5686         * gimple-fold.c (arith_overflowed_p): Use it.
5687         * tree.h (widest2_int_cst): New.
5688         * tree-vrp.c (wide_int_binop_overflow): Rename from
5689         vrp_int_const_binop.
5690         Rewrite to work on trees.
5691         (extract_range_from_multiplicative_op_1): Abstract code to...
5692         (wide_int_range_min_max): ...here.
5693         (wide_int_range_cross_product): ...and here.
5694         (extract_range_from_binary_expr_1): Abstract overflow code to...
5695         (wide_int_range_mult_wrapping): ...here.
5696         * tree-vrp.h (wide_int_range_cross_product): New.
5697         (wide_int_range_mult_wrapping): New.
5699 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
5700             Julia Koval  <julia.koval@intel.com>
5702         * config/i386/x86-tune-costs.h (skylake_memcpy,
5703         skylake_memset): Replace rep_prefix with unrolling for size 512.
5705 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
5707         PR middle-end/86544
5708         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
5709         comparision with EQ_EXPR in last stmt.
5711 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5713         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
5714         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
5715         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
5716         previously known as "PowerPC AltiVec Built-in Functions".  Move
5717         some material to new subsubsections "PowerPC AltiVec Built-in
5718         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
5719         ISA 2.07".
5720         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
5721         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
5722         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
5723         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
5725 2018-07-18  Richard Biener  <rguenther@suse.de>
5727         PR tree-optimization/86557
5728         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
5729         EXACT_DIV_EXPR.
5731 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
5733         * config/s390/s390.c (s390_function_profiler): Generate CFI.
5735 2018-07-17  Jeff Law  <law@redhat.com>
5737         * config/arm/arm.c (get_label_padding): Update for recent
5738         changes to label_to_alignment.
5740         PR tree-optimization/86010
5741         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
5743         * config/mips/mips.c (vr4130_align_insns): Update for recent
5744         changes to label_to_alignment.
5746         * config/frv/frv.c (frv_label_align): Update for recent changes
5747         to label_to_alignment.
5749         * config/nios2/nios2.c (nios2_label_align): Update for recent
5750         changes which dropped ALIGN_LABELS_LOG.
5752 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
5754         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
5755         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
5757 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
5759         * config/arc/arc.c (arc_label_align): Use align_labels instead of
5760         deprecated align_labels_log.
5762 2018-07-17  Richard Biener  <rguenther@suse.de>
5764         PR lto/86456
5765         * dwarf2out.c (init_sections_and_labels): Always generate
5766         a debug_line_str_section for early LTO debug.
5767         (dwarf2out_finish): Reset debug_line_str_hash output early.
5768         Bump counter for extra dwarf5 .debug_loc labels to not conflict
5769         with fat LTO part.
5770         (dwarf2out_early_finish): Output debug_line_str.
5772 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
5774         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
5775         index register on z196 or later.
5777 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
5779         * config/s390/s390.c (s390_default_align): Set default function
5780         alignment to 16.
5781         (s390_override_options_after_change): Call s390_default align.
5782         (s390_option_override_internal): Call s390_default_align.
5783         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
5785 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
5787         PR middle-end/86542
5788         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
5789         to _looptemp_ clauses, other than the first two.
5791 2018-07-17  Martin Liska  <mliska@suse.cz>
5793         * opts.c: Do not enable OPT_falign_* for -Os.
5795 2018-07-17  Martin Liska  <mliska@suse.cz>
5797         * align.h (MAX_CODE_ALIGN): New.
5798         (MAX_CODE_ALIGN_VALUE): New.
5799         * common/config/i386/i386-common.c (ix86_handle_option):
5800         (MAX_CODE_ALIGN): Moved to align.h.
5801         * final.c (MAX_CODE_ALIGN): Likewise.
5802         * opts.c (parse_and_check_align_values):
5803         (MAX_CODE_ALIGN): Likewise.
5804         (MAX_CODE_ALIGN_VALUE): Likewise.
5806 2018-07-17  Martin Liska  <mliska@suse.cz>
5808         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
5809         in order to fulfil coding style.
5810         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
5811         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5812         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
5813         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
5814         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
5815         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
5816         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
5817         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
5818         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
5820 2018-07-17  Martin Liska  <mliska@suse.cz>
5822         * align.h: New file.
5823         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
5824         directly.
5825         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
5826         align_flags of label_to_alignment.
5827         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
5828         align_flags class.
5829         * config/m68k/m68k.c: Do not use removed align_labels_value and
5830         align_loops_value.
5831         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
5832         (LOOP_ALIGN): Likewise.
5833         (LABEL_ALIGN): Likewise.
5834         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5835         Remove not used macro.
5836         (rs6000_loop_align): Change return type to align_flags.
5837         (rs6000_loop_align_max_skip): Remove.
5838         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
5839         Change return type to align_flags.
5840         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
5841         Remove not used macro.
5842         (rs6000_loop_align):  Change return type to align_flags.
5843         (rs6000_loop_align_max_skip): Remove.
5844         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
5845         * config/rx/rx-protos.h (rx_align_for_label): Make it
5846         static function.
5847         * config/rx/rx.c (rx_align_for_label): Change return type
5848         to align_flags.
5849         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5850         macro definitions.
5851         into align_flags class.
5852         (LABEL_ALIGN): Likewise.
5853         (LOOP_ALIGN): Likewise.
5854         * config/s390/s390.c (s390_label_align): Use align_flags
5855         class member.
5856         (s390_asm_output_function_label): Likewise.
5857         * config/sh/sh.c (sh_override_options_after_change):
5858         Use align_flags class directly without macros.
5859         (find_barrier): Likewise.
5860         (barrier_align): Likewise.
5861         (sh_loop_align): Likewise.
5862         * config/spu/spu.c (spu_option_override):
5863         Use align_flags_tuple::get_value instead of removed macros.
5864         (spu_sched_init): Likewise.
5865         * config/spu/spu.h (GTY): Likewise.
5866         * config/visium/visium.c (visium_option_override):
5867         Set "8" as default secondary alignment.
5868         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
5869         in order to guarantee secondary alignment of 8.
5870         * coretypes.h: Include align.h header file.
5871         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
5872         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
5873         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
5874         * doc/tm.texi.in: Likewise.
5875         * final.c (struct label_alignment): Remove not used structure.
5876         (LABEL_ALIGN): Change type to align_flags.
5877         (LOOP_ALIGN): Likewise.
5878         (JUMP_ALIGN): Likewise.
5879         (default_loop_align_max_skip): Remove.
5880         (default_label_align_max_skip): Likewise.
5881         (default_jump_align_max_skip): Likewise.
5882         (default_label_align_after_barrier_max_skip):
5883         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
5884         (LABEL_TO_MAX_SKIP): Remove.
5885         (label_to_alignment): Return align_flags type instead of integer.
5886         (label_to_max_skip): Remove.
5887         (align_fuzz): Use align_flags type.
5888         (compute_alignments): Use align_flags type and use align_flags::max
5889         to combine multiple alignments.
5890         (grow_label_align): Grow vec instead of C array.
5891         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
5892         (shorten_branches):  Use align_flags type and use align_flags::max
5893         to combine multiple alignments.
5894         (final_scan_insn_1): Remove usage of secondary alignment that comes
5895         from label alignment, but instead use proper secondary alignment
5896         which is computed in grow_label_align.
5897         * flags.h (struct align_flags_tuple): Move to align.h.
5898         (struct align_flags): Likewise.
5899         (state_align_loops): Rename to align_loops.
5900         (state_align_jumps): Rename to align_jumps.
5901         (state_align_labels): Rename to align_labels.
5902         (state_align_functions): Rename to align_functions.
5903         (align_loops_log): Remove.
5904         (align_jumps_log): Remove.
5905         (align_labels_log): Remove.
5906         (align_functions_log): Remove.
5907         (align_loops_max_skip): Remove.
5908         (align_jumps_max_skip): Remove.
5909         (align_labels_max_skip): Remove.
5910         (align_functions_max_skip): Remove.
5911         (align_loops_value): Remove.
5912         (align_jumps_value): Remove.
5913         (align_labels_value): Remove.
5914         (align_functions_value): Remove.
5915         * output.h (label_to_alignment): Change return type to align_flags.
5916         (label_to_max_skip): Remove.
5917         * target.def: Remove loop_align_max_skip, label_align_max_skip,
5918         jump_align_max_skip macros.
5919         * targhooks.h (default_loop_align_max_skip): Remove.
5920         (default_label_align_max_skip): Likewise.
5921         (default_jump_align_max_skip): Likewise.
5922         (default_label_align_after_barrier_max_skip): Remove.
5923         * toplev.c (read_log_maxskip): Use ::normalize function.
5924         (parse_N_M): Remove not used argument and also call ::normalize.
5925         (parse_alignment_opts): Do not pass unused arguments.
5926         * varasm.c (assemble_start_function): Use directly align_functions
5927         instead of removed macros.
5928         * system.h: Do not poison removed macros.
5930 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
5932         PR middle-end/86539
5933         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
5934         and cond temporaries don't have reference type if iterator has
5935         pointer type.  For init use &for_pre_body instead of pre_p if
5936         for_pre_body is non-empty.
5938 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
5940         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
5941         double-double modes to SFmode directly directly.
5942         (trunc<mode>sf2_fprs): Delete.
5944 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
5946         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
5947         for conversions between IFmode and the decimal floating point modes.
5948         (init_float128_ieee): Use the correct names for conversions between
5949         KFmode and the decimal floating point modes.
5951 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
5953         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
5954         for the conversions between TDmode and IFmode.
5955         (init_float128_ieee): Use more correct names for the conversions
5956         between TDmode and KFmode.
5958 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
5960         PR tree-optimization/86526
5961         * builtins.c (expand_builtin_memcmp): Formatting fixes.
5962         (inline_expand_builtin_string_cmp): Likewise.
5963         (inline_string_cmp): Likewise.  Use c_readstr instead of
5964         builtin_memcpy_read_str.  Add unit_mode temporary.
5966 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5968         PR middle-end/86528
5969         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
5970         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
5972 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
5974         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
5975         Alphabetize prototypes of built-in functions, separating out
5976         built-in functions that are listed in this section but should be
5977         described elsewhere.
5979 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
5981         PR target/86511
5982         * expmed.c (emit_store_flag): Do not emit setcc followed by a
5983         conditional move when trapping comparison was split to a
5984         non-trapping one (and vice versa).
5986 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
5988         * config/s390/s390.c (s390_function_profiler): Generate nops
5989         instead of profiler call sequences.
5990         * config/s390/s390.opt: Add the new option.
5992 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
5994         * config/s390/s390.c (s390_function_profiler): Generate
5995         __mcount_loc section.
5996         * config/s390/s390.opt: Add the new option.
5998 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
6000         * common.opt: Add the new warning.
6001         * config/s390/s390.c (s390_function_profiler): Emit "brasl
6002         %r0,__fentry__" when -mfentry is specified.
6003         (s390_option_override_internal): Disallow -mfentry for 31-bit
6004         CPUs.
6005         * config/s390/s390.opt: Add the new option.
6007 2018-07-16  Richard Biener  <rguenther@suse.de>
6009         PR lto/86523
6010         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
6011         for function-local FUNCTION_DECL and RESULT_DECL immediately.
6013 2018-07-16  Martin Liska  <mliska@suse.cz>
6015         PR ipa/86529
6016         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
6017         to ::get_create.
6019 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
6021         * config/arc/arcHS.md: Update ARCHS scheduling rules.
6023 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
6025         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
6026         for ARCHS4x.
6027         * config/arc/arc-cpus.def (hs4x): New cpu.
6028         (hs4xd): Likewise.
6029         * config/arc/arc-tables.opt: Regenerate.
6030         * config/arc/arc.c (arc_sched_issue_rate): New function.
6031         (TARGET_SCHED_ISSUE_RATE): Define.
6032         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
6033         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
6034         fpu_cvt.
6035         (attr tune): Add ARCHS4x tune values.
6036         (attr tune_dspmpy): Define.
6037         (*tst): Correct instruction type.
6038         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
6039         * config/arc/arcHS4x.md: New file.
6040         * config/arc/fpu.md: Update instruction type attributes.
6041         * config/arc/t-multilib: Regenerate.
6043 2018-07-16  Tom de Vries  <tdevries@suse.de>
6045         PR debug/86455
6046         * var-tracking.c (vt_initialize): Fix pre_dec handling.
6048 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
6050         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
6051         early clobber.
6053 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
6055         PR tree-optimization/86514
6056         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
6057         conversion to a boolean type from a type with greater precision.
6059 2018-07-16  Tom de Vries  <tdevries@suse.de>
6061         * var-tracking.c (vt_initialize): Print adjusted insn slim if
6062         dump_flags request TDF_SLIM.
6064 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
6066         * fold-const.c (int_const_binop_1): Abstract...
6067         (wide_int_binop): ...wide int code here.
6068         (poly_int_binop): ...poly int code here.
6069         Abstract the rest of int_const_binop_1 into int_const_binop.
6070         * fold-const.h (wide_int_binop): New.
6071         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
6072         Remove useless PLUS/MINUS_EXPR case.
6073         (zero_nonzero_bits_from_vr): Move wide int code...
6074         (zero_nonzero_bits_from_bounds): ...here.
6075         (extract_range_from_binary_expr_1): Move mask optimization code...
6076         (range_easy_mask_min_max): ...here.
6077         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
6078         (range_easy_mask_min_max): New.
6080 2018-07-15  Jeff Law  <law@redhat.com>
6082         PR target/85993
6083         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
6084         block.
6086 2018-07-14  Jim Wilson  <jimw@sifive.com>
6088         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
6090 2018-07-14  Paul Koning  <ni1d@arrl.net>
6092         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
6094 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
6096         * lto-streamer-out.c (copy_function_or_variable): Dump info about
6097         copying section.
6099 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
6100             Steve Munroe  <munroesj52@gmail.com>
6102         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
6103         (_mm_andnot_si128): Likewise.
6104         (_mm_or_si128): Likewise.
6105         (_mm_xor_si128): Likewise.
6107 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
6109         PR middle-end/78809
6110         * builtins.c (expand_builtin_memcmp): Inline the calls first
6111         when result_eq is false.
6112         (expand_builtin_strcmp): Inline the calls first.
6113         (expand_builtin_strncmp): Likewise.
6114         (inline_string_cmp): New routine. Expand a string compare
6115         call by using a sequence of char comparison.
6116         (inline_expand_builtin_string_cmp): New routine. Inline expansion
6117         a call to str(n)cmp/memcmp.
6118         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
6119         New option.
6120         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
6122 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
6124         * config/arm/driver-arm.c: Include arm-native.h.
6125         (host_detect_local_cpu): Use auto-generated data tables.
6126         (vendors, arm_cpu_table): Delete.  Move part information to ...
6127         * config/arm/arm-cpus.in: ... here.
6128         * config/arm/parsecpu.awk (gen_native): New function.
6129         (vendor, part): New CPU fields.
6130         (END): Add support for building the native CPU detection tables.
6131         * config/arm/t-arm (arm-native.h): Add build rule.
6132         (driver-arm.o): Add dependency on arm-native.h.
6134 2018-07-13  Richard Biener  <rguenther@suse.de>
6136         PR middle-end/85974
6137         * match.pd (addr1 - addr2): Allow either of the operand to
6138         have a conversion.
6140 2018-07-13  Tom de Vries  <tdevries@suse.de>
6142         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
6143         in remap_ssa_name.
6145 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
6147         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
6148         arrays instead of numbered variables.
6150 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
6152         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
6153         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
6155 2018-07-13  Richard Biener  <rguenther@suse.de>
6157         PR debug/86452
6158         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
6159         instead of get_context_die.
6161 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
6162             Richard Biener  <rguenther@suse.de>
6164         PR middle-end/86489
6165         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
6166         that the loop latch destination where phi is defined.
6168 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
6170         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
6171         (riscv_expand_epilogue): Add assertion to check interrupt mode.
6172         (riscv_set_current_function): Extract getting interrupt type to new
6173         function.
6174         (riscv_get_interrupt_type): New function.
6175         (riscv_merge_decl_attributes): New function, checking interrupt type is
6176         same.
6177         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
6179 2018-07-12  Paul Koning  <ni1d@arrl.net>
6181         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
6182         directive.
6184 2018-07-12  Paul Koning  <ni1d@arrl.net>
6186         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
6187         zero reference, add doloop_end instead.
6188         * doc/md.texi (decrement_and_branch_until_zero): Remove.
6189         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
6190         detail for doloop_end.
6192 2018-07-12  Martin Sebor  <msebor@redhat.com>
6194         PR c/86453
6195         * attribs.c (decl_attributes): Reject conflicting attributes before
6196         calling attribute handlers.
6198 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
6200         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
6201          parameter.
6202         (gcc::dump_manager::get_dump_file_name): likewise.
6203         (dump_begin): Likewise.
6204         * dumpfile.h (dump_begin): Update prototype.
6205         (gcc::dump_manager::get_dump_file_name,
6206         gcc::dump_manager::get_dump_file_name): Update prototype.
6208 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6210         * internal-fn.h (vectorizable_internal_fn_p): New function.
6211         * tree-vect-slp.c (compatible_calls_p): Likewise.
6212         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
6213         to internal functions.
6214         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
6216 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6218         * fold-const.h (inverse_conditions_p): Declare.
6219         * fold-const.c (inverse_conditions_p): New function.
6220         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
6221         that test the inverse condition of a conditional internal function.
6222         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
6223         * internal-fn.c (internal_fn_mask_index): Handle conditional
6224         internal functions.
6225         (vectorized_internal_fn_supported_p): New function.
6226         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
6227         (any_pred_load_store): Replace with...
6228         (need_to_predicate): ...this new variable.
6229         (redundant_ssa_names): New variable.
6230         (ifcvt_can_use_mask_load_store): Move initial checks to...
6231         (ifcvt_can_predicate): ...this new function.  Handle tree codes
6232         for which a conditional internal function exists.
6233         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
6234         instead of ifcvt_can_use_mask_load_store.  Update after variable
6235         name change.
6236         (predicate_load_or_store): New function, split out from
6237         predicate_mem_writes.
6238         (check_redundant_cond_expr): New function.
6239         (value_available_p): Likewise.
6240         (predicate_rhs_code): Likewise.
6241         (predicate_mem_writes): Rename to...
6242         (predicate_statements): ...this.  Use predicate_load_or_store
6243         and predicate_rhs_code.
6244         (combine_blocks, tree_if_conversion): Update after above name changes.
6245         (ifcvt_local_dce): Handle redundant_ssa_names.
6246         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
6247         general conditional functions.
6248         * tree-vect-stmts.c (vectorizable_call): Likewise.
6250 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6251             Alan Hayward  <alan.hayward@arm.com>
6252             David Sherwood  <david.sherwood@arm.com>
6254         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
6255         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
6256         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
6257         plus and minus and convert them into IFN_COND_FMA-based sequences.
6258         (convert_mult_to_fma): Handle conditional plus and minus.
6260 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6262         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
6263         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
6264         (cond_fnms_optab): New optabs.
6265         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
6266         internal functions.
6267         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
6268         * internal-fn.h (get_conditional_internal_fn): Declare.
6269         (get_unconditional_internal_fn): Likewise.
6270         * internal-fn.c (cond_ternary_direct): New macro.
6271         (expand_cond_ternary_optab_fn): Likewise.
6272         (direct_cond_ternary_optab_supported_p): Likewise.
6273         (FOR_EACH_COND_FN_PAIR): Likewise.
6274         (get_conditional_internal_fn): New function.
6275         (get_unconditional_internal_fn): Likewise.
6276         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
6277         (gimple_match_op::gimple_match_op): Add a new overload for 5
6278         operands.
6279         (gimple_match_op::set_op): Likewise.
6280         (gimple_resimplify5): Declare.
6281         * genmatch.c (decision_tree::gen): Generate simplifications for
6282         5 operands.
6283         * gimple-match-head.c (gimple_simplify): Define an overload for
6284         5 operands.  Handle calls with 5 arguments in the top-level overload.
6285         (convert_conditional_op): Handle conversions from unconditional
6286         internal functions to conditional ones.
6287         (gimple_resimplify5): New function.
6288         (build_call_internal): Pass a fifth operand.
6289         (maybe_push_res_to_seq): Likewise.
6290         (try_conditional_simplification): Try converting conditional
6291         internal functions to unconditional internal functions.
6292         Handle 3-operand unconditional forms.
6293         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
6294         Define ternary equivalents of the current rules for binary conditional
6295         internal functions.
6296         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
6297         ternary operations.
6298         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
6299         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
6300         (optab): Handle them.
6301         (SVE_COND_FP_TERNARY): New int iterator.
6302         (sve_fmla_op, sve_fmad_op): New int attributes.
6303         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
6304         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
6305         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
6307 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
6309         * target.def (preferred_else_value): New target hook.
6310         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
6311         * doc/tm.texi: Regenerate.
6312         * targhooks.h (default_preferred_else_value): Declare.
6313         * targhooks.c (default_preferred_else_value): New function.
6314         * internal-fn.h (conditional_internal_fn_code): Declare.
6315         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
6316         (get_conditional_internal_fn): Use it.
6317         (conditional_internal_fn_code): New function.
6318         * gimple-match.h (gimple_match_cond): New struct.
6319         (gimple_match_op): Add a cond member function.
6320         (gimple_match_op::gimple_match_op): Update all forms to take a
6321         gimple_match_cond.
6322         * genmatch.c (expr::gen_transform): Use the same condition as res_op
6323         for the suboperation, but don't specify a particular else_value.
6324         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
6325         (visit_nary_op, visit_reference_op_load): Pass
6326         gimple_match_cond::UNCOND to the gimple_match_op constructor.
6327         * gimple-match-head.c: Include tree-eh.h
6328         (convert_conditional_op): New function.
6329         (maybe_resimplify_conditional_op): Likewise.
6330         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
6331         (gimple_resimplify2): Likewise.
6332         (gimple_resimplify3): Likewise.
6333         (gimple_resimplify4): Likewise.
6334         (maybe_push_res_to_seq): Return null for conditional operations.
6335         (try_conditional_simplification): New function.
6336         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
6337         constructor.
6338         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6339         IFN_COND_* call.
6340         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
6341         function.
6342         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
6344 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
6346         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
6347         DECL_FCONTEXT
6348         (hash_tree): Do not hash DECL_FCONTEXT
6349         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
6350         Do not stream DECL_FCONTEXT.
6351         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
6352         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
6354 2018-07-12  Richard Biener  <rguenther@suse.de>
6356         PR debug/86462
6357         * dwarf2out.c (gen_block_die): Only output blocks when they have
6358         at least one !DECL_IGNORED_P variable.
6360 2018-07-12  Richard Biener  <rguenther@suse.de>
6362         PR target/84829
6363         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
6364         Remove -mieee-fp handling.
6366 2018-07-12  Richard Biener  <rguenther@suse.de>
6368         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
6369         left-over from last patch.
6371 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
6373         PR tree-optimization/86492
6374         * gimple-ssa-store-merging.c
6375         (imm_store_chain_info::coalesce_immediate_stores): Call
6376         check_no_overlap even for the merge_overlapping case.  Formatting fix.
6378 2018-07-12  Richard Biener  <rguenther@suse.de>
6380         PR middle-end/86479
6381         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
6382         move possibly trapping operations into the conditional.
6384 2018-07-12  Richard Biener  <rguenther@suse.de>
6386         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
6387         (vn_lookup_simplify_result): Remove recursion limit applied
6388         here.
6389         (vn_nary_build_or_lookup_1): Adjust.
6390         (try_to_simplify): Likewise.
6391         * gimple-match-head.c (gimple_resimplify1): Instead apply one
6392         here.
6393         (gimple_resimplify2): Likewise.
6394         (gimple_resimplify3): Likewise.
6395         (gimple_resimplify4): Likewise.
6397 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
6399         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
6400         Use __mmask64 type instead of __mmask8 for __M argument.
6401         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
6402         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
6403         __U argument.
6404         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
6405         __mmask16 for __M argument.
6406         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
6407         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
6408         to __mmask16 instead of __mmask8.
6409         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
6410         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
6411         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
6412         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
6413         instead of __mmask16 for __U argument.
6414         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
6415         __mmask16 instead of __mmask8 for __U argument.
6416         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
6417         __U argument.
6418         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6419         __mmask16.
6420         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6421         argument.
6422         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6423         __U argument.
6424         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6425         __mmask16.
6426         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6427         of __mmask16.
6428         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6429         __U argument.
6430         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6431         __U argument.
6432         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6433         __U argument.
6434         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
6435         __U argument.
6436         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
6437         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
6438         return type as well as __M argument type and all casts from __mmask8
6439         to __mmask32.
6440         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
6441         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
6442         return type as well as __M argument type and all casts from __mmask8
6443         to __mmask16.
6444         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
6445         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
6446         return type as well as __M argument type and all casts from __mmask8
6447         to __mmask32.
6448         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
6449         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
6450         return type as well as __M argument type and all casts from __mmask8
6451         to __mmask16.
6452         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6453         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
6454         __mmask16.
6456 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
6458         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
6459         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
6460         for __U argument.
6462 2018-07-11  Paul Koning  <ni1d@arrl.net>
6464         * doc/md.texi (define_subst): Document how multiple occurrences of
6465         the same argument in the replacement pattern are handled.
6467 2018-07-11  Paul Koning  <ni1d@arrl.net>
6469         * doc/extend.texi (Common Variable Attributes): Move "mode" into
6470         alphabetical order.
6471         (Common Type Attributes): Add "mode" attribute.
6473 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
6475         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
6476         stream DECL_ORIGINAL_TYPE.
6477         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
6478         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
6479         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
6480         Do not walk original type.
6481         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
6482         external decls.
6483         (write_ts_decl_non_common_tree_pointers): Do not stream
6484         DECL_ORIGINAL_TYPE
6485         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
6486         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
6488 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
6490         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
6491         thread twice from the same starting edge.
6493 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
6495         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
6496         * gimple.c (gimple_call_nonnull_result_p): ...here...
6497         (gimple_call_nonnull_arg): ...and here.
6498         * gimple.h (gimple_call_nonnull_result_p): New.
6499         (gimple_call_nonnull_arg): New.
6501 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
6503         * config/arm/arm-cpus.in: Move information from fpu field of each
6504         cpu definition to the isa field.
6505         * config/arm/parsecpu.awk (fpu): Delete match rule.
6506         (gen_comm_data): Don't add bits from the CPU's FPU entry.
6508 2018-07-11  Richard Biener  <rguenther@suse.de>
6510         PR debug/86457
6511         * dwarf2out.c (init_sections_and_labels): Use
6512         output_asm_line_debug_info consistently.
6513         (dwarf2out_early_finish): Likewise.
6514         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
6515         type units.
6517 2018-07-11  Richard Biener  <rguenther@suse.de>
6519         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
6520         Rework father_bb setting in a way to avoid propagating constants
6521         multiple times on a loop body.
6523 2018-07-10  Mark Wielaard  <mark@klomp.org>
6525         PR debug/86459
6526         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
6528 2018-07-10  Richard Biener  <rguenther@suse.de>
6530         * hash-map.h (hash_map::iterator::operator*): Return
6531         references to key and value.
6533 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
6535         PR c++/86443
6536         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
6537         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
6538         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
6539         (gimplify_omp_for): For composite loops, move outer
6540         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
6541         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
6542         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
6543         TREE_LIST for both the original class iterator and the "last" helper
6544         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
6545         loop, remember has_decl_expr from outer composite loops for the
6546         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
6548 2018-07-09  Martin Sebor  <msebor@redhat.com>
6550         PR middle-end/77357
6551         PR middle-end/86428
6552         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
6553         accessing implicitly initialized array elements.
6554         * expr.c (string_constant): Handle string initializers of
6555         character arrays within aggregates.
6556         * gimple-fold.c (fold_array_ctor_reference): Add argument.
6557         Store element offset.  As a special case, handle zero size.
6558         (fold_nonarray_ctor_reference): Same.
6559         (fold_ctor_reference): Add argument.  Store subobject offset.
6560         * gimple-fold.h (fold_ctor_reference): Add argument.
6562 2018-07-09  Paul Koning  <ni1d@arrl.net>
6564         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
6565         (pdp11_insn_cost): New function.
6566         (pdp11_md_asm_adjust): New function.
6567         (TARGET_INVALID_WITHIN_DOLOOP): Define.
6568         (pdp11_rtx_costs): Update to match machine better.
6569         (output_addr_const_pdp11): Correct format mismatch warnings.
6570         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
6571         * config/pdp11/pdp11.md: General change to add base_cost and/or
6572         length attributes for use by new pdp11_insn_cost function.
6573         (MIN_BRANCH): Correct definition.
6574         (MIN_SOB): Ditto.
6575         (doloop_end): Use standard pattern name for looping pattern.
6576         (doloop_end_nocc): New.
6577         (movsf): Add another constraint alternative.
6578         (zero_extendqihi2): Add constraint alternatives for not in place
6579         extend.
6580         (zero_extendhisi2): Remove.
6581         (shift patterns): Add CC handling variants.
6582         (bswaphi2): New.
6583         (bswapsi2): New.
6584         (rothi3): New.
6585         (define_peephole2): New peephole to recognize mov that sets CC for
6586         subsequent test.
6588 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
6590         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
6591         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
6592         wi::add.
6594 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
6596         PR c/86420
6597         * real.c (real_nextafter): Return true if result is denormal.
6599 2018-07-09  Martin Liska  <mliska@suse.cz>
6601         * common.opt: Add back wrongly removed attribute.
6603 2018-07-09  Richard Biener  <rguenther@suse.de>
6605         PR debug/86413
6606         * dwarf2out.c (gen_block_die): For an early generated DIE
6607         always output high/low PC attributes.
6609 2018-07-09  Tom de Vries  <tdevries@suse.de>
6611         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
6612         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
6613         onto VAR_DECL with abstract origin.
6615 2018-07-07  Jim Wilson  <jimw@sifive.com>
6617         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
6619 2018-07-07  Tom de Vries  <tdevries@suse.de>
6621         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
6623 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
6625         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
6626         overflow_type.
6627         (combine_bound): Use wide-int overflow calculation instead of
6628         rolling our own.
6629         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6630         overflow_type.
6631         * fold-const.c (int_const_binop_2): Same.
6632         (extract_muldiv_1): Same.
6633         (fold_div_compare): Same.
6634         (fold_abs_const): Same.
6635         * match.pd: Same.
6636         * poly-int.h (add): Same.
6637         (sub): Same.
6638         (neg): Same.
6639         (mul): Same.
6640         * predict.c (predict_iv_comparison): Same.
6641         * profile-count.c (slow_safe_scale_64bit): Same.
6642         * simplify-rtx.c (simplify_const_binary_operation): Same.
6643         * tree-chrec.c (tree_fold_binomial): Same.
6644         * tree-data-ref.c (split_constant_offset_1): Same.
6645         * tree-if-conv.c (idx_within_array_bound): Same.
6646         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
6647         * tree-ssa-phiopt.c (minmax_replacement): Same.
6648         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
6649         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
6650         * vr-values.c (vr_values::adjust_range_with_scev): Same.
6651         * wide-int.cc (wi::add_large): Same.
6652         (wi::mul_internal): Same.
6653         (wi::sub_large): Same.
6654         (wi::divmod_internal): Same.
6655         * wide-int.h: Change overflow type to overflow_type for neg, add,
6656         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
6657         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
6658         mul_internal, divmod_internal.
6659         (overflow_type): New enum.
6660         (accumulate_overflow): New.
6662 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
6664         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
6665         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
6667 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
6669         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
6670         argument is checked for zero before entering loop, avoid checking again.
6672 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
6674         * gimplify.h (generic_expr_could_trap_p): Set as global function.
6675         * gimplify.h (generic_expr_could_trap_p): Likwise.
6676         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
6678 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
6680         PR tree-optimization/86401
6681         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
6682         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
6683         (fold_bit_and_mask): ... here.  New helper function for match.pd.
6684         * fold-const.h (fold_bit_and_mask): Declare.
6685         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
6687 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
6689         PR target/86324
6690         * target.def (translate_mode_attribute): New hook.
6691         * targhooks.h (default_translate_mode_attribute): Declare.
6692         * targhooks.c (default_translate_mode_attribute): New function.
6693         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
6694         * doc/tm.texi: Regenerate.
6695         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
6696         (rs6000_translate_mode_attribute): New function.
6698 2018-07-06  Paul Koning  <ni1d@arrl.net>
6700         * doc/md.texi (define_split): Document DONE and FAIL.
6701         (define_peephole2): Ditto.
6703 2018-07-05  Jeff Law  <law@redhat.com>
6705         PR tree-optimization/86010
6706         * tree-ssa-dse.c (compute_trims): More aggressively trim at
6707         both the head and tail of mem* and str* calls.
6709 2018-07-05  Jim Wilson  <jimw@sifive.com>
6711         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
6713 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
6715         * config/aarch64/aarch64-simd.md: correct flags text for
6716         MIN_EXPR replacement.
6718 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
6720         * configure: Regenerated.
6722 2018-07-05  Carl Love  <cel@us.ibm.com>
6724         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
6725         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
6726         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
6727         VSX_BUILTIN_DOUBLEL_V4SF.
6729 2018-07-05  Martin Sebor  <msebor@redhat.com>
6731         PR c++/86400
6732         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
6733         than its domain to compute its the upper bound of a char array.
6735 2018-07-05  Nathan Sidwell  <nathan@acm.org>
6737         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
6738         * doc/cpp.texi: Update comment.
6739         * doc/tm.texi: Rebuilt.
6740         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
6741         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
6742         * doc/extend.texi (Backwards Compatibility): Clarify it is system
6743         headers affected by extern "C".
6744         * system.h: Poison NO_IMPLICIT_EXTERN_C.
6745         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
6746         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
6747         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
6748         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
6749         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
6750         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
6751         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
6752         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
6753         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
6754         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
6755         config/sparc/sp64-elf.h, config/spu/spu.h,
6756         config/stormy16/stormy16.h, config/v850/v850.h,
6757         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
6758         define NO_IMPLICIT_EXTERN_C.
6759         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
6761 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
6763         PR target/84711
6764         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
6765         instead of GET_MODE_SIZE when comparing Units.
6767 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
6769         PR target/84711
6770         * rtlanal.c (set_noop_p): Constrain on mode change,
6771         include hard-reg-set.h
6773 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
6775         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
6777 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
6779         Revert
6780         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
6782         PR sanitizer/84250
6783         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6784         libasan.
6785         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6787 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
6789         PR sanitizer/84250
6790         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6791         libasan.
6792         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6794 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
6796         PR middle-end/86380
6797         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
6799 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
6801         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
6802         neg_*_op* variables.
6804 2018-07-04  Martin Liska  <mliska@suse.cz>
6806         * tree-switch-conversion.c: Define
6807         max_ratio_for_speed and max_ratio_for_size constants.
6809 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
6810             Martin Liska  <mliska@suse.cz>
6812         PR middle-end/66240
6813         PR target/45996
6814         PR c/84100
6815         * common.opt: Rename align options with 'str_' prefix.
6816         * common/config/i386/i386-common.c (set_malign_value): New
6817         function.
6818         (ix86_handle_option): Use it to set -falign-* options/
6819         * config/aarch64/aarch64-protos.h (struct tune_params): Change
6820         type from int to string.
6821         * config/aarch64/aarch64.c: Update default values from int
6822         to string.
6823         * config/alpha/alpha.c (alpha_override_options_after_change):
6824         Likewise.
6825         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
6826         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6827         max skip conditionally.
6828         * config/i386/freebsd.h (SUBALIGN_LOG): New.
6829         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6830         max skip conditionally.
6831         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6832         max skip conditionally.
6833         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
6834         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6835         max skip conditionally.
6836         * config/i386/i386.c (struct ptt): Change type from int to
6837         string.
6838         (ix86_default_align): Set default values.
6839         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
6840         max skip conditionally.
6841         * config/i386/iamcu.h (SUBALIGN_LOG): New.
6842         (ASM_OUTPUT_MAX_SKIP_ALIGN):
6843         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
6844         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6845         max skip conditionally.
6846         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
6847         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
6848         * config/i386/x86-64.h (SUBALIGN_LOG): New.
6849         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
6850         max skip conditionally.
6851         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6852         * config/ia64/ia64.c (ia64_option_override): Set default values
6853         for alignment options.
6854         * config/m68k/m68k.c: Handle new str_align_* options.
6855         * config/mips/mips.c (mips_set_compression_mode): Change
6856         type of constants.
6857         (mips_option_override): Set default values for options.
6858         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6859         Likewise.
6860         * config/rs6000/rs6000.c (rs6000_option_override_internal):
6861         Likewise.
6862         * config/rx/rx.c (rx_option_override): Likewise.
6863         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
6864         (LABEL_ALIGN): Use align_labels_log.
6865         (LOOP_ALIGN): Use align_loops_align.
6866         * config/s390/s390.c (s390_asm_output_function_label): Use new
6867         macros.
6868         * config/sh/sh.c (sh_override_options_after_change):
6869         Change type of constants.
6870         * config/spu/spu.c (spu_sched_init): Likewise.
6871         * config/sparc/sparc.c (sparc_option_override): Set default
6872         values for options.
6873         * config/visium/visium.c (visium_option_override): Likewise.
6874         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
6875         emit p2align format with last argument if it's not needed.
6876         * doc/invoke.texi: Document extended format of -falign-*.
6877         * final.c: Use align_labels alignment.
6878         * flags.h (struct target_flag_state): Change type to use
6879         align_flags.
6880         (struct align_flags_tuple): New.
6881         (struct align_flags): Likewise.
6882         (align_loops_log): Redefine macro to use new types.
6883         (align_loops_max_skip): Redefine macro to use new types.
6884         (align_jumps_log): Redefine macro to use new types.
6885         (align_jumps_max_skip): Redefine macro to use new types.
6886         (align_labels_log): Redefine macro to use new types.
6887         (align_labels_max_skip): Redefine macro to use new types.
6888         (align_functions_log): Redefine macro to use new types.
6889         (align_loops): Redefine macro to use new types.
6890         (align_jumps): Redefine macro to use new types.
6891         (align_labels): Redefine macro to use new types.
6892         (align_functions): Redefine macro to use new types.
6893         (align_functions_max_skip): Redefine macro to use new types.
6894         (align_loops_value): New macro.
6895         (align_jumps_value): New macro.
6896         (align_labels_value): New macro.
6897         (align_functions_value): New macro.
6898         * function.c (invoke_set_current_function_hook): Propagate
6899         alignment values from flags to global variables default in
6900         topleev.h.
6901         * ipa-icf.c (sem_function::equals_wpa): Use
6902         cl_optimization_option_eq instead of memcmp.
6903         * lto-streamer.h (cl_optimization_stream_out): Support streaming
6904         of string types.
6905         (cl_optimization_stream_in): Likewise.
6906         * optc-save-gen.awk: Support strings in cl_optimization.
6907         * opth-gen.awk: Likewise.
6908         * opts.c (finish_options): Remove error checking of invalid
6909         value ranges.
6910         (MAX_CODE_ALIGN): Remove.
6911         (MAX_CODE_ALIGN_VALUE): Likewise.
6912         (parse_and_check_align_values): New function.
6913         (check_alignment_argument): Likewise.
6914         (common_handle_option): Use check_alignment_argument.
6915         * opts.h (parse_and_check_align_values): Declare.
6916         * toplev.c (init_alignments): Remove.
6917         (read_log_maxskip): New.
6918         (parse_N_M): Likewise.
6919         (parse_alignment_opts): Likewise.
6920         (backend_init_target): Remove usage of init_alignments.
6921         * toplev.h (parse_alignment_opts): Declare.
6922         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
6923         argument.
6924         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
6925         * tree.c (cl_option_hasher::equal): New.
6926         * varasm.c: Use new global macros.
6928 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
6930         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
6931         Use a simpler align directive also if MAXSKIP = ALIGN-1.
6932         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6933         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6934         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6935         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
6936         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
6937         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
6938         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
6939         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6940         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6941         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6942         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
6944 2018-07-04  Martin Liska  <mliska@suse.cz>
6945             Jonathan Wakely  <jwakely@redhat.com>
6947         * coverage.c: Use correct type.
6948         * doc/invoke.texi: Language correction.
6950 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
6952         PR target/85620
6953         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
6954         ENDBRANCH for non-tail call which may return via indirect branch.
6955         * doc/extend.texi: Document indirect_return attribute.
6957 2018-07-03  Martin Sebor  <msebor@redhat.com>
6959         PR tree-optimization/86274
6960         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
6961         precondition.
6962         (format_floating): Correct handling of infinities and NaNs.
6964 2018-07-03  Martin Sebor  <msebor@redhat.com>
6966         * print-tree.c (print_real_cst): New function.
6967         (print_node_brief): Call it.
6968         (print_node): Ditto.
6970 2018-07-03  Jeff Law  <law@redhat.com>
6972         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
6973         into a single pattern.
6975         * config/h8300/h8300.md (ors code_iterator): New.
6976         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
6977         a single pattern and single splitter.
6978         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
6979         (iorqi3_1, xorqi3_1): Likewise.
6980         (iorqi3, xorqi3 expanders): Similarly.
6982         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
6983         (movmd_internal) into a single pattern using the P mode iterator.
6984         (movmd splitters): Similarly.
6985         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
6986         (movsd splitters): Similarly.
6988         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
6989         ADDB, ADDW and ADDL into a single ADD attribute which selects the
6990         right table based on the size of the operand.
6991         * config/h8300/h8300.md (length_table): Corresponding changes. All
6992         references to "addb", "addw" and "addl" changed to "add".
6993         (btst patterns): Merge two variants into a single pattern.
6994         (tstqi, tsthi): Likewise.
6995         (addhi3_incdec, addsi3_incdec): Likewise.
6996         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
6997         (mulhi3, mulsi3): Likewise.
6998         (udivhi3, udivsi3): Likewise.
6999         (divhi3, divsi3): Likewise.
7000         (andorqi3, andorhi3, andorsi3): Likewise.
7002 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
7004         PR target/85694
7005         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
7006         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
7008 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7010         PR tree-optimization/85694
7011         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
7012         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
7013         UNSPEC_URHADD.
7014         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
7015         (<u>avg<mode>3_ceil): New patterns.
7017 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
7019         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
7020         scan-tree-dump directive.
7021         * gcc.dg/vect/slp-perm-2.c: Likewise.
7022         * gcc.dg/vect/slp-perm-3.c: Likewise.
7023         * gcc.dg/vect/slp-perm-5.c: Likewise.
7024         * gcc.dg/vect/slp-perm-6.c: Likewise.
7025         * gcc.dg/vect/slp-perm-7.c: Likewise.
7026         * gcc.dg/vect/slp-perm-8.c: Likewise.
7028 2018-07-03  Marek Polacek  <polacek@redhat.com>
7030         PR middle-end/86202
7031         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7033 2018-07-03  Richard Biener  <rguenther@suse.de>
7035         PR ipa/86389
7036         * tree-ssa-structalias.c (find_func_clobbers): Properly
7037         handle indirect calls.
7039 2018-07-03  Jeff Law  <law@redhat.com>
7041         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
7042         (shifts): New code iterator.
7043         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
7044         expander.  Fix HImode handling on H8/SX.
7045         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
7046         (subqi3, subhi3, subsi3 expanders): Likewise.
7047         (andqi3, andhi3, andsi3 expanders): Likewise.
7048         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
7049         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
7050         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
7051         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
7052         (zero_extendqihi2, zero_extendqisi2): Likewise.
7053         (extendqihi2, extendqisi2): Likewise.
7054         (rotlqi3, rotlhi3, rotlsi3): Likewise.
7055         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
7056         (rotlqi3_1, rotlhi3_1): Likewise.
7057         (logicalhi3_sn, logicalsi3_sn): Likewise.
7058         (logicalhi3, logicalsi3): Likewise.
7060 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7062         * tree-vect-patterns.c (vect_recog_rotate_pattern)
7063         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7064         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
7065         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
7066         type to append_pattern_def_seq instead of creating a stmt_vec_info
7067         directly.
7068         (build_mask_conversion): Likewise.  Remove vinfo argument.
7069         (vect_add_conversion_to_patterm): Likewise, renaming to...
7070         (vect_add_conversion_to_pattern): ...this.
7071         (vect_recog_mask_conversion_pattern): Update call to
7072         build_mask_conversion.  Pass the vector type to
7073         append_pattern_def_seq here too.
7074         (vect_recog_gather_scatter_pattern): Update call to
7075         vect_add_conversion_to_pattern.
7077 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7079         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
7080         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7081         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
7082         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
7083         STMT_VINFO_PATTERN_DEF_SEQ to null here.
7084         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
7085         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
7086         append_pattern_def_seq instead of new_pattern_def_seq.
7087         (vect_recog_divmod_pattern): Do both of the above.
7088         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
7089         is null.
7091 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7093         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
7094         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
7095         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
7096         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
7097         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
7098         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7099         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
7100         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
7101         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
7102         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
7103         parameter with a single stmt_vec_info.
7104         (vect_recog_func_ptr): Likewise.
7105         (vect_recog_gather_scatter_pattern): Likewise, folding in...
7106         (vect_try_gather_scatter_pattern): ...this.
7107         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
7108         the stmt_vec_info of the statement to be matched.  Don't clear
7109         STMT_VINFO_RELATED_STMT.
7110         (vect_pattern_recog): Update call accordingly.
7112 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7114         PR tree-optimization/85694
7115         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
7116         (uavgM3_ceil): Document new optabs.
7117         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
7118         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
7119         functions.
7120         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
7121         (savg_ceil_optab): New optabs.
7122         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
7123         (vect_vect_recog_func_ptrs): Add it.
7124         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
7125         constant directly from the associated lhs.
7127 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7129         * tree-vect-patterns.c (vect_split_statement): New function.
7130         (vect_convert_input): Use it to try to split an existing cast.
7132 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7134         * poly-int.h (print_hex): New function.
7135         * dumpfile.h (dump_dec, dump_hex): Declare.
7136         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
7137         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
7138         min_input_precision, operation_precision and operation_sign.
7139         * tree-vect-patterns.c (vect_get_range_info): New function.
7140         (vect_same_loop_or_bb_p, vect_single_imm_use)
7141         (vect_operation_fits_smaller_type): Delete.
7142         (vect_look_through_possible_promotion): Add an optional
7143         single_use_p parameter.
7144         (vect_recog_over_widening_pattern): Rewrite to use new
7145         stmt_vec_info infomration.  Handle one operation at a time.
7146         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
7147         (vect_truncatable_operation_p, vect_set_operation_type)
7148         (vect_set_min_input_precision): New functions.
7149         (vect_determine_min_output_precision_1): Likewise.
7150         (vect_determine_min_output_precision): Likewise.
7151         (vect_determine_precisions_from_range): Likewise.
7152         (vect_determine_precisions_from_users): Likewise.
7153         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
7154         (vect_vect_recog_func_ptrs): Put over_widening first.
7155         Add cast_forwprop.
7156         (vect_pattern_recog): Call vect_determine_precisions.
7158 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
7160         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
7161         statements that have been replaced by further pattern statements.
7162         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
7164 2018-07-03  Richard Biener  <rguenther@suse.de>
7166         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
7167         always set *dt.  Dump vectype in vectype overload.
7168         * dumpfile.h (dump_gimple_expr): New function.
7169         (dump_gimple_expr_loc): Likewise.
7170         * dumpfile.c (dump_gimple_expr): New function.
7171         (dump_gimple_expr_loc): Likewise.
7173 2018-07-02  Jeff Law  <law@redhat.com>
7175         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
7176         the H8/300, H8/300H and H8/S variants into a single pattern.
7177         (movhi_h8300, movqi_h8300hs): Similarly.
7178         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
7179         (QHI mode iterator): New.
7181         * config/h8300/h8300.md: Remove trailing whitespace.
7183 2018-07-02  Jim Wilson  <jimw@sifive.com>
7185         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
7186         instead of emit_insn for interrupt returns.
7187         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
7188         (riscv_sret, riscv_uret): Likewise.
7190 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
7192         * pretty-print.c (selftest::test_pp_format): Move save and restore
7193         of quotes to class auto_fix_quotes, and add an instance.
7194         * selftest.c: Include "intl.h".
7195         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
7196         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
7197         * selftest.h (selftest::auto_fix_quotes): New class.
7199 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
7201         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
7202         (aarch64_sve_prepare_conditional_op): Remove.
7203         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
7204         Allow aarch64_simd_reg_or_zero as select operand; remove
7205         the aarch64_sve_prepare_conditional_op call.
7206         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
7207         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7208         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
7209         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
7210         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
7211         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
7212         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
7213         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
7214         and a splitters to match all of the *_any patterns.
7215         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
7217         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
7218         (SVE_COND_FP_BINARY_REV): Remove.
7219         (sve_int_op_rev, sve_fp_op_rev): New.
7220         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
7221         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
7222         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
7223         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
7224         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
7225         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
7226         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
7227         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
7228         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
7230         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
7231         Remove match_dup 1 from the inner unspec.
7232         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
7234         * config/aarch64/aarch64.md (movprfx): New attr.
7235         (length): Default movprfx to 8.
7236         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
7237         (*madd<SVE_I>, *msub<SVE_I): Likewise.
7238         (*<su>mul<SVE_I>3_highpart): Likewise.
7239         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
7240         (*v<ASHIFT><SVE_I>3): Likewise.
7241         (*<su><MAXMIN><SVE_I>3): Likewise.
7242         (*<su><MAXMIN><SVE_F>3): Likewise.
7243         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
7244         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
7245         (*div<SVE_F>4): Likewise.
7247 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
7249         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
7250         in dump string.
7252 2018-07-02  Richard Biener  <rguenther@suse.de>
7254         PR tree-optimization/86363
7255         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
7256         memset argument refers to a non-variable address.
7258 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
7260         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
7261         {PLUS,MINUS}_EXPR code to...
7262         (adjust_symbolic_bound): ...here,
7263         (combine_bound): ...here,
7264         (set_value_range_with_overflow): ...and here.
7266 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
7268         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
7269         code...
7270         (extract_range_from_abs_expr): ...here.
7272 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
7274         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
7275         -fno-omit-frame-pointer when not optimizing.
7277 2018-07-02  Martin Liska  <mliska@suse.cz>
7279         PR ipa/86279
7280         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
7281         (propagate_nothrow): Likewise.
7283 2018-07-02  Martin Liska  <mliska@suse.cz>
7285         PR ipa/86323
7286         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
7288 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
7290         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
7291         function in r262149, changing "loc" param from source_location to
7292         const dump_location_t &.
7293         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
7294         declaration, as above.
7296 2018-07-01  Paul Koning  <ni1d@arrl.net>
7298         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7299         -munit-asm, -mgnu-asm, -mdec-asm.
7300         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
7301         (pdp11_output_labelref): New.
7302         (pdp11_output_def): New.
7303         (pdp11_output_addr_vec_elt): New.
7304         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
7305         %# and %@ format codes.
7306         (pdp11_option_override): New.
7307         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
7308         (pdp11_output_ident): New.
7309         (pdp11_asm_named_section): New.
7310         (pdp11_asm_init_sections): New.
7311         (pdp11_file_start): New.
7312         (pdp11_file_end): New.
7313         (output_ascii): Use .ascii/.asciz for -mdec-asm.
7314         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
7315         %o, like %c but octal.
7316         (pdp11_option_override): New.
7317         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
7318         -mdec-asm.
7319         (DATA_SECTION_ASM_OP): Ditto.
7320         (READONLY_DATA_SECTION_ASM_OP): New.
7321         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
7322         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
7323         (ASM_OUTPUT_LABELREF): Ditto.
7324         (ASM_OUTPUT_DEF): Ditto.
7325         (ASM_OUTPUT_EXTERNAL): New.
7326         (ASM_OUTPUT_SOURCE_FILENAME): New.
7327         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
7328         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
7329         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
7330         %# and %@ format codes.
7331         * config/pdp11/pdp11.opt (mgnu-asm): New.
7332         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
7333         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
7334         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
7336 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
7338         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
7339         dereferencing path[] beyond its length.
7340         (debug_path): New.
7341         (debug_all_paths): New.
7342         (rewire_first_differing_edge): New.
7343         (adjust_paths_after_duplication): New.
7344         (duplicate_thread_path): Call adjust_paths_after_duplication.
7345         Add new argument.
7346         (thread_through_all_blocks): Add new argument to
7347         duplicate_thread_path.
7349 2018-06-30  Jim Wilson  <jimw@sifive.com>
7351         * config/riscv/predicates.md (p2m1_shift_operand): New.
7352         (high_mask_shift_operand): New.
7353         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
7354         pattern using p2m1_shift_operand.
7355         (lshsi3_zero_extend_3+2): New combiner pattern using
7356         high_mask_shift_operand.
7358 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7360         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
7361         split out from...
7362         (vect_recog_rotate_pattern): ...here.
7363         (vect_convert_input): Try to insert casts of invariants in the
7364         preheader.
7365         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
7366         preheader to be empty.
7368 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7370         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
7371         vector type.  If given, install it in the new statement's
7372         STMT_VINFO_VECTYPE.
7373         (vect_element_precision): New function.
7374         (vect_unpromoted_value): New struct.
7375         (vect_unpromoted_value::vect_unpromoted_value): New function.
7376         (vect_unpromoted_value::set_op): Likewise.
7377         (vect_look_through_possible_promotion): Likewise.
7378         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
7379         (vect_widened_op_tree, vect_convert_input): Likewise.
7380         (vect_convert_inputs, vect_convert_output): Likewise.
7381         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
7382         to handle the optional cast of the multiplication result and
7383         vect_widened_op_tree to detect the widened multiplication itself.
7384         Do not require the input and output of promotion casts to have
7385         the same sign, but base the signedness of the operation on the
7386         input rather than the result.  If the pattern includes two
7387         promotions, check that those promotions have the same sign.
7388         Do not restrict the MULT_EXPR handling to a double-width result;
7389         handle quadruple-width results and wider.  Use vect_convert_inputs
7390         to convert the inputs to the common type.
7391         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
7392         to handle the optional cast of the ABS result.  Also allow a sign
7393         change or a sign extension between the ABS and MINUS.
7394         Use vect_widened_op_tree to detect the widened subtraction and use
7395         vect_convert_inputs to convert the inputs to the common type.
7396         (vect_handle_widen_op_by_const): Delete.
7397         (vect_recog_widen_op_pattern): New function.
7398         (vect_recog_widen_mult_pattern): Use it.
7399         (vect_recog_widen_shift_pattern): Likewise.
7400         (vect_recog_widen_sum_pattern): Use
7401         vect_look_through_possible_promotion to handle the promoted
7402         PLUS_EXPR operand.
7404 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7406         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
7407         the containing gimple_seq *.
7408         * gimple-iterator.h (gsi_for_stmt): Declare it.
7409         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
7410         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
7411         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
7412         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
7413         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
7414         checks.
7415         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
7416         split out from...
7417         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
7418         statement being replaced is part of an existing pattern
7419         definition sequence, inserting the new pattern statements before
7420         the original one.
7421         (vect_pattern_recog_1): Don't return a bool.  If the statement
7422         is already part of a pattern, instead apply pattern matching
7423         to the pattern definition statements.  Don't clear the
7424         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
7425         (vect_pattern_recog): Don't break after the first match;
7426         continue processing the pattern definition statements instead.
7427         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
7429 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7431         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
7432         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
7433         (vect_recog_widen_sum_pattern): Use it.
7435 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7437         * tree-vect-loop.c (vectorizable_reduction): Assert that the
7438         phi is not a pattern statement and has not been replaced by
7439         a pattern statement.
7440         * tree-vect-patterns.c (type_conversion_p): Don't check
7441         STMT_VINFO_IN_PATTERN_P.
7442         (vect_recog_vector_vector_shift_pattern): Likewise.
7443         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
7444         the pattern statement rather than the original statement; check
7445         directly for a WIDEN_MULT_EXPR here.
7446         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
7447         vect_is_simple_use to return the pattern statement rather
7448         than the original statement; use is_pattern_stmt_p to check
7449         for such a pattern statement.
7450         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
7451         to return the pattern statement rather than the original statement;
7452         don't do the same transformation here.
7453         (vect_is_simple_use): If the defining statement has been replaced
7454         by a pattern statement, return the pattern statement instead.
7455         Remove the corresponding (local) transformation from the vectype
7456         overload.
7458 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
7460         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
7461         end and default to null.
7462         * tree-vect-loop.c (vect_create_epilog_for_reduction)
7463         (vectorizable_reduction): Update calls accordingly, dropping the
7464         gimple ** argument if the passed-back statement isn't needed.
7465         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
7466         (vect_recog_rotate_pattern): Likewise.
7467         (vect_recog_mask_conversion_pattern): Likewise.
7468         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7469         (vect_mask_constant_operand_p): Likewise.
7470         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
7471         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
7472         (get_group_load_store_type, get_load_store_type): Likewise.
7473         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
7474         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
7475         (vectorizable_conversion, vectorizable_assignment): Likewise.
7476         (vectorizable_shift, vectorizable_operation): Likewise.
7477         (vectorizable_store, vect_is_simple_cond): Likewise.
7478         (vectorizable_condition, vectorizable_comparison): Likewise.
7479         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
7480         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
7481         and move it to the end.  Cope with null def_stmt_outs.
7483 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7485         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
7487 2018-06-29  Jeff Law  <law@redhat.com>
7489         * config/v850/v850.c (v850_legitimate_address_p): Handle large
7490         displacements for TARGET_V850E2V3 and newer.
7491         (TARGET_LRA_P): Remove.  Defaults to LRA now.
7492         * config/v850/v850.md (sign23byte_load): Remove.
7493         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
7494         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
7496 2018-06-29  Martin Liska  <mliska@suse.cz>
7498         PR lto/85759
7499         * coverage.c (coverage_init): Mangle full path name.
7500         * doc/invoke.texi: Document the change.
7501         * gcov-io.c (mangle_path): New.
7502         * gcov-io.h (mangle_path): Likewise.
7503         * gcov.c (mangle_name): Use mangle_path for path mangling.
7505 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7507         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
7508         if starting source register is not even.
7510 2018-06-29  Martin Liska  <mliska@suse.cz>
7512         PR tree-optimization/86263
7513         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
7514         Make edge redirection.
7516 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
7518         * dumpfile.c (dump_loc): Add indentation based on scope depth.
7519         (dump_scope_depth): New variable.
7520         (get_dump_scope_depth): New function.
7521         (dump_begin_scope): New function.
7522         (dump_end_scope): New function.
7523         * dumpfile.h (get_dump_scope_depth): New declaration.
7524         (dump_begin_scope): New declaration.
7525         (dump_end_scope): New declaration.
7526         (class auto_dump_scope): New class.
7527         (AUTO_DUMP_SCOPE): New macro.
7528         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
7529         AUTO_DUMP_SCOPE.
7531 2018-06-29  Richard Biener  <rguenther@suse.de>
7533         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
7534         compute_all_dependences succeeds.
7535         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
7536         exceed --param loop-max-datarefs-for-datadeps.
7538 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
7540         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
7542 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
7544         PR target/86348
7545         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
7546         alternative 0 in preferred_for_speed attribute.
7548 2018-06-28  Paul Koning  <ni1d@arrl.net>
7550         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
7551         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
7552         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
7553         * config/pdp11/pdp11.md: Correct "length" attribute calculation
7554         for shift insn patterns.
7556 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
7558         * cgraph.c (cgraph_node::get_body): Replace assignments to
7559         "dump_file" with calls to set_dump_file.
7560         * dumpfile.c (alt_dump_file): Make static, and group with...
7561         (alt_flags): ...this definition.
7562         (dumps_are_enabled): New variable.
7563         (refresh_dumps_are_enabled): New function.
7564         (set_dump_file): New function.
7565         (set_alt_dump_file): New function.
7566         (gcc::dump_manager::dump_start): Replace assignments to
7567         "dump_file" and "alt_dump_file" with calls to set_dump_file and
7568         set_alt_dump_file.
7569         (gcc::dump_manager::dump_finish): Likewise.
7570         * dumpfile.h (alt_dump_file): Delete decl.
7571         (dumps_are_enabled): New variable decl.
7572         (set_dump_file): New function decl.
7573         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
7574         global.
7575         * tree-nested.c (lower_nested_functions): Replace assignments to
7576         "dump_file" with calls to set_dump_file.
7578 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
7580         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
7581         goto_locus of each outgoing edge of each basic block.
7583 2018-06-28  Richard Biener  <rguenther@suse.de>
7585         * dwarf2out.c (decl_scope_table): Remove.
7586         (push_decl_scope): Likewise.
7587         (pop_decl_scope): Likewise.
7588         (gen_type_die_for_member): Do not call push/pop_decl_scope.
7589         (gen_struct_or_union_type_die): Likewise.
7590         (gen_tagged_type_die): Likewise.
7591         (dwarf2out_init): Do not initialize decl_scope_table.
7592         (dwarf2out_c_finalize): Do not free it.
7594 2018-06-28  Richard Biener  <rguenther@suse.de>
7596         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
7597         deciding whether to not re-use a DIE.
7599 2018-06-28  Richard Biener  <rguenther@suse.de>
7601         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
7602         DW_AT_abstract_origin attribute.
7604 2018-06-28  Martin Liska  <mliska@suse.cz>
7606         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
7607         Use newly introduced constants.
7608         * tree-switch-conversion.h (struct jump_table_cluster):
7609         Define max_ratio_for_size and max_ratio_for_speed.
7611 2018-06-28  Martin Liska  <mliska@suse.cz>
7613         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
7614         Add new checking assert to catch invalid state.
7615         (jump_table_cluster::can_be_handled): Handle single case
7616         clusters.
7617         (jump_table_cluster::is_beneficial): Bail out for such case.
7618         (bit_test_cluster::find_bit_tests):
7619         Add new checking assert to catch invalid state.
7620         (bit_test_cluster::can_be_handled): Handle single case
7621         clusters.
7622         (bit_test_cluster::is_beneficial): Bail out for such case.
7623         (switch_decision_tree::analyze_switch_statement):
7624         Fix comment.
7626 2018-06-28  Martin Liska  <mliska@suse.cz>
7628         * common.opt: Introduce -completion option.
7629         * gcc.c (driver_handle_option): Handle it.
7630         (driver::main): Print completions if completion
7631         is set.
7632         * opt-suggestions.c (option_proposer::get_completions):
7633         New function.
7634         (option_proposer::suggest_completion): Likewise.
7635         (option_proposer::find_param_completions): Likewise.
7636         (verify_autocompletions): Likewise.
7637         (test_completion_valid_options): Likewise.
7638         (test_completion_valid_params): Likewise.
7639         (in_completion_p): Likewise.
7640         (empty_completion_p): Likewise.
7641         (test_completion_partial_match): Likewise.
7642         (test_completion_garbage): Likewise.
7643         (opt_proposer_c_tests): Likewise.
7644         * opt-suggestions.h: Declare new functions.
7645         * opts.c (common_handle_option): Handle OPT__completion_.
7646         * selftest-run-tests.c (selftest::run_tests): Add
7647         opt_proposer_c_tests.
7648         * selftest.c (assert_str_startswith): New.
7649         * selftest.h (assert_str_startswith): Likewise.
7650         (opt_proposer_c_tests): New.
7651         (ASSERT_STR_STARTSWITH): Likewise.
7653 2018-06-28  Martin Liska  <mliska@suse.cz>
7655         * Makefile.in: Add opt-suggestions.o.
7656         * gcc-main.c: Include opt-suggestions.h.
7657         * gcc.c (driver::driver): Likewise.
7658         (driver::~driver): Remove m_option_suggestions.
7659         (driver::build_option_suggestions): Moved to option_proposer.
7660         (driver::suggest_option): Likewise.
7661         (driver::handle_unrecognized_options): Use option_proposer.
7662         * gcc.h (class driver): Add new memver m_option_proposer.
7663         * opt-suggestions.c: New file.
7664         * opt-suggestions.h: New file.
7666 2018-06-28  Martin Liska  <mliska@suse.cz>
7668         * vec.h (class auto_string_vec): New (moved from auto_argvec).
7669         (auto_string_vec::~auto_string_vec): Likewise.
7671 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
7673         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
7674         prevent_decl_creation_for_types fields up and add reset_location field.
7675         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
7676         statement if id->reset_location is true.
7677         (copy_edges_for_bb): Do not set goto_locus on the new edges if
7678         id->reset_location is true.
7679         (copy_phis_for_bb): Force input_location on the arguments if
7680         id->reset_location is true.
7681         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
7682         is set on the function to be inlined.
7684 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
7686         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
7688 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
7690         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
7691         registers for Pmode.
7692         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
7693         hard registers for the clobbered pseudo.
7695 2018-06-27  Paul Koning  <ni1d@arrl.net>
7697         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
7698         mutually exclusive options.
7699         * config/pdp11/constraints.md (h): New constraint.
7700         (O): Update definition to match shift code generation.
7701         (D): New constraint.
7702         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
7703         (CCFP): Remove.
7704         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
7705         function.
7706         (output_jump): Change arguments.
7707         (pdp11_fixed_cc_regs): New function.
7708         (pdp11_cc_mode): Ditto.
7709         (pdp11_expand_shift): Ditto.
7710         (pdp11_assemble_shift): Ditto.
7711         (pdp11_small_shift): Ditto.
7712         (pdp11_branch_cost): Remove.
7713         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
7714         from output.
7715         (pdp11_register_move_cost): Update for CC registers.
7716         (pdp11_rtx_costs): Add case for LSHIFTRT.
7717         (pdp11_output_jump): Add CCNZ mode conditional branches.
7718         (notice_update_cc_on_set): Remove.
7719         (pdp11_cc_mode): New function.
7720         (simple_memory_operand): Correct pre/post decrement case.
7721         (no_side_effect_operand): New function.
7722         (pdp11_regno_reg_class): Add CC_REGS class.
7723         (pdp11_fixed_cc_regs): New function.
7724         (pdp11_small_shift): New function.
7725         (pdp11_expand_shift): New function to expand shift insns.
7726         (pdp11_assemble_shift): New function to output shifts.
7727         (pdp11_branch_cost): Remove.
7728         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
7729         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
7730         (WCHAR_TYPE): Ditto.
7731         (PTRDIFF_TYPE): Ditto.
7732         (ADJUST_INSN_LENGTH): New macro.
7733         (FIXED_REGISTERS): Add CC registers.
7734         (CALL_USED_REGISTERS): Ditto.
7735         (reg_class): Ditto.
7736         (REG_CLASS_NAMES): Ditto.
7737         (REG_CLASS_CONTENTS): Ditto.
7738         (SELECT_CC_MODE): Use new function.
7739         (TARGET_FLAGS_REGNUM): New macro.
7740         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
7741         (cc0_reg_rtx): Remove.
7742         (CC_STATUS_MDEP): Remove.
7743         (CC_STATUS_MDEFP_INIT): Remove.
7744         (CC_IN_FPU): Remove.
7745         (NOTICE_UPDATE_CC): Remove.
7746         (REGISTER_NAMES): Add CC registers.
7747         (BRANCH_COST): Change to constant 1.
7748         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
7749         handling.
7750         * config/pdp11/pdp11.opt (mbcopy): Remove.
7751         (mbcopy-builtin): Remove.
7752         (mbranch-cheap): Remove.
7753         (mbranch-expensive): Remove.
7754         * config/pdp11/predicates.md (expand_shift_operand): Update to
7755         match shift code generation.
7756         (ccnz_operator): New predicate.
7757         * doc/invoke.texi (PDP-11 Options): Remove deleted options
7758         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
7759         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
7760         exclusive options.
7761         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
7762         description of O constraint.
7764 2018-06-27  Jeff Law  <law@redhat.com>
7765             Austin Law  <austinklaw@gmail.com>
7767         * config/v850/v850.md (addsi3_set_flags): New pattern.
7768         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
7769         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
7770         (zero_extendhisi2_v850_set_flags): Likewise.
7771         (zero_extendqisi2_v850_set_flags): Likewise.
7772         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
7773         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
7774         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
7776         * config/v850/v850-protos.h (notice_update_cc): Remove.
7777         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
7778         (v850_print_operand): Handle 'D' and "d".
7779         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
7780         Add handling of arithmetic/logical operations compared against zero.
7781         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
7782         Do not look at v850_compare_op, instead get mode from last argument.
7783         (v850_gen_compare): Remove
7784         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
7785         after reload for prologue insns.
7786         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
7787         patterns.
7788         (construct_save_jarl): Likewise.
7789         (TARGET_FLAGS_REGNUM): Define.
7790         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
7791         (NOTICE_UPDATE_CC): Remove.
7792         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
7793         than cc0.  Conditionalize on reload_completed.
7794         (cmpsi_insn, setfcc_insn): Likewise.
7795         (tst1 splitter): Turn into define_and_split which sets the flags
7796         after reload.
7797         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
7798         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
7799         (cstoresf4, cstoredf4): Clobber the flags.
7800         (cmpsi, cmpsf, cmpdf): Remove expanders.
7801         (setf_insn): Remove pattern.
7802         (addsi3): Turn into define_and_split which clobbers the flags after
7803         reload and a suitable pattern (addsi3_clobber_flags) for use after
7804         reload.
7805         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
7806         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
7807         (ashrsi3, ashrsi3_v850e2): Likewise.
7808         (bins): Clobber the flags.
7809         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
7810         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
7811         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
7812         (call_value_internal_short, call_value_internal_long): Likewise.
7813         (callt_save_interrupt, callt_return_interrupt): Likewise.
7814         (save_interrupt, return_interrupt): Likewise.
7815         (callt_save_all_interrupt, save_all_interrupt): Likewise.
7816         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
7817         (restore_all_interrupt, _restore_all_interrupt): Likewise.
7818         (All FP comparisons): Only allow after reload has completed.
7819         (trfsr): Likewise.
7820         (divh, divhu): Tweak output template.
7821         (branch_z_normal, branch_z_invert): Remove
7822         (branch_nz_normal, branch_nz_invert): Likewise.
7823         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
7825         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
7826         * config/v850/v850.c (notice_update_cc): Remove.
7827         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
7828         (CC_NO_CARRY): Likewise.
7829         (NOTICE_UPDATE_CC): Define to nothing.
7830         * config/v850/v850.md: Remove block comment on cc0 handling
7831         Remove "cc" attribute from all patterns.  Remove cc_status handling
7832         from all patterns.  Minor formatting fixes.
7834 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7836         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
7837         (cortex-a76.cortex-a55): Likewise.
7838         * config/aarch64/aarch64-tune.md: Regenerate.
7839         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
7840         cortex-a76.cortex-a55.
7842 2018-06-27  Jeff Law  <law@redhat.com>
7844         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
7845         (MULTILIB_DIRNAMES): Similarly.
7847 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
7849         * gimple.h (gimple_return_retbnd): Delete.
7850         (gimple_return_set_retbnd): Likewise.
7851         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
7852         gimple_return_set_retbnd.
7853         * gimple-pretty-print.c (dump_gimple_return): Remove call to
7854         gimple_return_retbnd and adjust.
7855         * tree-inline.h (struct copy_body_data): Remove retbnd field.
7856         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
7857         Explicitly return NULL in a couple more cases.  Move assertion
7858         on debug statements and remove unreachable code.
7859         (reset_debug_binding): Do not test id->retbnd.
7860         (expand_call_inline): Do not set it.
7862 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
7864         * configure.ac: Add --disable-gcov option.
7865         * configure: Regenerate.
7866         * Makefile.in: Honour @enable_gcov@.
7867         * doc/install.texi: Document --disable-gcov.
7869 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7871         * config/arm/arm-cpus.in (cortex-a76): New entry.
7872         (cortex-a76.cortex-a55): Likewise.
7873         * config/arm/arm-tables.opt: Regenerate.
7874         * config/arm/arm-tune.md: Likewise.
7875         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
7876         * doc/invoke.texi (ARM Options): Document cortex-a76 and
7877         cortex-a76.cortex-a55.
7879 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
7881         PR target/85769
7882         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
7884 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7886         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
7887         comment.
7888         (EPILOGUE_USES): Likewise.
7890 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
7892         * tree-inline.c (remap_location): New function extracted from...
7893         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
7894         (copy_phis_for_bb): ...here.  Call remap_location.
7895         (copy_cfg_body): Adjust call to copy_edges_for_bb.
7897 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
7899         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
7900         unaligned vsx for 16B memset.
7902 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
7904         PR target/86285
7905         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
7906         ieee128_float_type_node to long_double_type_node unless
7907         TARGET_LONG_DOUBLE_128 is set.
7909 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
7911         * cfgloop.c (get_loop_location): Convert return type from
7912         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
7913         by implicit construction from rtx_insn *, and using
7914         dump_user_location_t::from_function_decl for the fallback case.
7915         * cfgloop.h (get_loop_location): Convert return type from
7916         location_t to dump_user_location_t.
7917         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
7918         dump_printf_loc to pass in a dump_location_t rather than a
7919         location_t, via the gimple stmt.
7920         * coverage.c (get_coverage_counts): Update calls to
7921         dump_printf_loc to pass in dump_location_t rather than a
7922         location_t.
7923         * doc/optinfo.texi (Dump types): Convert example of
7924         dump_printf_loc from taking "locus" to taking "insn".  Update
7925         description of the "_loc" calls to cover dump_location_t.
7926         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
7927         "selftest.h".
7928         (dump_user_location_t::dump_user_location_t): New constructors,
7929         from gimple *stmt and rtx_insn *.
7930         (dump_user_location_t::from_function_decl): New function.
7931         (dump_loc): Make static.
7932         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
7933         const dump_location_t &.
7934         (dump_generic_expr_loc): Delete.
7935         (dump_printf_loc): Convert param "loc" from location_t to
7936         const dump_location_t &.
7937         (selftest::test_impl_location): New function.
7938         (selftest::dumpfile_c_tests): New function.
7939         * dumpfile.h: Include "profile-count.h".
7940         (class dump_user_location_t): New class.
7941         (struct dump_impl_location_t): New struct.
7942         (class dump_location_t): New class.
7943         (dump_printf_loc): Convert 2nd param from source_location to
7944         const dump_location_t &.
7945         (dump_generic_expr_loc): Delete.
7946         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
7947         const dump_location_t &.
7948         * gimple-fold.c (fold_gimple_assign): Update call to
7949         dump_printf_loc to pass in a dump_location_t rather than a
7950         location_t, via the gimple stmt.
7951         (gimple_fold_call): Likewise.
7952         * gimple-loop-interchange.cc
7953         (loop_cand::analyze_iloop_reduction_var): Update for change to
7954         check_reduction_path.
7955         (tree_loop_interchange::interchange): Update for change to
7956         find_loop_location.
7957         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
7958         change in return-type of find_loop_location.
7959         (graphite_regenerate_ast_isl): Likewise.
7960         * graphite-optimize-isl.c (optimize_isl): Likewise.
7961         * graphite.c (graphite_transform_loops): Likewise.
7962         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
7963         pass in a dump_location_t rather than a location_t, via the
7964         gimple stmt.
7965         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
7966         * ipa.c (walk_polymorphic_call_targets): Likewise.
7967         * loop-unroll.c (report_unroll): Convert "locus" param from
7968         location_t to dump_location_t.
7969         (decide_unrolling): Update for change to get_loop_location's
7970         return type.
7971         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
7972         location_t to dump_user_location_t.
7973         (grid_find_single_omp_among_assignments_1): Updates calls to
7974         dump_printf_loc to pass in a dump_location_t rather than a
7975         location_t, via the gimple stmt.
7976         (grid_parallel_clauses_gridifiable): Convert "tloc" from
7977         location_t to dump_location_t.  Updates calls to dump_printf_loc
7978         to pass in a dump_location_t rather than a location_t, via the
7979         gimple stmt.
7980         (grid_inner_loop_gridifiable_p): Likewise.
7981         (grid_dist_follows_simple_pattern): Likewise.
7982         (grid_gfor_follows_tiling_pattern): Likewise.
7983         (grid_target_follows_gridifiable_pattern): Likewise.
7984         (grid_attempt_target_gridification): Convert initialization
7985         of local "grid" from memset to zero-initialization; FIXME: does
7986         this require C++11?  Update call to dump_printf_loc to pass in a
7987         optinfo_location rather than a location_t, via the gimple stmt.
7988         * profile.c (read_profile_edge_counts): Updates call to
7989         dump_printf_loc to pass in a dump_location_t rather than a
7990         location_t
7991         (compute_branch_probabilities): Likewise.
7992         * selftest-run-tests.c (selftest::run_tests): Call
7993         dumpfile_c_tests.
7994         * selftest.h (dumpfile_c_tests): New decl.
7995         * tree-loop-distribution.c (pass_loop_distribution::execute):
7996         Update for change in return type of find_loop_location.
7997         * tree-parloops.c (parallelize_loops): Likewise.
7998         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
7999         "locus" from location_t to dump_user_location_t.
8000         (canonicalize_loop_induction_variables): Likewise.
8001         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
8002         for change in return type of find_loop_location.
8003         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
8004         to dump_printf_loc to pass in a dump_location_t rather than a
8005         location_t, via the stmt.
8006         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
8007         Likewise.
8008         * tree-vect-loop-manip.c (find_loop_location): Convert return
8009         type from source_location to dump_user_location_t.
8010         (vect_do_peeling): Update for above change.
8011         (vect_loop_versioning): Update for change in type of
8012         vect_location.
8013         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
8014         from location_t to dump_user_location_t.
8015         (vect_estimate_min_profitable_iters): Update for change in type
8016         of vect_location.
8017         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
8018         location_t to dump_location_t.
8019         (vect_slp_bb): Update for change in type of vect_location.
8020         * tree-vectorizer.c (vect_location): Convert from source_location
8021         to dump_user_location_t.
8022         (try_vectorize_loop_1): Update for change in vect_location's type.
8023         (vectorize_loops): Likewise.
8024         (increase_alignment): Likewise.
8025         * tree-vectorizer.h (vect_location): Convert from source_location
8026         to dump_user_location_t.
8027         (find_loop_location): Convert return type from source_location to
8028         dump_user_location_t.
8029         (check_reduction_path): Convert 1st param from location_t to
8030         dump_user_location_t.
8031         * value-prof.c (check_counter): Update call to dump_printf_loc to
8032         pass in a dump_user_location_t rather than a location_t; update
8033         call to error_at for change in type of "locus".
8034         (check_ic_target): Update call to dump_printf_loc to
8035         pass in a dump_user_location_t rather than a location_t, via the
8036         call_stmt.
8038 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
8040         * config/s390/s390.h (enum processor_flags): Do not use
8041         default tune parameter when -march was specified.
8043 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
8045         PR target/86314
8046         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
8047         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
8048         operands.
8050 2018-06-26  Richard Biener  <rguenther@suse.de>
8052         PR tree-optimization/86287
8053         PR bootstrap/86316
8054         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
8055         (vect_analyze_loop): Initialize n_stmts.
8057 2018-06-26  Richard Biener  <rguenther@suse.de>
8059         PR middle-end/86271
8060         * fold-const.c (fold_convertible_p): Pointer extension
8061         isn't valid.
8063 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
8065         PR debug/86064
8066         * dwarf2out.c (loc_list_has_views): Adjust comments.
8067         (dw_loc_list): Split single cross-partition range with
8068         nonzero locview.
8070 2018-06-25  Jeff Law  <law@redhat.com>
8072         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
8073         on -mbig-switch by default.
8075         * config/v850/predicates.md (const_float_1_operand): Fix match_code
8076         test.
8077         (const_float_0_operand): Remove unused predicate.
8078         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
8079         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
8080         (recipsf2): New expander.  Original pattern now called
8081         (recipsf2_insn).
8082         (recipdf2, recipdf2_insn): Similarly.
8083         (rsqrtsf2, rsqrtsf2_insn): Similarly
8084         (rsqrtdf2, rsqrtdf2_insn): Similarly
8086 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
8088         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
8089         Simplify logic for FreeBSD (twice).
8091 2018-06-25  Martin Sebor  <msebor@redhat.com>
8093         PR tree-optimization/86204
8094         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
8095         a strnlen result if it's less than the length of the string.
8097 2018-06-25  Martin Sebor  <msebor@redhat.com>
8099         PR tree-optimization/85700
8100         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
8101         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
8102         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
8104 2018-06-25  Martin Sebor  <msebor@redhat.com>
8106         * doc/extend.texi (Zero-length arrays): Update and clarify.
8108 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
8110         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
8111         added IEEE/IBM long double multilib support on PowerPC little
8112         endian Linux systems.
8113         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
8114         (MULTILIB_DEFAULTS): Likewise.
8115         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8116         Likewise.
8117         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
8118         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
8119         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
8121 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
8123         PR middle-end/86311
8124         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
8125         (REORDER_45): Likewise.
8127 2018-06-25  Jeff Law  <law@redhat.com>
8129         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
8130         dividend to 32 bits.  Adjust length.
8131         (udivmodhi4): Cleanup output template.  Fix length.
8133 2018-06-25  Carl Love  <cel@us.ibm.com>
8135         * config/rs6000/vsx.md: Change word selector to prefered location.
8137 2018-06-25  Richard Biener  <rguenther@suse.de>
8139         PR tree-optimization/86304
8140         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
8141         epilogue-if-converted loops as well.
8143 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
8145         * lto-section-out.c (lto_begin_section): Do not print section
8146         name for noaddr and unnumbered dumps.
8148 2018-06-25  Richard Biener  <rguenther@suse.de>
8150         * tree-vectorizer.h (struct vec_info_shared): New structure
8151         with parts split out from struct vec_info and loop_nest from
8152         struct _loop_vec_info.
8153         (struct vec_info): Adjust accordingly.
8154         (struct _loop_vec_info): Likewise.
8155         (LOOP_VINFO_LOOP_NEST): Adjust.
8156         (LOOP_VINFO_DATAREFS): Likewise.
8157         (LOOP_VINFO_DDRS): Likewise.
8158         (struct _bb_vec_info): Likewise.
8159         (BB_VINFO_DATAREFS): Likewise.
8160         (BB_VINFO_DDRS): Likewise.
8161         (struct _stmt_vec_info): Add dr_aux member.
8162         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
8163         (DR_MISALIGNMENT_UNINITIALIZED): New.
8164         (set_dr_misalignment): Adjust.
8165         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
8166         (vect_analyze_loop): Adjust prototype.
8167         (vect_analyze_loop_form): Likewise.
8168         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8169         Compute dependences lazily.
8170         (vect_record_base_alignments): Use shared datarefs/ddrs.
8171         (vect_verify_datarefs_alignment): Likewise.
8172         (vect_analyze_data_refs_alignment): Likewise.
8173         (vect_analyze_data_ref_accesses): Likewise.
8174         (vect_analyze_data_refs): Likewise.
8175         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
8176         constructor parameter for shared part.
8177         (vect_analyze_loop_form): Pass in shared part and adjust.
8178         (vect_analyze_loop_2): Pass in storage for the number of
8179         stmts.  Move loop nest finding to the caller.  Compute
8180         datarefs lazily.
8181         (vect_analyze_loop): Pass in shared part.
8182         (vect_transform_loop): Verify shared datarefs are unchanged.
8183         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
8184         constructor parameter for shared part.
8185         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
8186         (vect_slp_bb): Verify shared datarefs are unchanged before
8187         transform.
8188         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
8189         change.
8190         (new_stmt_vec_info): Initialize DR_AUX misalignment to
8191         DR_MISALIGNMENT_UNINITIALIZED.
8192         * tree-vectorizer.c (vec_info::vec_info): Add constructor
8193         parameter for shared part.
8194         (vec_info::~vec_info): Adjust.
8195         (vec_info_shared::vec_info_shared): New.
8196         (vec_info_shared::~vec_info_shared): Likewise.
8197         (vec_info_shared::save_datarefs): Likewise.
8198         (vec_info_shared::check_datarefs): Likewise.
8199         (try_vectorize_loop_1): Construct shared part live for analyses
8200         of a single loop for multiple vector sizes.
8201         * tree-parloops.c (gather_scalar_reductions): Adjust.
8203 2018-06-25  Richard Biener  <rguenther@suse.de>
8205         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
8206         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
8207         (vect_analyze_data_refs): Remove similar code from here and
8208         simplify accordingly.
8210 2018-06-25  Richard Biener  <rguenther@suse.de>
8212         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
8213         for reverse storage order accesses rather than asserting
8214         they cannot happen here.
8216 2018-06-25  Tom de Vries  <tdevries@suse.de>
8218         PR debug/86257
8219         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
8220         Use data16 instead of .byte for insn prefix.
8222 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
8224         PR C++/86082
8225         * parser.c (make_char_string_pack): Pass this literal chars
8226         through cpp_interpret_string.
8227         (cp_parser_userdef_numeric_literal): Check the result of
8228         make_char_string_pack.
8230 2018-06-24  Maya Rashish  <coypu@sdf.org>
8232         * ginclude/stddef.h: Simplify conditions around avoiding
8233         re-definition of __size_t.
8235 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8237         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
8238         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
8240 2018-06-22  Maya Rashish  <coypu@sdf.org>
8242         * doc/invoke.texi (mno-fancy-math-387): Update for changes
8243         made to OpenBSD and NetBSD through the years.
8245 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8247         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8248         behavior of vec_pack (vector double, vector double) to match
8249         behavior of vec_float2 (vector double, vector double).
8251 2018-06-22  Olivier Hainque  <hainque@adacore.com>
8253         * gimplify.c (gimplify_function_tree): Prevent creation
8254         of a trampoline for the address of the current function
8255         passed to entry/exit instrumentation hooks.
8257 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
8259         PR target/86222
8260         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
8261         correctly.
8263 2018-06-22  Martin Liska  <mliska@suse.cz>
8265         PR tree-optimization/86263
8266         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8267         Bail out if is_enabled is false.
8268         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
8269         New declaration.
8270         (jump_table_cluster::is_enabled): New function.
8272 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8274         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8275         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
8276         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
8277         (lto_input_ts_binfo_tree_pointers): Likewise.
8278         * tree-streamer-out.c (streamer_write_tree_bitfields,
8279         write_ts_binfo_tree_pointers): Likewise.
8280         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
8282 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
8284         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
8286 2018-06-22  Martin Liska  <mliska@suse.cz>
8288         * symbol-summary.h (get): Make it pure and inline move
8289         functionality from ::get function.
8290         (get): Remove and inline into ::get and ::get_create.
8291         (get_create): Move code from ::get function.
8293 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8295         PR target/85994
8296         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
8297         -x assembler-with-cpp.
8299 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
8301         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
8302         _FILE_OFFSET_BITS=64 for C++.
8304 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
8306         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
8307         conversion insn that shows up when pr85657-3.c is compiled using
8308         IEEE 128-bit long double.
8309         (neg<mode>2_internal): Use the correct mode to check whether the
8310         mode is IBM extended.
8311         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
8312         multiply and divide external functions from being created more
8313         than once.
8315 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8317         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
8318         functions.
8319         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
8320         the edge can be forwarded.
8321         (cfg_layout_merge_blocks): Likewise.
8323 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8325         * except.c (finish_eh_generation): Commit edge insertions only after
8326         the EH edges have been redirected from post-landing to landing pads.
8328 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8330         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
8331         create_tmp_var_for to create the FRAME decl.
8332         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
8334 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
8336         * tree-inline.c (copy_edges_for_bb): Minor tweak.
8337         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
8338         debug statement when resetting its value.
8339         (expand_call_inline): Copy the locus of the call onto the assignment
8340         of the return value, if any.  Use local variable in more cases.
8342 2018-06-21  Martin Liska  <mliska@suse.cz>
8344         * ipa-pure-const.c (propagate_nothrow): Use
8345         funct_state_summaries->get.
8346         (dump_malloc_lattice): Likewise.
8347         (propagate_malloc): Likewise.
8349 2018-06-21  Richard Biener  <rguenther@suse.de>
8351         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
8352         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
8353         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
8354         comment.
8355         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
8356         BLOCK_ABSTRACT_ORIGIN unconditionally.
8358 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
8360         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
8361         deleting it.
8362         * ipa-reference.c (ipa_reference_c_finalize): Delete
8363         ipa_ref_opt_sum_summaries and set it to NULL.
8365 2018-06-21  Tom de Vries  <tdevries@suse.de>
8367         PR tree-optimization/85859
8368         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
8369         test with comment from bb_no_side_effects_p.
8371 2018-06-21  Richard Biener  <rguenther@suse.de>
8373         PR tree-optimization/86232
8374         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
8375         max for constant niter.
8377 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8379         * config/aarch64/aarch64-simd.md
8380         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
8382 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
8384         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
8385         Make opernads of the unspec commutative.
8387 2018-06-21  Richard Biener  <rguenther@suse.de>
8389         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
8390         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
8391         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
8392         (vect_analyze_data_ref_dependence): Re-order checks to deal with
8393         NULL DR_STEP.
8394         (vect_record_base_alignments): Do not record base alignment
8395         for gathers or scatters.
8396         (vect_compute_data_ref_alignment): Drop return value that is always
8397         true.  Bail out early for gathers or scatters.
8398         (vect_enhance_data_refs_alignment): Bail out early for gathers
8399         or scatters.
8400         (vect_find_same_alignment_drs): Likewise.
8401         (vect_analyze_data_refs_alignment): Remove dead code.
8402         (vect_slp_analyze_and_verify_node_alignment): Likewise.
8403         (vect_analyze_data_refs): For possible gathers or scatters do
8404         not create an alternate DR, just check their possible validity
8405         and mark them.  Adjust DECL_NONALIASED handling to not rely
8406         on DR_BASE_ADDRESS.
8407         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
8408         update inits of gathers or scatters.
8409         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
8410         Also copy gather/scatter flag to pattern vinfo.
8412 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
8414         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
8415         behavior of vec_packsu (vector unsigned long long, vector unsigned
8416         long long) to match behavior of vec_packs with same signature.
8418 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
8419             Thomas Schwinge <thomas@codesourcery.com>
8420             Cesar Philippidis  <cesar@codesourcery.com>
8422         * gimplify.c (gimplify_scan_omp_clauses): Add support for
8423         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8424         (gimplify_adjust_omp_clauses): Likewise.
8425         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
8426         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
8427         (gimplify_omp_target_update): Update handling of acc update and
8428         enter/exit data.
8429         * omp-low.c (install_var_field): Remove unused parameter
8430         base_pointers_restrict.
8431         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
8432         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
8433         FINALIZE}
8434         (omp_target_base_pointers_restrict_p): Delete.
8435         (scan_omp_target): Update call to scan_sharing_clauses.
8436         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
8437         FINALIZE}.
8438         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
8439         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
8440         (convert_local_omp_clauses): Likewise.
8441         * tree-pretty-print.c (dump_omp_clause): Likewise.
8442         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
8443         FINALIZE}.
8444         (omp_clause_code_name): Likewise.
8446 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
8448         PR debug/86194
8449         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
8450         be narrowed.
8452         PR tree-optimization/86231
8453         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
8454         anti-range don't overwrite *vr0min before using it to compute *vr0max.
8456 2018-06-20  Tom de Vries  <tdevries@suse.de>
8458         PR tree-optimization/86097
8459         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
8460         iv type if signedness of iv type is not the same as that of *nit.
8462 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
8464         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
8465         EDGE_EH edges, verify they are all EDGE_EH.
8467 2018-06-20  Maya Rashish  <coypu@sdf.org>
8469         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
8471 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8473         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
8474         * config/aarch64/aarch64.c (xgene1_tunings): Add
8475         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
8476         (aarch64_mode_valid_for_sched_fusion_p):
8477         Allow 16-byte modes.
8478         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
8479         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
8480         128-bit modes.
8481         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
8482         New pattern.
8483         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
8484         * config/aarch64/iterators.md (VQ2): New mode iterator.
8486 2018-06-20  Martin Liska  <mliska@suse.cz>
8488         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8489         Change default ratio from 10 to 8.
8491 2018-06-20  Martin Liska  <mliska@suse.cz>
8493         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8494         New.
8495         (bit_test_cluster::find_bit_tests): Likewise.
8496         (switch_decision_tree::analyze_switch_statement): Find clusters.
8497         * tree-switch-conversion.h (struct jump_table_cluster): Document
8498         hierarchy.
8500 2018-06-20  Martin Liska  <mliska@suse.cz>
8502         * tree-switch-conversion.c (switch_conversion::collect):
8503         Record m_uniq property.
8504         (switch_conversion::expand): Bail out for special conditions.
8505         (group_cluster::~group_cluster): New.
8506         (group_cluster::group_cluster): Likewise.
8507         (group_cluster::dump): Likewise.
8508         (jump_table_cluster::emit): New.
8509         (switch_decision_tree::fix_phi_operands_for_edges): New.
8510         (struct case_node): Remove struct.
8511         (jump_table_cluster::can_be_handled): New.
8512         (case_values_threshold): Moved to header.
8513         (reset_out_edges_aux): Likewise.
8514         (jump_table_cluster::is_beneficial): New.
8515         (bit_test_cluster::can_be_handled): Likewise.
8516         (add_case_node): Remove.
8517         (bit_test_cluster::is_beneficial): New.
8518         (case_bit_test::cmp): New.
8519         (bit_test_cluster::emit): New.
8520         (expand_switch_as_decision_tree_p): Remove.
8521         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
8522         (fix_phi_operands_for_edge): Likewise.
8523         (switch_decision_tree::analyze_switch_statement): New.
8524         (compute_cases_per_edge): Move ...
8525         (switch_decision_tree::compute_cases_per_edge): ... here.
8526         (try_switch_expansion): Likewise.
8527         (switch_decision_tree::try_switch_expansion): Likewise.
8528         (record_phi_operand_mapping): Likewise.
8529         (switch_decision_tree::record_phi_operand_mapping): Likewise.
8530         (emit_case_decision_tree): Likewise.
8531         (switch_decision_tree::emit): Likewise.
8532         (balance_case_nodes): Likewise.
8533         (switch_decision_tree::balance_case_nodes): Likewise.
8534         (dump_case_nodes): Likewise.
8535         (switch_decision_tree::dump_case_nodes): Likewise.
8536         (emit_jump): Likewise.
8537         (switch_decision_tree::emit_jump): Likewise.
8538         (emit_cmp_and_jump_insns): Likewise.
8539         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
8540         (emit_case_nodes): Likewise.
8541         (switch_decision_tree::emit_case_nodes): Likewise.
8542         (conditional_probability): Remove.
8543         * tree-switch-conversion.h (enum cluster_type): New.
8544         (PRINT_CASE): New.
8545         (struct cluster): Likewise.
8546         (cluster::cluster): Likewise.
8547         (struct simple_cluster): Likewise.
8548         (simple_cluster::simple_cluster): Likewise.
8549         (struct group_cluster): Likewise.
8550         (struct jump_table_cluster): Likewise.
8551         (struct bit_test_cluster): Likewise.
8552         (struct min_cluster_item): Likewise.
8553         (struct case_tree_node): Likewise.
8554         (case_tree_node::case_tree_node): Likewise.
8555         (jump_table_cluster::case_values_threshold): Likewise.
8556         (struct case_bit_test): Likewise.
8557         (struct switch_decision_tree): Likewise.
8558         (struct switch_conversion): Likewise.
8559         (switch_decision_tree::reset_out_edges_aux): Likewise.
8561 2018-06-20  Martin Liska  <mliska@suse.cz>
8563         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
8564         (hoist_edge_and_branch_if_true): Likewise.
8565         (expand_switch_using_bit_tests_p): Likewise.
8566         (struct case_bit_test): Likewise.
8567         (case_bit_test_cmp): Likewise.
8568         (emit_case_bit_tests): Likewise.
8569         (switch_conversion::switch_conversion): New class.
8570         (struct switch_conv_info): Remove old struct.
8571         (collect_switch_conv_info): More to ...
8572         (switch_conversion::collect): ... this.
8573         (check_range): Likewise.
8574         (switch_conversion::check_range): Likewise.
8575         (check_all_empty_except_final): Likewise.
8576         (switch_conversion::check_all_empty_except_final): Likewise.
8577         (check_final_bb): Likewise.
8578         (switch_conversion::check_final_bb): Likewise.
8579         (create_temp_arrays): Likewise.
8580         (switch_conversion::create_temp_arrays): Likewise.
8581         (free_temp_arrays): Likewise.
8582         (gather_default_values): Likewise.
8583         (switch_conversion::gather_default_values): Likewise.
8584         (build_constructors): Likewise.
8585         (switch_conversion::build_constructors): Likewise.
8586         (constructor_contains_same_values_p): Likewise.
8587         (switch_conversion::contains_same_values_p): Likewise.
8588         (array_value_type): Likewise.
8589         (switch_conversion::array_value_type): Likewise.
8590         (build_one_array): Likewise.
8591         (switch_conversion::build_one_array): Likewise.
8592         (build_arrays): Likewise.
8593         (switch_conversion::build_arrays): Likewise.
8594         (gen_def_assigns): Likewise.
8595         (switch_conversion::gen_def_assigns): Likewise.
8596         (prune_bbs): Likewise.
8597         (switch_conversion::prune_bbs): Likewise.
8598         (fix_phi_nodes): Likewise.
8599         (switch_conversion::fix_phi_nodes): Likewise.
8600         (gen_inbound_check): Likewise.
8601         (switch_conversion::gen_inbound_check): Likewise.
8602         (process_switch): Use the newly created class.
8603         (switch_conversion::expand): New.
8604         (switch_conversion::~switch_conversion): New.
8605         * tree-switch-conversion.h: New file.
8607 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8609         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
8610         tree-vect-patterns.c.
8611         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
8612         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
8613         (vect_recog_sad_pattern): Likewise.
8614         (vect_recog_widen_sum_pattern): Likewise.
8615         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
8616         (vect_recog_widen_shift_pattern): Remove the type_in argument.
8617         (vect_recog_rotate_pattern): Likewise.
8618         (vect_recog_mult_pattern): Likewise.
8619         (vect_recog_vector_vector_shift_pattern): Likewise.
8620         (vect_recog_divmod_pattern): Likewise.
8621         (vect_recog_mixed_size_cond_pattern): Likewise.
8622         (vect_recog_bool_pattern): Likewise.
8623         (vect_recog_mask_conversion_pattern): Likewise.
8624         (vect_try_gather_scatter_pattern): Likewise.
8625         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
8626         (vect_recog_over_widening_pattern): Likewise.
8627         (vect_recog_gather_scatter_pattern): Likewise.
8628         (vect_recog_func_ptr): Move from tree-vectorizer.h
8629         (vect_vect_recog_func_ptrs): Move further down the file.
8630         (vect_recog_func): Likewise.  Remove the third argument.
8631         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
8632         (vect_pattern_recog_1): Expect the pattern function to do any
8633         necessary target tests.  Also expect it to provide a vector type.
8634         Remove the type_in handling.
8636 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8638         * tree-vect-patterns.c (vect_pattern_detected): New function.
8639         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
8640         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
8641         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
8642         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
8643         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8644         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8645         (vect_recog_mask_conversion_pattern)
8646         (vect_try_gather_scatter_pattern): Likewise.
8648 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8650         * tree-vect-patterns.c (vect_get_internal_def): New function.
8651         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8652         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
8653         (search_type_for_mask_1): Use it.
8655 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8657         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8658         redundant WIDEN_SUM_EXPR handling.
8659         (vect_recog_sad_pattern): Likewise.
8661 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8663         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
8664         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
8665         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
8666         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
8667         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
8669 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8671         * tree-vect-stmts.c (vectorizable_call): Make sure that we
8672         use the stmt_vec_info of the original bb statement for the
8673         new zero assignment, even if the call is part of a pattern.
8675 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8677         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
8678         that the sequence is attached to the original statement rather
8679         than the pattern statement.
8680         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
8681         PATTERN_DEF_SEQ from the original statement rather than
8682         the main pattern statement.
8683         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
8684         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
8685         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
8687 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
8689         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
8690         definition statements before the early exit for statements that aren't
8691         live or relevant.
8692         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
8693         split out from...
8694         (vect_transform_loop): ...here.  Process pattern definition
8695         statements without first checking whether the main pattern
8696         statement is live or relevant.
8698 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
8700         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
8701         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
8703 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
8705         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
8706         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
8707         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
8708         (expand_block_compare): Change select_block_compare_mode call.
8709         (expand_strncmp_align_check): Use new functions, fix comment.
8710         (emit_final_str_compare_gpr): New function.
8711         (expand_strn_compare): Refactor and clean up code.
8712         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
8714 2018-06-19  Tony Reix  <tony.reix@atos.com>
8715             Damien Bergamini  <damien.bergamini@atos.com>
8716             David Edelsohn  <dje.gcc@gmail.com>
8718         * collect2.c (static_obj): New variable.
8719         (static_libs): New variable.
8720         (is_in_list): Uncomment declaration.
8721         (main): Track AIX libraries linked statically.
8722         (is_in_list): Uncomment definition.
8723         (scan_prog_file): Don't add AIX shared libraries initializer
8724         to constructor list if linking statically.
8726 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
8728         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
8729         constant.
8730         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
8732 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
8734         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
8735         blocks.
8737 2018-06-19  Martin Liska  <mliska@suse.cz>
8739         * config/i386/i386.c (ix86_can_inline_p): Do not use
8740         ipa_fn_summaries::get_create.
8741         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
8742         get.
8743         (devirtualization_time_bonus): Likewise.
8744         (ipcp_propagate_stage): Likewise.
8745         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
8746         (edge_set_predicate): Likewise.
8747         (evaluate_conditions_for_known_args): Likewise.
8748         (evaluate_properties_for_edge): Likewise.
8749         (ipa_call_summary::reset): Tranform to ...
8750         (ipa_call_summary::~ipa_call_summary): ... this.
8751         (ipa_fn_summary::reset): Transform to ...
8752         (ipa_fn_summary::~ipa_fn_summary): ... this.
8753         (ipa_fn_summary_t::remove): Rename to ...
8754         (ipa_fn_summary_t::remove_callees): ... this.
8755         (ipa_fn_summary_t::duplicate): Use placement new
8756         instead of memory copy.
8757         (ipa_call_summary_t::duplicate): Likewise.
8758         (ipa_call_summary_t::remove): Remove.
8759         (dump_ipa_call_summary): Change get_create to get.
8760         (ipa_dump_fn_summary): Dump only when summary exists.
8761         (analyze_function_body): Use symbol_summary::get instead
8762         of get_create.
8763         (compute_fn_summary): Likewise.
8764         (estimate_edge_devirt_benefit): Likewise.
8765         (estimate_edge_size_and_time): Likewise.
8766         (inline_update_callee_summaries): Likewise.
8767         (remap_edge_change_prob): Likewise.
8768         (remap_edge_summaries): Likewise.
8769         (ipa_merge_fn_summary_after_inlining): Likewise.
8770         (write_ipa_call_summary): Likewise.
8771         (ipa_fn_summary_write): Likewise.
8772         (ipa_free_fn_summary): Likewise.
8773         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
8774         (struct ipa_call_summary): Likewise.
8775         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
8776         of get_create.
8777         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
8778         (estimate_size_after_inlining): Likewise.
8779         (estimate_growth): Likewise.
8780         (growth_likely_positive): Likewise.
8781         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
8782         (inline_call): Likewise.
8783         * ipa-inline.c (caller_growth_limits): Likewise.
8784         (can_inline_edge_p): Likewise.
8785         (can_inline_edge_by_limits_p): Likewise.
8786         (compute_uninlined_call_time): Likewise.
8787         (compute_inlined_call_time): Likewise.
8788         (want_inline_small_function_p): Likewise.
8789         (edge_badness): Likewise.
8790         (update_caller_keys): Likewise.
8791         (update_callee_keys): Likewise.
8792         (inline_small_functions): Likewise.
8793         (inline_to_all_callers_1): Likewise.
8794         (dump_overall_stats): Likewise.
8795         (early_inline_small_functions): Likewise.
8796         (early_inliner): Likewise.
8797         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
8798         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
8799         * ipa-pure-const.c (malloc_candidate_p): Likewise.
8800         * ipa-split.c (execute_split_functions): Likewise.
8801         * symbol-summary.h: Likewise.
8802         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
8804 2018-06-19  Richard Biener  <rguenther@suse.de>
8806         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
8807         (vectorize_loops): ... here.  Fix dbgcnt handling.
8808         (try_vectorize_loop): Wrap try_vectorize_loop_1.
8810 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
8812         PR target/86197
8813         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8814         ieee128 argument takes up only one (vector) register, not two (floating
8815         point) registers.
8817 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
8819         * gimplify.c (gimplify_init_constructor): Really never clear for an
8820         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
8822 2018-06-19  Richard Biener  <rguenther@suse.de>
8824         PR tree-optimization/86179
8825         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
8826         after failed recognition.
8828 2018-06-18  Martin Sebor  <msebor@redhat.com>
8830         PR middle-end/85602
8831         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
8832         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
8833         Handle integer subtraction.
8834         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
8835         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
8837 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
8839         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
8840         param from rtx to rtx_insn *.
8841         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
8842         param.
8843         (frv_ifcvt_modify_insn): Likwise.
8844         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
8845         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
8846         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
8847         as_a <rtx_insn *> cast to local "unprotected_region" once
8848         it's been established that it's not NULL or pc_rtx.
8849         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
8850         param "sethi" from rtx to rtx_insn *.
8851         (nds32_group_float_insns): Likewise for param "insn".
8852         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
8853         param.
8854         (vax_output_int_subtract): Likewise.
8855         * config/vax/vax.c (vax_output_int_add): Likewise for param
8856         "insn".
8857         (vax_output_int_subtract): Likewise.
8858         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
8859         (emit_pattern_after): Likewise for param "after".
8860         (emit_insn_after): Likewise.
8861         (emit_jump_insn_after): Likewise.
8862         (emit_call_insn_after): Likewise.
8863         (emit_debug_insn_after): Likewise.
8864         (emit_pattern_before): Likewise for param "before".
8865         (emit_insn_before): Likewise.
8866         (emit_jump_insn_before): Likewise.
8867         * final.c (get_insn_template): Likewise for param "insn", removing
8868         a cast.
8869         * output.h (get_insn_template): Likewise for 2nd param.
8870         * rtl.h (emit_insn_before): Likewise.
8871         (emit_jump_insn_before): Likewise.
8872         (emit_debug_insn_before_noloc): Likewise.
8873         (emit_insn_after): Likewise.
8874         (emit_jump_insn_after): Likewise.
8875         (emit_call_insn_after): Likewise.
8876         (emit_debug_insn_after): Likewise.
8877         (set_insn_deleted): Likewise for param.
8879 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
8881         PR target/85358
8882         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
8883         floating point modes, so that IFmode is numerically greater than
8884         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
8885         to declare the ordering.  This prevents IFmode from being
8886         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
8887         machine.  Include rs6000-modes.h to share the fractional values
8888         between genmodes* and the rest of the compiler.
8889         (IFmode): Likewise.
8890         (KFmode): Likewise.
8891         (TFmode): Likewise.
8892         * config/rs6000/rs6000-modes.h: New file.
8893         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
8894         meaning of rs6000_long_double_size so that 126..128 selects an
8895         appropriate 128-bit floating point type.
8896         (rs6000_option_override_internal): Likewise.
8897         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
8898         (TARGET_LONG_DOUBLE_128): Change the meaning of
8899         rs6000_long_double_size so that 126..128 selects an appropriate
8900         128-bit floating point type.
8901         (LONG_DOUBLE_TYPE_SIZE): Update comment.
8902         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
8903         source and destination to match the standard usage.
8904         (truncifkf2): Likewise.
8905         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
8906         ISA 2.07 to use an explicit clobber, instead of passing in a
8907         temporary.
8908         (copysign<mode>3_soft): Likewise.
8910 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
8912         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
8913         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
8914         (vect_slp_analyze_instance_dependence): Likewise.
8915         (vect_enhance_data_refs_alignment): Likewise.
8916         (vect_analyze_data_refs_alignment): Likewise.
8917         (vect_slp_analyze_and_verify_instance_alignment
8918         (vect_analyze_data_ref_accesses): Likewise.
8919         (vect_prune_runtime_alias_test_list): Likewise.
8920         (vect_analyze_data_refs): Likewise.
8921         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
8922         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
8923         (vect_analyze_scalar_cycles_1): Likewise.
8924         (vect_get_loop_niters): Likewise.
8925         (vect_analyze_loop_form_1): Likewise.
8926         (vect_update_vf_for_slp): Likewise.
8927         (vect_analyze_loop_operations): Likewise.
8928         (vect_analyze_loop): Likewise.
8929         (vectorizable_induction): Likewise.
8930         (vect_transform_loop): Likewise.
8931         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
8932         * tree-vect-slp.c (vect_analyze_slp): Likewise.
8933         (vect_make_slp_decision): Likewise.
8934         (vect_detect_hybrid_slp): Likewise.
8935         (vect_slp_analyze_operations): Likewise.
8936         (vect_slp_bb): Likewise.
8937         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
8938         (vectorizable_bswap): Likewise.
8939         (vectorizable_call): Likewise.
8940         (vectorizable_simd_clone_call): Likewise.
8941         (vectorizable_conversion): Likewise.
8942         (vectorizable_assignment): Likewise.
8943         (vectorizable_shift): Likewise.
8944         (vectorizable_operation): Likewise.
8945         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
8947 2018-06-18  Martin Sebor  <msebor@redhat.com>
8949         PR tree-optimization/81384
8950         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
8951         * builtins.c (expand_builtin_strnlen): New function.
8952         (expand_builtin): Call it.
8953         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
8954         * builtins.def (BUILT_IN_STRNLEN): New.
8955         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
8956         Warn for bounds in excess of maximum object size.
8957         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
8958         single-value ranges.  Handle strnlen.
8959         (handle_builtin_strlen): Handle strnlen.
8960         (strlen_check_and_optimize_stmt): Same.
8961         * doc/extend.texi (Other Builtins): Document strnlen.
8963 2018-06-18  Maya Rashish  <coypu@sdf.org>
8965         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
8966         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
8967         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
8969         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
8970         here to ...
8971         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
8973 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8975         * tree.c (escaped_string::escape): Replace cast to char * by
8976         const_cast<char *> (unescaped).
8978 2018-06-18  Nick Clifton  <nickc@redhat.com>
8980         PR 84195
8981         * tree.c (escaped_string): New class.  Converts an unescaped
8982         string into its escaped equivalent.
8983         (warn_deprecated_use): Use the new class to convert the
8984         deprecation message, if present.
8985         (test_escaped_strings): New self test.
8986         (test_c_tests): Add test_escaped_strings.
8987         * doc/extend.texi (deprecated): Add a note that the
8988         deprecation message is affected by the -fmessage-length
8989         option, and that control characters will be escaped.
8990         (#pragma GCC error): Document this pragma.
8991         (#pragma GCC warning): Likewise.
8992         * doc/invoke.texi (-fmessage-length): Document this option's
8993         effect on the #warning and #error preprocessor directives and
8994         the deprecated attribute.
8996 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
8998         * tree.c (decl_value_expr_lookup): Revert latest change.
8999         (decl_value_expr_insert): Likewise.
9001 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
9003         * gimplify.c (nonlocal_vlas): Delete.
9004         (nonlocal_vla_vars): Likewise.
9005         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
9006         referenced VLAs.
9007         (gimplify_body): Do not create and destroy nonlocal_vlas.
9008         * tree-nested.c: Include diagnostic.h.
9009         (use_pointer_in_frame): Tweak.
9010         (lookup_field_for_decl): Add assertion and declare the transformation.
9011         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
9012         internal error when the reference is in a wrong context.  Do not
9013         create a debug decl by default.
9014         (note_nonlocal_block_vlas): Delete.
9015         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
9016         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
9017         create a debug decl by default.
9018         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
9019         call is in a wrong context.
9020         (fixup_vla_decls): New function.
9021         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
9022         debug variables were created.
9023         * tree.c (decl_value_expr_lookup): Add checking assertion.
9024         (decl_value_expr_insert): Likewise.
9026 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
9028         PR middle-end/82479
9029         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
9030         * tree-scalar-evolution.c (interpret_expr): Likewise.
9031         (expression_expensive_p): Likewise.
9032         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
9033         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
9034         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
9035         (ssa_defined_by_minus_one_stmt_p): New.
9037 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
9039         PR middle-end/64946
9040         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
9041         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
9042         * dojump.c (do_jump): Likewise.
9043         * expr.c (expand_expr_real_2): Check operand type's sign.
9044         * fold-const.c (const_unop): Handle ABSU_EXPR.
9045         (fold_abs_const): Likewise.
9046         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
9047         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
9048         (strip_sign_op_1): Likesise.
9049         * match.pd: Add new pattern to generate ABSU_EXPR.
9050         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
9051         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
9052         * tree-eh.c (operation_could_trap_helper_p): Likewise.
9053         * tree-inline.c (estimate_operator_cost): Likewise.
9054         * tree-pretty-print.c (dump_generic_node): Likewise.
9055         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
9056         * tree.def (ABSU_EXPR): New.
9058 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
9060         PR middle-end/86095
9061         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
9062         documented as preserved for backward compatibility only.
9063         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
9065         PR rtl-optimization/86108
9066         * bb-reorder.c (create_forwarder_block): Renamed to ...
9067         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
9068         jump from new landing pad to the second part.
9069         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
9070         Adjust callers.
9072 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
9074         PR middle-end/85878
9075         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
9076         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
9077         Only call store_expr for halves if the mode is the same.
9079         PR middle-end/86123
9080         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
9081         Fix up comment formatting.
9083 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9085         * typed-splay-tree.h (typed_splay_tree::remove): New function.
9086         (typed_splay_tree::closure,
9087         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
9088         (typed_splay_tree::typed_splay_tree,
9089         typed_splay_tree::operator =): Declared private.
9090         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
9091         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
9092         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
9093         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
9094         typed_splay_tree::splay_tree_splay,
9095         typed_splay_tree::splay_tree_foreach_helper,
9096         typed_splay_tree::splay_tree_insert,
9097         typed_splay_tree::splay_tree_remove,
9098         typed_splay_tree::splay_tree_lookup,
9099         typed_splay_tree::splay_tree_predecessor,
9100         typed_splay_tree::splay_tree_successor,
9101         typed_splay_tree::splay_tree_min,
9102         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
9103         (typed_splay_tree::root, typed_splay_tree::comp,
9104         typed_splay_tree::delete_key,
9105         typed_splay_tree::delete_value): New data members.
9106         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
9107         typed_splay_tree::remove.
9109 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
9111         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
9112         -mginv and -mno-ginv to the assembler.
9113         * config/mips/mips.opt (-mcrc): New option.
9114         (-mginv): Likewise.
9115         * doc/invoke.text (-mcrc): Document.
9116         (-mginv): Likewise.
9118 2018-06-15  Nick Clifton  <nickc@redhat.com>
9120         PR 84195
9121         * tree.c (escaped_string): New class.  Converts an unescaped
9122         string into its escaped equivalent.
9123         (warn_deprecated_use): Use the new class to convert the
9124         deprecation message, if present.
9125         (test_escaped_strings): New self test.
9126         (test_c_tests): Add test_escaped_strings.
9127         * doc/extend.texi (deprecated): Add a note that the
9128         deprecation message is affected by the -fmessage-length
9129         option, and that control characters will be escaped.
9130         (#pragma GCC error): Document this pragma.
9131         (#pragma GCC warning): Likewise.
9132         * doc/invoke.texi (-fmessage-length): Document this option's
9133         effect on the #warning and #error preprocessor directives and
9134         the deprecated attribute.
9136 2018-06-15  Richard Biener  <rguenther@suse.de>
9138         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
9139         here, also noting vector size used.
9140         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
9141         size used in MSG_OPTIMIZED_LOCATIONS dump.
9142         (pass_slp_vectorize::execute): Adjust.
9144 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
9146         PR target/85968
9147         * config/arc/arc.c (arc_return_address_register): Fix
9148         if-condition.
9150 2018-06-15  Richard Biener  <rguenther@suse.de>
9152         PR middle-end/86159
9153         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
9154         leave useless conversion stripping to force_gimple_operand_gsi.
9155         (gimplify_build2): Likewise.
9156         (gimplify_build1): Likewise.
9158 2018-06-15  Richard Biener  <rguenther@suse.de>
9160         PR middle-end/86076
9161         * tree-cfg.c (move_stmt_op): unshare invariant addresses
9162         before adjusting their block.
9164 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9166         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
9167         multilibs for *-*-rtems*.
9168         * config/riscv/t-rtems: New file.
9170 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
9172         PR middle-end/86122
9173         * match.pd ((A +- CST1) +- CST2): Punt if last resort
9174         unsigned_type_for returns NULL.
9176         PR target/85945
9177         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
9178         subregs of multi-word pseudos unless the float mode has word size.
9180 2018-06-14  Richard Biener  <rguenther@suse.de>
9182         PR middle-end/86139
9183         * tree-vect-generic.c (build_word_mode_vector_type): Remove
9184         duplicate and harmful type_hash_canon.
9185         * tree.c (type_hash_canon): Assert we didn't find ourselves.
9187 2018-06-14  Richard Biener  <rguenther@suse.de>
9189         PR ipa/86124
9190         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
9191         NULL cgraph_node.
9193 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
9195         * config/rtems.h (STDINT_LONG32): Define.
9197 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
9198             Prachi Godbole  <prachi.godbole@imgtec.com>
9200         * config/mips/mips-cpus.def: Define P6600.
9201         * config/mips/mips-tables.opt: Regenerate.
9202         * config/mips/mips.c (mips_ucbranch_type): New enum.
9203         (mips_rtx_cost_data): Add support for P6600.
9204         (mips_issue_rate): Likewise.
9205         (mips_multipass_dfa_lookahead): Likewise.
9206         (mips_avoid_hazard): Likewise.
9207         (mips_reorg_process_insns): Likewise.
9208         (mips_classify_branch_p6600): New function.
9209         * config/mips/mips.h (TUNE_P6600): New define.
9210         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
9211         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
9212         * config/mips/mips.md: Include p6600.md.
9213         (processor): Add p6600.
9214         * config/mips/p6600.md: New file.
9215         * doc/invoke.texi: Add p6600 to supported architectures.
9217 2018-06-13  Martin Sebor  <msebor@redhat.com>
9219         PR tree-optimization/86114
9220         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
9221         of integer types.
9222         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
9224 2018-06-13  Richard Biener  <rguenther@suse.de>
9226         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
9227         Properly set vector type of the intermediate stmt.
9228         * tree-vect-stmts.c (vectorizable_operation): The destination
9229         var always has vectype_out type.
9231 2018-06-13  Jeff Law  <law@redhat.com>
9233         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
9234         integer 0 for argument to print_rtl_with_bb.
9235         (rl78_reorg): Likewise.
9237 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
9239         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
9240         from rtx to rtx_insn *.
9241         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
9242         "label".
9243         (add_sched_insns_for_speculation): Likewise for local "target",
9244         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9245         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
9246         from rtx_insn ** to rtx_code_label **.
9247         (reorg_emit_nops): Likewise.
9248         (c6x_reorg): Likewise for local "call_labels".
9249         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
9250         rtx to rtx_insn *.
9251         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
9252         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
9253         the loops over LABEL_REFS.
9254         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
9255         braf_label.
9256         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
9257         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
9258         (split_branches): Strengthen local "olabel" from rtx to
9259         rtx_insn *, adding a safe_as_a cast.
9260         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
9261         to "rtx_insn *".
9262         (add_insn_after): Likewise for first two params.
9263         (add_insn_before): Likewise.
9264         (remove_insn): Likewise for param.
9265         (emit_pattern_before_noloc): Likewise for second and third params.
9266         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
9267         (emit_call_insn_before_noloc): Likewise.
9268         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
9269         to "rtx_insn *".
9270         (emit_barrier_before): Likewise.
9271         (emit_label_before): Strengthen "label" param from "rtx" to
9272         "rtx_code_label *".  Strengthen "before" param from "rtx" to
9273         "rtx_insn *".
9274         (emit_insn_after_1): Strengthen "after" param from "rtx" to
9275         "rtx_insn *".
9276         (emit_pattern_after_noloc): Likewise.
9277         (emit_insn_after_noloc): Likewise.
9278         (emit_jump_insn_after_noloc): Likewise.
9279         (emit_call_insn_after_noloc): Likewise.
9280         (emit_debug_insn_after_noloc): Likewise.
9281         (emit_barrier_after): Likewise.
9282         (emit_label_after): Likewise for both params.
9283         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
9284         "loc" param from "int" to "location_t".
9285         (emit_insn_after_setloc): Likewise.
9286         (emit_jump_insn_after_setloc): Likewise.
9287         (emit_call_insn_after_setloc): Likewise.
9288         (emit_debug_insn_after_setloc): Likewise.
9289         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
9290         "loc" param from "int" to "location_t".
9291         (emit_pattern_before): Convert NULL_RTX to NULL.
9292         (emit_insn_before_setloc): Convert "loc" param from "int" to
9293         "location_t".
9294         (emit_jump_insn_before_setloc): Likewise.
9295         (emit_call_insn_before_setloc): Likewise.
9296         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
9297         rtx_insn *.  Convert "loc" param from "int" to "location_t".
9298         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
9299         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
9300         Convert 3rd param from "int" to "location_t".
9301         (emit_barrier_before, emit_barrier_after, next_real_insn):
9302         Strengthen param from rtx to rtx_insn *.
9303         (emit_label_before): Strengthen 1st param from "rtx" to
9304         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
9305         "rtx_insn *".
9306         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
9307         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
9308         Strengthen 2nd param from "rtx" to "rtx_insn *".
9309         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
9310         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
9311         Likewise. Convert 3rd param from "int" to "location_t".
9312         (emit_label_after): Strengthen 1st param from "rtx" to
9313         "rtx_code_label *".
9314         (next_real_insn, remove_insn): Strengthen param from "rtx" to
9315         "rtx_insn *".
9316         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
9317         from "rtx" to "rtx_insn *".
9319 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
9321         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
9322         bodies streamed in with -Q.
9323         * dumpfile.c (dump_files): Add lto-stream-out dump file.
9324         * dumpfile.h (tree_dump_index): Add lto_stream_out.
9325         * gimple-streamer-out.c: Include gimple-pretty-print.h
9326         (output_bb): Dump stmts streamed.
9327         * lto-section-out.c: Include print-tree.h
9328         (lto_begin_section): Dump sections created.
9329         (lto_output_decl_index): Dump decl encoded.
9330         * lto-streamer-out.c: Include print-tree.h
9331         (create_output_block): Dump output block created.
9332         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
9333         (output_function): Dump function output.
9334         (output_constructor): Dump constructor streamed.
9335         (write_global_stream): Output indexes encoded.
9336         (produce_asm_for_decls): Dump streams encoded.
9337         * lto-streamer.c (streamer_dump_file): New global var.
9338         * lto-streamer.h (streamer_dump_file): Declare.
9339         * passes.c (ipa_write_summaries): Initialize streamer dump.
9340         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
9341         in.
9343 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
9345         PR target/86048
9346         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
9347         offsets for register save directives.  Emit a second batch of save
9348         directives, if need be, when the function accesses prior frames.
9350 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9352         * config/arc/fpu.md (fmasf4): Force operand to register.
9353         (fnmasf4): Likewise.
9355 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9357         * config/arc/arc-protos.h (arc_pad_return): Remove.
9358         * config/arc/arc.c (machine_function): Remove force_short_suffix
9359         and size_reason.
9360         (arc_print_operand): Adjust printing of '&'.
9361         (arc_verify_short): Remove conditional printing of short suffix.
9362         (arc_final_prescan_insn): Remove reference to size_reason.
9363         (pad_return): New function.
9364         (arc_reorg): Call pad_return.
9365         (arc_pad_return): Remove.
9366         (arc_init_machine_status): Remove reference to force_short_suffix.
9367         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
9368         (attr length): When attribute iscompact is true force to 2
9369         regardless; in the case of maybe check if we want to force the
9370         instruction to have 4 bytes length.
9371         (nopv): Change it to generate 4 byte long nop as well.
9372         (blockage): New pattern.
9373         (simple_return): Remove call to arc_pad_return.
9374         (p_return_i): Likewise.
9376 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9378         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
9380 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9382         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
9383         ARC cores.
9385 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9387         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
9388         for ARC700 and ARCv2.
9390 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
9392         PR target/86076
9393         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
9394         operands[2] instead of operands[1].
9397 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
9399         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
9400         case, check whether the outer register overlaps an unallocatable
9401         register, not just whether it fits the required class.
9403 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
9405         * poly-int.h (can_div_trunc_p): Add new overload in which all values
9406         are poly_ints.
9407         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
9408         (memrefs_conflict_p): Likewise.
9409         (init_alias_analysis): Likewise.
9410         * cfgexpand.c (expand_debug_expr): Likewise.
9411         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
9412         * cse.c (fold_rtx): Likewise.
9413         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
9414         * expr.c (emit_block_move_hints): Likewise.
9415         (clear_storage_hints, push_block, emit_push_insn): Likewise.
9416         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
9417         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
9418         (emit_group_store): Likewise.
9419         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
9420         to read the PRE/POST_MODIFY increment.
9421         * calls.c (store_one_arg): Use strip_offset.
9422         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
9423         poly_int_rtx_p.
9424         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9425         by a VEC_SELECT.
9426         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9427         (simplify_binary_operation_1): Extend CONST_INT handling to
9428         poly_int_rtx_p.
9429         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
9430         than a HOST_WIDE_INT.
9431         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
9432         poly_int64.
9433         (adjust_mems, add_stores): Update accodingly.
9434         (vt_canonicalize_addr): Track polynomial offsets.
9435         (emit_note_insn_var_location): Likewise.
9436         (vt_add_function_parameter): Likewise.
9437         (vt_initialize): Likewise.
9439 2018-06-12  Jeff Law  <law@redhat.com>
9441         * config.gcc (alpha*-*-freebsd*): Remove.
9442         * config/alpha/freebsd.h: Remove.
9444 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
9446         PR other/69968
9447         * spellcheck-tree.c (levenshtein_distance): Rename to...
9448         (get_edit_distance): ...this, and update for underlying renaming.
9449         * spellcheck-tree.h (levenshtein_distance): Rename to...
9450         (get_edit_distance): ...this.
9451         * spellcheck.c (levenshtein_distance): Rename to...
9452         (get_edit_distance): ...this.  Convert from Levenshtein distance
9453         to Damerau-Levenshtein distance by supporting transpositions of
9454         adjacent characters.  Rename "v1" to "v_next" and "v0" to
9455         "v_one_ago".
9456         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
9457         (selftest::test_edit_distance_unit_test_oneway): ...this, and
9458         update for underlying renaming.
9459         (selftest::levenshtein_distance_unit_test): Rename to...
9460         (selftest::test_get_edit_distance_unit): ...this, and update for
9461         underlying renaming.
9462         (selftest::test_find_closest_string): Add example from PR 69968
9463         where transposition helps
9464         (selftest::test_metric_conditions): Update for renaming.
9465         (selftest::test_metric_conditions): Likewise.
9466         (selftest::spellcheck_c_tests): Likewise.
9467         * spellcheck.h (levenshtein_distance): Rename both overloads to...
9468         (get_edit_distance): ...this.
9469         (best_match::consider): Update for renaming.
9471 2018-06-12  Martin Sebor  <msebor@redhat.com>
9473         PR tree-optimization/85259
9474         * builtins.c (compute_objsize): Handle constant offsets.
9475         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
9476         true iff a warning has been issued.
9477         * gimple.h (gimple_nonartificial_location): New function.
9478         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
9479         gimple_nonartificial_location and handle -Wno-system-headers.
9480         (handle_builtin_stxncpy): Same.
9482 2018-06-12  Martin Sebor  <msebor@redhat.com>
9484         PR c/85931
9485         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
9487 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
9489         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
9490         BUILTIN_VEC_XST entries for pointer to double and long long.
9492 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
9494         PR target/85990
9495         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9496         Update comments.
9497         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9498         Likewise.
9500 2018-06-12  Martin Liska  <mliska@suse.cz>
9502         * doc/options.texi: Document IntegerRange.
9504 2018-06-12  Martin Liska  <mliska@suse.cz>
9506         * config/i386/i386.opt: Make MPX-related options as Deprecated.
9507         * opt-functions.awk: Handle Deprecated flag.
9508         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
9509         and report error.
9510         (read_cmdline_option): Report warning for a deprecated option.
9511         * opts.h (struct cl_option): Add new field cl_deprecated.
9512         (CL_ERR_DEPRECATED): New.
9514 2018-06-12  Martin Liska  <mliska@suse.cz>
9516         * doc/options.texi: Document Deprecated option flag.
9518 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9520         * config/arc/arc-arch.h (arc_extras): New enum.
9521         (arc_cpu_t):Add field extra.
9522         (arc_cpu_types): Consider the extras.
9523         * config/arc/arc-cpus.def: Add extras info.
9524         * config/arc/arc-opts.h (processor_type): Consider extra field.
9525         * config/arc/arc.c (arc_override_options): Handle extra field.
9527 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
9529         * config/arc/arc-arch.h: Update ARC_OPTX macro.
9530         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
9531         field.
9532         * config/arc/arc.c (arc_init): Update pic warning.
9533         (irq_range): Update irq range parsing warnings.
9534         (arc_override_options): Update various warning messages.
9535         (arc_handle_aux_attribute): Likewise.
9537 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
9539         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
9541 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9543         * doc/sourcebuild.texi: Document usage of line number 0 in verify
9544         compiler messages directives.
9546 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
9548         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
9549         * config/mips/mips-tables.opt: Regenerate.
9550         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
9551         mips64r6.
9552         * doc/invoke.texi: Document -march=i6500.
9554 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
9556         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
9557         (i6400_gpmul): Add cpu_unit.
9558         (i6400_gpdiv): Likewise.
9559         (i6400_msa_add_d): Update reservations.
9560         (i6400_msa_int_add) Likewise.
9561         (i6400_msa_short_logic3) Likewise.
9562         (i6400_msa_short_logic2) Likewise.
9563         (i6400_msa_short_logic) Likewise.
9564         (i6400_msa_move) Likewise.
9565         (i6400_msa_cmp) Likewise.
9566         (i6400_msa_short_float2) Likewise.
9567         (i6400_msa_div_d) Likewise.
9568         (i6400_msa_long_logic1) Likewise.
9569         (i6400_msa_long_logic2) Likewise.
9570         (i6400_msa_mult) Likewise.
9571         (i6400_msa_long_float2) Likewise.
9572         (i6400_msa_long_float4) Likewise.
9573         (i6400_msa_long_float5) Likewise.
9574         (i6400_msa_long_float8) Likewise.
9575         (i6400_fpu_fadd): Include frint type.
9576         (i6400_fpu_store): New define_insn_reservation.
9577         (i6400_fpu_load): Likewise.
9578         (i6400_fpu_move): Likewise.
9579         (i6400_fpu_fcmp): Likewise.
9580         (i6400_fpu_fmadd): Likewise.
9581         (i6400_int_mult): Include imul3nc type and update reservation.
9582         (i6400_int_div): Include idiv3 type and update reservation.
9583         (i6400_int_load): Update to check type not move_type.
9584         (i6400_int_store): Likewise.
9585         (i6400_int_prefetch): Set zero latency.
9587 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
9589         * gcc.c: Document new %@{...} sequence.
9590         (LINK_COMMAND_SPEC): Use it for the -L switches.
9591         (cpp_unique_options): Use it for the -I switches.
9592         (at_file_argbuf): New global variable.
9593         (in_at_file): Likewise.
9594         (alloc_args): Create at_file_argbuf.
9595         (clear_args): Truncate at_file_argbuf.
9596         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
9597         (open_at_file): New function.
9598         (close_at_file): Likewise.
9599         (create_at_file): Delete.
9600         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
9601         <'o'>: Likewise.
9602         <'@'>: New case.
9603         (validate_switches_from_spec): Deal with %@{...} sequence.
9604         (validate_switches): Likewise.
9605         (driver::finalize): Call clear_args.
9607 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
9609         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
9611 2018-06-11  Martin Sebor  <msebor@redhat.com>
9613         * doc/invoke.texi (-Wall): List -Wc++17-compat.
9614         (Wno-class-memaccess): Add @opindex.
9615         (Wno-templates, Wno-multiple-inheritance): Same.
9616         (Wno-virtual-inheritance, Wno-namespaces): Same.
9617         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
9618         (Wno-format-overflow, Wno-format-truncation): Same.
9619         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
9620         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
9621         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
9622         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
9623         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
9624         (Wno-misspelled-isr): Same.
9626 2018-06-11  Martin Sebor  <msebor@redhat.com>
9628         * PR tree-optimization/86083
9629         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
9631 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
9633         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
9635 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
9637         PR target/85755
9638         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
9639         on the correct operand.
9640         (*movdi_internal64): Ditto.
9642 2018-06-11  Martin Liska  <mliska@suse.cz>
9644         PR tree-optimization/86089
9645         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
9647 2018-06-11  Julia Koval  <julia.koval@intel.com>
9649         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
9650         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
9651         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
9653 2018-06-11  Olivier Hainque  <hainque@adacore.com>
9655         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
9656         for Ada with strict dwarf2.
9658 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
9660         PR target/85755
9661         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
9662         addresses.
9664 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
9666         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
9668 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
9670         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
9671         TARGET_ELF.
9673 2018-06-08  Martin Liska  <mliska@suse.cz>
9675         * tree-cfg.h (debug_function): Fix argument type to match
9676         implementation.
9678 2018-06-08  Martin Liska  <mliska@suse.cz>
9680         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
9681         Remove usage of MPX-related (and removed) fields.
9682         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
9684 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
9686         * cfg.c (debug): Use TDF_NONE rather than 0.
9687         * cfghooks.c (debug): Likewise.
9688         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
9689         (struct dump_option_value_info): Convert to...
9690         (struct kv_pair): ...this template type.
9691         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
9692         rather than 0.
9693         (optinfo_verbosity_options): Likewise.
9694         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
9695         OPTGROUP_NONE.
9696         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
9697         than int for "optgroup_flags" param.
9698         (dump_generic_expr_loc): Use dump_flags_t rather than int for
9699         "dump_kind" param.
9700         (dump_dec): Likewise.
9701         (dump_finish): Use TDF_NONE rather than 0.
9702         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
9703         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
9704         than 0.  Update for change to option_ptr.
9705         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
9706         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
9707         0.  Update for changes to optinfo_verbosity_options and
9708         optgroup_options.
9709         (opt_info_switch_p): Convert optgroup_flags from int to
9710         optgroup_flags_t.
9711         (dump_basic_block): Use dump_flags_t rather than int
9712         for "dump_kind" param.
9713         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
9714         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
9715         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
9716         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
9717         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
9718         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
9719         TDF_NONE): Convert from macros to...
9720         (enum dump_flag): ...this new enum.
9721         (dump_flags_t): Update to use enum.
9722         (operator|, operator&, operator~, operator|=, operator&=):
9723         Implement for dump_flags_t.
9724         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
9725         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
9726         Convert from macros to...
9727         (enum optgroup_flag): ...this new enum.
9728         (optgroup_flags_t): New typedef.
9729         (operator|, operator|=): Implement for optgroup_flags_t.
9730         (struct dump_file_info): Convert field "alt_flags" to
9731         dump_flags_t.  Convert field "optgroup_flags" to
9732         optgroup_flags_t.
9733         (dump_basic_block): Use dump_flags_t rather than int for param.
9734         (dump_generic_expr_loc): Likewise.
9735         (dump_dec): Likewise.
9736         (dump_register): Convert param "optgroup_flags" to
9737         optgroup_flags_t.
9738         (opt_info_enable_passes): Likewise.
9739         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
9740         than 0.
9741         * gimple-pretty-print.c (debug): Likewise.
9742         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
9743         (merged_store_group::apply_stores): Likewise.
9744         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
9745         * gimple.c (verify_gimple_pp): Likewise.
9746         * graphite-poly.c (print_pbb_body): Likewise.
9747         * passes.c (pass_manager::register_one_dump_file): Convert
9748         local "optgroup_flags" to optgroup_flags_t.
9749         * print-tree.c (print_node): Use TDF_NONE rather than 0.
9750         (debug): Likewise.
9751         (debug_body): Likewise.
9752         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
9753         to optgroup_flags_t.
9754         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
9755         than 0.
9756         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
9757         (convert_mult_to_fma): Likewise.
9758         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
9759         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
9760         * tree-vect-data-refs.c (dump_lower_bound): Convert param
9761         "dump_kind" to dump_flags_t.
9763 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
9765         * config/rs6000/rs6000.c (min, max): Delete.
9767 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
9769         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
9770         -mabi=spe and -mabi=no-spe.
9772 2018-06-08  Martin Liska  <mliska@suse.cz>
9774         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
9775         where we expect an existing summary.
9777 2018-06-08  Martin Liska  <mliska@suse.cz>
9779         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
9780         * ipa-inline.h (estimate_edge_growth): Likewise.
9782 2018-06-08  Martin Liska  <mliska@suse.cz>
9784         * cgraph.c (function_version_hasher::hash): Use
9785         cgraph_node::get_uid ().
9786         (function_version_hasher::equal):
9787         * cgraph.h (cgraph_node::get_uid): New method.
9788         * ipa-inline.c (update_caller_keys): Use
9789         cgraph_node::get_uid ().
9790         (update_callee_keys): Likewise.
9791         * ipa-utils.c (searchc): Likewise.
9792         (ipa_reduced_postorder): Likewise.
9793         * lto-cgraph.c (input_node): Likewise.
9794         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
9795         * symbol-summary.h (symtab_insertion): Likewise.
9796         (symtab_removal): Likewise.
9797         (symtab_duplication): Likewise.
9798         * tree-pretty-print.c (dump_function_header): Likewise.
9799         * tree-sra.c (convert_callers_for_node): Likewise.
9801 2018-06-08  Martin Liska  <mliska@suse.cz>
9803         * cgraph.c (symbol_table::create_edge): Always assign a new
9804         unique number.
9805         (symbol_table::free_edge): Do not recycle numbers.
9806         * cgraph.h (cgraph_edge::get): New method.
9807         * symbol-summary.h (symtab_removal): Use it.
9808         (symtab_duplication): Likewise.
9809         (call_summary::hashable_uid): Remove.
9811 2018-06-08  Martin Liska  <mliska@suse.cz>
9813         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
9814         (initialize_growth_caches): Remove.
9815         (free_growth_caches): Likewise.
9816         (do_estimate_edge_time): Use edge_growth_cache.
9817         (do_estimate_edge_size): Likewise.
9818         (do_estimate_edge_hints): Likewise.
9819         * ipa-inline.c (reset_edge_caches): Likewise.
9820         (recursive_inlining): Likewise.
9821         (inline_small_functions): Likewise.
9822         * ipa-inline.h (initialize_growth_caches): Remove.
9823         (estimate_edge_size): Likewise.
9824         (estimate_edge_time): Likewise.
9825         (estimate_edge_hints): Likewise.
9826         (reset_edge_growth_cache): Likewise.
9827         * symbol-summary.h (call_summary::remove): New method.
9829 2018-06-08  Martin Liska  <mliska@suse.cz>
9831         * ipa-cp.c (class edge_clone_summary): New summary.
9832         (grow_edge_clone_vectors): Remove.
9833         (ipcp_edge_duplication_hook): Remove.
9834         (class edge_clone_summary_t): New call_summary class.
9835         (ipcp_edge_removal_hook): Remove.
9836         (edge_clone_summary_t::duplicate): New function.
9837         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
9838         (create_specialized_node): Likewise.
9839         (ipcp_driver): Initialize edge_clone_summaries and do not
9840         register hooks.
9842 2018-06-08  Martin Liska  <mliska@suse.cz>
9844         * symbol-summary.h (get): New function.
9845         (call_summary::m_initialize_when_cloning): New class member.
9847 2018-06-08  Martin Liska  <mliska@suse.cz>
9849         * cgraph.c (cgraph_node::remove): Do not recycle uid.
9850         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
9851         (symbol_table::allocate_cgraph_symbol): Do not set uid.
9852         * passes.c (uid_hash_t): Record removed_nodes by their uids.
9853         (remove_cgraph_node_from_order): Use the removed_nodes set.
9854         (do_per_function_toporder): Likwise.
9855         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
9856         instead of summary_uid.
9857         (symtab_removal): Likewise.
9858         (symtab_duplication): Likewise.
9860 2018-06-08  Martin Liska  <mliska@suse.cz>
9862         * ipa-cp.c (ipcp_store_bits_results): Use
9863         ipcp_transformation_sum.
9864         (ipcp_store_vr_results): Likewise.
9865         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
9866         to ...
9867         (ipcp_transformation_initialize): ... this.
9868         (ipa_set_node_agg_value_chain):
9869         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
9870         (write_ipcp_transformation_info): Likewise.
9871         (read_ipcp_transformation_info): Likewise.
9872         (ipcp_update_bits): Likewise.
9873         (ipcp_update_vr): Likewise.
9874         (ipcp_transform_function): Likewise.
9875         * ipa-prop.h: Rename ipcp_transformation_summary to
9876         ipcp_transformation.
9877         (class ipcp_transformation_t): New function summary.
9878         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
9879         (ipa_get_agg_replacements_for_node): Likewise.
9881 2018-06-08  Martin Liska  <mliska@suse.cz>
9883         * ipa-pure-const.c (struct funct_state_d): Do it class instead
9884         of struct.
9885         (class funct_state_summary_t): New function_summary class.
9886         (has_function_state): Remove.
9887         (get_function_state): Likewise.
9888         (set_function_state): Likewise.
9889         (add_new_function): Likewise.
9890         (funct_state_summary_t::insert): New function.
9891         (duplicate_node_data): Remove.
9892         (remove_node_data): Remove.
9893         (funct_state_summary_t::duplicate): New function.
9894         (register_hooks): Create new funct_state_summaries.
9895         (pure_const_generate_summary): Use it.
9896         (pure_const_write_summary): Likewise.
9897         (pure_const_read_summary): Likewise.
9898         (propagate_pure_const): Likewise.
9899         (propagate_nothrow): Likewise.
9900         (dump_malloc_lattice): Likewise.
9901         (propagate_malloc): Likewise.
9902         (execute): Do not register hooks, just remove summary
9903         instead.
9904         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
9905         constructor.
9907 2018-06-08  Martin Liska  <mliska@suse.cz>
9909         * ipa-reference.c (remove_node_data): Remove.
9910         (duplicate_node_data): Likewise.
9911         (class ipa_ref_var_info_summary_t): New class.
9912         (class ipa_ref_opt_summary_t): Likewise.
9913         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
9914         (get_reference_optimization_summary): Use
9915         ipa_ref_opt_sum_summaries.
9916         (set_reference_vars_info): Remove.
9917         (set_reference_optimization_summary): Likewise.
9918         (ipa_init): Create summaries.
9919         (init_function_info): Use function summary.
9920         (ipa_ref_opt_summary_t::duplicate): New function.
9921         (ipa_ref_opt_summary_t::remove): New function.
9922         (get_read_write_all_from_node): Fix GNU coding style.
9923         (propagate): Use function summary.
9924         (write_node_summary_p): Fix GNU coding style.
9925         (stream_out_bitmap): Likewise.
9926         (ipa_reference_read_optimization_summary): Use function summary.
9927         (ipa_reference_c_finalize): Do not release hooks.
9929 2018-06-08  Martin Liska  <mliska@suse.cz>
9931         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
9932         (analyze_function_body): Extract multiple calls of get_create.
9933         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9934         * ipa-inline.c (recursive_inlining): Use ::get method.
9935         * ipa-inline.h (estimate_edge_growth): Likewise.
9937 2018-06-08  Martin Liska  <mliska@suse.cz>
9939         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
9940         HSA_INVALID.
9941         (hsa_function_summary::hsa_function_summary): Use the new enum
9942         value.
9943         (hsa_gpu_implementation_p): Use hsa_summaries::get.
9944         * hsa-gen.c (hsa_get_host_function): Likewise.
9945         (get_brig_function_name): Likewise.
9946         * ipa-hsa.c (process_hsa_functions): Likewise.
9947         (ipa_hsa_write_summary): Likewise.
9948         * symbol-summary.h (symtab_duplication): Use ::get function/
9949         (get): New function.
9951 2018-06-08  Martin Liska  <mliska@suse.cz>
9953         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
9954         of get.
9955         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
9956         (hsa_register_kernel): Likewise.
9957         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
9958         * hsa-gen.c (hsa_get_host_function): Likewise.
9959         (get_brig_function_name): Likewise.
9960         (generate_hsa): Likewise.
9961         (pass_gen_hsail::execute): Likewise.
9962         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
9963         (devirtualization_time_bonus): Likewise.
9964         (ipcp_propagate_stage): Likewise.
9965         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
9966         (edge_set_predicate): Likewise.
9967         (evaluate_conditions_for_known_args): Likewise.
9968         (evaluate_properties_for_edge): Likewise.
9969         (ipa_fn_summary::reset): Likewise.
9970         (ipa_fn_summary_t::duplicate): Likewise.
9971         (dump_ipa_call_summary): Likewise.
9972         (ipa_dump_fn_summary): Likewise.
9973         (analyze_function_body): Likewise.
9974         (compute_fn_summary): Likewise.
9975         (estimate_edge_devirt_benefit): Likewise.
9976         (estimate_edge_size_and_time): Likewise.
9977         (estimate_calls_size_and_time): Likewise.
9978         (estimate_node_size_and_time): Likewise.
9979         (inline_update_callee_summaries): Likewise.
9980         (remap_edge_change_prob): Likewise.
9981         (remap_edge_summaries): Likewise.
9982         (ipa_merge_fn_summary_after_inlining): Likewise.
9983         (ipa_update_overall_fn_summary): Likewise.
9984         (read_ipa_call_summary): Likewise.
9985         (inline_read_section): Likewise.
9986         (write_ipa_call_summary): Likewise.
9987         (ipa_fn_summary_write): Likewise.
9988         (ipa_free_fn_summary): Likewise.
9989         * ipa-hsa.c (process_hsa_functions): Likewise.
9990         (ipa_hsa_write_summary): Likewise.
9991         (ipa_hsa_read_section): Likewise.
9992         * ipa-icf.c (sem_function::merge): Likewise.
9993         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
9994         (do_estimate_edge_time): Likewise.
9995         (estimate_size_after_inlining): Likewise.
9996         (estimate_growth): Likewise.
9997         (growth_likely_positive): Likewise.
9998         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
9999         (inline_call): Likewise.
10000         * ipa-inline.c (caller_growth_limits): Likewise.
10001         (can_inline_edge_p): Likewise.
10002         (can_inline_edge_by_limits_p): Likewise.
10003         (compute_uninlined_call_time): Likewise.
10004         (compute_inlined_call_time): Likewise.
10005         (want_inline_small_function_p): Likewise.
10006         (edge_badness): Likewise.
10007         (update_caller_keys): Likewise.
10008         (update_callee_keys): Likewise.
10009         (recursive_inlining): Likewise.
10010         (inline_small_functions): Likewise.
10011         (inline_to_all_callers_1): Likewise.
10012         (dump_overall_stats): Likewise.
10013         (early_inline_small_functions): Likewise.
10014         (early_inliner): Likewise.
10015         * ipa-inline.h (estimate_edge_growth): Likewise.
10016         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10017         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10018         * ipa-prop.h (IPA_NODE_REF): Likewise.
10019         (IPA_EDGE_REF): Likewise.
10020         * ipa-pure-const.c (malloc_candidate_p): Likewise.
10021         (propagate_malloc): Likewise.
10022         * ipa-split.c (execute_split_functions): Likewise.
10023         * symbol-summary.h: Rename get to get_create.
10024         (get): Likewise.
10025         (get_create): Likewise.
10026         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10028 2018-06-08  Martin Liska  <mliska@suse.cz>
10030         * symbol-summary.h (release): Move definition out of class
10031         declaration.
10032         (symtab_removal): Likewise.
10033         (symtab_duplication): Likewise.
10035 2018-06-08  Martin Liska  <mliska@suse.cz>
10037         * symbol-summary.h (function_summary): Move constructor
10038         implementation out of class declaration.
10039         (release): Likewise.
10040         (symtab_insertion): Likewise.
10041         (symtab_removal): Likewise.
10042         (symtab_duplication): Likewise.
10043         (get): Likewise.
10045 2018-06-08  Martin Liska  <mliska@suse.cz>
10047         * Makefile.in: Remove support for MPX (macros, related functions,
10048         fields in cgraph_node, ...).
10049         * builtin-types.def (BT_BND): Likewise.
10050         (BT_FN_BND_CONST_PTR): Likewise.
10051         (BT_FN_CONST_PTR_BND): Likewise.
10052         (BT_FN_VOID_PTR_BND): Likewise.
10053         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
10054         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
10055         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
10056         (expand_builtin_mempcpy_with_bounds): Likewise.
10057         (expand_builtin_memset_with_bounds): Likewise.
10058         (expand_builtin_memset_args): Likewise.
10059         (std_expand_builtin_va_start): Likewise.
10060         (expand_builtin): Likewise.
10061         (expand_builtin_with_bounds): Likewise.
10062         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
10063         (DEF_LIB_BUILTIN_CHKP): Likewise.
10064         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
10065         (DEF_CHKP_BUILTIN): Likewise.
10066         (BUILT_IN_MEMCPY): Likewise.
10067         (BUILT_IN_MEMMOVE): Likewise.
10068         (BUILT_IN_MEMPCPY): Likewise.
10069         (BUILT_IN_MEMSET): Likewise.
10070         (BUILT_IN_STPCPY): Likewise.
10071         (BUILT_IN_STRCAT): Likewise.
10072         (BUILT_IN_STRCHR): Likewise.
10073         (BUILT_IN_STRCPY): Likewise.
10074         (BUILT_IN_STRLEN): Likewise.
10075         (BUILT_IN_MEMCPY_CHK): Likewise.
10076         (BUILT_IN_MEMMOVE_CHK): Likewise.
10077         (BUILT_IN_MEMPCPY_CHK): Likewise.
10078         (BUILT_IN_MEMSET_CHK): Likewise.
10079         (BUILT_IN_STPCPY_CHK): Likewise.
10080         (BUILT_IN_STRCAT_CHK): Likewise.
10081         (BUILT_IN_STRCPY_CHK): Likewise.
10082         * calls.c (store_bounds): Likewise.
10083         (emit_call_1): Likewise.
10084         (special_function_p): Likewise.
10085         (maybe_warn_nonstring_arg): Likewise.
10086         (initialize_argument_information): Likewise.
10087         (finalize_must_preallocate): Likewise.
10088         (compute_argument_addresses): Likewise.
10089         (expand_call): Likewise.
10090         * cfgexpand.c (expand_call_stmt): Likewise.
10091         (expand_return): Likewise.
10092         (expand_gimple_stmt_1): Likewise.
10093         (pass_expand::execute): Likewise.
10094         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
10095         (cgraph_node::remove): Likewise.
10096         (cgraph_node::dump): Likewise.
10097         (cgraph_node::verify_node): Likewise.
10098         * cgraph.h (chkp_function_instrumented_p): Likewise.
10099         (symtab_node::get_alias_target): Likewise.
10100         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
10101         (cgraph_local_p): Likewise.
10102         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
10103         (cgraph_edge::rebuild_references): Likewise.
10104         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
10105         (walk_polymorphic_call_targets): Likewise.
10106         (cgraph_node::expand_thunk): Likewise.
10107         (symbol_table::output_weakrefs): Likewise.
10108         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
10109         (ix86_handle_option): Likewise.
10110         * config/i386/constraints.md: Likewise.
10111         * config/i386/i386-builtin-types.def (BND): Likewise.
10112         (VOID): Likewise.
10113         (PVOID): Likewise.
10114         (ULONG): Likewise.
10115         * config/i386/i386-builtin.def (BDESC_END): Likewise.
10116         (BDESC_FIRST): Likewise.
10117         (BDESC): Likewise.
10118         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
10119         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
10120         * config/i386/i386.c (enum reg_class): Likewise.
10121         (ix86_target_string): Likewise.
10122         (ix86_option_override_internal): Likewise.
10123         (ix86_conditional_register_usage): Likewise.
10124         (ix86_valid_target_attribute_inner_p): Likewise.
10125         (ix86_set_indirect_branch_type): Likewise.
10126         (ix86_set_current_function): Likewise.
10127         (ix86_function_arg_regno_p): Likewise.
10128         (init_cumulative_args): Likewise.
10129         (ix86_function_arg_advance): Likewise.
10130         (ix86_function_arg): Likewise.
10131         (ix86_pass_by_reference): Likewise.
10132         (ix86_function_value_regno_p): Likewise.
10133         (ix86_function_value_1): Likewise.
10134         (ix86_function_value_bounds): Likewise.
10135         (ix86_return_in_memory): Likewise.
10136         (ix86_setup_incoming_vararg_bounds): Likewise.
10137         (ix86_va_start): Likewise.
10138         (indirect_thunk_need_prefix): Likewise.
10139         (print_reg): Likewise.
10140         (ix86_print_operand): Likewise.
10141         (ix86_expand_call): Likewise.
10142         (ix86_output_function_return): Likewise.
10143         (reg_encoded_number): Likewise.
10144         (BDESC_VERIFYS): Likewise.
10145         (ix86_init_mpx_builtins): Likewise.
10146         (ix86_init_builtins): Likewise.
10147         (ix86_emit_cmove): Likewise.
10148         (ix86_emit_move_max): Likewise.
10149         (ix86_expand_builtin): Likewise.
10150         (ix86_builtin_mpx_function): Likewise.
10151         (ix86_get_arg_address_for_bt): Likewise.
10152         (ix86_load_bounds): Likewise.
10153         (ix86_store_bounds): Likewise.
10154         (ix86_load_returned_bounds): Likewise.
10155         (ix86_store_returned_bounds): Likewise.
10156         (ix86_class_likely_spilled_p): Likewise.
10157         (ix86_hard_regno_mode_ok): Likewise.
10158         (x86_order_regs_for_local_alloc): Likewise.
10159         (ix86_mitigate_rop): Likewise.
10160         (ix86_bnd_prefixed_insn_p): Likewise.
10161         (ix86_mpx_bound_mode): Likewise.
10162         (ix86_make_bounds_constant): Likewise.
10163         (ix86_initialize_bounds): Likewise.
10164         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
10165         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
10166         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
10167         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
10168         (TARGET_CHKP_BOUND_MODE): Likewise.
10169         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
10170         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
10171         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
10172         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
10173         * config/i386/i386.h (TARGET_MPX): Likewise.
10174         (TARGET_MPX_P): Likewise.
10175         (VALID_BND_REG_MODE): Likewise.
10176         (FIRST_BND_REG): Likewise.
10177         (LAST_BND_REG): Likewise.
10178         (enum reg_class): Likewise.
10179         (BND_REG_P): Likewise.
10180         (BND_REGNO_P): Likewise.
10181         (BNDmode): Likewise.
10182         (ADJUST_INSN_LENGTH): Likewise.
10183         * config/i386/i386.md: Likewise.
10184         * config/i386/i386.opt: Likewise.
10185         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
10186         (defined): Likewise.
10187         (LINK_MPX): Likewise.
10188         (MPX_SPEC): Likewise.
10189         (LIBMPX_SPEC): Likewise.
10190         (LIBMPXWRAPPERS_SPEC): Likewise.
10191         (CHKP_SPEC): Likewise.
10192         * config/i386/predicates.md: Likewise.
10193         * dbxout.c (dbxout_type): Likewise.
10194         * doc/extend.texi: Likewise.
10195         * doc/invoke.texi: Likewise.
10196         * doc/md.texi: Likewise.
10197         * doc/tm.texi: Likewise.
10198         * doc/tm.texi.in: Likewise.
10199         * dwarf2out.c (is_base_type): Likewise.
10200         (gen_formal_types_die): Likewise.
10201         (gen_subprogram_die): Likewise.
10202         (gen_type_die_with_usage): Likewise.
10203         (gen_decl_die): Likewise.
10204         (dwarf2out_late_global_decl): Likewise.
10205         * expr.c (expand_assignment): Likewise.
10206         (emit_storent_insn): Likewise.
10207         (store_expr_with_bounds): Likewise.
10208         (store_expr): Likewise.
10209         (expand_expr_real_1): Likewise.
10210         * expr.h (store_expr_with_bounds): Likewise.
10211         * function.c (use_register_for_decl): Likewise.
10212         (struct bounds_parm_data): Likewise.
10213         (assign_parms_augmented_arg_list): Likewise.
10214         (assign_parm_find_entry_rtl): Likewise.
10215         (assign_parm_is_stack_parm): Likewise.
10216         (assign_parm_load_bounds): Likewise.
10217         (assign_bounds): Likewise.
10218         (assign_parms): Likewise.
10219         (expand_function_start): Likewise.
10220         * gcc.c (CHKP_SPEC): Likewise.
10221         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
10222         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
10223         (wrestrict_dom_walker::check_call): Likewise.
10224         * gimple.c (gimple_build_call_from_tree): Likewise.
10225         * gimple.h (enum gf_mask): Likewise.
10226         (gimple_call_with_bounds_p): Likewise.
10227         (gimple_call_set_with_bounds): Likewise.
10228         * gimplify.c (gimplify_init_constructor): Likewise.
10229         * ipa-cp.c (initialize_node_lattices): Likewise.
10230         (propagate_constants_across_call): Likewise.
10231         (find_more_scalar_values_for_callers_subset): Likewise.
10232         * ipa-hsa.c (process_hsa_functions): Likewise.
10233         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
10234         * ipa-icf.c (sem_function::merge): Likewise.
10235         * ipa-inline.c (early_inliner): Likewise.
10236         * ipa-pure-const.c (warn_function_noreturn): Likewise.
10237         (warn_function_cold): Likewise.
10238         (propagate_pure_const): Likewise.
10239         * ipa-ref.h (enum GTY): Likewise.
10240         * ipa-split.c (find_retbnd): Likewise.
10241         (consider_split): Likewise.
10242         (split_function): Likewise.
10243         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
10244         * ipa.c (walk_polymorphic_call_targets): Likewise.
10245         (symbol_table::remove_unreachable_nodes): Likewise.
10246         (process_references): Likewise.
10247         (cgraph_build_static_cdtor_1): Likewise.
10248         * lto-cgraph.c (lto_output_node): Likewise.
10249         (output_refs): Likewise.
10250         (compute_ltrans_boundary): Likewise.
10251         (input_overwrite_node): Likewise.
10252         (input_node): Likewise.
10253         (input_cgraph_1): Likewise.
10254         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
10255         * passes.c (pass_manager::execute_early_local_passes): Likewise.
10256         (class pass_chkp_instrumentation_passes): Likewise.
10257         (make_pass_chkp_instrumentation_passes): Likewise.
10258         * passes.def: Likewise.
10259         * rtl.h (struct GTY): Likewise.
10260         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
10261         * stor-layout.c (layout_type): Likewise.
10262         * symtab.c: Likewise.
10263         * target.def: Likewise.
10264         * targhooks.c (default_chkp_bound_type): Likewise.
10265         (default_chkp_bound_mode): Likewise.
10266         (default_builtin_chkp_function): Likewise.
10267         (default_chkp_function_value_bounds): Likewise.
10268         (default_chkp_make_bounds_constant): Likewise.
10269         (default_chkp_initialize_bounds): Likewise.
10270         * targhooks.h (default_chkp_bound_type): Likewise.
10271         (default_chkp_bound_mode): Likewise.
10272         (default_builtin_chkp_function): Likewise.
10273         (default_chkp_function_value_bounds): Likewise.
10274         (default_chkp_make_bounds_constant): Likewise.
10275         (default_chkp_initialize_bounds): Likewise.
10276         * toplev.c (compile_file): Likewise.
10277         (process_options): Likewise.
10278         * tree-core.h (DEF_BUILTIN): Likewise.
10279         (DEF_BUILTIN_CHKP): Likewise.
10280         * tree-inline.c (declare_return_variable): Likewise.
10281         (remap_gimple_stmt): Likewise.
10282         (copy_bb): Likewise.
10283         (initialize_inlined_parameters): Likewise.
10284         (expand_call_inline): Likewise.
10285         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
10286         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
10287         (make_pass_ipa_chkp_produce_thunks): Likewise.
10288         (make_pass_chkp): Likewise.
10289         (make_pass_chkp_opt): Likewise.
10290         (make_pass_chkp_instrumentation_passes): Likewise.
10291         * tree-pretty-print.c (dump_generic_node): Likewise.
10292         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
10293         * tree-ssa-dce.c (propagate_necessity): Likewise.
10294         (eliminate_unnecessary_stmts): Likewise.
10295         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
10296         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
10297         * tree-ssa-sccvn.h: Likewise.
10298         * tree-ssa-strlen.c (get_string_length): Likewise.
10299         (valid_builtin_call): Likewise.
10300         (adjust_last_stmt): Likewise.
10301         (handle_builtin_strchr): Likewise.
10302         (handle_builtin_strcpy): Likewise.
10303         (handle_builtin_stxncpy): Likewise.
10304         (handle_builtin_memcpy): Likewise.
10305         (handle_builtin_strcat): Likewise.
10306         (strlen_check_and_optimize_stmt): Likewise.
10307         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
10308         * tree-streamer-in.c: Likewise.
10309         * tree-streamer.c (record_common_node): Likewise.
10310         * tree.c (tree_code_size): Likewise.
10311         (wide_int_to_tree_1): Likewise.
10312         (type_contains_placeholder_1): Likewise.
10313         (build_common_tree_nodes): Likewise.
10314         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
10315         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
10316         (POINTER_BOUNDS_P): Likewise.
10317         (BOUNDED_TYPE_P): Likewise.
10318         (BOUNDED_P): Likewise.
10319         (CALL_WITH_BOUNDS_P): Likewise.
10320         (pointer_bounds_type_node): Likewise.
10321         * value-prof.c (gimple_ic): Likewise.
10322         * var-tracking.c (vt_add_function_parameters): Likewise.
10323         * varasm.c (make_decl_rtl): Likewise.
10324         (assemble_start_function): Likewise.
10325         (output_constant): Likewise.
10326         (maybe_assemble_visibility): Likewise.
10327         * varpool.c (ctor_for_folding): Likewise.
10328         * chkp-builtins.def: Remove.
10329         * ipa-chkp.c: Remove.
10330         * ipa-chkp.h: Remove.
10331         * rtl-chkp.c: Remove.
10332         * rtl-chkp.h: Remove.
10333         * tree-chkp-opt.c: Remove.
10334         * tree-chkp.c: Remove.
10335         * tree-chkp.h: Remove.
10337 2018-06-07  Carl Love  <cel@us.ibm.com>
10339         * config/rs6000/vsx.md (vextract_fp_from_shorth,
10340         vextract_fp_from_shortl): Add BE support.
10342 2018-06-07  Paul Koning  <ni1d@arrl.net>
10344         * compare-elim.c (try_merge_compare): Don't merge compare if
10345         address contains a side effect.
10346         (try_eliminate_compare): Likewise.
10348 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
10350         * config.gcc: Support "tremont".
10351         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
10352         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
10353         PROCESSOR_TREMONT.
10354         * config/i386/i386.c (m_TREMONT): Define.
10355         (processor_target_table): Add "tremont".
10356         (PTA_TREMONT): Define.
10357         (ix86_lea_outperforms): Add TARGET_TREMONT.
10358         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
10359         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
10360         and M_INTEL_GOLDMONT_PLUS.
10361         (fold_builtin_cpu): Add "tremont".
10362         (ix86_add_stmt_cost): Add TARGET_TREMONT.
10363         (ix86_option_override_internal): Add "tremont".
10364         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
10365         (processor_type): Add PROCESSOR_TREMONT.
10366         * config/i386/x86-tune.def: Add m_TREMONT.
10367         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
10369 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10371         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
10372         symbol defined for msp430i* devices to be lower case.
10374 2018-06-07  Richard Biener  <rguenther@suse.de>
10376         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
10377         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
10378         Properly wrap signed arithmetic if overflow wraps.
10380 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
10382         PR tree-optimization/69615
10383         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
10384         of a cast from a same precision integral SSA_NAME in a bb dominated
10385         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
10386         cast to utype if rhs2 has already a compatible type.
10388 2018-06-07  Richard Biener  <rguenther@suse.de>
10390         PR tree-optimization/85935
10391         * graphite-scop-detection.c (find_params_in_bb): Analyze
10392         condition operands with respect to the correct loop.  Assert
10393         the analysis doesn't fail.
10395 2018-06-04  Carl Love  <cel@us.ibm.com>
10397         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
10398         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
10399         as it is slightly cheaper.
10400         (first_match_or_eos_index_<mode>):
10401         Calculate index using natural element order.
10402         (first_match_index_<mode>):
10403         Calculate index using natural element order.
10404         (first_match_or_eos_index_<mode>):
10405         Calculate index using natural order.
10406         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
10407         for BE and LE modes.
10408         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
10409         P9V_BUILTIN_VCLZLSBB_V16QI.
10410         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
10411         specific.
10413 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10415         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
10416         indentation and line wrap for many prototypes.  Add missing
10417         @smallexample directives around block of prototypes for vec_xl and
10418         vec_xst.
10420 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
10422         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
10423         track if we pass or return IEEE 128-bit floating point.
10424         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
10425         C++ mangling that is compatible with GCC 8.1.
10426         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
10427         (init_cumulative_args): Note if we pass or return IEEE 128-bit
10428         floating point types.
10429         (rs6000_function_arg_advance_1): Likewise.
10430         (rs6000_mangle_type): Optionally generate mangled names that match
10431         what GCC 8.1 generated for IEEE 128-bit floating point types.
10432         (rs6000_globalize_decl_name): If we have an external function that
10433         passes or returns IEEE 128-bit types, generate a weak reference
10434         from the mangled name used in GCC 8.1 to the current mangled
10435         name.
10436         (rs6000_init_builtins): Make __ibm128 use the long double type if
10437         long double is IBM extended double.  Make __float128 use the long
10438         double type if long double is IEEE 128-bit.
10440         PR target/85657
10441         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
10442         macro for __ibm128 built-in functions.
10443         (PACK_IF): Add __ibm128 pack/unpack functions.
10444         (UNPACK_IF): Likewise.
10445         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
10446         enable long double built-in functions if long double is IEEE
10447         128-bit floating point.
10448         (rs6000_invalid_builtin): Update long double built-in function
10449         error message.
10450         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
10451         functions, adjust the built-in function to use the long double
10452         built-in function if __ibm128 and long double are the same type.
10453         * doc/extend.texi (PowerPC builtins): Update documention for
10454         __builtin_{,un}pack_longdouble.  Add documentation for
10455         __builtin_{,un}pack_ibm128.
10457 2018-06-06  Jim Wilson  <jimw@sifive.com>
10459         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
10460         (struct machine_function): New field interrupt_mode.
10461         (riscv_handle_type_attribute): New function.  Add forward declaration.
10462         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
10463         (riscv_expand_epilogue): Check interrupt_mode field.
10464         (riscv_set_current_function): Check interrupt attribute args and
10465         set interrupt_mode field.
10466         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
10467         (riscv_sret, riscv_uret): New.
10468         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
10469         new arguments to interrupt attribute.
10471 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
10473         PR target/63177
10474         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
10475         Don't handle -mcpu=power8 if -mpower9-vector is also used.
10477 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10479         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
10480         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
10481         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
10482         several redundant entries.
10484 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
10486         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
10487         type from "rtx" to "rtx_insn *".
10488         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
10489         for local "call_insn", removing cast.
10490         (ix86_expand_call): Likewise, introducing a "call_insn" local.
10492 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
10494         PR tree-optimization/86066
10495         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
10496         for BIT_INSERT_EXPR stores.
10498 2018-06-06  Richard Biener  <rguenther@suse.de>
10500         PR tree-optimization/86062
10501         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
10502         component refs ontop
10503         of to be offsetted base.
10505 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10507         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
10508         to be static and remove check on interrupt attribute name.
10510 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
10512         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
10513         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
10515 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
10517         PR target/79924
10518         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
10519         second argument.
10520         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
10521         Remove second argument, change how error is called.
10522         (aarch64_layout_arg): Remove second argument from
10523         aarch64_err_no_fpadvsimd call.
10524         (aarch64_init_cumulative_args): Ditto.
10525         (aarch64_gimplify_va_arg_expr): Ditto.
10526         * config/aarch64/aarch64.md (mov<mode>): Ditto.
10528 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
10530         * config/i386/i386.md (simple_return_indirect_internal): New expander.
10531         (*simple_return_indirect_internal<mode>): Rename from
10532         simple_return_indirect_internal.  Use W mode iterator.
10533         (rstorssp): New expander.
10534         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
10535         (clrssbsy): New expander.
10536         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
10538 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
10540         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
10541         __typeof__.
10542         (cmse_check_pointed_object): Likewise.
10544 2018-06-05  Martin Liska  <mliska@suse.cz>
10546         PR gcov-profile/47618
10547         * doc/invoke.texi: Document how -fprofile-dir format
10548         is extended.
10550 2018-06-05  Richard Biener  <rguenther@suse.de>
10552         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
10553         removal pretend DOM info isn't available so we do not update
10554         it and only remove edges, not dominated blocks.  Actually free
10555         DOM info in case we removed something.  Remove unreachable blocks.
10556         (mfb_keep_latches): Work with either DOM info or marked backedges.
10557         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
10558         first.  Mark backedges if DOM info isn't available.
10559         (Re-)compute DOM info after cleanup_control_flow_pre.
10561 2018-06-05  Richard Biener  <rguenther@suse.de>
10563         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
10564         (locus_discrim_hasher::hash): Adjust.
10565         (locus_discrim_hasher::equal): Likewise.
10566         (next_discriminator_for_locus): Work on line directly.
10567         (same_line_p): Pass in expanded locus1 as well.
10568         (assign_discriminators): Avoid redundant location expansions.
10570 2018-06-05  Richard Biener  <rguenther@suse.de>
10572         PR tree-optimization/86046
10573         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
10574         if required after clearing TREE_ADDRESSABLE.
10576 2018-06-05  Richard Biener  <rguenther@suse.de>
10578         PR tree-optimization/86047
10579         * tree-ssa-loop.c (for_each_index): Glob handling of all
10580         decls and constants and really handle all of them.
10582 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10584         PR target/81497
10585         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
10586         qualifier_void_pointer and qualifier_const_void_pointer.
10587         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
10588         (arm_init_builtins): Handle the above.
10589         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
10590         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
10591         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
10592         void intrinsics.
10594 2018-06-05  Martin Liska  <mliska@suse.cz>
10596         * auto-profile.c (read_autofdo_file): Do not use
10597         gcov_ctr_summary struct.
10598         (afdo_callsite_hot_enough_for_early_inline): Likewise.
10599         * coverage.c (struct counts_entry): Likewise.
10600         (read_counts_file): Read just single summary entry.
10601         (get_coverage_counts): Use gcov_summary struct.
10602         * coverage.h (get_coverage_counts): Likewise.
10603         * gcov-dump.c (dump_working_sets): Likewise.
10604         (tag_summary): Dump just single summary.
10605         * gcov-io.c (gcov_write_summary): Write just histogram
10606         summary.
10607         (gcov_read_summary): Read just single summary.
10608         (compute_working_sets): Use gcov_summary struct.
10609         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
10610         of GCOV_COUNTERS_SUMMABLE.
10611         (GCOV_COUNTERS_SUMMABLE): Remove.
10612         (GCOV_FIRST_VALUE_COUNTER): Replace with
10613         GCOV_COUNTER_V_INTERVAL.
10614         (struct gcov_ctr_summary): Remove.
10615         (struct gcov_summary): Directly use fields of former
10616         gcov_ctr_summary.
10617         (compute_working_sets): Use gcov_summary struct.
10618         * gcov.c (read_count_file): Do not use ctrs fields.
10619         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
10620         struct.
10621         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
10622         struct.
10623         * profile.c: Likewise.
10624         * profile.h: Likewise.
10626 2018-06-05  Martin Liska  <mliska@suse.cz>
10628         PR gcov-profile/84846
10629         * gcov.c (output_lines): Print working directory only
10630         in intermediate format.
10632 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
10634         * config/s390/s390-builtin-types.def: Add void function type.
10635         * config/s390/s390-builtins.def: Use the function type for the
10636         tbeginc builtin.
10638 2018-06-04  Jim Wilson  <jimw@sifive.com>
10640         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
10641         to int.
10642         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
10643         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
10644         handle EH_RETURN_DATA_REGNO registers properly.
10645         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
10646         (riscv_expand_epilogue): Update comment.  Change argument name and
10647         type.  Update code to use new name and type.  Pass new args to
10648         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
10649         EXCEPTION_RETURN.
10650         * config/riscv/riscv.md (NORMAL_RETURN): New.
10651         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
10652         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
10653         (eh_return): Call gen_eh_return_internal and emit barrier.
10654         (eh_return_internal): Call riscv_expand_epilogue.
10656 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
10658         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
10659         bit_insertion field and declare can_be_merged_into method.
10660         (merged_store_group::can_be_merged_into): New method.
10661         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
10662         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
10663         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
10665 2018-06-04  Richard Biener  <rguenther@suse.de>
10667         PR tree-optimization/85955
10668         * builtins.c (fold_builtin_sincos): Convert pointers to
10669         destination to appropriate type before dereferencing.
10671 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
10673         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
10675 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
10677         * expr.c (expand_expr_real_1): Force the operand into memory if
10678         its TYPE_MODE is BLKmode and if there is no integer mode for
10679         the number of bits being extracted.
10681 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
10683         PR target/85832
10684         PR target/86036
10685         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
10686         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
10688 2018-06-04  Richard Biener  <rguenther@suse.de>
10690         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
10691         (cleanup_tree_cfg_noloop): ... single caller.  Do
10692         start_recording_case_labels later.
10694 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
10696         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
10697         to _IMMINTRIN_H_INCLUDED.
10698         * config/i386/pconfigintrin.h: Ditto.
10699         * config/i386/waitpkgintrin.h: Ditto.
10700         * config/i386/immintrin.h: Add includes for sgxintrin.h,
10701         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
10702         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
10703         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
10704         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
10705         waitpkgintrin.h and cldemoteintrin.h.
10707 2018-06-04  Richard Biener  <rguenther@suse.de>
10709         PR tree-optimization/86038
10710         * tracer.c (find_best_successor): Check probability for
10711         being initialized, bail out if not.
10713 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
10715         PR target/86003
10716         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
10717         of bits to ignore when comparing architectures.
10719 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
10721         PR tree-optimization/69615
10722         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
10723         maximum or minimum of the type, try to merge it also as if
10724         range1 is + [-, x - 1] or + [x + 1, -].
10726         PR c++/86025
10727         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
10729 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
10731         PR tree-optimization/86034
10732         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
10733         the unsigned bitfield type in a bit insertion sequence if it does not
10734         have a larger precision than the bitfield size.
10735         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
10737 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
10739         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
10741 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
10743         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
10744         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
10745         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
10746         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
10748 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
10750         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
10751         Disable -fdelete-null-pointer-checks for ELF toolchain.
10753 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
10754             Kito Cheng  <kito.cheng@gmail.com>
10756         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
10757         (nds32le-*-*, nds32be-*-*): Integrate checking process.
10758         (nds32*-*-*): Add glibc and uclibc conditions.
10759         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
10760         (TARGET_EXCEPT_UNWIND_INFO): Define.
10761         * config/nds32/elf.h: New file.
10762         * config/nds32/linux.h: New file.
10763         * config/nds32/nds32-elf.opt: New file.
10764         * config/nds32/nds32-linux.opt: New file.
10765         * config/nds32/nds32-fp-as-gp.c
10766         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
10767         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
10768         TARGET_LINUX_ABI.
10769         (nds32_asm_file_end): Ditto.
10770         (nds32_print_operand): Ditto.
10771         (nds32_insert_attributes): Ditto.
10772         (nds32_init_libfuncs): New function.
10773         (TARGET_HAVE_TLS): Define.
10774         (TARGET_INIT_LIBFUNCS): Define.
10775         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
10776         spec content.
10777         (TARGET_ELF): Apply different mcmodel setting.
10778         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
10779         been migrated into elf.h and linux.h files.
10780         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
10781         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
10782         (mcmodel): The content has been migrated into nds32-elf.opt and
10783         nds32-linux.opt files.
10784         * config/nds32/t-elf: New file.
10785         * config/nds32/t-linux: New file.
10787 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
10788             Shiva Chen  <shiva0217@gmail.com>
10790         * config/nds32/constants.md (unspec_volatile_element): Add
10791         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
10792         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
10793         optimization.
10794         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
10795         (make_pass_nds32_fp_as_gp): Declare.
10796         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
10797         optmization pass.
10798         (nds32_asm_function_end_prologue): Remove unused asm output.
10799         (nds32_asm_function_begin_epilogue): Remove unused asm output.
10800         (nds32_asm_file_start): Output necessary fp_as_gp information.
10801         (nds32_option_override): Adjust register usage.
10802         (nds32_expand_prologue): Consider fp_as_gp situation.
10803         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
10804         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
10805         (epilogue): Ditto.
10806         (return): Ditto.
10807         (simple_return): Ditto.
10808         (omit_fp_begin): Output special directive for fp_as_gp.
10809         (omit_fp_end): Output special directive for fp_as_gp.
10810         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
10811         mforbid-fp-as-gp): New options.
10813 2018-06-01  Mark Wielaard  <mark@klomp.org>
10815         * dwarf2out.c (dwarf2out_finish): Remove generation of
10816         DW_AT_loclists_base.
10818 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
10820         * gimple-ssa-store-merging.c: Include gimple-fold.h.
10821         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
10822         (struct merged_store_group): Add bit_insertion field.
10823         (dump_char_array): Use standard hexadecimal format.
10824         (merged_store_group::merged_store_group): Set bit_insertion to false.
10825         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
10826         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
10827         also print the mask in the dump file.
10828         (pass_store_merging::gate): Minor tweak.
10829         (imm_store_chain_info::coalesce_immediate): Fix wrong association
10830         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
10831         stores with INTEGER_CST stores.
10832         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
10833         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
10834         and use it throughout.  Generate bit insertion sequences if need be.
10835         (pass_store_merging::process_store): Remove redundant condition.
10836         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
10838 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
10840         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
10841         the 128-bit floating point types.  Fix function comment.
10843 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
10845         * config/aarch64/aarch64-simd.md
10846         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
10847         mnemonics.
10848         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
10849         mnemonics.
10851 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
10853         PR tree-optimization/85989
10854         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
10855         variable.
10856         (backprop::intersect_uses): Check it when deciding whether this
10857         is a backedge reference.
10858         (backprop::process_block): Add each phi to m_visited_phis
10859         after visiting it, then clear it at the end.
10861 2018-06-01  Richard Biener  <rguenther@suse.de>
10863         * tree-vectorizer.h (vect_dr_stmt): New function.
10864         (vect_get_load_cost): Adjust.
10865         (vect_get_store_cost): Likewise.
10866         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
10867         Use vect_dr_stmt instead of DR_SMTT.
10868         (vect_record_base_alignments): Likewise.
10869         (vect_calculate_target_alignment): Likewise.
10870         (vect_compute_data_ref_alignment): Likewise and make static.
10871         (vect_update_misalignment_for_peel): Likewise.
10872         (vect_verify_datarefs_alignment): Likewise.
10873         (vector_alignment_reachable_p): Likewise.
10874         (vect_get_data_access_cost): Likewise.  Pass down
10875         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
10876         (vect_get_peeling_costs_all_drs): Likewise.
10877         (vect_peeling_hash_get_lowest_cost): Likewise.
10878         (vect_enhance_data_refs_alignment): Likewise.
10879         (vect_find_same_alignment_drs): Likewise.
10880         (vect_analyze_data_refs_alignment): Likewise.
10881         (vect_analyze_group_access_1): Likewise.
10882         (vect_analyze_group_access): Likewise.
10883         (vect_analyze_data_ref_access): Likewise.
10884         (vect_analyze_data_ref_accesses): Likewise.
10885         (vect_vfa_segment_size): Likewise.
10886         (vect_small_gap_p): Likewise.
10887         (vectorizable_with_step_bound_p): Likewise.
10888         (vect_prune_runtime_alias_test_list): Likewise.
10889         (vect_analyze_data_refs): Likewise.
10890         (vect_supportable_dr_alignment): Likewise.
10891         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10892         (vect_gen_prolog_loop_niters): Likewise.
10893         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10894         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
10895         modify DR_STMT.
10896         (vect_recog_mask_conversion_pattern): Likewise.
10897         (vect_try_gather_scatter_pattern): Likewise.
10898         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
10899         to vect_get_store_cost.
10900         (vect_get_store_cost): Get stmt_info instead of DR.
10901         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
10902         (vect_get_load_cost): Get stmt_info instead of DR.
10904 2018-06-01  Richard Biener  <rguenther@suse.de>
10906         PR middle-end/86017
10907         * gimple-fold.c (var_decl_component_p): Also allow offsetted
10908         vars wrapped in MEM_REFs.
10910 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
10912         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
10913         Fix subreg tests so that we only return a choice between
10914         GENERAL_REGS and FP_REGS if the original classes included both.
10916 2018-06-01  Richard Biener  <rguenther@suse.de>
10918         PR ipa/85960
10919         * tree-ssa-structalias.c (get_function_part_constraint):
10920         Handle NULL fi->decl.
10921         (find_func_aliases_for_call): Properly handle indirect
10922         fi from direct call.
10923         (find_func_clobbers): Likewise.
10924         (ipa_pta_execute): Likewise.
10925         (create_variable_info_for): For functions that are ifunc_resolver
10926         resolve to a varinfo that contains the result of the resolver call.
10927         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
10928         aliases.
10930 2018-05-31  Michael Collison  <michael.collison@arm.com>
10932         * config/aarch64/aarch64.md:
10933         (*fix_to_zero_extenddfdi2): New pattern.
10934         * gcc.target/aarch64/fix_extend1.c: New testcase.
10936 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
10938         PR middle-end/78809
10939         PR middle-end/83026
10940         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
10941         and BUILT_IN_STRNCMP_EQ.
10942         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
10943         BUILT_IN_STRNCMP_EQ.
10944         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
10945         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10946         (gimple_fold_builtin): Likewise.
10947         * tree-ssa-strlen.c (compute_string_length): New function.
10948         (determine_min_obsize): New function.
10949         (handle_builtin_string_cmp): New function to handle calls to
10950         string compare functions.
10951         (strlen_optimize_stmt): Add handling to builtin string compare
10952         calls.
10953         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
10954         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
10955         * tree.c (build_common_builtin_nodes): Add new defines of
10956         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
10958 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
10960         PR target/85984
10961         * bb-reorder.c (pass_partition_blocks::gate): Return false for
10962         functions with naked attribute.
10964 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
10966         * config/i386/sse.md (avx_vec_concat<mode>):
10967         Substitute concat_tg_mode mode attribute with xtg_mode.
10968         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
10969         (concat_tg_mode): Remove mode attribute.
10971 2018-05-31  Martin Sebor  <msebor@redhat.com>
10973         PR c/82063
10974         * calls.c (alloc_max_size): Correct a logic error/typo.
10975         Treat excessive arguments as infinite.  Warn for invalid arguments.
10976         * doc/invoke.texi (-Walloc-size-larger-than): Update.
10978 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
10980         PR target/85829
10981         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
10982         and movx for Haswell.
10984 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
10985             Cesar Philippidis  <cesar@codesourcery.com>
10987         PR middle-end/85879
10988         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
10989         when emitting error on private/firstprivate reductions.
10990         * omp-low.c (lower_omp_target): Avoid reference-type processing
10991         on pointers for firstprivate clause.
10993 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
10995         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
10996         (st1x2): Likewise.
10997         (st1x3): Likewise.
10998         * config/aarch64/aarch64-simd.md
10999         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
11000         (aarch64_ld1_x3_<mode>): Likewise
11001         (aarch64_st1x2<VALLDIF:mode>): Likewise
11002         (aarch64_st1_x2_<mode>): Likewise
11003         (aarch64_st1x3<VALLDIF:mode>): Likewise
11004         (aarch64_st1_x3_<mode>): Likewise
11005         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
11006         (vld1_s8_x3): Likewise.
11007         (vld1_u16_x3): Likewise.
11008         (vld1_s16_x3): Likewise.
11009         (vld1_u32_x3): Likewise.
11010         (vld1_s32_x3): Likewise.
11011         (vld1_u64_x3): Likewise.
11012         (vld1_s64_x3): Likewise.
11013         (vld1_f16_x3): Likewise.
11014         (vld1_f32_x3): Likewise.
11015         (vld1_f64_x3): Likewise.
11016         (vld1_p8_x3): Likewise.
11017         (vld1_p16_x3): Likewise.
11018         (vld1_p64_x3): Likewise.
11019         (vld1q_u8_x3): Likewise.
11020         (vld1q_s8_x3): Likewise.
11021         (vld1q_u16_x3): Likewise.
11022         (vld1q_s16_x3): Likewise.
11023         (vld1q_u32_x3): Likewise.
11024         (vld1q_s32_x3): Likewise.
11025         (vld1q_u64_x3): Likewise.
11026         (vld1q_s64_x3): Likewise.
11027         (vld1q_f16_x3): Likewise.
11028         (vld1q_f32_x3): Likewise.
11029         (vld1q_f64_x3): Likewise.
11030         (vld1q_p8_x3): Likewise.
11031         (vld1q_p16_x3): Likewise.
11032         (vld1q_p64_x3): Likewise.
11033         (vst1_s64_x2): Likewise.
11034         (vst1_u64_x2): Likewise.
11035         (vst1_f64_x2): Likewise.
11036         (vst1_s8_x2): Likewise.
11037         (vst1_p8_x2): Likewise.
11038         (vst1_s16_x2): Likewise.
11039         (vst1_p16_x2): Likewise.
11040         (vst1_s32_x2): Likewise.
11041         (vst1_u8_x2): Likewise.
11042         (vst1_u16_x2): Likewise.
11043         (vst1_u32_x2): Likewise.
11044         (vst1_f16_x2): Likewise.
11045         (vst1_f32_x2): Likewise.
11046         (vst1_p64_x2): Likewise.
11047         (vst1q_s8_x2): Likewise.
11048         (vst1q_p8_x2): Likewise.
11049         (vst1q_s16_x2): Likewise.
11050         (vst1q_p16_x2): Likewise.
11051         (vst1q_s32_x2): Likewise.
11052         (vst1q_s64_x2): Likewise.
11053         (vst1q_u8_x2): Likewise.
11054         (vst1q_u16_x2): Likewise.
11055         (vst1q_u32_x2): Likewise.
11056         (vst1q_u64_x2): Likewise.
11057         (vst1q_f16_x2): Likewise.
11058         (vst1q_f32_x2): Likewise.
11059         (vst1q_f64_x2): Likewise.
11060         (vst1q_p64_x2): Likewise.
11061         (vst1_s64_x3): Likewise.
11062         (vst1_u64_x3): Likewise.
11063         (vst1_f64_x3): Likewise.
11064         (vst1_s8_x3): Likewise.
11065         (vst1_p8_x3): Likewise.
11066         (vst1_s16_x3): Likewise.
11067         (vst1_p16_x3): Likewise.
11068         (vst1_s32_x3): Likewise.
11069         (vst1_u8_x3): Likewise.
11070         (vst1_u16_x3): Likewise.
11071         (vst1_u32_x3): Likewise.
11072         (vst1_f16_x3): Likewise.
11073         (vst1_f32_x3): Likewise.
11074         (vst1_p64_x3): Likewise.
11075         (vst1q_s8_x3): Likewise.
11076         (vst1q_p8_x3): Likewise.
11077         (vst1q_s16_x3): Likewise.
11078         (vst1q_p16_x3): Likewise.
11079         (vst1q_s32_x3): Likewise.
11080         (vst1q_s64_x3): Likewise.
11081         (vst1q_u8_x3): Likewise.
11082         (vst1q_u16_x3): Likewise.
11083         (vst1q_u32_x3): Likewise.
11084         (vst1q_u64_x3): Likewise.
11085         (vst1q_f16_x3): Likewise.
11086         (vst1q_f32_x3): Likewise.
11087         (vst1q_f64_x3): Likewise.
11088         (vst1q_p64_x3): Likewise.
11090 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11092         * config/msp430/msp430.c (msp430_output_labelref): Prepend
11093         user_label_prefix to name.
11095         * tree-core.h: Update comment about the format of NAME string
11096         passed to handler in attribute_spec.
11098         * config/msp430/msp430.md: Remove erroneous subreg expression from
11099         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
11100         zero_extend{q,h}isi2.
11102 2018-05-30  Borislav Petkov  <bp@suse.de>
11104         * doc/extend.texi: Document some architecture specific
11105         constraints and sort entries.
11107 2018-05-30  Martin Sebor  <msebor@redhat.com>
11109         PR middle-end/85369
11110         * builtins.c (expand_builtin_stpcpy_1): New function.
11111         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
11112         only if the former succeeds.
11114 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
11116         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
11117         in saphira.
11119 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11121         * doc/invoke.texi (-flinker-output): Document
11123 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11125         * passes.c (ipa_write_summaries): Only modify statements if body
11126         is in memory.
11127         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
11128         incrementally linking.
11129         (ipa_passes): Likewise.
11130         * lto-cgraph.c (lto_output_node): When incrementally linking do not
11131         pass down resolution info.
11132         * common.opt (flag_incremental_link): Update info.
11133         * gcc.c (plugin specs): Turn flinker-output=* to
11134         -plugin-opt=-linker-output-known
11135         * toplev.c (compile_file): Also cut compilation when doing incremental
11136         link.
11137         * flag-types. (enum lto_partition_model): Add
11138         LTO_LINKER_OUTPUT_NOLTOREL.
11139         (invoke.texi): Add -flinker-output docs.
11140         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
11141         link same way as WPA; do not stream in dead initializers.
11143         * dwarf2out.c (dwarf2out_die_ref_for_decl,
11144         darf2out_register_external_decl): Support incremental link.
11146 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11148         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
11150 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
11152         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
11153         it down to simple_object_copy_lto_debug_sections.
11154         (run_gcc): Determine incremental LTO link time and configure
11155         lto1 into non-wpa mode, disable renaming of debug sections.
11157 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11159         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11160         descriptions of various incorrectly documented functions.
11162 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11164         Revert:
11165         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11166         address check not strict.
11168 2018-05-30  Richard Biener  <rguenther@suse.de>
11170         PR tree-optimization/85964
11171         * tracer.c (better_p): Drop initialized count check, we only
11172         call the function with initialized counts now.
11173         (find_best_successor): Do find a best edge if one
11174         has uninitialized count.
11175         (find_best_predecessor): Likewise.  Do BB frequency check only
11176         if count is initialized.
11178 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
11180         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
11181         (aarch64_ldrstr_offset_compare): New.
11182         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
11183         load/store orderings.
11184         (aarch64_gen_adjusted_ldpstp): Likewise.
11186 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
11188         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
11189         Check for subset of GENERAL_REGS and FP_REGS.
11190         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
11191         r=w alternative.
11193 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
11195         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
11196         and wi::to_poly_offset.  Add the current offset and then check
11197         whether the sum fits, rather than using an unchecked addition of
11198         a checked term.  Check for a shwi rather than a uhwi.
11199         * expr.c (get_bit_range): Use tree_to_poly_uint64.
11200         (store_constructor): Use poly_int_tree_p.
11201         (expand_expr_real_1): Likewise.
11202         * function.c (assign_temp): Likewise.
11203         * fold-const.c (const_binop): Use poly_int_tree_p and
11204         wi::to_poly_offset.
11205         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
11206         division.
11207         * ipa-icf-gimple.c (func_checker::compare_operand): Use
11208         to_poly_offset for MEM offsets.
11209         * ipa-icf.c (sem_variable::equals): Likewise.
11210         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
11211         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
11212         wi::to_poly_offset for BIT_FIELD_REF offsets.
11213         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
11214         wi::to_poly_offset.
11215         * var-tracking.c (emit_note_insn_var_location): Use
11216         tree_to_poly_uint64.
11218 2018-05-29  Jim Wilson  <jimw@sifive.com>
11220         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
11222 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
11224         PR target/85950
11225         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
11226         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
11227         sequence.
11228         (sse4_1_round<mode>2): Use nonimmediate_operand
11229         for operand 1 predicate.
11231 2018-05-29  Martin Sebor  <msebor@redhat.com>
11232             Richard Biener  <rguenther@suse.de>
11234         PR testsuite/85888
11235         * calls.c (get_size_range): Call determine_value_range instead
11236         of get_value_range..
11237         * tree-vrp.h (determine_value_range): Declared new function.
11238         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
11240 2018-05-29  Richard Biener  <rguenther@suse.de>
11242         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
11243         sure to use non-pattern stmts for get_earlier_stmt arguments.
11244         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
11245         called on pattern stmts.
11246         (get_later_stmt): Likewise.
11248 2018-05-29  Martin Liska  <mliska@suse.cz>
11250         PR gcov-profile/85759
11251         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
11252         env variables.
11254 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
11256         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11257         VEC_UNPACK_*_EXPR.
11258         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
11259         VEC_PACK_*_EXPR.
11261         PR target/85918
11262         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11263         VEC_PACK_FLOAT_EXPR): New tree codes.
11264         * tree-pretty-print.c (op_code_prio): Handle
11265         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11266         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11267         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11268         * tree-inline.c (estimate_operator_cost): Likewise.
11269         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
11270         * fold-const.c (const_binop): Likewise.
11271         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
11272         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
11273         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11274         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
11275         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
11276         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
11277         * expr.c (expand_expr_real_2): Likewise.
11278         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
11279         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
11280         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
11281         optabs.
11282         * optabs.c (expand_widen_pattern_expr): For
11283         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
11284         sign from result type rather than operand's type.
11285         (expand_binop_directly): For vec_packu_float_optab and
11286         vec_packs_float_optab allow result type to be different from operand's
11287         type.
11288         * optabs-tree.c (optab_for_tree_code): Handle
11289         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11290         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
11291         * tree-vect-generic.c (expand_vector_operations_1):  Handle
11292         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
11293         VEC_PACK_FLOAT_EXPR.
11294         * tree-vect-stmts.c (supportable_widening_operation): Handle
11295         FIX_TRUNC_EXPR.
11296         (supportable_narrowing_operation): Handle FLOAT_EXPR.
11297         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
11298         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
11299         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
11300         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
11301         mode attributes.
11302         (vec_pack<floatprefix>_float_<mode>): New expander.
11303         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
11304         attributes.
11305         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
11306         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
11307         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
11308         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
11309         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
11310         Document.
11311         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
11312         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
11313         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
11314         VEC_PACK_FLOAT_EXPR): Document.
11316 2018-05-29  Richard Biener  <rguenther@suse.de>
11318         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
11319         member.
11320         (stmt_vec_info_vec): Make pointer.
11321         (init_stmt_vec_info_vec): Remove.
11322         (free_stmt_vec_info_vec): Likewise.
11323         (set_stmt_vec_info_vec): New function.
11324         (free_stmt_vec_infos): Likewise.
11325         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
11326         (set_vinfo_for_stmt): Likewise.
11327         (get_earlier_stmt): Likewise.
11328         (get_later_stmt): Likewise.
11329         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
11330         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
11331         (vec_info::~vec_info): Free stmt_vec_infos.
11332         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
11333         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
11334         (pass_slp_vectorize::execute): Likewise.
11335         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
11336         (free_stmt_vec_info_vec): Likewise.
11337         (set_stmt_vec_info_vec): New function.
11338         (free_stmt_vec_infos): Likewise.
11339         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
11340         the global stmt_vec_info_vec.
11341         * tree-parloops.c (gather_scalar_reductions): Use
11342         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
11343         vector.
11345 2018-05-29  Richard Biener  <rguenther@suse.de>
11347         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
11349 2018-05-29  Martin Liska  <mliska@suse.cz>
11350             David Malcolm  <dmalcolm@redhat.com>
11352         * vec.c (test_reverse): New.
11353         (vec_c_tests): Add new test.
11354         * vec.h (vl_ptr>::reverse): New function.
11356 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
11358         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
11360         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
11361         and later.
11363 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11365         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
11367 2018-05-28  Richard Biener  <rguenther@suse.de>
11369         PR tree-optimization/85933
11370         * tree-vect-data-refs.c (vect_record_base_alignments): Only
11371         look at stmts marked as vectorizable.
11373 2018-05-28  Richard Biener  <rguenther@suse.de>
11375         PR tree-optimization/85934
11376         * tree-vect-generic.c (expand_vector_operations_1): Hoist
11377         vector boolean check before scalar optimization.
11379 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
11381         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
11382         for armv5te.
11384 2018-05-28  Mark Wielaard  <mark@klomp.org>
11386         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
11387         if it is an expression containing a minus sign.
11389 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
11391         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
11393 2018-05-27  Paul Koning  <ni1d@arrl.net>
11395         * config/pdp11/pdp11.md (truncsihi2): Remove.
11397 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
11398             Chung-Ju Wu  <jasonwucj@gmail.com>
11400         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
11401         implementation.
11402         (unaligned_store_dw): Ditto.
11403         * config/nds32/nds32-memory-manipulation.c
11404         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
11405         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
11406         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
11407         (emit_setmem_word_loop): Rename to ...
11408         (emit_setmem_doubleword_loop): ... this.
11409         (nds32_gen_dup_4_byte_to_word_value): New function.
11410         (nds32_gen_dup_8_byte_to_double_word_value): New function.
11411         (nds32_expand_setmem_loop): Refine implementation.
11412         (nds32_expand_setmem_loop_v3m): Ditto.
11413         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
11414         pattern.
11416 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
11418         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
11420 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
11422         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
11423         (nds32_init_machine_status): Initialize machine->attr_naked_p and
11424         machine->attr_no_prologue_p.
11425         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
11426         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
11427         (nds32_expand_epilogue): Consider attr_naked_p.
11428         (nds32_expand_epilogue_v3pop): Likewise.
11429         (nds32_can_use_return_insn): Likewise.
11430         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
11431         attr_no_prologue_p fields.
11432         * config/nds32/nds32.opt (mret-in-naked-func): New option.
11434 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
11436         PR target/85918
11437         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11438         attributes.
11439         * config/i386/sse.md
11440         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11441         Rename to ...
11442         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11443         ... this.
11444         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11445         Rename to ...
11446         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
11447         ... this.
11448         (*<floatsuffix>floatv2div2sf2): Rename to ...
11449         (*float<floatunssuffix>v2div2sf2): ... this.
11450         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
11451         (float<floatunssuffix>v2div2sf2_mask): ... this.
11452         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
11453         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
11454         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
11455         to ...
11456         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11457         ... this.
11458         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11459         Rename to ...
11460         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11461         ... this.
11462         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11463         Rename to ...
11464         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11465         ... this.
11466         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
11467         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
11468         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
11469         gen_ufix_truncv8dfv8si2.
11470         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
11471         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
11472         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
11473         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
11474         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
11475         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
11476         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
11477         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
11479 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
11481         PR target/85900
11482         PR target/85345
11483         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
11485 2018-05-25  Jim Wilson  <jimw@sifive.com>
11487         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
11488         * config/riscv/riscv.c (struct machine_function): Add
11489         interrupt_handler_p and attribute_checked_p fields.
11490         (riscv_attribute_table): Add interrupt.
11491         (riscv_interrupt_type_p): New.
11492         (riscv_save_reg_p): Save extra regs for interrupt handler.
11493         (riscv_use_save_libcall): Return false  for interrupt handler.
11494         (riscv_first_stack_step): Add forward declaration.
11495         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
11496         for interrupt handler with large frame.  Use it for saved reg list.
11497         (riscv_expand_prologue): Move flag_stack_usage_info support to
11498         eliminate duplication.
11499         (riscv_expand_epilogue): Generate mret for interrupt handler.
11500         (riscv_epilogue_uses): New.
11501         (riscv_can_use_return_insn): Return false for interrupt handler.
11502         (riscv_function_ok_for_sibcall): Likewise.
11503         (riscv_set_current_function): Add interrupt handler support.
11504         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
11505         * config/riscv/riscv.md (UNSPECV_MRET): New.
11506         (GP_REGNUM): New.
11507         (riscv_frflags, riscv_fsflags): Use tab after opcode.
11508         (riscv_mret): New.
11509         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
11511 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
11513         PR tree-optimization/85712
11514         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
11515         this candidate has already been replaced in-situ by a copy.
11517 2018-05-25  Jason Merrill  <jason@redhat.com>
11519         PR c++/80485 - inline function non-zero address.
11520         * symtab.c (nonzero_address): Check DECL_COMDAT.
11522 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
11524         PR target/83628
11525         * config/alpha/alpha.md (ashlsi3): New insn pattern.
11526         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
11527         extension of SImode operation.  Use const123_operand predicate.
11528         (*saddsi_1): Remove.
11529         (*saddl_se_1): Ditto.
11530         (*ssubsi_1): Ditto.
11531         (*ssubl_se_1): Ditto.
11532         * config/alpha/predicates.md (const123_operand): New predicate.
11533         * config/alpha/constraints.md (P): Use IN_RANGE.
11535 2018-05-25  Richard Biener  <rguenther@suse.de>
11537         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
11538         defaulted to true.
11539         (ref_maybe_used_by_stmt_p): Likewise.
11540         (stmt_may_clobber_ref_p): Likewise.
11541         (stmt_may_clobber_ref_p_1): Likewise.
11542         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
11543         and pass it along.
11544         (ref_maybe_used_by_stmt_p): Likewise.
11545         (stmt_may_clobber_ref_p): Likewise.
11546         (stmt_may_clobber_ref_p_1): Likewise.
11547         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
11548         the alias oracle to disambiguate DRs with stmts DR analysis
11549         couldn't handle.
11550         (vect_analyze_data_refs): Do not give up on not analyzable
11551         DRs for BB vectorization.  Remove code truncating the dataref
11552         vector.
11554 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
11556         PR target/85832
11557         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11558         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
11559         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
11561 2018-05-25  Richard Biener  <rguenther@suse.de>
11563         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
11564         function, combining stmt data ref gathering and fatal analysis
11565         parts.
11566         (vect_analyze_data_refs): Remove now redudnant code and simplify.
11567         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
11568         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
11569         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
11570         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
11572 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
11574         PR tree-optimization/85720
11575         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
11576         SCC if all partitions are builtins.
11577         (version_loop_by_alias_check): New parameter.  Generate cancelable
11578         runtime alias check if all partitions are builtins.
11579         (distribute_loop): Update call to above function.
11581 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
11583         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
11584         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
11585         (parm_default_def_partition_arg): Ditto.
11586         (set_parm_default_def_partition): Ditto.
11587         (get_parm_default_def_partitions): Ditto and make it static.
11588         (get_undefined_value_partitions): Ditto and make it static.
11589         (remove_ssa_form): Refactor call to init_var_map here.
11590         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
11591         computation for loop region.
11592         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
11593         (register_default_def): Delete.
11594         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
11595         (parm_default_def_partition_arg): Ditto.
11596         (set_parm_default_def_partition): Ditto.
11597         (get_parm_default_def_partitions): Ditto and make it static.
11598         (get_undefined_value_partitions): Ditto and make it static.
11599         (coalesce_with_default, coalesce_with_default): Update comment.
11600         (create_coalesce_list_for_region): New func factored out from
11601         create_outofssa_var_map.
11602         (populate_coalesce_list_for_outofssa): New func factored out from
11603         create_outofssa_var_map and coalesce_ssa_name.
11604         (create_outofssa_var_map): Delete.
11605         (coalesce_ssa_name): Refactor to support live range computation.
11606         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
11607         (get_parm_default_def_partitions): Delete.
11608         (get_undefined_value_partitions): Ditto.
11609         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
11610         computation for loop region.
11611         (new_tree_live_info, loe_visit_block): Ditto.
11612         (live_worklist, set_var_live_on_entry): Ditto.
11613         (calculate_live_on_exit, verify_live_on_entry): Ditto.
11614         * tree-ssa-live.h (struct _var_map): New fields.
11615         (init_var_map): Change decl.
11616         (region_contains_p): New.
11618 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
11620         * tree-ssa-live.h (live_merge_and_clear): Delete.
11622 2018-05-25  Richard Biener  <rguenther@suse.de>
11624         PR c++/85912
11625         * tree-dump.c (dequeue_and_dump): Remove access to removed
11626         operand 2 of a SWITCH_EXPR.
11628 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11630         * doc/sourcebuild.texi (vect_double_cond_arith): Include
11631         multiplication and division.
11632         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
11633         (cond_udiv@var{m}, cond_umod@var{m}): Document.
11634         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
11635         (cond_udiv_optab, cond_umod_optab): New optabs.
11636         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
11637         (IFN_COND_RDIV): New internal functions.
11638         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
11639         TRUNC_MOD_EXPR and RDIV_EXPR.
11640         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
11641         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
11642         New unspecs.
11643         (SVE_INT_BINARY): Include mult.
11644         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
11645         (optab, sve_int_op): Handle mult.
11646         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
11647         UNSPEC_COND_DIV.
11648         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
11649         for SVE_INT_BINARY_SD.
11651 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11653         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
11654         (optab, sve_int_op): Handle div and udiv.
11655         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
11656         for SVE_INT_BINARY_SD.
11657         (*<optab><mode>3): New insn for the same.
11659 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11661         * tree-vect-patterns.c: Include predict.h.
11662         (vect_recog_divmod_pattern): Restrict check for division support
11663         to when optimizing for size.
11665 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11667         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
11668         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
11669         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
11670         (gimple_match_op::set_op): Likewise.
11671         (gimple_resimplify4): Declare.
11672         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
11673         (expr::gen_transform): Likewise.
11674         (decision_tree::gen): Generate a simplification routine for 4 operands.
11675         * gimple-match-head.c (gimple_simplify): Add an overload for
11676         4 operands.  In the top-level function, handle up to 4 call
11677         arguments and call gimple_resimplify4.
11678         (gimple_resimplify4): New function.
11679         (build_call_internal): Pass a fourth operand.
11680         (maybe_push_to_seq): Likewise.
11681         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
11682         Fold VEC_COND_EXPRs of an operation and a default value into
11683         an IFN_COND_* function if possible.
11684         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
11685         New unspecs.
11686         (SVE_COND_FP_BINARY): Include them.
11687         (optab, sve_fp_op): Handle them.
11688         (SVE_INT_BINARY_REV): New code iterator.
11689         (SVE_COND_FP_BINARY_REV): New int iterator.
11690         (commutative): New int attribute.
11691         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
11692         Declare.
11693         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
11694         function.
11695         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
11696         (*cond_<optab><mode>): New patterns for reversed operands.
11698 2018-05-25  Richard Biener  <rguenther@suse.de>
11700         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
11701         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
11702         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
11703         (STMT_VINFO_GROUPED_ACCESS): Adjust.
11704         * tree-vect-data-refs.c (everywhere): Adjust users.
11705         * tree-vect-loop.c (everywhere): Likewise.
11706         * tree-vect-slp.c (everywhere): Likewise.
11707         * tree-vect-stmts.c (everywhere): Likewise.
11708         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11710 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
11712         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
11713         Rename to...
11714         (gcc_cv_as_section_exclude): ... this.
11715         Try Solaris as #exclude syntax.
11716         * configure: Regenerate.
11717         * config.in: Regenerate.
11718         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
11719         SECTION_EXCLUDE.
11720         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
11721         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
11723         * varasm.c (default_elf_asm_named_section): Don't check if
11724         HAVE_GAS_SECTION_EXCLUDE is defined.
11726 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11728         * doc/md.texi: Update the documentation of the cond_* optabs
11729         to mention the new final operand.  Fix GET_MODE_NUNITS call.
11730         Describe the scalar case too.
11731         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
11732         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
11733         instead of 2.
11734         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
11735         (get_conditional_internal_fn): Update comment.
11736         * tree-vect-loop.c (vectorizable_reduction): Pass the original
11737         accumulator value as a final argument to conditional functions.
11738         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
11739         a define_expand and add an "else" operand.  Assert for now that
11740         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
11741         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
11742         (*cond_<optab><mode>): New patterns.
11743         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
11744         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
11745         (UNSPEC_COND_EOR): Delete.
11746         (optab): Remove associated mappings.
11747         (SVE_INT_BINARY): New code iterator.
11748         (sve_int_op): Remove int attribute and add "minus" to the code
11749         attribute.
11750         (SVE_COND_INT_OP): Delete.
11751         (SVE_COND_FP_OP): Rename to...
11752         (SVE_COND_FP_BINARY): ...this.
11754 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
11756         * optabs.c (can_reuse_operands_p): New function.
11757         (maybe_legitimize_operands): Try to reuse the results for
11758         earlier operands.
11760 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
11762         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
11763         Add {q} suffix to insn mnemonic.
11765 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11767         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
11768         (msp430_warn_func_return): New.
11770 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
11772         * fold-const.c (tree_nonzero_bits): New function.
11773         * fold-const.h (tree_nonzero_bits): Likewise.
11774         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
11775         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
11777 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
11779         PR target/85900
11780         PR target/85345
11781         * varasm.c (assemble_alias): Check ifunc_resolver only on
11782         FUNCTION_DECL.
11784 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
11786         PR target/85903
11787         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
11788         when memory input operand is handled.
11790 2018-05-24  Luis Machado  <luis.machado@linaro.org>
11792         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
11793         global.
11794         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
11796 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
11798         * match.pd: Delay FMA folds until after vectorization.
11800 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11802         PR target/83009
11803         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
11804         address check not strict.
11806 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
11808         * gimple-match.h (gimple_match_op): New class.
11809         (mprts_hook): Replace parameters with a gimple_match_op *.
11810         (maybe_build_generic_op): Likewise.
11811         (gimple_simplified_result_is_gimple_val): Replace parameters with
11812         a const gimple_match_op *.
11813         (gimple_simplify): Replace code_helper * and tree * parameters with
11814         a gimple_match_op * parameter.
11815         (gimple_resimplify1): Replace code_helper *, tree and tree *
11816         parameters with a gimple_match_op * parameter.
11817         (gimple_resimplify2): Likewise.
11818         (gimple_resimplify3): Likewise.
11819         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
11820         parameters with a gimple_match_op * parameter.
11821         * gimple-match-head.c (gimple_simplify): Change prototypes of
11822         auto-generated functions to take a gimple_match_op * instead of
11823         separate code_helper * and tree * parameters.  Make the same
11824         change in the top-level overload and update calls to the
11825         gimple_resimplify routines.  Update calls to the auto-generated
11826         functions and to maybe_push_res_to_seq in the publicly-facing
11827         operation-specific gimple_simplify overloads.
11828         (gimple_match_op::MAX_NUM_OPS): Define.
11829         (gimple_resimplify1): Replace rcode and ops with a single res_op
11830         parameter.  Update call to gimple_simplify.
11831         (gimple_resimplify2): Likewise.
11832         (gimple_resimplify3): Likewise.
11833         (mprts_hook): Replace parameters with a gimple_match_op *.
11834         (maybe_build_generic_op): Likewise.
11835         (build_call_internal): Replace type, nargs and ops with
11836         a gimple_match_op *.
11837         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
11838         with a single gimple_match_op *.  Update calls to mprts_hook,
11839         build_call_internal and gimple_simplified_result_is_gimple_val.
11840         Factor out code that is common to the tree_code and combined_fn cases.
11841         * genmatch.c (expr::gen_transform): Replace tem_code and
11842         tem_ops with a gimple_match_op called tem_op.  Update calls
11843         to the gimple_resimplify functions and maybe_push_res_to_seq.
11844         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
11845         res_ops.  Update call to the gimple_resimplify functions.
11846         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
11847         (decision_tree::gen): Make the functions take a gimple_match_op *
11848         called res_op instead of separate res_code and res_ops parameters.
11849         Update call accordingly.
11850         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
11851         and ops with a single res_op parameter.  Update calls to
11852         maybe_build_generic_op and maybe_push_res_to_seq.
11853         (fold_stmt_1): Update calls to gimple_simplify and
11854         replace_stmt_with_simplification.
11855         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
11856         and gimple_simplified_result_is_gimple_val.
11857         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
11858         gimple_simplify.
11859         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
11860         with a gimple_match_op *.
11861         (vn_nary_build_or_lookup): Likewise.  Update call to
11862         vn_nary_build_or_lookup_1.
11863         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
11864         gimple_match_op *.  Update calls to the gimple_resimplify routines
11865         and to gimple_simplified_result_is_gimple_val.
11866         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
11867         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
11868         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
11869         (visit_nary_op): Likewise.
11870         (visit_reference_op_load): Likewise.
11872 2018-05-23  Luis Machado  <luis.machado@linaro.org>
11874         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
11875         modifier for printing the step amount.
11877 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
11879         PR target/78849
11880         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
11881         types.
11883 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
11885         * doc/sourcebuild.texi (Endianness): New subsubsection.
11887 2018-05-23  Luis Machado  <luis.machado@linaro.org>
11889         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11890         <prefetch_dynamic_strides>: New const bool field.
11891         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11892         prefetch_dynamic_strides.
11893         (exynosm1_prefetch_tune): Likewise.
11894         (thunderxt88_prefetch_tune): Likewise.
11895         (thunderx_prefetch_tune): Likewise.
11896         (thunderx2t99_prefetch_tune): Likewise.
11897         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
11898         false.
11899         (aarch64_override_options_internal): Update to set
11900         PARAM_PREFETCH_DYNAMIC_STRIDES.
11901         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
11902         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
11903         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
11904         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
11905         prefetch-dynamic-strides setting.
11907 2018-05-23  Luis Machado  <luis.machado@linaro.org>
11909         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
11910         <minimum_stride>: New const int field.
11911         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
11912         minimum_stride field defaulting to -1.
11913         (exynosm1_prefetch_tune): Likewise.
11914         (thunderxt88_prefetch_tune): Likewise.
11915         (thunderx_prefetch_tune): Likewise.
11916         (thunderx2t99_prefetch_tune): Likewise.
11917         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
11918         <default_opt_level>: Set to 3.
11919         (aarch64_override_options_internal): Update to set
11920         PARAM_PREFETCH_MINIMUM_STRIDE.
11921         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
11922         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
11923         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
11924         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
11925         stride is constant and is below the minimum stride threshold.
11927 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11929         * config/arm/arm-cpus.in (mode26): Delete.
11930         (armv4): Delete mode26 reference.
11931         * config/arm/arm.c (arm_configure_build_target): Delete use of
11932         isa_bit_mode26.
11934 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
11936         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11937         New insn pattern.
11938         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11939         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
11940         for non-SSE modes.
11941         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
11942         (floatunsdidf2): Ditto.
11944 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
11946         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
11947         (fixuns_trunc<mode>si2_avx512f): Ditto.
11948         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
11949         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
11950         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
11952 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
11954         PR rtl-optimization/79985
11955         * df-scan.c (df_insn_refs_collect): Remove special case for
11956         global registers and asm statements.
11958 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
11960         * extend.texi (Global Register Variables): Rewrite the bullet list.
11961         Note that the register is available for allocation. Note that access
11962         via inline asm must use constraints. Add note about async-signal
11963         handlers. Remove paragraph about automagic register selection.
11965 2018-05-23  Richard Biener  <rguenther@suse.de>
11967         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
11968         of fixed offset from memset VN.
11970 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
11972         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
11973         first_interp field.
11974         (alloc_cand_and_find_basis): Initialize first_interp field.
11975         (slsr_process_mul): Modify first_interp field.
11976         (slsr_process_add): Likewise.
11977         (slsr_process_cast): Modify first_interp field for each new
11978         interpretation.
11979         (slsr_process_copy): Likewise.
11980         (dump_candidate): Dump first_interp field.
11981         (replace_mult_candidate): Process all interpretations, not just
11982         subsequent ones.
11983         (replace_rhs_if_not_dup): Likewise.
11984         (replace_one_candidate): Likewise.
11986 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
11988         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
11989         Add new boolean.
11990         (aarch64_needs_frame_chain): New function.
11991         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
11993 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
11995         PR target/84882
11996         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
11997         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
11998         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
11999         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
12000         as true for strict-align.
12001         (aarch64_can_inline_p): Perform checks even when callee has no
12002         attributes to check for strict alignment.
12003         * doc/extend.texi (AArch64 Function Attributes): Document
12004         no-strict-align.
12005         * doc/invoke.texi: (AArch64 Options): Likewise.
12007 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
12009         PR tree-optimization/85853
12010         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
12011         the handling of the root of the node to...
12012         (vect_slp_analyze_node_operations_1): ...this new function,
12013         and run the whole thing with the child nodes' def types
12014         set according to their SLP node's def type.
12016 2018-05-23  Richard Biener  <rguenther@suse.de>
12018         PR middle-end/85874
12019         * tree-data-ref.c (create_runtime_alias_checks): Defer
12020         and ignore overflow warnings.
12022 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
12024         PR tree-optimization/85822
12025         * tree-vrp.c (is_masked_range_test): Fix handling of negative
12026         constants.
12028 2018-05-23  Richard Biener  <rguenther@suse.de>
12030         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12031         memset constants via native_interpret_expr.
12033 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
12035         PR target/85345
12036         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
12037         attribute.
12038         (cgraph_node::create_alias): Likewise.
12039         (cgraph_node::get_availability): Check ifunc_resolver instead
12040         of looking up ifunc attribute.
12041         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
12042         * varasm.c (do_assemble_alias): Likewise.
12043         (assemble_alias): Likewise.
12044         (default_binds_local_p_3): Likewise.
12045         * cgraph.h (cgraph_node): Add ifunc_resolver.
12046         (cgraph_node::only_called_directly_or_aliased_p): Return false
12047         for IFUNC resolver.
12048         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
12049         attribute.
12050         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
12051         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
12052         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
12053         instead of looking up ifunc attribute.
12055 2018-05-22  Luis Machado  <luis.machado@linaro.org>
12057         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
12059 2018-05-22  Martin Sebor  <msebor@redhat.com>
12061         PR middle-end/85359
12062         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
12063         only when expasion succeeds.
12064         (expand_builtin_strcmp): Same.
12065         (expand_builtin_strncmp): Same.
12067 2018-05-22  Martin Sebor  <msebor@redhat.com>
12069         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
12071 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
12072             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12074         * config/aarch64/aarch64-ldpstp.md: Replace uses of
12075         aarch64_mem_pair_operand with memory_operand and delete operand swapping
12076         code.
12077         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12078         Add check for legitimate_address.
12079         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
12080         (aarch64_swap_ldrstr_operands): New.
12081         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
12082         Define prototype.
12084 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
12085             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12087         * config/aarch64/aarch64.md: New patterns to generate stp
12088         and ldp.
12089         (store_pair_sw, store_pair_dw): New patterns to generate stp for
12090         single words and double words.
12091         (load_pair_sw, load_pair_dw): Likewise.
12092         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
12093         Delete.
12094         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
12095         Delete.
12096         * config/aarch64/aarch64-ldpstp.md: Modify peephole
12097         for different mode ldpstp and add peephole for merged zero stores.
12098         Likewise for loads.
12099         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
12100         Add size check.
12101         (aarch64_gen_store_pair): Rename calls to match new patterns.
12102         (aarch64_gen_load_pair): Rename calls to match new patterns.
12103         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
12104         (load_pair<DREG:mode><DREG2:mode>): ... This.
12105         (store_pair<mode>): Rename to...
12106         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
12107         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
12108         New mode iterators.
12109         (V_INT_EQUIV): Handle SImode.
12110         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
12111         New predicate.
12113 2018-05-22  Martin Sebor  <msebor@redhat.com>
12115         PR c/85623
12116         * calls.c (maybe_warn_nonstring_arg): Use string length to set
12117         or ajust the presumed bound on an operation to avoid unnecessary
12118         warnings.
12120 2018-05-22  Martin Sebor  <msebor@redhat.com>
12122         PR tree-optimization/85826
12123         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
12124         assuming that a DECL necesarily has a constant size.
12126 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
12128         PR middle-end/85862
12129         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
12131 2018-05-22  Richard Biener  <rguenther@suse.de>
12133         PR tree-optimization/85834
12134         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
12135         non-constant and non-zero memset arguments.
12137 2018-05-22  Martin Liska  <mliska@suse.cz>
12139         PR ipa/85607
12140         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
12142 2018-05-22  Richard Biener  <rguenther@suse.de>
12144         PR tree-optimization/85863
12145         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
12146         comparisons when vectype is specified.
12147         (vectorizable_condition): Do not specify vectype for
12148         vect_is_simple_cond when SLP vectorizing.
12150 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
12152         PR target/85657
12153         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
12154         define __ibm128 as long double.
12155         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
12156         as a distinct type when IEEE 128-bit support is enabled.
12157         (init_float128_ieee): Fix up conversions between IFmode and IEEE
12158         128-bit types to use the correct functions.
12159         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
12160         convert between 128-bit floating point types that have different
12161         modes but the same representation, instead of using gen_lowpart to
12162         makean alias.
12163         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12164         KFmode.
12165         (IFKF_reg): New attributes to give the register constraints for
12166         IFmode and KFmode.
12167         (extend<mode>tf2_internal): New insns to mark an explicit
12168         conversion between 128-bit floating point types that have a
12169         different mode but share the same representation.
12171 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
12173         PR tree-optimization/85814
12174         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
12175         a null return from get_strinfo when unsharing the next
12176         strinfo in the chain.
12178 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
12180         PR gcc/84923
12181         * varasm.c (weak_finish): Clean up weak_decls.
12183 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12185         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
12186         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
12187         UNSPEC_UADALP values.
12188         * config/aarch64/iterators.md (ABAL): New int iterator.
12189         (ABDL2): Likewise.
12190         (ADALP): Likewise.
12191         (sur): Add mappings for the above.
12192         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
12193         New define_insn.
12194         (aarch64_<sur>abal<mode>_4): Likewise.
12195         (aarch64_<sur>adalp<mode>_3): Likewise.
12196         (<sur>sadv16qi): New define_expand.
12198 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
12200         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
12201         (*movdf_internal): Ditto.
12202         (*rcpsf2_sse): Ditto.
12203         (*rsqrtsf2_sse): Ditto.
12204         (*sqrt<mode>2_sse): Ditto.
12206 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
12208         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
12209         eor3q<mode>4.
12210         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
12211         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
12212         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
12213         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
12214         vbcaxq_s64): New.
12215         * config/aarch64/arm_neon.h: Likewise.
12216         * config/aarch64/iterators.md (VQ_I): New.
12218 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
12220         * config.gcc: Add arc/t-multilib-linux to tmake_file for
12221         arc*-*-linux*.
12222         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
12223         MULTILIB_DIRNAMES
12225 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
12227         * config/nds32/constraints.md (S): New constraint.
12228         * config/nds32/nds32.md (call_internal): Use constraint S.
12229         (call_value_internal): Likewise.
12230         (sibcall_internal): Likewise.
12231         (sibcall_value_internal): Likewise.
12233 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
12234             Chung-Ju Wu  <jasonwucj@gmail.com>
12236         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
12237         into consideration.
12239 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
12240             Chung-Ju Wu  <jasonwucj@gmail.com>
12242         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
12243         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
12244         (nds32_rtx_costs_impl): Simplify.
12245         (nds32_address_cost_impl): Simplify.
12246         (nds32_init_rtx_costs): New function.
12247         (nds32_rtx_costs_speed_prefer): Likewise.
12248         (nds32_rtx_costs_size_prefer): Likewise.
12249         (nds32_address_cost_speed_prefer): Likewise.
12250         (nds32_address_cost_speed_fwprop): Likewise.
12251         (nds32_address_cost_size_prefer): Likewise.
12252         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
12253         * config/nds32/nds32.c (nds32_option_override): Use
12254         nds32_init_rtx_costs function.
12256 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
12258         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
12259         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
12260         (TARGET_PIPELINE_N8): Likewise.
12261         (TARGET_PIPELINE_N10): Likewise.
12262         (TARGET_PIPELINE_N13): Likewise.
12263         (TARGET_PIPELINE_GRAYWOLF): Likewise.
12265 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
12267         * config/nds32/nds32-fpu.md: Update copyright year.
12269 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12271         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
12273 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12275         * config/nds32/nds32.c
12276         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
12277         * config/nds32/nds32.opt (minline-asm-r15): New option.
12279 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12281         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
12282         MASK_HW_ABS.
12283         * config/nds32/nds32.md (abssi2): New pattern.
12285 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
12287         * config/i386/i386.md (rex64namesuffix): New mode attribute.
12288         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
12289         Merge insn pattern from sse_cvtsi2ss<round_name> and
12290         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
12291         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
12292         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
12293         using SWI48 mode iterator.
12294         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
12295         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
12296         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
12297         pattern from sse_cvttss2si<round_saeonly_name>
12298         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
12299         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
12300         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
12301         using SWI48 mode iterator.
12302         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
12303         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
12304         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
12305         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
12306         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
12307         using SWI48 mode iterator.
12308         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
12309         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
12310         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
12311         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
12312         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
12313         SWI48 mode iterator.
12314         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
12315         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
12316         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
12317         pattern from sse_cvttsd2si<round_saeonly_name>
12318         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
12320 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12322         * config/nds32/nds32-md-auxiliary.c
12323         (nds32_valid_smw_lwm_base_p): Refine.
12324         (nds32_output_smw_single_word): Refine.
12325         (nds32_output_smw_double_word): New.
12326         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
12328 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12330         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
12331         (nds32_output_stack_pop): Refine.
12332         (nds32_expand_unaligned_load): Refine.
12333         (nds32_expand_unaligned_store): Refine.
12335 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
12336             Chung-Ju Wu  <jasonwucj@gmail.com>
12338         * config/nds32/constants.md: Add TP_REGNUM constant.
12339         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
12340         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
12341         UNSPEC_ADD32.
12342         * config/nds32/nds32-doubleword.md: Consider flag_pic.
12343         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
12344         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
12345         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
12346         and PIC code generation.
12347         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
12348         code generation.
12349         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
12350         optimization.
12351         * config/nds32/nds32.md: Support TLS and PIC.
12352         * config/nds32/nds32.c: Support TLS and PIC.
12353         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
12354         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
12355         predicate.
12357 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
12359         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
12360         mode with E_ prefix.
12362 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
12363             Chung-Ju Wu  <jasonwucj@gmail.com>
12365         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
12366         * config/nds32/nds32-md-auxiliary.c
12367         (symbolic_reference_mentioned_p): New.
12368         (nds32_legitimize_ict_address): New.
12369         (nds32_expand_ict_move): New.
12370         (nds32_indirect_call_referenced_p): New.
12371         (nds32_symbol_binds_local_p): Delete.
12372         (nds32_long_call_p): Modify.
12373         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
12374         * config/nds32/nds32-protos.h
12375         (symbolic_reference_mentioned_p): Declare.
12376         (nds32_legitimize_ict_address): Declare.
12377         (nds32_expand_ict_move): Declare.
12378         (nds32_indirect_call_referenced_p): Declare.
12379         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
12380         (nds32_relax_group): Use nds32_ict_const_p as condition.
12381         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
12382         (nds32_asm_file_start): Output ict_model directive in asm code.
12383         (nds32_legitimate_address_p): Consider indirect call.
12384         (nds32_print_operand): Consider indirect call.
12385         (nds32_print_operand_address): Consider indirect call.
12386         (nds32_insert_attributes): Handle "indirect_call" attribute.
12387         (TARGET_LEGITIMATE_ADDRESS_P): Define.
12388         (TARGET_LEGITIMATE_CONSTANT_P): Define.
12389         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
12390         (TARGET_DELEGITIMIZE_ADDRESS): Define.
12391         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
12392         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
12393         (TARGET_ICT_MODEL_SMALL): Define.
12394         (TARGET_ICT_MODEL_LARGE): Define.
12395         * config/nds32/nds32.md (movsi): Consider ict model.
12396         (call, call_value): Consider ict model.
12397         (sibcall, sibcall_value): Consider ict model.
12398         * config/nds32/nds32.opt (mict-model): New option.
12399         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
12400         model.
12402 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
12403             Monk Chiang  <sh.chiang04@gmail.com>
12404             Jim Wilson <jimw@sifive.com>
12406         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
12407         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
12408         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
12409         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
12410         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
12411         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
12412         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
12413         compute save_libcall_adjustment properly.
12414         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
12415         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
12416         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
12417         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
12418         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
12419         (ABI_SPEC): Handle mabi=ilp32e.
12420         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
12421         (RVE): Add RVE mask.
12422         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
12423         <-march>: Add rv32e as an example.
12425 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
12427         PR c++/82899
12428         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
12429         (intra_create_variable_infos): Handle C++ constructors.
12431 2018-05-18  Martin Liska  <mliska@suse.cz>
12433         * passes.def: Remove a redundant pass.
12435 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
12437         PR bootstrap/85838
12438         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
12440 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12442         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
12443         (ARMv4): Update.
12444         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
12445         (ARMv6m): Update.
12446         (armv2, armv2a, armv3, armv3m): Delete architectures.
12447         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
12448         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
12449         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
12450         Delete cpus.
12451         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
12452         (*mulsidi3adddi): Likewise.
12453         (mulsidi3): Likewise.
12454         (*mulsidi3_nov6): Likewise.
12455         (umulsidi3): Likewise.
12456         (umulsidi3_nov6): Likewise.
12457         (umaddsidi4): Likewise.
12458         (*umulsidi3adddi): Likewise.
12459         (smulsi3_highpart): Likewise.
12460         (*smulsi3_highpart_nov6): Likewise.
12461         (umulsi3_highpart): Likewise.
12462         (*umulsi3_highpart_nov6): Likewise.
12463         * config/arm/arm.h (arm_arch3m): Delete.
12464         * config/arm/arm.c (arm_arch3m): Delete.
12465         (arm_option_override_internal): Update armv3-related comment.
12466         (arm_configure_build_target): Delete use of isa_bit_mode32.
12467         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
12468         (arm_rtx_costs_internal): Delete check of arm_arch3m.
12469         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
12470         (mulsa3): Likewise.
12471         (mulusa3): Likewise.
12472         * config/arm/arm-protos.h (arm_arch3m): Delete.
12473         * config/arm/arm-tables.opt: Regenerate.
12474         * config/arm/arm-tune.md: Likewise.
12475         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
12476         deleted architectures.
12478 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12480         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
12481         (armv5t, armv5te): New features.
12482         (ARMv5, ARMv5e): Delete fgroups.
12483         (ARMv5t, ARMv5te): Adjust for above changes.
12484         (ARMv6m): Likewise.
12485         (armv5, armv5e): Delete arches.
12486         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
12487         arm_arch5.
12488         (*call_reg_arm): Likewise.
12489         (*call_value_reg_armv5): Likewise.
12490         (*call_value_reg_arm): Likewise.
12491         (*call_symbol): Likewise.
12492         (*call_value_symbol): Likewise.
12493         (*sibcall_insn): Likewise.
12494         (*sibcall_value_insn): Likewise.
12495         (clzsi2): Likewise.
12496         (prefetch): Likewise.
12497         (define_split and define_peephole2 dependent on arm_arch5):
12498         Likewise.
12499         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
12500         arm_arch5e.
12501         (TARGET_ARM_QBIT): Likewise.
12502         (TARGET_DSP_MULTIPLY): Likewise.
12503         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
12504         (arm_arch5, arm_arch5e): Delete.
12505         (arm_arch5t, arm_arch5te): Declare.
12506         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
12507         (arm_arch5t): Declare.
12508         (arm_option_reconfigure_globals): Update for the above.
12509         (arm_options_perform_arch_sanity_checks): Update comment, replace
12510         use of arm_arch5 with arm_arch5t.
12511         (use_return_insn): Likewise.
12512         (arm_emit_call_insn): Likewise.
12513         (output_return_instruction): Likewise.
12514         (arm_final_prescan_insn): Likewise.
12515         (arm_coproc_builtin_available): Likewise.
12516         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
12517         arm_arch5e with arm_arch5t and arm_arch5te.
12518         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
12519         (arm_arch5t, arm_arch5te): Declare.
12520         * config/arm/arm-tables.opt: Regenerate.
12521         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
12522         * config/arm/t-multilib: Likewise.
12523         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
12524         instead of arm_arch5.
12525         (*call_reg_thumb1): Likewise.
12526         (*call_value_reg_thumb1_v5): Likewise.
12527         (*call_value_reg_thumb1): Likewise.
12528         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
12529         unreachable path.
12530         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
12532 2018-05-18  Martin Liska  <mliska@suse.cz>
12534         PR gcov-profile/84846
12535         * doc/gcov.texi: Document -t option of gcov tool.
12537 2018-05-18  Martin Liska  <mliska@suse.cz>
12539         PR gcov-profile/84846
12540         * gcov.c (print_usage): Add new -t option.
12541         (process_args): Handle the option.
12542         (generate_results): Use stdout as output when requested by
12543         the option.
12545 2018-05-18  Martin Liska  <mliska@suse.cz>
12547         PR gcov-profile/84846
12548         * coverage.c (coverage_init): Write PWD to .gcno file.
12549         * doc/gcov.texi: Document how working directory is printed.
12550         * gcov-dump.c (dump_gcov_file): Print PWD.
12551         * gcov.c (output_intermediate_file): Likewise.
12552         (read_graph_file): Read PWD string.
12553         (output_lines): Print PWD.
12555 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
12557         PR middle-end/85817
12558         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
12559         for retval and return false if all args to phi are zero.
12561 2018-05-18  Richard Biener  <rguenther@suse.de>
12563         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
12564         method.
12565         (evrp_dom_walker::before_dom_children): Call it.
12567 2018-05-18  Richard Biener  <rguenther@suse.de>
12569         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
12570         results when processing array refs with variable index.
12572 2018-05-18  Toon Moene  <toon@moene.org>
12574         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
12575         directly after that of -floop-interchange. Indicate that both
12576         options are enabled by default when specifying -O3.
12578 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12580         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
12581         iterator.  Delete separate integer-mode vec_set<mode> expander.
12582         (aarch64_simd_vec_setv2di): Delete.
12583         (vec_setv2di): Delete.
12584         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
12585         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
12586         the "w, r" alternative.
12588 2018-05-18  Martin Liska  <mliska@suse.cz>
12590         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
12591         * tree-pass.h (make_pass_lower_switch_O0): New function.
12592         * tree-switch-conversion.c (node_has_low_bound): Remove.
12593         (node_has_high_bound): Likewise.
12594         (node_is_bounded): Likewise.
12595         (class pass_lower_switch): Make it a template type and create
12596         two instances.
12597         (pass_lower_switch::execute): Add template argument.
12598         (make_pass_lower_switch): New function.
12599         (make_pass_lower_switch_O0): New function.
12600         (do_jump_if_equal): Remove.
12601         (emit_case_nodes): Simplify to just handle all 3 cases and leave
12602         all the hard work to tree optimization passes.
12604 2018-05-18  Martin Liska  <mliska@suse.cz>
12606         * dbgcnt.c (limit_low): Renamed from limit.
12607         (limit_high): New variable.
12608         (dbg_cnt_is_enabled): Check for upper limit.
12609         (dbg_cnt): Adjust dumping.
12610         (dbg_cnt_set_limit_by_index): Add new argument for high
12611         value.
12612         (dbg_cnt_set_limit_by_name): Likewise.
12613         (dbg_cnt_process_single_pair): Parse new format.
12614         (dbg_cnt_process_opt): Use strtok.
12615         (dbg_cnt_list_all_counters): Remove 'value' and add
12616         'limit_high'.
12617         * doc/invoke.texi: Document changes.
12619 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
12621         * doc/sourcebuild.texi (scalar_all_fma): Document.
12622         * tree.def (FMA_EXPR): Delete.
12623         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
12624         * internal-fn.c (ternary_direct): New macro.
12625         (expand_ternary_optab_fn): Likewise.
12626         (direct_ternary_optab_supported_p): Likewise.
12627         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
12628         * builtins.c (fold_builtin_fma): Delete.
12629         (fold_builtin_3): Don't call it.
12630         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
12631         * expr.c (expand_expr_real_2): Likewise.
12632         * fold-const.c (operand_equal_p): Likewise.
12633         (fold_ternary_loc): Likewise.
12634         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
12635         * gimple.c (DEFTREECODE): Likewise.
12636         * gimplify.c (gimplify_expr): Likewise.
12637         * optabs-tree.c (optab_for_tree_code): Likewise.
12638         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
12639         * tree-eh.c (operation_could_trap_p): Likewise.
12640         (stmt_could_throw_1_p): Likewise.
12641         * tree-inline.c (estimate_operator_cost): Likewise.
12642         * tree-pretty-print.c (dump_generic_node): Likewise.
12643         (op_code_prio): Likewise.
12644         * tree-ssa-loop-im.c (stmt_cost): Likewise.
12645         * tree-ssa-operands.c (get_expr_operands): Likewise.
12646         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
12647         * fold-const-call.h (fold_fma): Delete.
12648         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
12649         CFN_FNMA and CFN_FNMS.
12650         (fold_fma): Delete.
12651         * genmatch.c (combined_fn): New enum.
12652         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
12653         (commutative_op): New function.
12654         (commutate): Use it.  Handle more than 2 operands.
12655         (dt_operand::gen_gimple_expr): Use commutative_op.
12656         (parser::parse_expr): Allow :c to be used with non-binary
12657         operators if the commutative operand is known.
12658         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
12659         CFN_FMS, CFN_FNMA and CFN_FNMS.
12660         (backprop::process_assign_use): Remove FMA_EXPR handling.
12661         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
12662         (gen_hsa_fma): New function.
12663         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
12664         IFN_FNMA and IFN_FNMS.
12665         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
12666         * gimple-fold.h (follow_all_ssa_edges): Declare.
12667         * gimple-fold.c (follow_all_ssa_edges): New function.
12668         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
12669         gimple_build interface and use follow_all_ssa_edges to fold the result.
12670         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
12671         instead of checking for optabs directly.
12672         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
12673         rather than FMA_EXPRs.
12674         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
12675         call to IFN_FMA instead of an FMA_EXPR.
12677 2018-05-17  Jim Wilson  <jimw@sifive.com>
12679         * expr.c (do_tablejump): When converting index to Pmode, if we have a
12680         sign extended promoted subreg, and the range does not have the sign bit
12681         set, then do a sign extend.
12683         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
12684         test, check for sign extended subreg and/or constant operands, and
12685         do a sign extend in that case.
12687 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
12689         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
12690         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
12691         Add untyped.
12692         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
12693         Change logics_shift_reg to logics_shift_imm.
12694         (thunderx2t99_fp_loadpair_basic): Delete.
12695         (thunderx2t99_fp_storepair_basic): Delete.
12696         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
12697         (thunderx2t99_asimd_polynomial): Delete.
12698         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
12699         and neon_fp_mul_d_scalar_q.
12700         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
12701         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
12702         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
12703         (thunderx2t99_asimd_lut): Add missing tbl types.
12704         (thunderx2t99_asimd_ext): Delete.
12705         (thunderx2t99_asimd_load1_1_mult): Delete.
12706         (thunderx2t99_asimd_load1_2_mult): Delete.
12707         (thunderx2t99_asimd_load1_ldp): New.
12708         (thunderx2t99_asimd_load1): New.
12709         (thunderx2t99_asimd_load2): Add missing *load2* types.
12710         (thunderx2t99_asimd_load3): New.
12711         (thunderx2t99_asimd_load4): New.
12712         (thunderx2t99_asimd_store1_1_mult): Delete.
12713         (thunderx2t99_asimd_store1_2_mult): Delete.
12714         (thunderx2t99_asimd_store2_mult): Delete.
12715         (thunderx2t99_asimd_store2_onelane): Delete.
12716         (thunderx2t99_asimd_store_stp): New.
12717         (thunderx2t99_asimd_store1): New.
12718         (thunderx2t99_asimd_store2): New.
12719         (thunderx2t99_asimd_store3): New.
12720         (thunderx2t99_asimd_store4): New.
12722 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
12724         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
12725         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
12727 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
12728             Segher Boessenkool  <segher@kernel.crashing.org>
12730         PR target/85698
12731         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
12732         operand.
12734 2018-05-17  Richard Biener  <rguenther@suse.de>
12736         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
12737         for pruning loop and prune defs feeding only already visited PHIs.
12739 2018-05-17  Richard Biener  <rguenther@suse.de>
12741         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
12743 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
12744             Richard Biener  <rguenther@suse.de>
12746         PR tree-optimization/85793
12747         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
12748         for VMAT_ELEMENTWISE.
12750 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
12752         * internal-fn.h (lookup_internal_fn): Declare
12753         * internal-fn.c (lookup_internal_fn): New function.
12754         * gimple.c (gimple_build_call_from_tree): Handle calls to
12755         internal functions.
12756         * gimple-pretty-print.c (dump_gimple_call): Print "." before
12757         internal function names.
12758         * tree-pretty-print.c (dump_generic_node): Likewise.
12759         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
12761 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
12763         * gimple-fold.h (gimple_build): Make the function forms take
12764         combined_fn rather than built_in_function.
12765         (gimple_simplify): Likewise.
12766         * gimple-match-head.c (gimple_simplify): Likewise.
12767         * gimple-fold.c (gimple_build): Likewise.
12768         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
12769         rather than gimple_build_call_internal.
12770         (get_initial_defs_for_reduction): Likewise.
12771         (vect_create_epilog_for_reduction): Likewise.
12772         (vectorizable_live_operation): Likewise.
12774 2018-05-17  Martin Liska  <mliska@suse.cz>
12776         * gimple-ssa-sprintf.c (format_directive): Do not use
12777         space in between 'G_' and '('.
12779 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
12781         PR target/85323
12782         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12783         even if the mask is not all ones.
12785         PR target/85323
12786         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12787         vector.
12788         (ix86_gimple_fold_builtin): Likewise.
12790         PR target/85323
12791         * config/i386/i386.c: Include tree-vector-builder.h.
12792         (ix86_vector_shift_count): New function.
12793         (ix86_fold_builtin): Fold shift builtins by scalar count.
12794         (ix86_gimple_fold_builtin): Likewise.
12796         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
12797         _mm512_setzero): New intrinsics.
12799 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
12800             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12802         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
12803         code generation for cases where splatting a value is not useful.
12804         * simplify-rtx.c (simplify_ternary_operation): Simplify
12805         vec_merge across a vec_duplicate and a paradoxical subreg forming
12806         a vector mode to a vec_concat.
12808 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
12810         * config.gcc: Support "goldmont-plus".
12811         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
12812         "goldmont-plus".
12813         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12814         PROCESSOR_GOLDMONT_PLUS.
12815         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
12816         (processor_target_table): Add "goldmont-plus".
12817         (PTA_GOLDMONT_PLUS): Define.
12818         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
12819         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
12820         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
12821         (fold_builtin_cpu): Add "goldmont-plus".
12822         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
12823         (ix86_option_override_internal): Add "goldmont-plus".
12824         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
12825         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
12826         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
12827         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
12829 2018-05-17  Richard Biener  <rguenther@suse.de>
12831         PR tree-optimization/85757
12832         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
12833         remove defs that only feed that PHI from further processing.
12835 2018-05-16  Jim Wilson  <jimw@sifive.com>
12837         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
12838         asterisk to name.
12839         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
12841 2018-05-16  Mark Wielaard  <mark@klomp.org>
12843         * dwarf2out.c (count_index_strings): New function.
12844         (output_indirect_strings): Call count_index_strings and generate
12845         header for dwarf_version >= 5.
12847 2018-05-16  Mark Wielaard  <mark@klomp.org>
12849         * dwarf2out.c (dwarf_FORM): New function.
12850         (set_indirect_string): Use dwarf_FORM.
12851         (reset_indirect_string): Likewise.
12852         (size_of_die): Likewise.
12853         (value_format): Likewise.
12854         (output_die): Likewise.
12855         (add_skeleton_AT_string): Likewise.
12856         (output_macinfo_op): Likewise.
12857         (index_string): Likewise.
12858         (output_index_string_offset): Likewise.
12859         (output_index_string): Likewise.
12860         (count_index_strings): Likewise.
12862 2018-05-16  Carl Love  <cel@us.ibm.com>
12864         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
12865         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
12867 2018-05-16  Martin Jambor  <mjambor@suse.cz>
12869         * ipa-prop.c (ipa_free_all_edge_args): Remove.
12870         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
12872 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
12874         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
12875         (fnma<mode>4): Likewise.
12876         (fms<mode>4): Likewise.
12877         (fnms<mode>4): Likewise.
12878         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
12879         (aarch64_fnma<mode>4): Likewise.
12880         (aarch64_fms<mode>4): Likewise.
12881         (aarch64_fnms<mode>4): Likewise.
12882         (aarch64_fnmadd<mode>4): Likewise.
12884 2018-05-16  Jason Merrill  <jason@redhat.com>
12886         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
12888 2018-05-16  Richard Biener  <rguenther@suse.de>
12890         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
12891         (dump_stmt_cost): Declare.
12892         (add_stmt_cost): Dump cost we add.
12893         (add_stmt_costs): New function.
12894         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
12895         No longer exported.
12896         (vect_analyze_stmt): Adjust prototype.
12897         (vectorizable_condition): Likewise.
12898         (vectorizable_live_operation): Likewise.
12899         (vectorizable_reduction): Likewise.
12900         (vectorizable_induction): Likewise.
12901         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
12902         cost vector to pass to vectorizable_ and record afterwards.
12903         (vect_model_reduction_cost): Take cost vector argument and adjust.
12904         (vect_model_induction_cost): Likewise.
12905         (vectorizable_reduction): Likewise.
12906         (vectorizable_induction): Likewise.
12907         (vectorizable_live_operation): Likewise.
12908         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
12909         SLP_TREE_NUMBER_OF_VEC_STMTS.
12910         (vect_analyze_slp_cost_1): Remove.
12911         (vect_analyze_slp_cost): Likewise.
12912         (vect_slp_analyze_node_operations): Take visited args and
12913         a target cost vector.  Avoid processing already visited stmt sets.
12914         (vect_slp_analyze_operations): Use a local cost vector to gather
12915         costs and register those of non-discarded instances.
12916         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
12917         (vect_schedule_slp_instance): Remove copying of
12918         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
12919         zero.
12920         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
12921         adding cost.  Record cost entry location.
12922         (vect_prologue_cost_for_slp_op): Function to compute cost of
12923         a constant or invariant generated for SLP vect in the prologue,
12924         split out from vect_analyze_slp_cost_1.
12925         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
12926         (vect_model_promotion_demotion_cost): Likewise.
12927         (vect_model_store_cost): Likewise, make static.
12928         (vect_model_load_cost): Likewise.
12929         (vectorizable_bswap): Add cost vector arg and adjust.
12930         (vectorizable_call): Likewise.
12931         (vectorizable_simd_clone_call): Likewise.
12932         (vectorizable_conversion): Likewise.
12933         (vectorizable_assignment): Likewise.
12934         (vectorizable_shift): Likewise.
12935         (vectorizable_operation): Likewise.
12936         (vectorizable_store): Likewise.
12937         (vectorizable_load): Likewise.
12938         (vectorizable_condition): Likewise.
12939         (vectorizable_comparison): Likewise.
12940         (can_vectorize_live_stmts): Likewise.
12941         (vect_analyze_stmt): Likewise.
12942         (vect_transform_stmt): Adjust calls to vectorizable_*.
12943         * tree-vectorizer.c: Include gimple-pretty-print.h.
12944         (dump_stmt_cost): New function.
12946 2018-05-16  Richard Biener  <rguenther@suse.de>
12948         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
12949         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
12950         * tree-ssa-dse.c: Include tree-ssa-loop.h.
12951         (check_name): New callback.
12952         (dse_classify_store): Track cycles via a visited bitmap of PHI
12953         defs and simplify handling of in-loop and across loop dead stores
12954         and properly fail for loop-variant refs.  Handle byte-tracking with
12955         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
12956         limiting the walk.
12958 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
12960         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
12961         (vect_get_mask_type_for_stmt): Likewise.
12962         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
12963         split out from...
12964         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
12965         to determine the statement's vector type and the vector type that
12966         should be used for calculating nunits.  Deal with cases in which
12967         the type has to be deferred.
12968         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
12969         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
12970         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
12971         (vect_determine_vf_for_stmt): New functions, split out from...
12972         (vect_determine_vectorization_factor): ...here.
12973         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
12974         (vect_get_mask_type_for_stmt): New functions, split out from
12975         vect_determine_vectorization_factor.
12977 2018-05-16  Richard Biener  <rguenther@suse.de>
12979         * tree-cfg.c (verify_gimple_assign_ternary): Properly
12980         verify the [VEC_]COND_EXPR embedded comparison.
12982 2018-05-15  Martin Sebor  <msebor@redhat.com>
12984         PR tree-optimization/85753
12985         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
12986         RECORD_TYPE in addition to ARRAY_TYPE.
12988 2018-05-15  Martin Sebor  <msebor@redhat.com>
12990         PR middle-end/85643
12991         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
12993 2018-05-15  Richard Biener  <rguenther@suse.de>
12995         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
12996         add by_clobber_p one.  Change algorithm to collect all defs
12997         representing uses we need to walk and try reducing them to
12998         a single one before failing.
12999         (dse_dom_walker::dse_optimize_stmt): Adjust.
13001 2018-05-13  Mark Wielaard  <mark@klomp.org>
13003         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
13004         (size_of_loc_descr): Likewise.
13005         (output_loc_operands): Likewise.
13006         (output_loc_operands_raw): Likewise.
13007         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
13008         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
13009         (hash_loc_operands): Likewise.
13010         (compare_loc_operands): Likewise.
13012 2018-05-14  Mark Wielaard  <mark@klomp.org>
13014         * dwarf2out.c (count_index_addrs): New function.
13015         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
13017 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13019         PR tree-optimization/83648
13020         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
13021         return value as malloc candidate.
13023 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13025         PR ipa/85734
13026         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
13027         param as true in call to suggest_attribute.
13029 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
13031         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
13032         -mreadonly-in-sdata.
13034 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13036         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
13037         New pattern.
13038         (aarch64_crypto_aesd_fused): Likewise.
13040 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
13042         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
13043         (movsi_aarch64): Likewise.
13044         (load_pairsi): Likewise.
13045         (load_pairdi): Likewise.
13046         (store_pairsi): Likewise.
13047         (store_pairdi): Likewise.
13048         (load_pairsf): Likewise.
13049         (load_pairdf): Likewise.
13050         (store_pairsf): Likewise.
13051         (store_pairdf): Likewise.
13052         (zero_extend): Likewise.
13053         (trunc): Swap alternatives.
13054         (fcvt_target): Add '?' to prefer w over r.
13056 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
13058         PR target/85756
13059         * config/i386/i386.md: Disallow non-commutative arithmetics in
13060         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
13061         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
13062         in the peephole2 before it.
13064 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
13066         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
13067         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
13068         (ix86_handle_option): Handle -mcldemote.
13069         * config.gcc: New header.
13070         * config/i386/cldemoteintrin.h: New file.
13071         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
13072         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
13073         -mcldemote.
13074         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13075         OPTION_MASK_ISA_CLDEMOTE.
13076         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
13077         (ix86_valid_target_attribute_inner_p): Ditto.
13078         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
13079         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
13080         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
13081         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
13082         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
13083         (cldemote): New.
13084         * config/i386/i386.opt: Add -mcldemote.
13085         * config/i386/x86intrin.h: New header.
13086         * doc/invoke.texi: Add -mcldemote.
13088 2018-05-14  Richard Biener  <rguenther@suse.de>
13090         * doc/match-and-simplify.texi: Adjust :s documentation.
13092 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
13094         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
13095         intended memcpy size.
13096         (REORDER_45): Likewise.
13098 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
13100         * sort.cc: New file.
13101         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
13102         * vec.c (qsort_chk): Use gcc_qsort.
13103         * Makefile.in (OBJS-libcommon): Add sort.o.
13104         (build/sort.o): New target.  Use it...
13105         (BUILD_RTL): ... here, and...
13106         (build/gencfn-macros): ... here, and...
13107         (build/genmatch): ... here.
13109 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13110             Chung-Ju Wu  <jasonwucj@gmail.com>
13112         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
13113         * config/nds32/nds32-graywolf.md: New file.
13114         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
13115         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
13116         pipeline.
13117         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
13118         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
13119         * config/nds32/nds32.md (pipeline_model): Add graywolf.
13120         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
13121         * config/nds32/pipelines.md: Include n15 settings.
13123 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13124             Chung-Ju Wu  <jasonwucj@gmail.com>
13126         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
13127         * config/nds32/nds32-n13.md: New file.
13128         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
13129         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
13130         pipeline.
13131         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
13132         * config/nds32/nds32.md (pipeline_model): Add n13.
13133         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
13134         * config/nds32/pipelines.md: Include n13 settings.
13136 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
13137             Chung-Ju Wu  <jasonwucj@gmail.com>
13139         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
13140         * config/nds32/nds32-n10.md: New file.
13141         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
13142         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
13143         pipeline.
13144         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
13145         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
13146         * config/nds32/nds32.md (pipeline_model): Add n10.
13147         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
13148         * config/nds32/pipelines.md: Include n10 settings.
13150 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
13151             Kito Cheng  <kito.cheng@gmail.com>
13152             Chung-Ju Wu  <jasonwucj@gmail.com>
13154         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
13155         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
13156         Add enum values for DSP extension instructions.
13157         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
13158         New constraints.
13159         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
13160         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
13161         New code iterators.
13162         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
13163         * config/nds32/nds32-dspext.md: New file for DSP implementation.
13164         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
13165         * config/nds32/nds32-intrinsic.md: Likewise.
13166         * config/nds32/nds32_intrinsic.h: Likewise.
13167         * config/nds32/nds32-md-auxiliary.c: Likewise.
13168         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
13169         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
13170         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
13171         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
13172         * config/nds32/nds32-protos.h: New declarations for DSP extension.
13173         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
13174         TYPE_DMAC in switch statement.
13175         * config/nds32/nds32.c: New checking and implementation for DSP
13176         extension instructions.
13177         * config/nds32/nds32.h: Likewise.
13178         * config/nds32/nds32.md: Likewise.
13179         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
13180         * config/nds32/predicates.md: Implement new predicates for DSP
13181         extension.
13183 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
13185         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
13186         Reformat alternatives and attributes so it is easier to identify
13187         which constraints/attributes go with which instruction.
13188         (mov<mode>_hardfloat32, FMOVE64): Likewise.
13189         (mov<mode>_softfloat32, FMOVE64): Likewise.
13190         (mov<mode>_hardfloat64, FMOVE64): Likewise.
13191         (mov<mode>_softfloat64, FMOVE64): Likewise.
13193 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13195         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13196         subsection.
13197         (Basic PowerPC Built-in Functions): The new name of the
13198         subsection previously known as "PowerPC Built-in Functions".
13199         (Basic PowerPC Built-in Functions Available on all Configurations):
13200         New subsubsection.
13201         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
13202         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
13203         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
13204         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
13206 2018-05-11  Martin Jambor  <mjambor@suse.cz>
13208         PR ipa/85655
13209         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13210         single const.
13212 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
13214         PR target/85733
13215         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
13217 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
13219         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
13220         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
13221         (ix86_handle_option): Handle -mwaitpkg.
13222         * config.gcc: New header.
13223         * config/i386/cpuid.h (bit_WAITPKG): New bit.
13224         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
13225         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
13226         function type.
13227         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13228         OPTION_MASK_ISA_WAITPKG.
13229         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
13230         (ix86_option_override_internal): Add PTA_WAITPKG.
13231         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
13232         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
13233         IX86_BUILTIN_TPAUSE.
13234         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
13235         __builtin_ia32_umwait and __builtin_ia32_tpause.
13236         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
13237         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
13238         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
13239         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
13240         UNSPECV_TPAUSE): New.
13241         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
13242         * config/i386/i386.opt: Add -mwaitpkg.
13243         * config/i386/waitpkgintrin.h: New file.
13244         * config/i386/x86intrin.h: New header.
13245         * doc/invoke.texi: Add -mwaitpkg.
13247 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
13249         PR target/85606
13250         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
13251         equivalent.
13252         (cortex-m0): Use armv6s-m isa.
13253         (cortex-m0plus): Likewise.
13254         (cortex-m1): Likewise.
13255         (cortex-m0.small-multiply): Likewise.
13256         (cortex-m0plus.small-multiply): Likewise.
13257         (cortex-m1.small-multiply): Likewise.
13259 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
13260             Jakub Jelinek  <jakub@redhat.com>
13262         PR tree-optimization/85692
13263         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
13264         source permute as well.
13266 2018-05-11  Martin Liska  <mliska@suse.cz>
13268         PR sanitizer/85556
13269         * doc/extend.texi: Document LLVM style format for no_sanitize
13270         attribute.
13272 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
13274         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
13275         mode_supports_vsx_dform_quad to mode_supports_dq_form.
13276         (mode_supports_vsx_dform_quad): Likewise.
13277         (mode_supports_vmx_dform): Move these functions to be next to the
13278         other mode_supports functions.
13279         (mode_supports_dq_form): Likewise.
13280         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
13281         mode_supports_dq_form.
13282         (reg_offset_addressing_ok_p): Likewise.
13283         (offsettable_ok_by_alignment): Likewise.
13284         (rs6000_legitimate_offset_address_p): Likewise.
13285         (legitimate_lo_sum_address_p): Likewise.
13286         (rs6000_legitimize_address): Likewise.
13287         (rs6000_legitimize_reload_address): Likewise.
13288         (rs6000_secondary_reload_inner): Likewise.
13289         (rs6000_preferred_reload_class): Likewise.
13290         (rs6000_output_move_128bit): Likewise.
13292 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
13294         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
13295         Generate SImode target register for null target.
13296         <case IX86_BUILTIN_XGETBV>: Ditto.
13297         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
13298         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
13300 2018-05-10  Carl Love  <cel@us.ibm.com>
13302         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
13303         dcbtt and dcbtstt if operands[2] is 0.
13305 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
13307         PR target/85693
13308         * config/i386/sse.md (usadv64qi): New expander.
13310 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
13312         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
13313         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
13314         -maltivec=be support.
13315         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
13316         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
13317         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
13318         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
13319         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
13320         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
13321         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
13322         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
13323         altivec_vsumsws): Adjust.
13324         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
13325         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
13326         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
13327         support.
13328         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
13329         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
13330         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
13331         (altivec_lve<VI_char>x): Delete expand.
13332         (*altivec_lve<VI_char>x_internal): Rename to...
13333         (altivec_lve<VI_char>x): ... this.
13334         (altivec_lvxl_<mode>): Delete expand.
13335         (*altivec_lvxl_<mode>_internal): Rename to ...
13336         (altivec_lvxl_<mode>): ... this.
13337         (altivec_stvxl_<mode>): Delete expand.
13338         (*altivec_stvxl_<mode>_internal): Rename to ...
13339         (altivec_stvxl_<mode>): ... this.
13340         (altivec_stve<VI_char>x): Delete expand.
13341         (*altivec_stve<VI_char>x_internal): Rename to ...
13342         (altivec_stve<VI_char>x): ... this.
13343         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
13344         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
13345         reduc_plus_scal_<mode>): Adjust.
13346         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
13347         comment.
13348         (rs6000_cpu_cpp_builtins): Adjust.
13349         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
13350         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
13351         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
13352         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
13353         -maltivec=be support.
13354         (rs6000_split_vec_extract_var): Adjust.
13355         (rs6000_split_v4si_init): Adjust.
13356         (swap_selector_for_mode): Delete.
13357         (altivec_expand_lvx_be, altivec_expand_stvx_be,
13358         altivec_expand_stvex_be): Delete.
13359         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
13360         -maltivec=be support.
13361         (rs6000_gimple_fold_builtin): Ditto.
13362         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
13363         Adjust.
13364         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
13365         (TARGET_DIRECT_MOVE_64BIT): Adjust.
13366         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
13367         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
13368         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
13369         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
13370         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
13371         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
13372         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
13373         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
13374         anonymous split): Adjust.
13375         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
13376         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
13378 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
13380         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
13381         when --with-gxx-include-dir is also specified.
13382         * configure: Regenerate.
13384 2018-05-09  Jim Wilson  <jimw@sifive.com>
13386         PR target/84797
13387         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
13388         * config/riscv/t-withmultilib: New.
13389         * config/riscv/withmultilib.h: New.
13390         * doc/install.texi: Document RISC-V --with-multilib-list support.
13392 2018-05-09  Richard Biener  <rguenther@suse.de>
13394         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
13395         vector.
13396         (vect_bb_vectorization_profitable_p): Adjust.  Compute
13397         actual scalar cost using the cost vector and the add_stmt_cost
13398         machinery.
13400 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13402         PR rtl-optimization/85645
13403         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
13404         in the REG_CFA_REGISTER note for LR, don't leave it empty.
13406 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13408         PR rtl-optimization/85645
13409         * shrink-wrap.c (spread_components): Return a boolean saying if
13410         anything was changed.
13411         (try_shrink_wrapping_separate): Iterate spread_components until
13412         nothing changes anymore.
13414 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13416         PR rtl-optimization/85645
13417         * regrename.c (build_def_use): Also kill the chains that include the
13418         destination of a REG_CFA_REGISTER note.
13420 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
13422         PR rtl-optimization/85645
13423         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
13424         insn that has a REG_CFA_REGISTER note.
13426 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
13428         * cfgexpand.c (expand_clobber): New function.
13429         (expand_gimple_stmt_1): Use it.
13430         * tree-vect-stmts.c (vect_clobber_variable): New function,
13431         split out from...
13432         (vectorizable_simd_clone_call): ...here.
13433         (vectorizable_store): Emit a clobber either side of an
13434         IFN_STORE_LANES sequence.
13435         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
13437 2018-05-09  Tom de Vries  <tom@codesourcery.com>
13439         PR target/85626
13440         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
13441         (define_insn "trap_if_false"): Add exit after trap.
13443 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
13445         PR rtl-optimization/85638
13446         * bb-reorder.c: Include common/common-target.h.
13447         (create_forwarder_block): New function extracted from...
13448         (fix_up_crossing_landing_pad): ...here.  Rename into...
13449         (dw2_fix_up_crossing_landing_pad): ...this.
13450         (sjlj_fix_up_crossing_landing_pad): New function.
13451         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
13452         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
13453         from both partitions and exit the loop after one iteration.
13455 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13457         Revert:
13458         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13459         subsection.
13460         (Basic PowerPC Built-in Functions): The new name of the
13461         subsection previously known as "PowerPC Built-in Functions".
13462         (Basic PowerPC Built-in Functions Available on all Configurations):
13463         New subsubsection.
13464         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13465         subsubsection.
13466         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13467         subsubsection.
13468         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13469         subsubsection.
13470         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13471         subsubsection.
13473 2018-05-08  Jim Wilson  <jimw@sifive.com>
13475         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
13476         (LD_EMUL_SUFFIX): New.
13477         (LINK_SPEC): Use it.
13479 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
13481         * doc/extend.texi (PowerPC Built-in Functions): Rename this
13482         subsection.
13483         (Basic PowerPC Built-in Functions): The new name of the
13484         subsection previously known as "PowerPC Built-in Functions".
13485         (Basic PowerPC Built-in Functions Available on all Configurations):
13486         New subsubsection.
13487         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13488         subsubsection.
13489         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13490         subsubsection.
13491         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13492         subsubsection.
13493         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13494         subsubsection.
13496 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
13498         PR target/85683
13499         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
13500         after cmpelim optimization.
13502 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
13504         * config.gcc: Support "goldmont".
13505         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
13506         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
13507         PROCESSOR_GOLDMONT.
13508         * config/i386/i386.c (m_GOLDMONT): Define.
13509         (processor_target_table): Add "goldmont".
13510         (PTA_GOLDMONT): Define.
13511         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
13512         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
13513         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
13514         (fold_builtin_cpu): Add "goldmont".
13515         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
13516         (ix86_option_override_internal): Add "goldmont".
13517         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
13518         (processor_type): Add PROCESSOR_GOLDMONT.
13519         * config/i386/i386.md: Add CPU "glm".
13520         * config/i386/glm.md: New file.
13521         * config/i386/x86-tune.def: Add m_GOLDMONT.
13522         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
13524 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
13526         PR target/85572
13527         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
13528         E_V4DImode.
13529         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
13530         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
13531         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
13533         PR target/85317
13534         * config/i386/i386.c (ix86_fold_builtin): Handle
13535         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
13537         PR target/85480
13538         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
13539         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
13541 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
13543         PR target/85658
13544         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
13545         (check_arch): Likewise.
13546         (check_fpu): Return the result rather than printing it.
13547         (end arch): Fix operator precedence.
13548         (end cpu): Likewise.
13549         (END): Print the result from check_fpu.
13551 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
13552             Alan Hayward  <alan.hayward@arm.com>
13553             David Sherwood  <david.sherwood@arm.com>
13555         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
13556         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
13557         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
13558         (*fcmuo<mode>_and): New patterns.
13560 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
13562         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
13563         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
13564         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
13565         (cmp_op, sve_imm_con): New code attributes.
13566         (SVE_COND_INT_CMP, imm_con): Delete.
13567         (cmp_op): Remove above unspecs from int attribute.
13568         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
13569         to...
13570         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
13571         comparison-specific unspecs.
13572         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
13573         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
13574         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
13575         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
13576         (*vec_fcm<cmp_op><mode>): Rename to...
13577         (*fcm<cmp_op><mode>): ...this and adjust likewise.
13578         (*vec_fcmuo<mode>): Rename to...
13579         (*fcmuo<mode>): ...this and adjust likewise.
13580         (*pred_fcm<cmp_op><mode>): New pattern.
13581         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
13582         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
13583         functions.
13584         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
13585         and UNORDERED.
13586         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
13587         (aarch64_emit_sve_predicated_cond): New function.
13588         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
13589         (aarch64_emit_unspec_cond_or): Replace with...
13590         (aarch64_emit_sve_or_conds): ...this new function.  Use
13591         aarch64_emit_sve_ptrue_op for the individual comparisons and
13592         aarch64_emit_binop to OR them together.
13593         (aarch64_emit_inverted_unspec_cond): Replace with...
13594         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
13595         aarch64_emit_sve_ptrue_op for the comparison and
13596         aarch64_emit_unop to invert the result.
13597         (aarch64_expand_sve_vec_cmp_float): Update after the above
13598         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
13600 2018-05-07  Nathan Sidwell  <nathan@acm.org>
13602         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
13603         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
13604         (Backwards Compatibility): Likewise.
13606 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13608         PR bootstrap/85681
13609         Revert:
13610         2018-05-07  Luis Machado  <luis.machado@linaro.org>
13612         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13613         <prefetch_dynamic_strides>: New const bool field.
13614         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13615         prefetch_dynamic_strides.
13616         (exynosm1_prefetch_tune): Likewise.
13617         (thunderxt88_prefetch_tune): Likewise.
13618         (thunderx_prefetch_tune): Likewise.
13619         (thunderx2t99_prefetch_tune): Likewise.
13620         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13621         to false.
13622         (aarch64_override_options_internal): Update to set
13623         PARAM_PREFETCH_DYNAMIC_STRIDES.
13624         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13625         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13626         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13627         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13628         prefetch-dynamic-strides setting.
13630         2018-05-07  Luis Machado  <luis.machado@linaro.org>
13632         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13633         <minimum_stride>: New const int field.
13634         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13635         minimum_stride field.
13636         (exynosm1_prefetch_tune): Likewise.
13637         (thunderxt88_prefetch_tune): Likewise.
13638         (thunderx_prefetch_tune): Likewise.
13639         (thunderx2t99_prefetch_tune): Likewise.
13640         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13641         (aarch64_override_options_internal): Update to set
13642         PARAM_PREFETCH_MINIMUM_STRIDE.
13643         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13644         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13645         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13646         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13647         stride is constant and is below the minimum stride threshold.
13649 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13651         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
13652         to 512.
13654 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13656         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13657         <prefetch_dynamic_strides>: New const bool field.
13658         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13659         prefetch_dynamic_strides.
13660         (exynosm1_prefetch_tune): Likewise.
13661         (thunderxt88_prefetch_tune): Likewise.
13662         (thunderx_prefetch_tune): Likewise.
13663         (thunderx2t99_prefetch_tune): Likewise.
13664         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
13665         to false.
13666         (aarch64_override_options_internal): Update to set
13667         PARAM_PREFETCH_DYNAMIC_STRIDES.
13668         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13669         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13670         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13671         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13672         prefetch-dynamic-strides setting.
13674 2018-05-07  Luis Machado  <luis.machado@linaro.org>
13676         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13677         <minimum_stride>: New const int field.
13678         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13679         minimum_stride field.
13680         (exynosm1_prefetch_tune): Likewise.
13681         (thunderxt88_prefetch_tune): Likewise.
13682         (thunderx_prefetch_tune): Likewise.
13683         (thunderx2t99_prefetch_tune): Likewise.
13684         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
13685         (aarch64_override_options_internal): Update to set
13686         PARAM_PREFETCH_MINIMUM_STRIDE.
13687         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13688         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13689         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13690         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13691         stride is constant and is below the minimum stride threshold.
13693 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
13695         PR c++/85659
13696         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
13697         the type is addressable.  Don't force op into register if it has
13698         BLKmode.
13700 2018-05-05  Roland McGrath  <mcgrathr@google.com>
13702         PR other/77609
13703         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
13704         any section for which we don't know a specific type it should have,
13705         regardless of name.  Previously this was done only for the exact
13706         names ".init_array", ".fini_array", and ".preinit_array".
13707         (default_elf_asm_named_section): Add comment about
13708         relationship with default_section_type_flags and SECTION_NOTYPE.
13709         (get_section): Don't consider it a type conflict if one side has
13710         SECTION_NOTYPE and the other doesn't, as long as neither has the
13711         SECTION_BSS et al used in the default_section_type_flags logic.
13713 2018-05-05  Tom de Vries  <tom@codesourcery.com>
13715         PR target/85653
13716         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
13717         (workaround_barsyncs): New function.
13718         (nvptx_reorg): Use workaround_barsyncs.
13719         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
13720         (define_expand "nvptx_membar_cta"): New define_expand.
13721         (define_insn "*nvptx_membar_cta"): New insn.
13723 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
13725         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
13726         To improve optimization opportunities.
13727         * builtin-types.def: The new needed builtin types for the above.
13729 2018-05-04  Richard Biener  <rguenther@suse.de>
13731         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
13732         * gimple-ssa-store-merging.c
13733         (imm_store_chain_info::output_merged_store): Remove redundant create,
13734         release split_store vector contents on failure.
13735         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
13736         scalar stmt vector on cache hit.
13738 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
13740         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
13741         Xilinx FP support.
13742         * config.gcc (powerpc-xilinx-eabi*): Remove.
13743         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13744         support.
13745         (fusion_addis_mem_combo_load): Ditto.
13746         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
13747         FP support.
13748         (rs6000_cpu_cpp_builtins): Ditto.
13749         * config/rs6000/rs6000-linux.c
13750         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
13751         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
13752         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
13753         support.
13754         (rs6000_setup_reg_addr_masks): Ditto.
13755         (rs6000_init_hard_regno_mode_ok): Ditto.
13756         (rs6000_option_override_internal): Ditto.
13757         (legitimate_lo_sum_address_p): Ditto.
13758         (rs6000_legitimize_address): Ditto.
13759         (rs6000_legitimize_reload_address): Ditto.
13760         (rs6000_legitimate_address_p): Ditto.
13761         (abi_v4_pass_in_fpr): Ditto.
13762         (setup_incoming_varargs): Ditto.
13763         (rs6000_gimplify_va_arg): Ditto.
13764         (rs6000_split_multireg_move): Ditto.
13765         (rs6000_savres_strategy): Ditto.
13766         (rs6000_emit_prologue_components): Ditto.
13767         (rs6000_emit_epilogue_components): Ditto.
13768         (rs6000_emit_prologue): Ditto.
13769         (rs6000_emit_epilogue): Ditto.
13770         (rs6000_elf_file_end): Ditto.
13771         (rs6000_function_value): Ditto.
13772         (rs6000_libcall_value): Ditto.
13773         * config/rs6000/rs6000.h: Ditto.
13774         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
13775         (TARGET_MINMAX): ... this.  New.
13776         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
13777         * config/rs6000/rs6000.md: Remove Xilinx FP support.
13778         (*movsi_internal1_single): Delete.
13779         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
13780         mfpu=, mxilinx-fpu): Delete.
13781         * config/rs6000/singlefp.h: Delete.
13782         * config/rs6000/sysv4.h: Remove Xilinx FP support.
13783         * config/rs6000/t-rs6000: Ditto.
13784         * config/rs6000/t-xilinx: Delete.
13785         * config/rs6000/titan.md: Adjust for fp_type removal.
13786         * config/rs6000/vsx.md: Remove Xilinx FP support.
13787         (VStype_simple): Delete.
13788         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
13789         * config/rs6000/xfpu.h: Delete.
13790         * config/rs6000/xfpu.md: Delete.
13791         * config/rs6000/xilinx.h: Delete.
13792         * config/rs6000/xilinx.opt: Delete.
13793         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
13794         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
13796 2018-05-04  Tom de Vries  <tom@codesourcery.com>
13798         PR libgomp/85639
13799         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
13800         if ignore == 0.
13802 2018-05-04  Richard Biener  <rguenther@suse.de>
13804         PR middle-end/85627
13805         * tree-complex.c (update_complex_assignment): We are always in SSA form.
13806         (expand_complex_div_wide): Likewise.
13807         (expand_complex_operations_1): Likewise.
13808         (expand_complex_libcall): Preserve EH info of the original stmt.
13809         (tree_lower_complex): Handle removed blocks.
13810         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
13811         on complex multiplication and division libcall builtins.
13813 2018-05-04  Richard Biener  <rguenther@suse.de>
13815         PR middle-end/85574
13816         * fold-const.c (negate_expr_p): Restrict negation of operand
13817         zero of a division to when we know that can happen without
13818         overflow.
13819         (fold_negate_expr_1): Likewise.
13821 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
13823         PR libstdc++/85466
13824         * real.h (real_nextafter): Declare.
13825         * real.c (real_nextafter): New function.
13826         * fold-const-call.c (fold_const_nextafter): New function.
13827         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
13828         CASE_CFN_NEXTTOWARD.
13829         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
13830         even when arg1_mode is different from arg0_mode.
13832 2018-05-03  Nathan Sidwell  <nathan@acm.org>
13834         * doc/extend.texi (Deprecated Features): Remove
13835         -ffriend-injection.
13836         (Backwards Compatibility): Likewise.
13837         * doc/invoke.texi (C++ Language Options): Likewise.
13838         (C++ Dialect Options): Likewise.
13840 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
13842         PR target/85530
13843         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
13844         _mm512_mask_mullox_epi64): New intrinsics.
13846 2018-05-03  Tom de Vries  <tom@codesourcery.com>
13848         PR testsuite/85106
13849         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13850         dump files): Add offload-tree.
13852 2018-05-03  Richard Biener  <rguenther@suse.de>
13854         PR tree-optimization/85615
13855         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
13856         to loops not nested in BBs loop father to avoid creating multi-entry
13857         loops.
13859 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13861         PR tree-optimization/70291
13862         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
13863         arguments.  Change return type to tree.  Emit libcall as a new
13864         statement rather than replacing existing one when inplace_p is true.
13865         (expand_complex_multiplication_components): New function.
13866         (expand_complex_multiplication): Expand floating-point complex
13867         multiplication using the above.
13868         (expand_complex_division): Rename inner_type parameter to type.
13869         Update expand_complex_libcall call-site.
13870         (expand_complex_operations_1): Update expand_complex_multiplication
13871         and expand_complex_division call-sites.
13873 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
13875         PR target/85582
13876         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13877         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13878         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
13879         the highest significant bit of the shift count mask is clear.  In
13880         check whether and[sq]i3 is needed verify that all significant bits
13881         of the shift count other than the highest are set.
13883 2018-05-02  Tom de Vries  <tom@codesourcery.com>
13885         PR libgomp/82428
13886         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
13887         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
13888         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
13889         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
13890         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
13891         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
13892         __builtin_goacc_parlevel_size.
13894 2018-05-02  Richard Biener  <rguenther@suse.de>
13896         PR tree-optimization/85597
13897         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
13898         do not use split vect_get_vec_defs call but call vect_get_slp_defs
13899         directly.
13901 2018-05-02  Tom de Vries  <tom@codesourcery.com>
13903         PR testsuite/85106
13904         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13905         dump files): Add ltrans-tree.
13907 2018-05-02  Tom de Vries  <tom@codesourcery.com>
13909         PR testsuite/85106
13910         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
13911         dump files): Add wpa-ipa.
13913 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
13915         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
13916         powerpc*-*-linux*paired* target.
13917         * config/rs6000/750cl.h: Delete.
13918         * config/rs6000/paired.h: Delete.
13919         * config/rs6000/paired.md: Delete.
13920         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
13921         float support.
13922         * config/rs6000/rs6000-builtin.def: Remove paired float support.
13923         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
13924         comment.  Remove paired float support.
13925         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
13926         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
13927         VECTOR_PAIRED.
13928         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13929         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
13930         declarations.
13931         * config/rs6000/rs6000.c: Remove paired float support.
13932         (paired_expand_vector_init, paired_expand_vector_move,
13933         paired_emit_vector_compare, paired_emit_vector_cond_expr,
13934         (paired_expand_lv_builtin, paired_expand_stv_builtin,
13935         paired_expand_builtin, paired_expand_predicate_builtin,
13936         paired_init_builtins): Delete.
13937         * config/rs6000/rs6000.h: Remove paired float support.
13938         * config/rs6000/rs6000.md: Remove paired float support.
13939         (move_from_CR_ov_bit): Delete.
13940         * config/rs6000/rs6000.opt (mpaired): Delete.
13941         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
13942         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
13944 2018-05-02  Richard Biener  <rguenther@suse.de>
13946         PR middle-end/85567
13947         * gimplify.c (gimplify_save_expr): When in SSA form allow
13948         SAVE_EXPRs to compute to SSA vars.
13950 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
13952         PR target/85582
13953         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
13954         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
13955         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
13956         clobber operands[2], instead use a new pseudo.  Formatting fixes.
13958 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
13960         PR tree-optimization/85586
13961         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
13962         exit early for statements in the same group if the accesses are
13963         not strided.
13965 2018-05-02  Tom de Vries  <tom@codesourcery.com>
13967         PR lto/85451
13968         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
13969         error message.
13971 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
13973         PR tree-optimization/85143
13974         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
13976 2018-05-01  Tom de Vries  <tom@codesourcery.com>
13978         PR lto/85451
13979         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
13980         not found" error message.
13982 2018-05-01  Tom de Vries  <tom@codesourcery.com>
13984         PR other/83786
13985         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
13986         * vec.c (test_ordered_remove_if): New function.
13987         (vec_c_tests): Call test_ordered_remove_if.
13988         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
13989         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
13990         * tree-vect-patterns.c (vect_pattern_recog_1): Use
13991         VEC_ORDERED_REMOVE_IF.
13993 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13995         PR tree-optimization/82665
13996         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
13997         pointer subtraction where arguments come from a memchr call.
13999 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
14001         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
14002         --push-state --as-needed and --pop-state instead of --as-needed and
14003         --no-as-needed if ld supports it.
14004         * configure: Regenerated.
14006         PR web/85578
14007         * doc/install.texi2html: Replace _002d with - and _002a with * in
14008         generated html files using sed.
14010 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14012         PR c++/85523
14013         * gcc-rich-location.c (blank_line_before_p): New function.
14014         (use_new_line): New function.
14015         (gcc_rich_location::add_fixit_insert_formatted): New function.
14016         * gcc-rich-location.h
14017         (gcc_rich_location::add_fixit_insert_formatted): New function.
14019 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14021         * selftest.c (assert_streq): Rename "expected" and "actual" to
14022         "val1" and "val2".  Extend NULL-handling to cover both inputs
14023         symmetrically, while still requiring both to be non-NULL for a pass.
14024         * selftest.h (assert_streq): Rename "expected" and "actual" to
14025         "val1" and "val2".
14026         (ASSERT_EQ): Likewise.
14027         (ASSERT_EQ_AT): Likewise.
14028         (ASSERT_KNOWN_EQ): Likewise.
14029         (ASSERT_KNOWN_EQ_AT): Likewise.
14030         (ASSERT_NE): Likewise.
14031         (ASSERT_MAYBE_NE): Likewise.
14032         (ASSERT_MAYBE_NE_AT): Likewise.
14033         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
14034         the assertion to pass.
14035         (ASSERT_STREQ_AT): Likewise.
14037 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
14039         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
14040         interaction with -pie.
14042 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14044         * selftest.h: Fix alphabetization of per-source-file selftest
14045         declarations.
14047 2018-04-30  Jason Merrill  <jason@redhat.com>
14049         PR c++/61982 - dead stores to destroyed objects.
14050         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
14051         of clobber.
14053 2018-04-30  Jason Merrill  <jason@redhat.com>
14055         * tree.c (build_clobber): New.
14056         * tree.h: Declare it.
14057         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
14059 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14061         * diagnostic-show-locus.c (layout::layout): Update for
14062         location_get_source_line returning a char_span.
14063         (struct char_span): Move to input.h.
14064         (struct correction): Update for fields in char_span becoming
14065         private.
14066         (struct source_line): Update for location_get_source_line
14067         returning a char_span.
14068         (layout::print_line): Likewise.
14069         * edit-context.c (edited_file::print_content): Likewise.
14070         (edited_file::print_diff_hunk): Likewise.
14071         (edited_file::print_run_of_changed_lines): Likewise.
14072         (edited_file::get_num_lines): Likewise.
14073         (edited_line::edited_line): Likewise.
14074         * final.c (asm_show_source): Likewise.
14075         * input.c (location_get_source_line): Convert return type
14076         from const char * to char_span, losing the final "line_len"
14077         param.
14078         (dump_location_info): Update for the above.
14079         (get_substring_ranges_for_loc): Likewise.  Use a char_span
14080         when handling the literal within the line.
14081         (test_reading_source_line): Update for location_get_source_line
14082         returning a char_span.
14083         * input.h (class char_span): Move here from
14084         diagnostic-show-locus.c, converting from a struct to a class.
14085         Make data members private.
14086         (char_span::operator bool): New.
14087         (char_span::length): New.
14088         (char_span::get_buffer): New.
14089         (char_span::operator[]): New.
14090         (char_span::subspan): Make const.
14091         (char_span::xstrdup): New.
14092         (location_get_source_line): Convert return type from const char *
14093         to char_span, losing the final "line_size" param.
14095 2018-04-30  Jan Hubicka  <jh@suse.cz>
14097         * lto-wrapper.c (ltrans_priorities): New static var.
14098         (cmp_priority): New.
14099         (run_gcc): Read priorities and if doing parallel build order
14100         the Makefile by them.
14102 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
14104         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
14106 2018-04-30  Richard Biener  <rguenther@suse.de>
14108         * tree-cfg.c (verify_address): Remove base argument, add
14109         flag whether to check TREE_ADDRESSABLE and do that.
14110         (verify_expr): Remove.
14111         (verify_types_in_gimple_reference): Add pieces from verify_expr.
14112         (verify_gimple_assign_single): Likewise.
14113         (verify_gimple_switch): Likewise.
14114         (verify_expr_location_1): Dereference tp once.  Add (disabled)
14115         piece from verify_expr.
14116         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
14118 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14120         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
14122 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14124         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
14125         (small_data_pattern): Likewise.
14126         (arc_rewrite_small_data): Likewise.
14127         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
14128         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
14129         (get_symbol_alignment): New function.
14130         (legitimate_small_data_address_p): Likewise.
14131         (legitimate_scaled_address): Update, call
14132         legitimate_small_data_address_p.
14133         (output_sdata): New static variable.
14134         (arc_print_operand): Update how we handle small data operands.
14135         (arc_print_operand_address): Likewise.
14136         (arc_legitimate_address_p): Update, use
14137         legitimate_small_data_address_p.
14138         (arc_rewrite_small_data_p): Remove.
14139         (arc_rewrite_small_data_1): Likewise.
14140         (arc_rewrite_small_data): Likewise.
14141         (small_data_pattern): Likewise.
14142         (compact_sda_memory_operand): Update to use
14143         legitimate_small_data_address_p and get_symbol_alignment.
14144         (prepare_move_operands): Don't rewite sdata pattern.
14145         (prepare_extend_operands): Remove.
14146         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
14147         pattern.
14148         (zero_extendqisi2): Likewise.
14149         (zero_extendhisi2): Likewise.
14150         (extendqihi2): Likewise.
14151         (extendqisi2): Likewise.
14152         (extendhisi2): Likewise.
14153         (addsi3): Likewise.
14154         (subsi3): Likewise.
14155         (andsi3): Likewise.
14156         * config/arc/constraints.md (Usd): Change it to memory constraint.
14158 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
14160         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
14161         as source of std instructions.
14162         * config/arc/arc.md (movsi_insn): Update pattern predicate to
14163         allow 6-bit constants as source for store instructions.
14164         (movdi_insn): Update instruction pattern to allow 6-bit constants
14165         as source for store instructions.
14167 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
14169         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
14171 2018-04-30  Nathan Sidwell  <nathan@acm.org>
14172             Sandra Loosemore <sandra@codesourcery.com>
14174         * dumpfile.c (dump_open): Allow '-' for stdout.
14175         * doc/invoke.texi (Developer Options): Document dump filename
14176         determination early.  Document stdin/stdout selection.
14178 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
14180         Microblaze Target: PIC data text relative
14182         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
14183         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
14184         Add declaration.
14185         * config/microblaze/microblaze.h (microblaze_constant_address_p):
14186         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
14187         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
14188         New addressing mode for data-text relative position indepenedent code.
14189         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
14190         'ADDRESS_SYMBOLIC_TXT_REL'.
14191         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
14192         (microblaze_legitimate_pic_operand): Exclude function calls from
14193         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
14194         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
14195         addresses cases.
14196         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14197         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
14198         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
14199         for 'address + offset'.
14200         (microblaze_expand_prologue): Add new function prologue call for
14201         'r20' assignation.
14202         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
14203         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
14204         table in case of TARGET_PIC_DATA_TEXT_REL.
14205         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
14206         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
14207         Add new macros 'UNSPEC_TEXT',
14208         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
14209         + exclude function calls from 'UNSPEC_PLT' in case of data text
14210         relative mode.
14211         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14212         new target hook for generating address diff vector tables in case of
14213         flag_pic.
14214         * doc/tm.texi : Regenerate.
14215         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
14216         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
14217         of addr diff vector generation.
14218         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
14219         target hook definition.
14220         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
14221         Add default function for generate_pic_addr_diff_vec -> flag_pic.
14222         * doc/invoke.texi (Add new pic option): Add new microblaze pic
14223         option for data text relative.
14225 2018-04-30  Richard Biener  <rguenther@suse.de>
14227         * tree-chrec.h (evolution_function_is_constant_p): Remove
14228         redundant check.
14229         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14231 2018-04-30  Richard Biener  <rguenther@suse.de>
14233         PR bootstrap/85571
14234         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
14236 2018-04-30  Richard Biener  <rguenther@suse.de>
14238         PR tree-optimization/28364
14239         PR tree-optimization/85275
14240         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
14241         copying first exit test.
14243 2018-04-28  Mark Wielaard  <mark@klomp.org>
14245         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
14246         dwarf_version >= 5.
14247         (dwarf_AT): Handle DW_AT_addr_base.
14248         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
14250 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
14252         PR target/84431
14253         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
14254         (*ashl<dwi>3_doubleword_mask_1): Ditto.
14255         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
14256         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
14258 2018-04-28  Richard Biener  <rguenther@suse.de>
14260         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
14261         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
14262         to reflect use.  Only add interesting stmts.
14264 2018-04-27  Martin Jambor  <mjambor@suse.cz>
14266         PR ipa/85549
14267         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
14268         the jump function allows for passing through aggregate values.
14270 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
14272         * input.h (in_system_header_at): Convert from macro to inline
14273         function.
14274         (from_macro_expansion_at): Likewise.
14275         (from_macro_definition_at): Likewise.
14277 2018-04-27  Jeff Law  <law@redhat.com>
14279         * config.gcc: Mark tile* targets as deprecated/obsolete.
14281 2018-04-27  Richard Biener  <rguenther@suse.de>
14283         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
14284         fix for ILP32.
14286 2018-04-27  Richard Biener  <rguenther@suse.de>
14288         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
14290 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
14292         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
14293         with Yd constraint. Set "preferred_for_speed" attribute from
14294         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
14295         with Yd constraint.
14296         (*movdi_internal): Ditto.
14297         (movti_interunit splitters): Remove
14298         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
14299         (movdi_interunit splitters): Ditto.
14300         * config/i386/constraints.md (Ye): Remove.
14301         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
14303 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14305         PR target/85512
14306         * config/aarch64/constraints.md (Usg): Limit to 31.
14307         (Usj): Limit to 63.
14309 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
14311         PR tree-optimization/85529
14312         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
14313         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
14314         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
14315         zero extension or masking of the MSB bit.
14316         (optimize_range_tests): Add FIRST_BB argument, pass it through
14317         to optimize_range_tests_var_bound.
14318         (maybe_optimize_range_tests, reassociate_bb): Adjust
14319         optimize_range_tests callers.
14321 2018-04-26  Richard Biener  <rguenther@suse.de>
14322             Jakub Jelinek  <jakub@redhat.com>
14324         * cgraph.h (symbol_table): Just declare debug method here.
14325         * symtab.c (symbol_table::debug): Define.
14327 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
14329         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
14331 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
14333         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
14334         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
14335         (*movdi_internal): Substitute Yi and Yj constraint with x
14336         and Ym and Yn constraint with y constraint.  Update "isa"
14337         attribute and set "preferred_for_speed" attribute from
14338         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
14339         (*movsi_internal): Ditto.
14340         (*movdf_internal): Ditto.
14341         (*movsf_internal): Ditto.
14342         (*zero_extendsidi2): Ditto.
14343         * config/i386/sse.md (vec_set<mode>_0): Ditto.
14344         (sse2_loadld): Ditto.
14345         (*vec_extract<ssevecmodelower>_0): Ditto.
14346         (*vec_extractv4si_0_zext_sse4): Ditto.
14347         (vec_concatv2di): Ditto.
14348         (*vec_dup<mode>): Ditto.
14349         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
14350         * config/i386/constraints.md (Yi): Remove.
14351         (Yj): Remove.
14352         (Ym): Remove.
14353         (Yn): Remove.
14355 2018-04-26  Nathan Sidwell  <nathan@acm.org>
14357         * dumpfile.c (dump_open): New.
14358         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
14359         (dump_finish): Detect stdio/stderr by value not name.
14361 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
14363         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
14365 2018-04-26  Tom de Vries  <tom@codesourcery.com>
14367         PR target/84952
14368         * config/nvptx/nvptx.c (verify_neutering_jumps)
14369         (verify_neutering_labels): New function
14370         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
14372 2018-04-26  Tom de Vries  <tom@codesourcery.com>
14374         PR target/84025
14375         * config/nvptx/nvptx.c (needs_neutering_p): New function.
14376         (nvptx_single): Use needs_neutering_p to skip over insns that do not
14377         need neutering.
14379 2018-04-26  Richard Biener <rguenther@suse.de>
14380             Tom de Vries  <tom@codesourcery.com>
14382         PR lto/85422
14383         * lto-streamer-out.c (output_function): Fixup loops if required to match
14384         discovery done in the reader.
14386 2018-04-26  Richard Biener  <rguenther@suse.de>
14388         PR tree-optimization/85116
14389         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
14390         have a loop exit from the single latch predecessor.  Remove
14391         case of header with just condition.
14392         (ch_base::copy_headers): Exclude infinite loops from any
14393         processing.
14394         (pass_ch::execute): Record exits.
14396 2018-04-26  Richard Biener  <rguenther@suse.de>
14398         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
14399         prologue cost vector and pass it to vect_get_load_cost.
14400         (vect_get_peeling_costs_all_drs): Likewise.
14401         (vect_peeling_hash_get_lowest_cost): Likewise.
14402         (vect_enhance_data_refs_alignment): Likewise.
14404 2018-04-26  Richard Biener  <rguenther@suse.de>
14406         PR middle-end/85450
14407         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
14408         checking of integer<->pointer conversions.
14409         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
14410         sign-/zero-extending pointer types.
14411         (expand_omp_for_static_chunk): Likewise.
14413 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
14414             Jean Lee  <xiaoyur347@gmail.com>
14416         * config/mips/mips.c (mips_asan_shadow_offset): New function.
14417         (TARGET_ASAN_SHADOW_OFFSET): Define.
14418         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
14419         true for -fsanitize=address.
14421 2018-04-25  Mark Wielaard  <mark@klomp.org>
14423         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
14424         shorter ones.
14426 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
14428         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
14429         than "alu", remove explicit "memory" and "imm_disp" attributes.
14430         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
14432         PR middle-end/85414
14433         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
14434         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
14435         gen_lowpart_no_emit.
14437 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
14439         PR target/85473
14440         * config/i386/i386.c (ix86_expand_builtin): Change memory
14441         operand to XI, extend p0 to Pmode.
14442         * config/i386/i386.md: Change unspec volatile and operand
14443         1 mode to XI, change operand 0 mode to P.
14445 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14447         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
14448         GET_MODE_MASK before any checking.
14449         (nds32_can_use_bset_p): Likewise.
14450         (nds32_can_use_btgl_p): Likewise.
14452 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14454         * config/nds32/nds32-doubleword.md: New define_split pattern for
14455         illegal register number.
14457 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14459         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
14461 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
14463         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
14465 2018-04-25  Richard Biener  <rguenther@suse.de>
14467         * lto-streamer.h (LTO_major_version): Bump to 8.
14469 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
14471         * BASE-VER: Set to 9.0.0.
14473 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
14475         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
14476         in __abskf2 and __powikf2.
14478 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14480         PR target/85512
14481         * config/aarch64/constraints.md (Usg, Usj): New constraints.
14482         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
14483         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
14484         Use the above on operand 2.  Reindent.
14485         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
14487 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
14489         PR target/85485
14490         * common/config/i386/i386-common.c (ix86_handle_option): Don't
14491         handle OPT_mcet.
14492         * config/i386/i386.opt (mcet): Removed.
14493         * doc/install.texi: Remove -mcet documentation.
14494         * doc/invoke.texi: Likewise.
14496 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
14498         PR target/85485
14499         * doc/install.texi: Remove -mcet from bootstrap-cet.
14501 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14503         PR target/85511
14504         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14505         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
14506         if TARGET_64BIT.
14508         PR target/85503
14509         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
14510         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
14511         containing a CONST_VECTOR.
14513 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
14515         * doc/install.texi: Update newlib dependency for nvptx.
14517 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14519         PR target/85508
14520         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
14521         instead of INTVAL when shifting x left.
14523 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
14525         PR tree-optimization/85478
14526         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
14527         vect_grouped_store_supported for single element vectors.
14529 2018-04-24  Richard Biener  <rguenther@suse.de>
14531         PR target/85491
14532         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
14533         load cost increase to the case of non-constant step.
14535 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
14537         PR target/84828
14538         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
14539         destination if any_malformed_asm.
14541 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
14543         PR middle-end/85496
14544         * expr.c (store_field): In the bitfield case, if the value comes from
14545         a function call and is returned in registers by means of a PARALLEL,
14546         do not change the mode of the temporary unless BLKmode and VOIDmode.
14548 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
14550         PR rtl-optimization/85423
14551         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
14552         dependencies to debug insns when the previous insn is non-debug.
14554 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
14556         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
14557         enums into a single definition.
14558         (fls): Fix predicates and printing.
14559         (seti): Likewise.
14561 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
14563         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
14564         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
14565         and short u6 immediate.
14566         (check_if_valid_sleep_operand): Remove.
14567         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
14569 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14571         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
14572         flag_always_save_lp condition.
14573         * config/nds32/nds32.opt (malways-save-lp): New option.
14575 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
14577         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
14578         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
14579         * config/nds32/nds32.h
14580         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
14581         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
14583 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
14585         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
14586         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
14588 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
14589             Chung-Ju Wu  <jasonwucj@gmail.com>
14591         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
14592         Declare.
14593         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
14594         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
14596 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14598         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
14600 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14602         * config/nds32/nds32-protos.h (nds32_data_alignment,
14603         nds32_local_alignment): Declare.
14604         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
14605         nds32_local_alignment): New functions.
14606         (TARGET_CONSTANT_ALIGNMENT): Define.
14607         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
14609 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14611         * config/nds32/nds32.c
14612         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
14613         (TARGET_MODES_TIEABLE_P): Likewise.
14615 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
14617         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
14618         level Ofast and Og.
14620 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
14621             Chung-Ju Wu  <jasonwucj@gmail.com>
14623         * config/nds32/constants.md (unspec_volatile_element): Add enum values
14624         for unaligned access.
14625         * config/nds32/nds32-intrinsic.c: Implementation of expanding
14626         unaligned access.
14627         * config/nds32/nds32-intrinsic.md: Likewise.
14628         * config/nds32/nds32_intrinsic.h: Likewise.
14629         * config/nds32/nds32.h (nds32_builtins): Likewise.
14630         * config/nds32/nds32.opt (munaligned-access): New option.
14631         * config/nds32/nds32.c (nds32_asm_file_start): Display
14632         flag_unaligned_access status.
14634 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
14636         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
14637         -mno-relax is present.
14638         * config/riscv/linux.h (LINK_SPEC): Ditto.
14640 2018-04-20  Martin Sebor  <msebor@redhat.com>
14642         PR c/85365
14643         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
14644         for null pointers.
14645         (gimple_fold_builtin_stxcpy_chk): Same.
14646         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
14648 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
14650         PR target/85456
14651         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
14652         __powikf2 when long double is IEEE 128-bit.
14654 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
14656         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
14657         step to make sure stack always aligned.
14659 2018-04-20  Carl Love  <cel@us.ibm.com>
14661         PR target/83402
14662         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
14663         size check for arg0.
14665 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
14666             Tom de Vries  <tom@codesourcery.com>
14668         PR target/85445
14669         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
14670         Emit insns for calls too.
14671         (nvptx_find_par): Always look for worker-level predecessor insn.
14672         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
14673         calls.
14674         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
14675         (nvptx_process_pars): Propagate frames for calls.
14677 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
14679         PR target/85469
14680         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
14681         Removed.
14682         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
14683         (ix86_handle_option): Don't handle OPT_mibt.
14684         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
14685         __SHSTK__.
14686         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
14687         has_ibt and ibt.
14688         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
14689         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
14690         (ix86_target_macros): Define __CET__ with flag_cf_protection
14691         for -fcf-protection.
14692         * config/i386/i386.c (isa2_opts): Remove -mibt.
14693         * config/i386/i386.h (TARGET_IBT): Removed.
14694         (TARGET_IBT_P): Likewise.
14695         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
14696         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
14697         * config/i386/i386.opt (mcet): Update help message.
14698         (mshstk): Likewise.
14699         (mibt): Removed.
14700         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
14701         -mcet as an alias for -mshstk.
14703 2018-04-20  Richard Biener <rguenther@suse.de>
14705         PR middle-end/85475
14706         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
14707         complexity by forcing a single use of the multiply operand.
14709 2018-04-20  Martin Jambor  <mjambor@suse.cz>
14711         ipa/85449
14712         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
14713         recursion dependency to only apply to non-clones.
14715 2018-04-20  Martin Jambor  <mjambor@suse.cz>
14717         ipa/85447
14718         * ipa-cp.c (create_specialized_node): Check that clones of
14719         self-recursive edges exist during IPA-CP.
14721 2018-04-19  Toon Moene  <toon@moene.org>
14723         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
14724         by -O3.
14726 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
14728         PR tree-optimization/85467
14729         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
14730         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
14731         VECTOR_CST element to type.
14733 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
14735         PR target/85397
14736         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
14737         * config/i386/i386.md (builtin_setjmp_setup): Removed.
14738         (builtin_longjmp): Likewise.
14739         (save_stack_nonlocal): New pattern.
14740         (restore_stack_nonlocal): Likewise.
14742 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
14744         PR target/85404
14745         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14746         Replace ASM_OUTPUT_LABEL with fprintf.
14748 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
14750         PR target/85417
14751         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
14752         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
14753         * config/i386/i386-c.c (ix86_target_macros_internal): Also
14754         define __IBT__ and __SHSTK__ for -fcf-protection.
14755         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
14756         TARGET_IBT.
14757         (ix86_trampoline_init): Likewise.
14758         (x86_output_mi_thunk): Likewise.
14759         (ix86_notrack_prefixed_insn_p): Likewise.
14760         (ix86_option_override_internal): Don't disallow -fcf-protection.
14761         * config/i386/i386.md (rdssp<mode>): Also enable for
14762         -fcf-protection.
14763         (incssp<mode>): Likewise.
14764         (nop_endbr): Likewise.
14765         * config/i386/i386.opt (mcet): Change help message to built-in
14766         functions only.
14767         (mibt): Likewise.
14768         (mshstk): Likewise.
14769         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
14770         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
14771         enable CET built-in functions.
14773 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
14775         * common/config/i386/i386-common.c
14776         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
14777         OPTION_MASK_ISA_MOVDIRI_UNSET,
14778         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
14779         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
14780         * config.gcc (movdirintrin.h): New header.
14781         * config/i386/cpuid.h (bit_MOVDIRI,
14782         bit_MOVDIR64B): New bits.
14783         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
14784         and -mmvodir64b.
14785         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
14786         (VOID, PVOID, PCVOID)): New function types.
14787         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
14788         __builtin_ia32_directstoreu_u64,
14789         __builtin_ia32_movdir64b): New builtins.
14790         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
14791         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
14792         and -mmovdiri.
14793         (ix86_valid_target_attribute_inner_p): Ditto.
14794         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
14795         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
14796         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
14797         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
14798         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
14799         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
14800         (movdiri<mode>, movdir64b_<mode>): New.
14801         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
14802         * config/i386/immintrin.h: Include movdirintrin.h.
14803         * config/i386/movdirintrin.h: New file.
14804         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
14806 2018-04-19  Richard Biener  <rguenther@suse.de>
14808         PR middle-end/85455
14809         * cfg.c (clear_bb_flags): When loop state says we have
14810         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
14812 2018-04-19  Richard Biener  <rguenther@suse.de>
14814         PR tree-optimization/84737
14815         * tree-vect-data-refs.c (vect_copy_ref_info): New function
14816         copying restrict info.
14817         (vect_setup_realignment): Use it.
14818         * tree-vectorizer.h (vect_copy_ref_info): Declare.
14819         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
14820         the first DR to all generated stores.
14821         (vectorizable_load): Likewise for loads.
14823 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
14825         PR tree-optimization/85446
14826         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
14827         the integral and pointer types to have the same precision.
14829         * doc/install.texi: Document --disable-cet being the default and
14830         --enable-cet=auto.
14832 2018-04-18  Martin Liska  <mliska@suse.cz>
14834         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14835         style.
14837 2018-04-18  Martin Liska  <mliska@suse.cz>
14839         Revert
14840         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
14842         PR ipa/83983
14843         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
14844         arguments if they are comparable.
14846 2018-04-18  Martin Liska  <mliska@suse.cz>
14848         Revert
14849         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
14851         PR lto/84805
14852         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14853         incomplete types.
14855 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
14857         PR target/85388
14858         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
14859         ENDBR after calling __morestack.
14861 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
14863         PR jit/85384
14864         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
14865         by using gcc_base_ver to generate a gcc_driver_version, and use
14866         it when generating GCC_DRIVER_NAME.
14867         * configure: Regenerate.
14869 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
14871         PR target/81084
14872         * config.gcc: Obsolete powerpc*-*-*spe*.
14874 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
14876         PR debug/84637
14877         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
14878         (stabstr_D): Change type of unum from unsigned int to
14879         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
14880         type.
14882 2018-04-17  Jim Wilson  <jimw@sifive.com>
14884         PR 84856
14885         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
14886         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
14887         Set arg_pointer_offset after using pretend_args_size.
14889 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
14891         PR rtl-optimization/85431
14892         * dse.c (record_store): Ignore zero width stores.
14894         PR sanitizer/85230
14895         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
14896         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
14897         __builtin_stack_restore rather than after it.
14898         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
14899         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
14900         argument instead of virtual_dynamic_stack_rtx.
14902 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14904         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
14905         New prototype.
14906         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14907         Add note to error message to explain internal mapping of overloaded
14908         built-in function name to non-overloaded built-in function name.
14909         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
14910         function.
14912 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
14914         PR target/85424
14915         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
14916         where the inputs overlap with the output.
14918 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
14920         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
14921         (=v, v) alternative and explicit "memory" attribute.
14922         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
14923         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14924         attributes.
14925         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
14926         "sselog1" type instead of "sselog".
14927         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
14928         "sselog".  Remove explicit "memory" attribute.
14929         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
14930         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14931         attributes.
14932         (vec_extract_hi_v32hi): Merge all alternatives into one, use
14933         "sselog1" type instead of "sselog".  Remove explicit "memory"
14934         attribute.
14935         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14936         use "sselog1" type instead of "sselog".  Remove explicit "memory"
14937         attribute.
14938         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14939         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14940         attributes.
14941         (vec_extract_hi_v64qi): Merge all alternatives into one, use
14942         "sselog1" type instead of "sselog".  Remove explicit "memory"
14943         attribute.
14944         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14945         use "sselog1" type instead of "sselog".  Remove explicit "memory"
14946         attribute.
14948         PR target/85430
14949         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
14951         PR middle-end/85414
14952         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
14953         on a SUBREG.
14955 2018-04-17  Martin Jambor  <mjambor@suse.cz>
14957         PR ipa/85421
14958         * ipa-cp.c (create_specialized_node): Call
14959         expand_all_artificial_thunks if necessary.
14961 2018-04-17  Martin Liska  <mliska@suse.cz>
14963         PR lto/85405
14964         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
14965         in message, remote space in between '_G' and '('.
14967 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
14969         PR target/85281
14970         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
14971         avx512f_vmcmp<mode>3<round_saeonly_name>,
14972         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
14973         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14974         avx512f_rndscale<mode><round_saeonly_name>,
14975         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
14976         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
14977         Use %<iptr>2 instead of %2 for -masm=intel.
14978         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
14979         avx512f_vcvttss2usi<round_saeonly_name>,
14980         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
14981         -masm=intel.
14982         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
14983         avx512f_vcvttsd2usi<round_saeonly_name>,
14984         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
14985         Use %q1 instead of %1 for -masm=intel.
14986         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
14987         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
14988         of %3 for -masm=intel.
14989         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
14990         -masm=intel.
14991         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14992         -masm=intel.
14993         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
14994         -masm=intel.
14995         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
14996         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
14997         %g1.
14998         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
14999         -masm=intel.
15000         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
15001         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
15002         %g1 and one with %0 and %1.
15003         (avx512er_vmrcp28<mode><round_saeonly_name>,
15004         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
15005         %1 for -masm=intel.
15006         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
15007         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
15008         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
15009         of %0 and %{%4%} for -masm=intel.
15010         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
15011         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
15012         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
15013         order of %0 and %{%5%}%{z%} for -masm=intel.
15015 2018-04-17  Jan Hubicka  <jh@suse.cz>
15017         PR lto/85405
15018         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15020 2018-04-17  Martin Liska  <mliska@suse.cz>
15022         PR ipa/85329
15023         * multiple_target.c (create_dispatcher_calls): Set apostrophes
15024         for target_clone error message.  Make default implementation
15025         clone to be a local declaration.
15026         (separate_attrs): Add new argument and check for an empty
15027         string.
15028         (expand_target_clones): Handle it.
15029         (ipa_target_clone): Make redirection just for target_clones
15030         functions.
15032 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
15033             Tom de Vries  <tom@codesourcery.com>
15035         PR middle-end/84955
15036         * omp-expand.c (expand_oacc_for): Add dummy false branch for
15037         tiled basic blocks without omp continue statements.
15039 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
15041         PR target/83660
15042         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15043         vec_extract expression as having side effects to make sure it gets
15044         a cleanup point.
15046 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
15048         PR target/85403
15049         * config/i386/i386.c (get_builtin_code_for_version): Check
15050         error_mark_node.
15052 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
15054         PR target/84331
15055         * config.gcc: Support "skylake".
15056         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15057         PROCESSOR_SKYLAKE.
15058         * config/i386/i386.c (m_SKYLAKE): Define.
15059         (processor_target_table): Add "skylake".
15060         (ix86_option_override_internal): Add "skylake".
15061         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
15062         PROCESSOR_CANNONLAKE.
15063         (get_builtin_code_for_version): Fix priority for
15064         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
15065         PROCESSOR_SKYLAKE-AVX512.
15066         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
15067         (processor_type): Add PROCESSOR_SKYLAKE.
15069 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
15070             Jason Merrill  <jason@redhat.com>
15072         PR c++/85112
15073         * convert.c (convert_to_integer_1): Use direct recursion for
15074         enumeral types and types with a precision less than the number
15075         of bits in their mode.
15077 2018-04-16  Julia Koval  <julia.koval@intel.com>
15079         PR target/84413
15080         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
15081         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
15083 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
15085         PR target/85293
15086         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
15087         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
15088         and -mno-direct-move.
15090 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
15092         PR target/83402
15093         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
15094         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
15095         Ensure negative shifts result in {0}.
15097 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
15099         PR rtl-optimization/79916
15100         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
15101         regs (if any) to define how to gnerate SD moves when LRA is in
15102         progress.
15104 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
15106         PR rtl-optimization/85393
15107         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
15108         * except.c (expand_dw2_landing_pad_for_region): Make static.
15109         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
15110         a label and unconditional jump to old_bb, rather than
15111         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
15112         basic block.
15114         PR rtl-optimization/85376
15115         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
15116         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
15117         instead of a specific value.
15119 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
15120             Bin Cheng  <bin.cheng@arm.com>
15122         PR tree-optimization/82965
15123         PR tree-optimization/83991
15124         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
15125         by_profile_only parameter.
15126         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
15127         information if the loop was predicted to iterate too many times.
15128         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
15130 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
15132         PR lto/71991
15133         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
15134         always inline.
15136 2018-04-13  Martin Liska  <mliska@suse.cz>
15137             Jakub Jelinek  <jakub@redhat.com>
15139         PR middle-end/81657
15140         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
15141         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
15142         * builtins.c (expand_builtin_memory_copy_args): Use
15143         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
15144         handle dest_addr == pc_rtx.
15146 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
15148         PR target/85291
15149         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
15150         asked to not generate direct moves.
15151         (fix_trunc<mode>si2_stfiwx): Similar.
15152         (fix_trunc<mode>si2_internal): Similar.
15154 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15156         PR debug/83157
15157         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
15158         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
15159         lookup if dest in some wider mode is known to be const0_rtx and
15160         if so, record permanent equivalence for it to be ZERO_EXTEND of
15161         the narrower mode destination.
15163 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
15165         * lto-streamer-out.c (output_function): Revert 259346.
15166         * omp-expand.c (expand_oacc_for): Likewise.
15168 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
15170         PR rtl-optimization/85354
15171         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
15172         * sel-sched.c (sel_global_init): ... here.
15174 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
15176         PR target/85238
15177         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
15178         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
15179         mode for PE-COFF targets.
15180         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
15181         (i386_pe_asm_lto_end): Likewise.
15182         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
15183         (TARGET_ASM_LTO_END): Likewise.
15184         * config/i386/winnt.c (saved_debug_info_level): New static variable.
15185         (i386_pe_asm_lto_start): New function.
15186         (i386_pe_asm_lto_end): Likewise.
15188 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
15189             Richard Biener  <rguenther@suse.de>
15191         PR middle-end/84955
15192         * lto-streamer-out.c (output_function): Fix CFG loop state before
15193         streaming out.
15194         * omp-expand.c (expand_oacc_for): Handle calls to internal
15195         functions like regular functions.
15197 2018-04-12  Richard Biener  <rguenther@suse.de>
15199         PR lto/85371
15200         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
15201         for the early LTO debug to properly generate references to it
15202         during DIE emission.  Do not re-use that for the skeleton for
15203         split-dwarf.
15204         (dwarf2out_early_finish): Likewise.
15206 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15208         PR target/85328
15209         * config/i386/sse.md
15210         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
15211         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
15212         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
15213         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
15214         and output is a reg, avoid creating invalid lowpart subreg, but
15215         instead split into a 512-bit move.  Don't split if not AVX512VL,
15216         input is xmm16+ reg and output is a mem.
15217         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
15218         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
15219         xmm16+ reg and output is a mem.
15221 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15223         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
15224         also for flag_dwarf2_cfi_asm.
15226 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
15228         PR rtl-optimization/85342
15229         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
15230         a bool scalar var inside of the loop instead.  Don't try to update
15231         recog_data.operand after failed apply_change_group.
15233 2018-04-12  Tom de Vries  <tom@codesourcery.com>
15235         PR target/85296
15236         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
15237         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
15238         array with flexible array member as array without given dimension.
15239         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
15240         argument for undefined param to true.
15242 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
15244         PR target/85321
15245         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15246         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
15247         from PowerPC section.
15248         * config/rs6000/sysv4.opt (mcall-): Improve help text.
15249         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
15250         help text that is too long.
15251         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
15252         help text that is too long.
15253         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
15254         help text that is too long.
15256 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
15258         * config/alpha/alpha.md (stack_probe_internal): Rename
15259         from "probe_stack".  Update all callers.
15261 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15263         PR rtl-optimization/84566
15264         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
15265         sched_macro_fuse_insns.
15267 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15269         PR target/84301
15270         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
15271         (compute_block_dependences): ... from here.
15273 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15275         PR tree-optimization/85331
15276         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
15277         from int to HOST_WIDE_INT.
15279 2018-04-11  Martin Jambor  <mjambor@suse.cz>
15281         PR ipa/84149
15282         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
15283         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
15284         not the same as the source val.
15285         (cgraph_edge_brings_value_p): New parameter.
15286         (gather_edges_for_value): Pass destination value to
15287         cgraph_edge_brings_value_p.
15288         (perhaps_add_new_callers): Likewise.
15289         (get_info_about_necessary_edges): Likewise and exclude values brought
15290         only by self-recursive edges.
15291         (create_specialized_node): Redirect only clones of self-calling edges.
15292         (+self_recursive_pass_through_p): New function.
15293         (find_more_scalar_values_for_callers_subset): Use it.
15294         (find_aggregate_values_for_callers_subset): Likewise.
15295         (known_aggs_to_agg_replacement_list): Removed.
15296         (decide_whether_version_node): Re-calculate known constants for all
15297         remaining context clones.
15299 2018-04-11  Richard Biener  <rguenther@suse.de>
15301         PR lto/85339
15302         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
15303         from early DWARF output.
15304         (dwarf2out_early_finish): Output line info unconditionally into
15305         early DWARF and add reference to it.
15307 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15309         PR target/85281
15310         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
15311         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
15312         other than V2DFmode using iptr mode attribute.
15313         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
15315 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
15317         PR rtl-optimization/84659
15318         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
15320 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
15322         PR debug/85302
15323         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
15324         SIZEP is NULL.
15325         (output_loc_list): Pass address of a dummy size variable even in the
15326         locview handling loop.
15327         (index_location_lists): Add comment on why skip_loc_list_entry can't
15328         call size_of_locs.
15330 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
15332         PR target/85261
15333         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15334         into register.
15336 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
15338         PR target/85321
15339         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
15340         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
15341         and -mstring-compare-inline-limit.
15343 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15345         PR target/85287
15346         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
15347         for stack clash protection in a register whenever we need it to be in
15348         a register.
15350 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15352         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
15353         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
15355 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
15357         PR target/85321
15358         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
15359         the help text.
15360         (mlong-double-): Ditto.
15361         * config/rs6000/sysv4.opt (msdata=): Ditto.
15362         (mtls-size=): Ditto.
15364 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15366         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
15367         erroneous entries for
15368         "vector int vec_ldl (int, long int *)", and
15369         "vector unsigned int vec_ldl (int, unsigned long int *)".
15370         Add comments and entries for
15371         "vector bool char vec_ldl (int, bool char *)",
15372         "vector bool short vec_ldl (int, bool short *)",
15373         "vector bool int vec_ldl (int, bool int *)",
15374         "vector bool long long vec_ldl (int, bool long long *)",
15375         "vector pixel vec_ldl (int, pixel *)",
15376         "vector long long vec_ldl (int, long long *)",
15377         "vector unsigned long long vec_ldl (int, unsigned long long *)".
15378         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
15379         type tree bool_long_long_type_node and correct definition of
15380         bool_V2DI_type_node to make reference to this new type tree.
15381         (rs6000_mangle_type): Replace erroneous reference to
15382         bool_long_type_node with bool_long_long_type_node.
15383         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
15384         comments to emphasize sign distinctions for char and int types and
15385         replace RS6000_BTI_bool_long constant with
15386         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
15387         use of RS6000_BTI_pixel.
15388         (bool_long_type_node): Remove this macro definition.
15389         (bool_long_long_type_node): New macro definition
15391 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
15393         PR rtl-optimization/85300
15394         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
15395         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
15396         simplify_unary_operation fails.
15398 2018-04-10  Martin Liska  <mliska@suse.cz>
15400         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
15401         cgraph_edge and ipa_ref.
15403 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
15405         PR target/85177
15406         PR target/85255
15407         * config/i386/sse.md
15408         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
15409         computation of the VEC_MERGE selector from mask.
15410         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
15411         Fix decoding of the VEC_MERGE selector into mask.
15413 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
15415         PR tree-optimization/85286
15416         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
15418 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
15420         * final.c (final_1): Set insn_last_address as well as
15421         insn_current_address.
15423 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15425         PR target/85173
15426         * explow.c (emit_stack_probe): Call validize_mem on memory location
15427         before passing it to gen_probe_stack.  Create address operand and
15428         legitimize it for the probe_stack_address case.
15430 2018-04-09  Jan Hubicka  <jh@suse.cz>
15432         PR lto/85078
15433         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
15434         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
15435         * tree.c (free_lang_data_in_type): Fix handling of binfos;
15436         walk basetypes.
15437         (free_lang_data): Rebuild type inheritance graph.
15439 2018-04-09  Martin Sebor  <msebor@redhat.com>
15441         * invoke.texi (-finline-small-functions): Mention other optimization
15442         options.
15443         (-findirect-inlining, -fpartial-inlining): Same.
15444         (-finline-functions-called-once): Same.
15445         (-freorder-blocks-and-partition): Same.
15447 2018-04-09  Jan Hubicka  <jh@suse.cz>
15449         PR rtl/84058
15450         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
15451         jumps; choose last target that matches the criteria (i.e.
15452         no partition changes for non-crossing jumps).
15453         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
15454         support for redirecting crossing jumps to non-crossing.
15456 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
15458         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
15459         also for naked functions.
15461 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
15463         * config/arc/arc.md (add_shift): New pattern.
15464         (add_shift2): Likewise.
15465         (sub_shift): Likewise.
15466         (sub_shift_cmp0_noout): Likewise.
15467         (compare_si_ashiftsi): Likewise.
15468         (xbfu_cmp0_noout): New combine pattern.
15469         (xbfu_cmp0"): Likewise.
15470         (movsi_set_cc_insn): Place the predicable variant first.
15471         (commutative_binary_cmp0_noout): Remove clobber.
15472         (commutative_binary_cmp0): New pattern.
15473         (noncommutative_binary_cmp0): Likewise.
15474         (noncommutative_binary_cmp0_noout): Likewise.
15475         (noncommutative_binary_comparison_result_used): Removed.
15476         (rsub_cmp0): New pattern.
15477         (rsub_cmp0_noout): Likewise.
15478         (extzvsi): Changed, keep only meaningful variants.
15479         (SQH, SEZ): New iterators.
15480         (SQH_postfix): New mode attribute.
15481         (SEZ_prefix): New code attribute.
15482         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
15483         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
15484         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
15485         of numerical value.
15486         (noncommutative_operator): Check the availability of barrel
15487         shifter option.
15489 2018-04-09  Richard Biener  <rguenther@suse.de>
15491         PR tree-optimization/85284
15492         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
15493         Only use the niter constraining form of simple_iv when the exit
15494         is always executed.
15496 2018-04-09  Tom de Vries  <tom@codesourcery.com>
15498         PR target/84041
15499         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15500         (define_expand "*memory_barrier"): New define_expand.
15501         (define_insn "memory_barrier"): New insn.
15503 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15505         PR rtl-optimization/80463
15506         PR rtl-optimization/83972
15507         PR rtl-optimization/83480
15509         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
15510         correct producer for the insn.
15511         (tidy_control_flow): Fixup seqnos in case of debug insns.
15513 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15515         PR rtl-optimization/83913
15517         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
15518         different sched-times when merging exprs.
15520 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15522         PR rtl-optimization/83962
15524         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
15525         tidy_fallthru_edge and tidy_control_flow.
15527 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
15529         PR rtl-optimization/83530
15531         * sel-sched.c (force_next_insn): New global variable.
15532         (remove_insn_for_debug): When force_next_insn is true, also leave only
15533         next insn in the ready list.
15534         (sel_sched_region): When the region wasn't scheduled, make another pass
15535         over it with force_next_insn set to 1.
15537 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
15539         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
15540         into tm_file.
15541         * config/nds32/constants.md (unspec_volatile_element): Add enum values
15542         for interrupt control.
15543         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15544         functions for interrupt control.
15545         * config/nds32/nds32-intrinsic.md: Likewise.
15546         * config/nds32/nds32_intrinsic.h: Likewise.
15547         * config/nds32/nds32.h (nds32_builtins): Likewise.
15549 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
15551         * config/nds32/nds32.c (nds32_init_machine_status,
15552         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
15553         strict_aligned_p field.
15554         (nds32_expand_to_rtl_hook): New function.
15555         (TARGET_EXPAND_TO_RTL_HOOK): Define.
15556         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
15558 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15559             Chung-Ju Wu  <jasonwucj@gmail.com>
15561         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
15562         * config/nds32/nds32-n7.md: New file.
15563         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
15564         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
15565         pipeline.
15566         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
15567         * config/nds32/nds32.md (pipeline_model): Add n7.
15568         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
15569         * config/nds32/pipelines.md: Include n7 settings.
15571 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15572             Chung-Ju Wu  <jasonwucj@gmail.com>
15574         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
15575         * config/nds32/nds32-e8.md: New file.
15576         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
15577         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
15578         pipeline.
15579         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
15580         * config/nds32/nds32.md (pipeline_model): Add e8.
15581         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
15582         * config/nds32/pipelines.md: Include e8 settings.
15584 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15585             Chung-Ju Wu  <jasonwucj@gmail.com>
15587         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
15588         * config/nds32/nds32-n8.md: New file.
15589         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
15590         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
15591         pipeline.
15592         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
15593         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
15594         * config/nds32/nds32.md (pipeline_model): Add n8.
15595         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
15596         * config/nds32/pipelines.md: Include n8 settings.
15598 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
15599             Chung-Ju Wu  <jasonwucj@gmail.com>
15601         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
15602         * config/nds32/nds32-n9-2r1w.md: New file.
15603         * config/nds32/nds32-n9-3r2w.md: New file.
15604         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
15605         nds32_register_ports): New or modify for cpu n9.
15606         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
15607         pipeline.
15608         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
15609         * config/nds32/nds32-utils.c: New file.
15610         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
15611         TARGET_MUL_SLOW): Define.
15612         * config/nds32/nds32.md (pipeline_model): New attribute.
15613         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
15614         New options that support cpu n9.
15615         * config/nds32/pipelines.md: Include n9 settings.
15616         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
15618 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
15620         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
15621         information if necessary.
15622         (output_cond_branch_compare_zero): Likewise.
15623         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
15624         (nds32_target_alignment): Refine for alignment.
15625         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
15626         (FUNCTION_BOUNDARY): Modify.
15627         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
15628         align case.
15629         * config/nds32/nds32.opt (malways-align, malign-functions): New.
15631 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
15633         * config/nds32/constants.md (unspec_volatile_element): Add values for
15634         TLB operation and data prefetch.
15635         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
15636         functions for TLB operation and data prefetch.
15637         * config/nds32/nds32-intrinsic.md: Likewise.
15638         * config/nds32/nds32_intrinsic.h: Likewise.
15639         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
15640         (nds32_print_operand): Likewise.
15641         * config/nds32/nds32.h (nds32_builtins): Likewise.
15643 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
15644         Andrew Pinski <pinsika@gcc.gnu.org>
15646         PR middle-end/82976
15647         * match.pd: Use constant_boolean_node of correct type instead of
15648         boolean_true_node or boolean_false_node for simplifying
15649         pointer comparisons to zero.
15651 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
15653         PR tree-optimization/80021
15654         * tree.c (verify_type_variant): Make error call in verify_variant_match
15655         translatable and remove final full stop.
15657 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
15659         * config/nds32/constants.md (unspec_volatile_element): Add
15660         UNSPEC_VOLATILE_EH_RETURN.
15661         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
15662         nds32_output_stack_pop): Support dwarf exception handling process.
15663         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
15664         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
15665         exception handling process.
15666         (nds32_compute_stack_frame): Likewise.
15667         (nds32_return_addr_rtx): Likewise.
15668         (nds32_initial_elimination_offset): Likewise.
15669         (nds32_expand_prologue): Likewise.
15670         (nds32_expand_epilogue): Likewise.
15671         (nds32_dynamic_chain_address): New function.
15672         * config/nds32/nds32.h (machine_function): Add fields for dwarf
15673         exception handling.
15674         (DYNAMIC_CHAIN_ADDRESS): Define.
15675         (EH_RETURN_DATA_REGNO): Define.
15676         (EH_RETURN_STACKADJ_RTX): Define.
15677         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
15678         patterns for dwarf exception handling.
15680 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
15682         * config/nds32/nds32.h: Clean up obsolete macros.
15684 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
15686         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15687         Add enum values for particular instructions.
15688         * config/nds32/nds32-intrinsic.c: Implementation of expanding
15689         particular intrinsic functions.
15690         * config/nds32/nds32-intrinsic.md: Likewise.
15691         * config/nds32/nds32_intrinsic.h: Likewise.
15692         * config/nds32/nds32.h (nds32_builtins): Likewise.
15693         * config/nds32/nds32.md (type): Add pbsad and pbsada.
15694         (btst, ave): New patterns for particular instructions.
15696 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
15698         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15699         Add enum values for atomic load/store and memory sync.
15700         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
15701         and memory sync.
15702         * config/nds32/nds32-intrinsic.md: Likewise.
15703         * config/nds32/nds32_intrinsic.h: Likewise.
15704         * config/nds32/nds32.h (nds32_builtins): Likewise.
15706 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
15708         PR tree-optimization/85257
15709         * fold-const.c (native_encode_vector): If not all elts could fit
15710         and off is -1, return 0 rather than offset.
15711         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
15712         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
15713         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
15714         adjust buffer in native_interpret_expr call.
15716 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
15718         * config/nds32/constants.md (unspec_volatile_element): Add cache
15719         control enum values.
15720         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
15721         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
15722         * config/nds32/nds32.c (nds32_cctl_names): New.
15723         (nds32_print_operand): Handle cache control register names.
15724         * config/nds32/nds32.h (nds32_builtins): New enum values.
15725         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
15726         macros.
15727         * config/nds32/nds32.md (type): Add mmu.
15728         * config/nds32/pipelines.md (simple_insn): Add mmu.
15730 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
15732         * config/nds32/nds32.md (type): Remove call.
15733         * config/nds32/pipelines.md (simple_insn): Likewise.
15735 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
15737         * config/nds32/constants.md (unspec_volatile_element): Add
15738         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
15739         UNSPEC_VOLATILE_FMFCFG.
15740         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
15741         description for fmfcfg and fmfcsr.
15742         (bdesc_1arg): Add fmtcsr.
15743         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
15744         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
15745         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
15746         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
15747         unspec_fmfcfg): New patterns.
15748         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
15749         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
15750         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
15751         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
15752         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
15753         __nds32__fmfcfg): Define.
15755 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
15757         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
15758         intrinsic register names.
15759         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
15760         intrinsic register enum values and macros.
15762 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
15764         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
15765         for load/store addressing form.
15766         (nds32_print_operand_address): Likewise.
15768 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
15770         PR target/85196
15771         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
15772         based on LABEL_REF.  Remove useless assertion.
15773         (pic_address_needs_scratch): Fix formatting.
15774         (sparc_legitimize_pic_address): Minor tweaks.
15775         (sparc_delegitimize_address): Adjust assertion accordingly.
15776         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
15777         into symbolic_operand.
15778         (movsi_high_pic_label_ref): Likewise.
15779         (movsi_lo_sum_pic_label_ref): Likewise.
15780         (movdi_pic_label_ref): Likewise.
15781         (movdi_high_pic_label_ref): Likewise.
15782         (movdi_lo_sum_pic_label_ref): Likewise.
15784 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
15786         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
15787         custom LIB_SPEC setup.
15789 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
15790             Kito Cheng  <kito.cheng@gmail.com>
15792         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
15793         * config/riscv/freebsd.h: New.
15795 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
15797         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
15798         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
15799         file.
15801 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
15802             Kito Cheng  <kito.cheng@gmail.com>
15804         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
15805         nds32_output_call, nds32_symbol_binds_local_p): New functions.
15806         * config/nds32/nds32-protos.h (nds32_output_call,
15807         nds32_output_return): Declare.
15808         * config/nds32/nds32.md: Refine all the call and return patterns.
15810 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
15812         PR debug/85252
15813         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
15814         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
15816         PR rtl-optimization/84872
15817         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
15818         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
15819         EDGE_CROSSING edge.
15821 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
15823         * expr.c (copy_blkmode_to_reg): Revert 254862.
15824         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
15826 2018-04-06  Richard Biener  <rguenther@suse.de>
15828         PR middle-end/85244
15829         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
15830         after seeing a component reference with an adjacent field.  Treat
15831         refs to arrays at struct end of external decls similar to
15832         refs to unconstrained commons.
15834 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
15836         PR sanitizer/85213
15837         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
15838         look through SAVE_EXPRs with non-side-effects argument.  Adjust
15839         recursive calls.
15840         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
15841         save_p here.
15843 2018-04-06  Richard Biener  <rguenther@suse.de>
15845         PR middle-end/85180
15846         * alias.c (find_base_term): New wrapper around find_base_term
15847         unwinding CSELIB_VAL_PTR changes.
15848         (find_base_term): Do not restore CSELIB_VAL_PTR during the
15849         recursion.
15851 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15853         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
15854         instructions.
15855         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
15856         constant definitions.
15857         ("nop"): lr 0,0 -> nopr r0
15858         ("nop_lr0", "nop_lr1"): New insn definitions.
15860 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
15862         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
15863         NDS32_V3PUSH_AVAILABLE_P macro.
15865 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
15866             Chung-Ju Wu  <jasonwucj@gmail.com>
15868         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
15869         (nds32*-*-*): Add float and fpu_config into supported_defaults.
15870         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
15871         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
15872         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
15873         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
15874         * config/nds32/constraints.md: New constraints and checking for hard
15875         float configuration.
15876         * config/nds32/iterators.md: New mode iterator and attribute for hard
15877         float configuration.
15878         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
15879         patterns.
15880         * config/nds32/nds32-fpu.md: New file.
15881         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
15882         deal with hard float code generation.
15883         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
15884         ARCH_V3S.
15885         (abi_type, float_reg_number): New enum type.
15886         * config/nds32/nds32-predicates.c: New predicates for hard float.
15887         * config/nds32/nds32-protos.h: Declare functions for hard float.
15888         * config/nds32/nds32.c: Implementation for hard float configuration.
15889         * config/nds32/nds32.h: Definitions for hard float configuration.
15890         * config/nds32/nds32.md: Include hard float machine description and
15891         modify patterns for hard float configuration.
15892         * config/nds32/nds32.opt: New options for hard float configuration.
15893         * config/nds32/predicates.md: New predicates for hard float
15894         configuration.
15896 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
15898         * common/config/nds32/nds32-common.c
15899         (nds32_option_optimization_table): Enable -mreleax-hint by default.
15901 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
15903         PR middle-end/85195
15904         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
15905         CONSTRUCTOR_ELT (ctor, ...)->value.
15907 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
15909         PR target/85193
15910         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15912 2018-04-05  Tom de Vries  <tom@codesourcery.com>
15914         PR target/85204
15915         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
15916         cond jump.
15918 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
15919             Kito Cheng  <kito.cheng@gmail.com>
15921         * config/nds32/constraints.md (U33): Fine-tune checking condition.
15922         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
15923         * config/nds32/nds32.h (nds32_16bit_address_type): Add
15924         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
15926 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
15927             Kito Cheng  <kito.cheng@gmail.com>
15929         * config/nds32/constraints.md (Ufe): New memory constraint.
15930         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
15931         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
15932         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
15933         operands.
15934         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
15935         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
15937 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15939         * config/nds32/nds32.md: Use optimize_size in the condition for
15940         alu-shift instructions.
15942 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15944         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
15946 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15948         * config/nds32/nds32.md (negsi2): Refine pattern.
15950 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
15951             Chung-Ju Wu  <jasonwucj@gmail.com>
15953         * config/nds32/iterators.md (shift_rotate): New code iterator.
15954         (shift): New code attribute.
15955         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
15956         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
15957         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
15958         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
15959         bit-wise operations.
15960         (andsi3, *andsi3): Ditto.
15961         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
15962         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
15963         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
15964         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
15965         nds32_ior_operand, nds32_xor_operand): New predicates.
15967 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15969         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
15970         (addsi3, subsi3): ... this.
15972 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15974         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
15976 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15978         * config/nds32/nds32.md: Adjust indention.
15980 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
15982         * config/nds32/nds32.md (feature): New attribute.
15984 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
15986         * config/nds32/nds32.md (subtype): New attribute.
15988 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
15990         PR target/85203
15991         * config/arm/arm-builtins.c (arm_expand_builtin): Change
15992         expansion to perform a bitwise AND of the argument followed by a
15993         boolean negation of the result.
15995 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
15997         PR rtl-optimization/84878
15998         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
15999         the basic block.  Assert the use reference is not artificial and that
16000         it has an associated insn.
16002 2018-04-04  Michael Matz  <matz@suse.de>
16004         * builtins.c (compute_objsize): Pass correct operand
16005         to array_at_struct_end_p.
16007 2018-04-04  Richard Biener  <rguenther@suse.de>
16009         PR lto/85176
16010         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
16011         from contexts for DINFO_LEVEL_TERSE and below.
16013 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16015         * config/nds32/nds32-doubleword.md (move_<mode>): Require
16016         resiter_operand condition.
16017         * config/nds32/nds32.md (*move<mode>): Ditto.
16019 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16020             Monk Chiang  <sh.chiang04@gmail.com>
16022         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
16024 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16026         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
16028 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16029             Kito Cheng  <kito.cheng@gmail.com>
16031         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
16032         nds32_cond_code_str, output_cond_branch,
16033         output_cond_branch_compare_zero, nds32_expand_cbranch,
16034         nds32_expand_cstore, nds32_expand_movcc,
16035         nds32_output_cbranchsi4_equality_zero,
16036         nds32_output_cbranchsi4_equality_reg,
16037         nds32_output_cbranchsi4_equality_reg_or_const_int,
16038         nds32_output_cbranchsi4_greater_less_zero: New functions.
16039         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
16040         nds32_expand_cstore, nds32_expand_movcc,
16041         nds32_output_cbranchsi4_equality_zero,
16042         nds32_output_cbranchsi4_equality_reg,
16043         nds32_output_cbranchsi4_equality_reg_or_const_int,
16044         nds32_output_cbranchsi4_greater_less_zero): Declare.
16045         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
16046         nds32_rimm11s_operand): New predicates.
16047         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
16048         * config/nds32/nds32.md: Rewrite all the branch and conditional move
16049         patterns.
16051 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
16053         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
16054         * config/nds32/nds32.md: Ditto.
16055         * config/nds32/pipelines.md: Ditto.
16057 2018-04-04  Richard Biener  <rguenther@suse.de>
16059         PR tree-optimization/85168
16060         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
16061         propagating abnormals.
16063 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16065         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
16067 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
16068             Kito Cheng  <kito.cheng@gmail.com>
16070         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
16071         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
16072         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
16073         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
16074         * config/nds32/nds32.md (sibcall_internal): New.
16075         (sibcall_register): Remove.
16076         (sibcall_immediate): Remove.
16077         (sibcall_value_internal): New.
16078         (sibcall_value_register): Remove.
16079         (sibcall_value_immediate): Remove.
16080         * config/nds32/predicates.md (nds32_general_register_operand): New.
16081         (nds32_call_address_operand): New.
16083 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
16085         PR rtl-optimization/85167
16086         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
16087         bb_defs if *split_p, instead preinitialize it to NULL.
16089         PR tree-optimization/85156
16090         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
16091         evaluating the argument multiple times.
16093 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
16095         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
16096         than vector.
16097         (_mm_cvtpd_ps): Likewise.
16098         (_mm_cvttpd_epi32): Likewise.
16099         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
16100         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
16101         vector, pixel, and bool following altivec.h include.
16103 2018-04-03  Martin Sebor  <msebor@redhat.com>
16105         * doc/extend.texi (Common Function Attributes): Clarify.
16106         (const attribute): Likewise.
16107         (pure attribute): Likewise.
16109 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
16111         PR target/85169
16112         * config/i386/i386.c (ix86_expand_vector_set): Use
16113         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
16115 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
16117         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
16118         instructions when changing rounding bits to preserve precision bits
16119         in the x87 control word.
16121 2018-04-03  Martin Liska  <mliska@suse.cz>
16123         PR tree-optimization/82491
16124         * rtl.h (strip_offset_and_add): Replace += suboffset with
16125         poly_uint64 () + suboffset.
16127 2018-03-29  Martin Liska  <mliska@suse.cz>
16128             Martin Jambor  <mjambor@suse.cz>
16130         PR ipa/84947
16131         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
16132         param_type is not an integral or pointer type.
16134 2018-04-03  Richard Biener  <rguenther@suse.de>
16136         * sese.h (recompute_all_dominators): Remove.
16138 2018-04-02  Martin Sebor  <msebor@redhat.com>
16140         * doc/invoke.texi (-Wrestrict): Fix typos.
16142 2018-04-02  Jim Wilson  <jimw@sifive.com>
16144         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
16145         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
16146         (<optab>di3, <optab>si3_extend): Likewise.
16147         (<optab>si3_mask, <optab>si3_mask_1): New.
16148         (<optab>di3_mask, <optab>di3_mask_1): New.
16149         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
16150         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
16151         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
16153 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
16155         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
16156         example.
16158 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
16160         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
16161         (nds32_canonicalize_comparison): New function.
16163 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16164             Kito Cheng  <kito.cheng@gmail.com>
16165             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
16167         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
16168         * config/nds32/constants.md (unspec_volatile_element): Add
16169         UNSPEC_VOLATILE_RELAX_GROUP.
16170         * config/nds32/nds32-relax-opt.c: New file.
16171         * config/nds32/nds32-predicates.c
16172         (nds32_symbol_load_store_p): New function.
16173         * config/nds32/nds32-protos.h
16174         (nds32_symbol_load_store_p): Declare function.
16175         (make_pass_nds32_relax_opt): Declare new rtl pass function.
16176         * config/nds32/nds32.c
16177         (nds32_register_pass): New function to register pass.
16178         (nds32_register_passes): New function to register passes.
16179         * config/nds32/nds32.md (relax_group): New pattern.
16180         * config/nds32/nds32.opt (mrelax-hint): New option.
16181         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
16183 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
16185         * config/nds32/t-nds32: Modify files dependency.
16187 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16189         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
16190         (PROFILE_HOOK): Define its implementation.
16192 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
16194         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
16195         type and 32-bit size.
16197 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
16199         PR middle-end/85090
16200         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
16201         (V_128_256): New mode iterator.
16202         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
16203         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
16204         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
16205         of V.
16206         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
16207         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
16209 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
16211         PR target/83315
16212         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
16213         NaN inputs correctly.
16215 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
16217         PR target/80546
16218         * config/rs6000/vsx.md (??r): New mode attribute.
16219         (*vsx_mov<mode>_64bit): Use it.
16220         (*vsx_mov<mode>_32bit): Likewise.
16222 2018-03-30  Martin Sebor  <msebor@redhat.com>
16224         PR tree-optimization/84818
16225         * builtins.c (check_access): Use warning_n.
16227 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
16229         PR target/83822
16230         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16231         condition.
16232         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16233         condition.
16235 2018-03-30  Julia Koval  <julia.koval@intel.com>
16237         PR target/84413
16238         * x86-tune.def (movx, partial_reg_dependency): Enable for
16239         m_SKYLAKE_AVX512.
16241 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
16243         PR inline-asm/84985
16244         * lra-constraints.c (process_alt_operands): Move setting
16245         this_alternative_matches below.
16247 2018-03-29  Martin Liska  <mliska@suse.cz>
16249         PR lto/84995.
16250         * doc/invoke.texi: Document how LTO works with debug info.
16251         Describe auto-load support of binutils.  Mention 'x86-64'
16252         as valid option value of -march option.
16254 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
16256         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
16258         PR c/85094
16259         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
16260         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
16261         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
16262         checking.
16264 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
16266         PR target/84912
16267         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
16268         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
16269         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
16270         for RS6000_BTM_POWERPC64.
16271         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
16272         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
16273         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
16274         definition.
16275         (DIVDE): Use it.
16276         (DIVDEU): Likewise.
16278 2018-03-28  Carl Love  <cel@us.ibm.com>
16280         Revert
16281         2017-09-27  Carl Love  <cel@us.ibm.com>
16283         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
16284         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
16285         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
16286         fctiw instruction.
16288 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
16290         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
16291         instead of __vector bool.
16292         (_mm_max_pu8): Likewise.
16293         (_mm_min_pi16): Likewise.
16295 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
16297         PR target/84912
16298         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
16299         (DIVWEUO): Likewise.
16300         (DIVDEO): Likewise.
16301         (DIVDEUO): Likewise.
16302         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
16303         DIVWEUO and DIVDEUO.
16304         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
16305         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
16306         (div_extend): Likewise.
16307         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
16308         builtin function.
16309         (__builtin_divweuo): Likewise.
16310         (__builtin_divdeo): Likewise.
16311         (__builtin_divdeuo): Likewise.
16313 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
16315         PR target/85095
16316         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
16317         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
16319         PR tree-optimization/82004
16320         * gimple-match-head.c (optimize_pow_to_exp): New function.
16321         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
16322         Don't fold to exp if optimize_pow_to_exp is false.
16324 2018-03-28  Martin Liska  <mliska@suse.cz>
16326         PR other/84819
16327         * calls.c (initialize_argument_information): Fix trailing space.
16328         * common.opt: Fix typo and provide better explanation for
16329         -fsanitize-coverage option.
16330         * config/i386/i386.opt: Fix typo.
16332 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
16333             Martin Liska  <mliska@suse.cz>
16335         PR sanitizer/85081
16336         * gimplify.c (asan_poison_variable): Don't do the check for
16337         gimplify_omp_ctxp here.
16338         (gimplify_decl_expr): Do it here.
16339         (gimplify_target_expr): Likewise.
16341 2018-03-28  Martin Liska  <mliska@suse.cz>
16343         PR target/84988
16344         * config/i386/i386.c (ix86_function_arg_advance): Do not call
16345         chkp_type_bounds_count if MPX is not enabled.
16347 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
16349         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
16351 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
16353         PR target/84914
16354         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
16355         function to create the function decl for complex long double
16356         multiply and divide for -mabi=ieeelongdouble.
16357         (init_float128_ieee): Call it.
16359 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
16361         PR target/85044
16362         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
16363         -fcf-protection=branch -mibt.
16364         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
16366 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16368         PR target/81863
16369         * config/arm/arm.c (arm_valid_symbolic_address): Handle
16370         arm_word_relocations.
16372 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
16374         PR target/85056
16375         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
16376         extern array declarations.
16378 2018-03-27  Richard Biener  <rguenther@suse.de>
16380         PR middle-end/84067
16381         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
16382         explicit single_use checks.
16384 2018-03-27  Richard Biener  <rguenther@suse.de>
16386         PR tree-optimization/85082
16387         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
16388         Valueize the VUSE.
16390 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
16392         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
16393         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
16394         Turn on fasynchronous-unwind-tables and funwind-tables.
16396 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
16398         PR target/85073
16399         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
16400         (*bmi_blsr_<mode>_ccz): Ditto.
16402 2018-03-26  Tom de Vries  <tom@codesourcery.com>
16404         PR tree-optimization/85063
16405         * omp-general.c (offloading_function_p): New function.  Factor out
16406         of ...
16407         * omp-offload.c (pass_omp_target_link::gate): ... here.
16408         * omp-general.h (offloading_function_p): Declare.
16409         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
16410         with attribute omp declare target for offloading functions.
16412 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
16414         PR tree-optimization/84005
16415         * tree-data-ref.h (get_base_for_alignment): Declare.
16416         * tree-data-ref.c (get_base_for_alignment_1): New function.
16417         (get_base_for_alignment): Likewise.
16418         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
16419         get_base_for_alignment to find a suitable base object, instead
16420         of always using drb->base_address.
16422 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
16424         PR inline-asm/85022
16425         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
16426         known size by default.
16428 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
16430         PR inline-asm/85030
16431         * lra-constraints.c (process_alt_operands): Don't match BLKmode
16432         and non BLKmode operands.
16434 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16436         PR target/85026
16437         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16438         Clean up attributes.
16440 2018-03-23  Richard Biener  <rguenther@suse.de>
16442         PR debug/85020
16443         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
16444         we are going to emit early debug for LTO.
16446 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
16448         PR inline-asm/85034
16449         * function.c (match_asm_constraints_1): Don't optimize if input
16450         doesn't satisfy general_operand predicate for output's mode.
16452         PR inline-asm/85022
16453         * alias.c (write_dependence_p): Don't require for x_canonicalized
16454         non-VOIDmode if x has VOIDmode.
16456         PR sanitizer/85029
16457         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
16458         just don't try to optimize it rather than assert it never happens.
16460 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16462         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
16463         macro expansions for definition of ST_INTERNAL_<mode> and
16464         LD_INTERNAL_<mode> builtins.
16465         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
16466         Remove prototype.
16467         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
16468         function.
16469         (altivec_expand_st_builtin): Likewise.
16470         (altivec_expand_builtin): Remove calls to deleted functions.
16471         (rs6000_address_for_altivec): Delete this function.
16472         * config/rs6000/vector.md: Remove expands for
16473         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
16475 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
16477         PR target/84826
16478         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
16479         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
16480         re-computing once computed.
16481         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
16482         (arm_init_machine_status): Initialize
16483         machine->static_chain_stack_bytes.
16485 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16487         PR target/84760
16488         * doc/extend.texi: Add four new prototypes for vec_ld.
16489         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
16490         definitions for more logical presentation.
16491         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
16492         entries for V1TI variants of __builtin_altivec_ld builtin.
16493         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
16494         handling of V1TI variant of LVX icode pattern.
16495         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
16496         (rs6000_gimple_fold_builtin): Likewise.
16497         (altivec_init_builtins): Add code to define
16498         __builtin_altivec_lvx_v1ti function.
16500 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
16502         PR inline-asm/84941
16503         * function.c (match_asm_constraints_1): Don't do the optimization
16504         if input isn't a REG, SUBREG, MEM or constant.
16506 2018-03-22  Tom de Vries  <tom@codesourcery.com>
16508         PR tree-optimization/84956
16509         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
16510         bb_has_abnormal_pred.
16512 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
16514         PR sanitizer/85018
16515         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
16516         DECL_INITIAL (decl) to decl at the end.
16517         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
16518         adjust the comment.
16520 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
16522         * doc/extend.texi (__builtin_tgmath): Document when complex
16523         integer types are treated as _Complex _Float64.
16525 2018-03-21  Tom de Vries  <tom@codesourcery.com>
16527         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
16529 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16531         PR tree-optimization/84960
16532         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
16533         if it is ENTRY block, move them into single succ of ENTRY in that case.
16535 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
16537         PR tree-optimization/84811
16538         * poly-int.h (poly_span_traits): Remove the T3 parameter and
16539         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
16540         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
16541         (known_subrange_p): Update accordingly.  Cast each value involved
16542         in the size comparison, rather than casting the result of the
16543         subtraction.
16545 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16547         PR tree-optimization/84982
16548         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
16549         by flipping the least significant bit rather than all bits from
16550         bitpos to bitpos + bitsize - 1.
16552 2018-03-21  Nathan Sidwell  <nathan@acm.org>
16554         * doc/extend.texi (Deprecated Features): Remove mention of
16555         long-deleted deprecations.
16557 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16559         PR jit/84288
16560         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
16561         * configure: Regenerate.
16563 2018-03-21  Tom de Vries  <tom@codesourcery.com>
16565         PR tree-optimization/83126
16566         * tree-parloops.c (num_phis): New function.
16567         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
16569 2018-03-21  Nathan Sidwell  <nathan@acm.org>
16571         * doc/extend.texi (Deprecated Features): Update deprecated flags,
16572         mention anon-struct/union members and trailing attributes.
16574 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
16576         PR tree-optimization/84969
16577         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
16578         builtin memset partitions if they set different rhs values.
16580 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
16582         PR rtl-optimization/84989
16583         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
16584         VEC_DUPLICATE with scalar result mode.
16586 2018-03-21  Martin Liska  <mliska@suse.cz>
16588         PR ipa/84963
16589         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
16590         not intended return statement.
16592 2018-03-21  Martin Liska  <mliska@suse.cz>
16594         PR target/84988
16595         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
16596         (chkp_find_bound_slots_1): Limit number of iterations.
16598 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
16600         PR target/84838
16601         * Minor grammar fixes for x86 options.
16603 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
16605         PR debug/84875
16606         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
16607         holding REG_CFA_RESTORE notes, instead turn them into a USE.
16609 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
16611         PR target/83789
16612         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
16613         (altivec_lvx_<mode>_1op): Likewise.
16614         (altivec_stvx_<mode>_2op): Likewise.
16615         (altivec_stvx_<mode>_1op): Likewise.
16616         (altivec_lvx_<VM2:mode>): New define_expand.
16617         (altivec_stvx_<VM2:mode>): Likewise.
16618         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
16619         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16620         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
16621         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
16622         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
16623         (rs6000_gen_lvx): Likewise.
16624         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
16625         (altivec_expand_stv_builtin): Likewise.
16626         (altivec_expand_builtin): Likewise.
16627         * config/rs6000/vector.md: Likewise.
16629 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16631         PR target/82518
16632         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16633         BYTES_BIG_ENDIAN.
16635 2018-03-20  Richard Biener  <rguenther@suse.de>
16637         PR target/84986
16638         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16639         sign-conversions as zero, fall back to standard scalar_stmt
16640         cost for the rest.
16642 2018-03-20  Martin Liska  <mliska@suse.cz>
16644         PR ipa/84825
16645         * predict.c (rebuild_frequencies): Handle case when we have
16646         PROFILE_ABSENT, but flag_guess_branch_prob is false.
16648 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
16650         PR target/84990
16651         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
16652         flag_section_anchors.
16653         * varasm.c (use_blocks_for_decl_p): Remove hack for
16654         dw2_force_const_mem.
16656         PR target/84845
16657         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
16658         to ...
16659         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
16660         be created, use lowpart_subreg of operands[0] rather than operands[0]
16661         itself.
16662         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
16663         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
16664         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
16665         and n constraint instead of aarch64_shift_imm_di and Usd.
16666         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
16667         (*aarch64_<optab>_reg_minus<mode>3): ... this.
16669 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
16671         PR target/82989
16672         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
16673         to favor GPR over NEON registers.
16674         (<shift>di3_neon): Likewise.
16676 2018-03-20  Tom de Vries  <tom@codesourcery.com>
16678         PR target/84952
16679         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
16680         (nvptx_process_pars): Emit bar.sync asap and alap.
16682 2018-03-20  Tom de Vries  <tom@codesourcery.com>
16684         PR target/84954
16685         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
16686         seen_label if seen_label is already set.
16688 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
16690         PR target/84945
16691         * config/i386/i386.c (fold_builtin_cpu): For features above 31
16692         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
16693         Use 1U instead of 1.  Formatting fixes.
16695         PR c/84953
16696         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
16697         instead of TREE_TYPE (s1) for the return value.
16699 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
16701         PR tree-optimization/84946
16702         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
16703         bitsize + bitsize in poly_uint64 rather than poly_int64.
16705         PR sanitizer/78651
16706         * dwarf2asm.c: Include fold-const.c.
16707         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
16708         of decl rather than decl itself.
16710         PR rtl-optimization/84643
16711         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
16713 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
16715         PR sanitizer/78651
16716         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
16717         calling assemble_variable.
16719 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
16721         PR target/81647
16722         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
16723         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
16725 2018-03-19  Jim Wilson  <jimw@sifive.com>
16727         PR bootstrap/84856
16728         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
16729         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
16730         (riscv_first_stack_step): Likewise.
16731         (riscv_option_override): Use STACK_BOUNDARY instead of
16732         MIN_STACK_BOUNDARY.
16733         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
16734         MIN_STACK_BOUNDARY.
16735         (BIGGEST_ALIGNMENT): Set to 128.
16736         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
16737         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
16738         STACK_BOUNDARY.
16740 2018-03-19  Richard Biener  <rguenther@suse.de>
16742         PR tree-optimization/84933
16743         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
16744         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
16746 2018-03-19  Richard Biener  <rguenther@suse.de>
16748         PR tree-optimization/84859
16749         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
16750         (cond_if_else_store_replacement): Perform sinking operation on
16751         single-store BBs regardless of MAX_STORES_TO_SINK setting.
16752         Generalize what a BB with a single eligible store is.
16754 2018-03-19  Richard Biener  <rguenther@suse.de>
16756         PR tree-optimization/84929
16757         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
16758         chrec_is_positive against non-chrec arg.
16760 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
16762         PR target/84711
16763         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
16765 2018-03-18  Martin Liska  <mliska@suse.cz>
16767         PR rtl-optimization/84635
16768         * regrename.c (build_def_use): Use matches_mode only when
16769         matches >= 0.
16771 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
16773         PR tree-optimization/84913
16774         * tree-vect-loop.c (vectorizable_reduction): Don't try to
16775         vectorize chains of COND_EXPRs.
16777 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
16779         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
16781 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
16783         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
16785 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
16787         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
16789 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
16790             Kito Cheng  <kito.cheng@gmail.com>
16792         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
16793         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
16794         (nds32_adjust_reg_alloc_order): New function.
16795         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
16797 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
16799         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
16800         nds32_print_operand, nds32_print_operand_address): Use
16801         HOST_WIDE_INT_PRINT_DEC instead.
16803 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
16805         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
16807 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
16809         PR target/84902
16810         * config/i386/i386.c (initial_ix86_tune_features,
16811         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
16812         unsigned long long.
16813         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
16814         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
16815         rather than 1u << ix86_tune.  Formatting fix.
16816         (ix86_option_override_internal): Change ix86_arch_mask from
16817         unsigned int to unsigned HOST_WIDE_INT, initialize to
16818         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
16819         (ix86_function_specific_restore): Likewise.
16821 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
16823         PR target/84899
16824         * postreload.c (reload_combine_recognize_pattern): Perform
16825         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
16826         truncate_int_for_mode the result for the destination's mode.
16828         PR c/84909
16829         * hsa-gen.c (mem_type_for_type): Fix comment typo.
16830         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16831         Likewise.
16832         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16833         Likewise.
16835 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
16837         PR target/84876
16838         * lra-assigns.c (lra_split_hard_reg_for): Don't use
16839         regno_allocno_class_array and sorted_pseudos.
16840         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
16841         insns where regno is used.
16843 2018-03-16  Martin Liska  <mliska@suse.cz>
16845         PR ipa/84833
16846         * multiple_target.c (create_dispatcher_calls): Redirect
16847         reference in the symbol table.
16849 2018-03-16  Martin Liska  <mliska@suse.cz>
16851         PR ipa/84722
16852         * multiple_target.c (create_dispatcher_calls): Redirect also
16853         an alias.
16855 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
16857         PR c++/79937
16858         PR c++/82410
16859         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
16860         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
16861         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
16863 2018-03-16  Julia Koval  <julia.koval@intel.com>
16865         * doc/invoke.texi (Skylake Server): Add CLWB.
16866         Cannonlake): Remove CLWB.
16868 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
16870         PR tree-optimization/84841
16871         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
16872         1 << 3.
16873         (FLOAT_ONE_CONST_TYPE): Define.
16874         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
16875         (sort_by_operand_rank): Put entries with higher constant_type last
16876         rather than first to match comments.
16878 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
16880         * config/nios2/nios2.md (movsi_internal): Fix thinko in
16881         split predicate.
16883 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
16885         PR c++/79085
16886         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
16887         check and use address of target always.
16889 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
16891         PR target/84574
16892         * config/i386/i386.c (indirect_thunk_needed): Update comments.
16893         (indirect_thunk_bnd_needed): Likewise.
16894         (indirect_thunks_used): Likewise.
16895         (indirect_thunks_bnd_used): Likewise.
16896         (indirect_return_needed): New.
16897         (indirect_return_bnd_needed): Likewise.
16898         (output_indirect_thunk_function): Add a bool argument for
16899         function return.
16900         (output_indirect_thunk_function): Don't generate alias for
16901         function return thunk.
16902         (ix86_code_end): Call output_indirect_thunk_function to generate
16903         function return thunks.
16904         (ix86_output_function_return): Set indirect_return_bnd_needed
16905         and indirect_return_needed instead of indirect_thunk_bnd_needed
16906         and indirect_thunk_needed.
16908 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
16910         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
16911         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
16912         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
16914 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
16915             Paul Hua <paul.hua.gm@gmail.com>
16917         PR c/84852
16918         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
16920 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
16922         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
16923         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
16924         resp. SFmode cases.
16926 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
16928         PR target/84711
16929         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
16930         instead of GET_MODE_SIZE when comparing Units.
16932 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
16934         PR target/68256
16935         * varasm.c (hash_section): Return an unchangeble hash value
16936         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
16937         Return !aarch64_can_use_per_function_literal_pools_p ().
16939 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
16941         PR target/84860
16942         * optabs.c (emit_conditional_move): Pass address of cmode's copy
16943         rather than address of cmode as last argument to prepare_cmp_insn.
16945 2018-03-15  Julia Koval  <julia.koval@intel.com>
16947         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
16948         F_AVX512VNNI, F_AVX512BITALG): New.
16950 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
16952         PR target/83451
16953         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
16954         insn for floating-point loads and stores.
16956 2018-03-14  Carl Love  <cel@us.ibm.com>
16958         * config/rs6000/rs6000-c.c: Add macro definitions for
16959         ALTIVEC_BUILTIN_VEC_PERMXOR.
16960         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
16961         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
16962         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
16963         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
16964         UNSPEC_VPERMXOR.
16965         * config/doc/extend.texi: Add prototypes for vec_permxor.
16967 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
16969         PR c/84852
16970         * diagnostic-show-locus.c (class layout_point): Convert m_line
16971         from int to linenum_type.
16972         (line_span::comparator): Use linenum "compare" function when
16973         comparing line numbers.
16974         (test_line_span): New function.
16975         (layout_range::contains_point): Convert param "row" from int to
16976         linenum_type.
16977         (layout_range::intersects_line_p): Likewise.
16978         (layout::will_show_line_p): Likewise.
16979         (layout::print_source_line): Likewise.
16980         (layout::should_print_annotation_line_p): Likewise.
16981         (layout::print_annotation_line): Likewise.
16982         (layout::print_leading_fixits): Likewise.
16983         (layout::annotation_line_showed_range_p): Likewise.
16984         (struct line_corrections): Likewise for field m_row.
16985         (line_corrections::line_corrections): Likewise for param "row".
16986         (layout::print_trailing_fixits): Likewise.
16987         (layout::get_state_at_point): Likewise.
16988         (layout::get_x_bound_for_row): Likewise.
16989         (layout::print_line): Likewise.
16990         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
16991         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
16992         * input.c (selftest::test_linenum_comparisons): New function.
16993         (selftest::input_c_tests): Call it.
16994         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
16995         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
16996         * selftest.h (ASSERT_GT): New macro.
16997         (ASSERT_GT_AT): New macro.
16998         (ASSERT_LT): New macro.
16999         (ASSERT_LT_AT): New macro.
17001 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
17003         PR rtl-optimization/84780
17004         * combine.c (distribute_links): Don't make a link based on pc_rtx.
17006 2018-03-14  Martin Liska  <mliska@suse.cz>
17008         * tree.c (record_node_allocation_statistics): Use
17009         get_stats_node_kind.
17010         (get_stats_node_kind): New function extracted from
17011         record_node_allocation_statistics.
17012         (free_node): Use get_stats_node_kind.
17014 2018-03-14  Richard Biener  <rguenther@suse.de>
17016         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
17017         that the value-set of ANTIC_IN doesn't grow.
17019         Revert
17020         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17021         member.
17022         (BB_VISITED_WITH_VISITED_SUCCS): New define.
17023         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17025 2018-03-14  Julia Koval  <julia.koval@intel.com>
17027         * config.gcc (icelake-client, icelake-server): New.
17028         (icelake): Remove.
17029         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
17030         (initial_ix86_arch_features): Ditto.
17031         (PTA_SKYLAKE): Add SGX.
17032         (PTA_ICELAKE): Remove.
17033         (PTA_ICELAKE_CLIENT): New.
17034         (PTA_ICELAKE_SERVER): New.
17035         (ix86_option_override_internal): Split up icelake on icelake client and
17036         icelake server.
17037         (get_builtin_code_for_version): Ditto.
17038         (fold_builtin_cpu): Ditto.
17039         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
17040         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
17041         * config/i386/i386.h (processor_type): Ditto.
17042         * doc/invoke.texi: Ditto.
17044 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
17046         PR sanitizer/83392
17047         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
17048         INTEGER_CST offset, add it together with bitpos / 8 and
17049         sign extend based on POINTER_SIZE.
17051         PR target/84844
17052         Revert
17053         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
17055         PR target/78090
17056         * config/i386/constraints.md (Yc): New register constraint.
17057         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
17058         Use Yc constraint for alternative 2 of operand 0.  Remove
17059         preferred_for_speed attribute.
17061 2018-03-14  Richard Biener  <rguenther@suse.de>
17063         PR tree-optimization/84830
17064         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
17065         with the old one to avoid oscillations.
17067 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
17069         PR target/83712
17070         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
17071         pseudos.
17072         (assign_by_spills): Return a flag of reload assignment failure.
17073         Do not process the reload assignment failures.  Do not spill other
17074         reload pseudos if they has the same reg class.  Update n if
17075         necessary.
17076         (lra_assign): Add a return arg.  Set up from the result of
17077         assign_by_spills call.
17078         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17079         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17080         usage_insns if it is not NULL.
17081         (spill_hard_reg_in_range): New function.
17082         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17083         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17084         function prototypes.
17085         (lra_assign): Change prototype.
17086         * lra.c (lra): Add code to deal with fails by splitting hard reg
17087         live ranges.
17089 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
17091         * config/riscv/riscv.opt (mrelax): New option.
17092         * config/riscv/riscv.c (riscv_file_start): Emit ".option
17093         "norelax" when riscv_mrelax is disabled.
17094         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
17096 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17098         PR target/84743
17099         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
17100         reassociation for int modes.
17102 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17104         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
17105         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
17106         for big-endian.
17107         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
17108         * config/aarch64/aarch64-sve.md
17109         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
17110         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
17111         (*extend<mode><Vwide>2): Rename to...
17112         (aarch64_sve_extend<mode><Vwide>2): ...this.
17113         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
17114         renaming the old pattern to...
17115         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
17116         unsigned packs.
17117         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
17118         define_expand, renaming the old pattern to...
17119         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
17120         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
17121         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
17122         account when deciding which SVE instruction the optab should use.
17123         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
17125 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17127         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
17128         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
17129         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
17130         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
17131         (tlsdesc_small_<mode>): Turn a define_expand and use
17132         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
17133         (tlsdesc_small_advsimd_<mode>): ...this.
17134         (tlsdesc_small_sve_<mode>): New pattern.
17136 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
17138         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
17139         (UNSPEC_UMUL_HIGHPART): New constants.
17140         (MUL_HIGHPART): New int iteraor.
17141         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
17142         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
17143         define_expand.
17144         (*<su>mul<mode>3_highpart): New define_insn.
17146 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
17148         PR lto/84805
17149         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17150         incomplete types.
17152 2018-03-13  Martin Liska  <mliska@suse.cz>
17154         PR ipa/84658.
17155         * (sem_item_optimizer::sem_item_optimizer): Initialize new
17156         vector.
17157         (sem_item_optimizer::~sem_item_optimizer): Release it.
17158         (sem_item_optimizer::merge_classes): Register variable aliases.
17159         (sem_item_optimizer::fixup_pt_set): New function.
17160         (sem_item_optimizer::fixup_points_to_sets): Likewise.
17161         * ipa-icf.h: Declare new variables and functions.
17163 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
17165         PR middle-end/84834
17166         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
17167         integer_pow2p@2 and test integer_pow2p in condition.
17168         (A < 0 ? C : 0): Similarly for @1.
17170         PR middle-end/84831
17171         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
17172         characters starting at p contain '\0' character, don't look beyond
17173         that.
17175         PR target/84827
17176         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17177         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
17179         PR target/84828
17180         * reg-stack.c (change_stack): Change update_end var from int to
17181         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
17182         also call set_block_for_insn on the newly added insns and rescan.
17184         PR target/84786
17185         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17186         on the last operand.
17188         PR c++/84704
17189         * tree.c (stabilize_reference_1): Return save_expr (e) for
17190         STATEMENT_LIST even if it doesn't have side-effects.
17192 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
17194         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
17196 2018-03-12  Renlin Li  <renlin.li@arm.com>
17198         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
17199         aarch64_output_scalar_simd_mov_immediate.
17201 2018-03-12  Martin Sebor  <msebor@redhat.com>
17203         PR tree-optimization/83456
17204         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
17205         for perfectly overlapping calls to memcpy.
17206         (gimple_fold_builtin_memory_chk): Same.
17207         (gimple_fold_builtin_strcpy): Handle no-warning.
17208         (gimple_fold_builtin_stxcpy_chk): Same.
17209         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
17211 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
17213         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
17214         parameter.  Use it for SFmode.
17215         (rs6000_function_arg_advance_1): Adjust.
17216         (rs6000_function_arg): Adjust.
17217         (rs6000_gimplify_va_arg): Pass false for that new parameter.
17219 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
17221         PR rtl-optimization/84169
17222         PR rtl-optimization/84780
17223         * combine.c (can_combine_p): Check for a 2-insn combination whether
17224         the destination register is used between the two insns, too.
17226 2018-03-12  Richard Biener  <rguenther@suse.de>
17228         PR tree-optimization/84803
17229         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
17230         for refs DR analysis didn't process.
17232 2018-03-12  Richard Biener  <rguenther@suse.de>
17234         PR tree-optimization/84777
17235         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
17236         force-vectorize loops ignore whether we are optimizing for size.
17238 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
17240         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
17241         (TARGET_MD_ASM_ADJUST): Define.
17243 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
17244             Kito Cheng  <kito.cheng@gmail.com>
17245             Chung-Ju Wu  <jasonwucj@gmail.com>
17247         * config/nds32/nds32.c (nds32_compute_stack_frame,
17248         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
17249         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
17250         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
17251         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
17252         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
17253         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
17254         * config/nds32/nds32.md (prologue, epilogue): Use macro
17255         NDS32_V3PUSH_AVAILABLE_P to do checking.
17257 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
17259         PR debug/58150
17260         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
17261         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
17262         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
17263         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
17264         addition of most attributes on !orig_type_die or the attribute not
17265         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
17267 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17268             Chung-Ju Wu  <jasonwucj@gmail.com>
17270         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
17271         __NDS32_VH__ macro.
17272         * config/nds32/nds32.opt (mvh): New option.
17274 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17275             Chung-Ju Wu  <jasonwucj@gmail.com>
17277         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
17278         function.
17279         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17280         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
17281         definition.
17283 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
17284             Chung-Ju Wu  <jasonwucj@gmail.com>
17286         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
17287         function.
17288         * config/nds32/nds32-multiple.md (strlensi): New pattern.
17289         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
17291 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
17292             Kito Cheng  <kito.cheng@gmail.com>
17293             Chung-Ju Wu  <jasonwucj@gmail.com>
17295         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
17296         UNSPEC_FFMISM and UNSPEC_FLMISM.
17297         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
17298         for ffb, ffmism and flmism.
17299         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
17300         (unspec_ffmism): Ditto.
17301         (unspec_flmism): Ditto.
17302         (nds32_expand_builtin_impl): Check if string extension is available.
17303         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
17304         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
17306 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
17308         Reverting patch:
17309         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
17311         PR target/83712
17312         * lra-assigns.c (assign_by_spills): Return a flag of reload
17313         assignment failure.  Do not process the reload assignment
17314         failures.  Do not spill other reload pseudos if they has the same
17315         reg class.
17316         (lra_assign): Add a return arg.  Set up from the result of
17317         assign_by_spills call.
17318         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17319         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17320         usage_insns if it is not NULL.
17321         (spill_hard_reg_in_range): New function.
17322         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17323         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17324         function prototypes.
17325         (lra_assign): Change prototype.
17326         * lra.c (lra): Add code to deal with fails by splitting hard reg
17327         live ranges.
17329 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
17331         PR target/84807
17332         * config/i386/i386.opt: Replace Enforcment with Enforcement.
17334 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
17336         PR debug/84620
17337         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
17338         (dw_val_node): Add val_symbolic_view.
17339         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
17340         (symview_upper_bound): New.
17341         (new_line_info_table): Initialize symviews_since_reset.
17342         (dwarf2out_source_line): Count symviews_since_reset and set
17343         symview_upper_bound.
17344         (dw_val_equal_p): Handle symview.
17345         (add_AT_symview): New.
17346         (print_dw_val): Handle symview.
17347         (attr_checksum, attr_checksum_ordered): Likewise.
17348         (same_dw_val_p, size_of_die): Likewise.
17349         (value_format, output_die): Likewise.
17350         (add_high_low_attributes): Use add_AT_symview for entry_view.
17351         (dwarf2out_finish): Reset symview_upper_bound, clear
17352         zero_view_p.
17354 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
17356         PR target/83969
17357         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
17358         Add strict argument and use it.
17359         (rs6000_split_multireg_move): Update for new strict argument.
17360         (mem_operand_gpr): Disallow all non-offsettable addresses.
17361         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
17363 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
17365         PR target/84772
17366         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
17367         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
17368         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
17370         PR c++/84767
17371         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
17372         decl, use remap_type if we want to use the type.
17374 2018-03-09  Martin Sebor  <msebor@redhat.com>
17376         PR tree-optimization/84526
17377         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
17378         Remove dead code.
17379         (builtin_access::generic_overlap): Be prepared to handle non-array
17380         base objects.
17382 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
17384         PR rtl-optimization/84682
17385         * lra-constraints.c (process_address_1): Check is_address flag
17386         for address constraints.
17387         (process_alt_operands): Likewise.
17388         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
17389         preprocess_constraints.
17390         * recog.h (preprocess_constraints): Add oploc parameter.
17391         Adjust callers.
17392         * recog.c (preprocess_constraints): Test address_operand for
17393         CT_ADDRESS constraints.
17395 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
17397         PR target/83712
17398         * lra-assigns.c (assign_by_spills): Return a flag of reload
17399         assignment failure.  Do not process the reload assignment
17400         failures.  Do not spill other reload pseudos if they has the same
17401         reg class.
17402         (lra_assign): Add a return arg.  Set up from the result of
17403         assign_by_spills call.
17404         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
17405         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
17406         usage_insns if it is not NULL.
17407         (spill_hard_reg_in_range): New function.
17408         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
17409         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
17410         function prototypes.
17411         (lra_assign): Change prototype.
17412         * lra.c (lra): Add code to deal with fails by splitting hard reg
17413         live ranges.
17415 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17417         PR target/83193
17418         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
17419         Accept complain bool parameter.  Only emit errors if it is true.
17420         (arm_parse_cpu_option_name): Likewise.
17421         (arm_target_thumb_only): Adjust callers of the above.
17422         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
17423         prototype to take a default true bool parameter.
17424         (arm_parse_arch_option_name): Likewise.
17426 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
17427             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
17429         PR jit/64089
17430         PR jit/84288
17431         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
17432         * configure: Regenerate.
17433         * configure.ac ("linker --version-script option"): New.
17434         ("linker soname option"): New.
17436 2018-03-09  Richard Biener  <rguenther@suse.de>
17438         PR tree-optimization/84775
17439         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
17440         immediate uses of predicate stmts and mark them modified.
17442         Revert
17443         PR tree-optimization/84178
17444         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17445         to caller.
17446         (version_loop_for_if_conversion): Delay update_ssa call.
17447         (tree_if_conversion): Delay update_ssa until after predicate
17448         insertion.
17450 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
17452         PR target/84763
17453         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
17454         when the function accesses prior frames.
17456 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
17458         PR debug/84456
17459         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
17460         gen_llsym, otherwise call maybe_gen_llsym.
17462         PR inline-asm/84742
17463         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
17464         has ',' character inside of it.
17466 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17468         PR target/84748
17469         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
17470         as clobbering CC_REGNUM.
17472 2018-03-08  Richard Biener  <rguenther@suse.de>
17474         PR middle-end/84552
17475         * tree-scalar-evolution.c: Include tree-into-ssa.h.
17476         (follow_copies_to_constant): Do not follow SSA names registered
17477         for update.
17479 2018-03-08  Richard Biener  <rguenther@suse.de>
17481         PR tree-optimization/84178
17482         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17483         to caller.
17484         (version_loop_for_if_conversion): Delay update_ssa call.
17485         (tree_if_conversion): Delay update_ssa until after predicate
17486         insertion.
17488 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
17490         PR tree-optimization/84178
17491         * tree-if-conv.c (release_bb_predicate): Remove the
17492         the assertion that the stmts have NULL use_ops.
17493         Discard the statements, asserting that they haven't
17494         yet been added to a BB.
17496 2018-03-08  Richard Biener  <rguenther@suse.de>
17498         PR tree-optimization/84746
17499         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
17500         (phi_translate): Pass in destination ANTIC_OUT set.
17501         (phi_translate_1): Likewise.  For a simplified result lookup
17502         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
17503         (phi_translate_set): Adjust.
17504         (do_pre_regular_insertion): Likewise.
17505         (do_pre_partial_partial_insertion): Likewise.
17507 2018-03-08  Martin Liska  <mliska@suse.cz>
17509         PR gcov-profile/84735
17510         * doc/gcov.texi: Document usage of profile files.
17511         * gcov-io.h: Document changes in the format.
17513 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
17515         PR debug/84404
17516         PR debug/84408
17517         * dwarf2out.c (struct dw_line_info_table): Update comments for
17518         view == -1.
17519         (FORCE_RESET_NEXT_VIEW): New.
17520         (FORCE_RESETTING_VIEW_P): New.
17521         (RESETTING_VIEW_P): Check for -1 too.
17522         (ZERO_VIEW_P): Likewise.
17523         (new_line_info_table): Force-reset next view.
17524         (dwarf2out_begin_function): Likewise.
17525         (dwarf2out_source_line): Simplify zero_view_p initialization.
17526         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
17527         view directly.  Omit view when omitting .loc at line 0.
17529 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
17531         PR tree-optimization/84740
17532         * tree-switch-conversion.c (process_switch): Call build_constructors
17533         only if info.phi_count is non-zero.
17535         PR tree-optimization/84739
17536         * tree-tailcall.c (find_tail_calls): Check call arguments against
17537         DECL_ARGUMENTS (current_function_decl) rather than
17538         DECL_ARGUMENTS (func) when checking for tail recursion.
17540 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
17542         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
17543         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
17544         Volker Reichelt's entry and add entries for people that perform
17545         GCC fuzzy testing and report numerous bugs.
17547 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
17549         PR target/82411
17550         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
17551         readonly data in sdata, if that is disabled.
17552         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
17553         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
17554         -mreadonly-in-sdata option.
17556 2018-03-07  Martin Sebor  <msebor@redhat.com>
17558         PR tree-optimization/84468
17559         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
17560         basic block when looking for nul assignment.
17562 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
17564         PR target/84277
17565         * except.h (output_function_exception_table): Adjust prototype.
17566         * except.c (output_function_exception_table): Remove FNNAME parameter
17567         and add SECTION parameter.  Ouput one part of the table at a time.
17568         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
17569         the first part of the exception table and emit unwind directives.
17570         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
17571         (i386_pe_seh_cold_init): Likewise.
17572         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
17573         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
17574         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
17575         (ix86_output_call_insn): Emit a nop in one more case for SEH.
17576         * config/i386/winnt.c: Include except.h.
17577         (struct seh_frame_state): Add reg_offset, after_prologue and
17578         in_cold_section fields.
17579         (i386_pe_seh_end_prologue): Set seh->after_prologue.
17580         (i386_pe_seh_cold_init): New function.
17581         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
17582         to seh->in_cold_section.
17583         (seh_emit_push): Record the offset of the push.
17584         (seh_emit_save): Record the offet of the save.
17585         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
17586         Test seh->after_prologue to disregard the epilogue.
17587         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
17588         (i386_pe_end_cold_function): New function.
17590 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
17592         PR fortran/84565
17593         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17594         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17596         PR c++/84704
17597         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
17598         on tmp_var.
17599         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
17600         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
17602         PR middle-end/84723
17603         * multiple_target.c: Include tree-inline.h and intl.h.
17604         (expand_target_clones): Diagnose and fail if node->definition and
17605         !tree_versionable_function_p (node->decl).
17607 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
17609         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
17610         sprint_ul.
17611         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
17612         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
17613         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
17615 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
17617         PR target/84710
17618         * combine.c (try_combine): Use reg_or_subregno instead of handling
17619         just paradoxical SUBREGs and REGs.
17621 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
17623         * config/arc/arc.c (arc_finalize_pic): Remove function.
17624         (arc_must_save_register): We use single base PIC register, remove
17625         checks to save/restore the PIC register.
17626         (arc_expand_prologue): Likewise.
17627         * config/arc/arc-protos.h (arc_set_default_type_attributes):
17628         Remove.
17629         (arc_verify_short): Likewise.
17630         (arc_attr_type): Likewise.
17631         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
17632         (walk_stores): Likewise.
17633         (arc_address_cost): Make it static.
17634         (arc_verify_short): Likewise.
17635         (branch_dest): Likewise.
17636         (arc_attr_type): Likewise.
17637         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
17638         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
17639         (arc_final_prescan_insn): Remove inserting the nops due to
17640         hardware hazards.  It is done in reorg step.
17641         (insn_length_variant_t): Remove.
17642         (insn_length_parameters_t): Likewise.
17643         (arc_insn_length_parameters): Likewise.
17644         (arc_get_insn_variants): Likewise.
17645         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
17647 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
17649         PR inline-asm/84683
17650         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
17651         assertion failure.
17653         PR tree-optimization/84687
17654         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
17655         on new_node->decl.
17656         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
17658 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17660         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
17661         Rename to ppc_speculation_barrier.
17662         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17663         __builtin_ppc_speculation_barrier.
17665 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
17667         PR target/84700
17668         * combine.c (combine_simplify_rtx): Don't try to simplify if
17669         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
17670         are equal to x.
17672 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
17674         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
17675         to 32 bytes when compiling for POWER9.
17677 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
17679         PR target/84564
17680         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
17681         regparm >= 3 with no arg reg available also for calls with
17682         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
17684         PR target/84524
17685         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17686         orig,vex.
17687         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
17689 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
17691         PR target/84264
17692         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
17694 2018-03-05  Richard Biener  <rguenther@suse.de>
17696         PR tree-optimization/84486
17697         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
17698         When inserting a __builtin_assume_aligned call set the LHS
17699         SSA name alignment info accordingly.
17701 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
17703         PR tree-optimization/84114
17704         * config/aarch64/aarch64.c (aarch64_reassociation_width)
17705         Avoid reassociation of FLOAT_MODE addition.
17707 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
17709         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
17710         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
17711         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
17712         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
17713         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
17714         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
17715         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
17716         and -mwbnoinvd.
17717         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
17718         __builtin_ia32_wbinvd): New builtins.
17719         (SPECIAL_ARGS2): New.
17720         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
17721         (SPECIAL_ARGS2): New.
17722         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
17723         (ix86_valid_target_attribute_inner_p): Ditto.
17724         (ix86_init_mmx_sse_builtins): Add special_args2.
17725         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
17726         TARGET_WBNOINVD_P): New.
17727         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
17728         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
17729         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
17730         * config/i386/immintrin.h (_wbinvd): New intrinsic.
17731         * config/i386/pconfigintrin.h: New file.
17732         * config/i386/wbnoinvdintrin.h: Ditto.
17733         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
17734         wbnoinvdintrin.h.
17735         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
17737 2018-03-05  Richard Biener  <rguenther@suse.de>
17739         PR tree-optimization/84670
17740         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
17741         member.
17742         (BB_VISITED_WITH_VISITED_SUCCS): New define.
17743         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
17744         (compute_antic_aux): Only assert the number of values in ANTIC_IN
17745         doesn't grow if all successors (recursively) were visited at least
17746         once.
17748 2018-03-05  Richard Biener  <rguenther@suse.de>
17750         PR tree-optimization/84650
17751         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
17752         if executed in the loop pipeline.
17754 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
17756         * doc/configfiles.texi (Configuration Files): Move info about
17757         conditionalizing $target-protos.h to...
17758         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
17759         differs from $target-protos.h.
17761 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
17762             Chung-Ju Wu  <jasonwucj@gmail.com>
17764         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
17765         * config/nds32/nds32-multiple.md (setmemsi): Define.
17766         * config/nds32/nds32-memory-manipulation.c
17767         (nds32_gen_dup_4_byte_to_word_value): New.
17768         (emit_setmem_word_loop): New.
17769         (emit_setmem_byte_loop): New.
17770         (nds32_expand_setmem_loop): New.
17771         (nds32_expand_setmem_loop_v3m): New.
17772         (nds32_expand_setmem_unroll): New.
17773         (nds32_expand_setmem): New.
17775 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
17776             Chung-Ju Wu  <jasonwucj@gmail.com>
17778         * config/nds32/nds32-memory-manipulation.c
17779         (nds32_emit_load_store): New.
17780         (nds32_emit_post_inc_load_store): New.
17781         (nds32_emit_mem_move): New.
17782         (nds32_emit_mem_move_block): New.
17783         (nds32_expand_movmemsi_loop_unknown_size): New.
17784         (nds32_expand_movmemsi_loop_known_size): New.
17785         (nds32_expand_movmemsi_loop): New.
17786         (nds32_expand_movmemsi_unroll): New.
17787         (nds32_expand_movmemqi): Rename ...
17788         (nds32_expand_movmemsi): ... to this.
17789         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
17790         (movmemsi): ... to this.
17791         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
17792         (nds32_expand_movmemsi): ... to this.
17794 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
17795             Monk Chiang  <sh.chiang04@gmail.com>
17796             Chung-Ju Wu  <jasonwucj@gmail.com>
17798         * config/nds32/nds32-protos.h
17799         (nds32_expand_load_multiple): New arguments.
17800         (nds32_expand_store_multiple): Ditto.
17801         (nds32_valid_multiple_load_store): Rename ...
17802         (nds32_valid_multiple_load_store_p): ... to this.
17803         * config/nds32/nds32-memory-manipulation.c
17804         (nds32_expand_load_multiple): Refine implementation.
17805         (nds32_expand_store_multiple): Ditto.
17806         * config/nds32/nds32-multiple.md
17807         (load_multiple): Update nds32_expand_load_multiple interface.
17808         (store_multiple): Update nds32_expand_store_multiple interface.
17809         * config/nds32/nds32-predicates.c
17810         (nds32_valid_multiple_load_store): Rename ...
17811         (nds32_valid_multiple_load_store_p): ... to this and refine
17812         implementation.
17813         * config/nds32/predicates.md
17814         (nds32_load_multiple_and_update_address_operation): New predicate.
17815         (nds32_store_multiple_and_update_address_operation): New predicate.
17817 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
17818             Chung-Ju Wu  <jasonwucj@gmail.com>
17820         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
17821         (combo): New attribute.
17822         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
17824 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17826         * config/nds32/nds32.opt: Change -mcmodel= default value.
17828 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
17829             Monk Chiang  <sh.chiang04@gmail.com>
17830             Chung-Ju Wu  <jasonwucj@gmail.com>
17832         * config/nds32/constants.md (unspec_element): New enum.
17833         * config/nds32/constraints.md (Umw): New constraint.
17834         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
17835         * config/nds32/nds32-intrinsic.md: Likewise.
17836         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
17837         (nds32_valid_smw_lwm_base_p): New.
17838         (nds32_output_smw_single_word): New.
17839         (nds32_output_lmw_single_word): New.
17840         (nds32_expand_unaligned_load): New.
17841         (nds32_expand_unaligned_store): New.
17842         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
17843         (nds32_output_smw_single_word): Declare.
17844         (nds32_output_lmw_single_word): Declare.
17845         (nds32_expand_unaligned_load): Declare.
17846         (nds32_expand_unaligned_store): Declare.
17847         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
17848         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
17849         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
17850         NDS32_BUILTIN_UASTORE_DW.
17851         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
17852         predicate.
17854 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
17855             Kito Cheng  <kito.cheng@gmail.com>
17856             Chung-Ju Wu  <jasonwucj@gmail.com>
17858         * config/nds32/nds32-intrinsic.c
17859         (nds32_expand_builtin_null_ftype_reg): Delete.
17860         (nds32_expand_builtin_reg_ftype_imm): Ditto.
17861         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
17862         (nds32_read_argument): New.
17863         (nds32_legitimize_target): Ditto.
17864         (nds32_legitimize_argument): Ditto.
17865         (nds32_check_constant_argument): Ditto.
17866         (nds32_expand_unop_builtin): Ditto.
17867         (nds32_expand_unopimm_builtin): Ditto.
17868         (nds32_expand_binop_builtin): Ditto.
17869         (nds32_builtin_decl_impl): Ditto.
17870         (builtin_description): Ditto.
17871         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
17872         (nds32_init_builtins_impl): Ditto.
17873         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
17874         (nds32_builtin_decl): New.
17875         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
17876         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
17878 2018-03-02  Jeff Law  <law@redhat.com>
17880         * reorg.c (stop_search_p): Handle DEBUG_INSN.
17881         (redundant_insn, fill_simple_delay_slots): Likewise.
17882         (fill_slots_from_thread): Likewise.
17883         * resource.c (mark_referenced_resources): Likewise.
17884         (mark_set_resources, find_dead_or_set_registers): Likewise.
17886 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
17888         * substring-locations.h (format_warning_va): Formatting fix for
17889         ATTRIBUTE_GCC_DIAG.
17890         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
17891         argument.
17892         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
17893         * substring-locations.c: Include intl.h.
17894         (format_warning_va): Turned into small wrapper around
17895         format_warning_n_va, renamed to ...
17896         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
17897         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
17898         use ngettext.
17899         (format_warning_at_substring_n): New function.
17900         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
17901         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
17902         format_warning_at_substring with just a shorter name instead of
17903         const function pointer.
17904         (fmtwarn_n): New function.
17905         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
17906         appropriate, get rid of all the fmtstr temporaries, move conditionals
17907         with G_() wrapped string literals directly into fmtwarn arguments,
17908         cast dir.len to (int), formatting fixes.
17910 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
17912         * doc/invoke.texi: Remove "Cilk Plus" references.
17914 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
17915             Richard Biener  <rguenther@suse.de>
17917         PR ipa/84628
17918         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
17919         for error or warning attributes if CALL_FROM_THUNK_P is set.
17920         Formatting fixes.
17922 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
17924         PR target/56540
17925         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17926         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
17928         PR target/56540
17929         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
17930         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
17932         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
17933         instead of -1U in last predictors element's probability member.
17935 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
17937         PR ipa/83983
17938         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17939         arguments if they are comparable.
17941 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
17943         PR tree-optimization/84634
17944         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
17945         masks and masked_loop_p with a single loop_masks, making sure it's
17946         null for bb vectorization.
17948 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
17950         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
17951         (vect_analyze_data_ref_access): Use loop->safe_len rather than
17952         loop->force_vectorize to check whether there is no alias.
17954 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
17956         PR target/84614
17957         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17958         prototypes.
17959         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
17960         comments.
17961         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
17962         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
17963         instead of a loop around prev_real_insn.
17964         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
17965         prev_real_insn.
17967         PR inline-asm/84625
17968         * config/i386/i386.c (ix86_print_operand): Use conditional
17969         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
17970         zero vector.
17972 2018-03-02  Richard Biener  <rguenther@suse.de>
17974         PR tree-optimization/84427
17975         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
17976         (bitmap_set_subtract_values): Rewrite to handle multiple
17977         exprs per value.
17978         (clean): Likewise.
17979         (prune_clobbered_mems): Likewise.
17980         (phi_translate): Take edge instead of pred/phiblock.
17981         (phi_translate_1): Likewise.
17982         (phi_translate_set): Likewise.  Insert all translated
17983         exprs for a value into the set, keeping possibly multiple
17984         expressions per value.
17985         (compute_antic_aux): Adjust for phi_translate changes.
17986         When intersecting union the expressions and prune those
17987         not in the final value set, keeping possibly multiple
17988         expressions per value.  Do not use value-insertion
17989         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
17990         all expressions.  Add verification that the value-sets
17991         only shrink during iteration.
17992         (compute_partial_antic_aux): Adjust for the phi_translate changes.
17993         (do_pre_regular_insertion): Likewise.
17994         (do_pre_partial_partial_insertion): Likewise.
17996 2018-03-02  Richard Biener  <rguenther@suse.de>
17998         PR target/82005
17999         * config/darwin.c (saved_debug_info_level): New static global.
18000         (darwin_asm_lto_start): Disable debug info generation for LTO out.
18001         (darwin_asm_lto_end): Restore debug info generation settings.
18003 2018-03-01  Martin Liska  <mliska@suse.cz>
18005         PR sanitizer/82484
18006         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18007         volatile arguments.
18009 2018-03-01  Richard Biener  <rguenther@suse.de>
18011         PR debug/84645
18012         * dwarf2out.c (gen_variable_die): Properly handle late VLA
18013         type annotation with LTO when debug was disabled at compile-time.
18015 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
18017         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
18018         XINT with INTVAL.
18019         (mips_final_postscan_insn): Likewise.
18021 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
18023         PR rtl-optimization/84528
18024         * alias.c (init_alias_target): Add commentary.
18025         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
18026         a unique base value if the frame pointer is not eliminated
18027         to the stack pointer.
18029 2018-03-01  Tom de Vries  <tom@codesourcery.com>
18031         PR rtl-optimization/83327
18032         * lra-int.h (hard_regs_spilled_into): Declare.
18033         * lra.c (hard_regs_spilled_into): Define.
18034         (init_reg_info): Init hard_regs_spilled_into.
18035         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
18036         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
18037         (process_bb_lives): Handle hard_regs_spilled_into.
18038         (lra_create_live_ranges_1): Before doing liveness propagation, clear
18039         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
18041 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
18043         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
18044         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
18045         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
18046         * config/rs6000/aix72.h: New file.
18048 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
18050         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
18051         instead of warning_at with conditional singular and plural messages
18052         where possible.
18054         PR target/52991
18055         * stor-layout.c (update_alignment_for_field): For
18056         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
18057         && !DECL_PACKED (field), do the alignment update, just use
18058         only desired_align instead of MAX (type_align, desired_align)
18059         as the alignment.
18060         (place_field): Don't do known_align < desired_align handling
18061         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
18062         is non-NULL, instead do it after rli->prev_field handling and
18063         only if not within a bitfield word.  For DECL_PACKED (field)
18064         use type_align of BITS_PER_UNIT.
18066 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
18068         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
18069         superfluous parentheses and trailing spaces.
18071 2018-02-28  Richard Biener  <rguenther@suse.de>
18073         PR tree-optimization/84584
18074         * graphite-scop-detection.c (scop_detection::add_scop): Discard
18075         SCoPs with fake exit edge.
18077 2018-02-28  Martin Liska  <mliska@suse.cz>
18079         PR testsuite/84597
18080         * timevar.c (timer::print): Fix format to properly print 100%
18081         values.
18083 2018-02-28  Richard Biener  <rguenther@suse.de>
18085         PR middle-end/84607
18086         * genmatch.c (capture_info::walk_match): Do not mark
18087         captured expressions without operands as expr_p given
18088         they act more like predicates and should be subject to
18089         "lost tail" side-effect preserving.
18091 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
18093         PR rtl-optimization/81611
18094         * auto-inc-dec.c (attempt_change): Move dead note from
18095         mem_insn if it's the next use of regno
18096         (find_address): Take address use of reg holding
18097         non-incremented value.  Add parm to limit search to the named
18098         reg only.
18099         (merge_in_block): Attempt to use a mem insn that is the next
18100         use of the original regno.
18102 2018-02-27  Martin Sebor  <msebor@redhat.com>
18104         PR c++/83871
18105         * doc/invoke.texi (-Wmissing-attributes): New option.
18106         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
18108 2018-02-27  Martin Sebor  <msebor@redhat.com>
18110         PR translation/84207
18111         * diagnostic-core.h (warning_n, error_n, inform_n): Change
18112         n argument to unsigned HOST_WIDE_INT.
18113         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
18114         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
18115         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
18116         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
18118 2018-02-27  Richard Biener  <rguenther@suse.de>
18120         PR tree-optimization/84512
18121         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
18122         Do not use the estimate returned from record_stmt_cost for
18123         the scalar iteration cost but sum properly using add_stmt_cost.
18125 2018-02-27  Richard Biener  <rguenther@suse.de>
18127         PR tree-optimization/84466
18128         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18129         Adjust last change to less strictly validate use operands.
18131 2018-02-27  Martin Liska  <mliska@suse.cz>
18133         PR gcov-profile/84548
18134         * gcov.c (process_file): Allow partial overlap and consider it
18135         also as group functions.
18136         (output_lines): Properly calculate range of lines for a group.
18138 2018-02-27  Martin Liska  <mliska@suse.cz>
18140         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
18141         'ggc' suffixes.  Change first column width.
18142         (timer::print): Fix formatting of the column.
18144 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
18146         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
18147         preserve inline entry blocks for the sake of debug inline
18148         entry point markers alone.
18149         (remove_unused_locals): Suggest in comments a better place to
18150         force the preservation of inline entry blocks that are
18151         otherwise unused, but do not preserve them.
18153 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18155         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
18157 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18159         PR target/84039
18160         * config/i386/constraints.md (Bs): Replace
18161         ix86_indirect_branch_register with
18162         TARGET_INDIRECT_BRANCH_REGISTER.
18163         (Bw): Likewise.
18164         * config/i386/i386.md (indirect_jump): Likewise.
18165         (tablejump): Likewise.
18166         (*sibcall_memory): Likewise.
18167         (*sibcall_value_memory): Likewise.
18168         Peepholes of indirect call and jump via memory: Likewise.
18169         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
18170         (*sibcall_value_GOT_32): Likewise.
18171         * config/i386/predicates.md (indirect_branch_operand): Likewise.
18172         (GOT_memory_operand): Likewise.
18173         (call_insn_operand): Likewise.
18174         (sibcall_insn_operand): Likewise.
18175         (GOT32_symbol_operand): Likewise.
18176         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
18178 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
18180         PR rtl-optimization/83496
18181         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
18182         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
18183         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
18184         redundant insn, if any.
18185         (relax_delay_slots): Likewise.
18186         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
18188 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
18190         PR tree-optimization/83965
18191         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
18192         that grouped statements are part of a reduction chain.  Return
18193         true if the statement is not marked as a reduction itself but
18194         is part of a group.
18195         (vect_recog_dot_prod_pattern): Don't check whether the statement
18196         is part of a group here.
18197         (vect_recog_sad_pattern): Likewise.
18198         (vect_recog_widen_sum_pattern): Likewise.
18200 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
18202         PR debug/84545
18203         * final.c (rest_of_clean_state): Also look for calls inside sequences.
18205 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
18207         PR target/84530
18208         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
18209         the bool argument.
18210         (ix86_output_indirect_function_return): New prototype.
18211         (ix86_split_simple_return_pop_internal): Likewise.
18212         * config/i386/i386.c (indirect_return_via_cx): New.
18213         (indirect_return_via_cx_bnd): Likewise.
18214         (indirect_thunk_name): Handle return va CX_REG.
18215         (output_indirect_thunk_function): Create alias for
18216         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
18217         (ix86_output_indirect_jmp): Remove the bool argument.
18218         (ix86_output_indirect_function_return): New function.
18219         (ix86_split_simple_return_pop_internal): Likewise.
18220         * config/i386/i386.md (*indirect_jump): Don't pass false
18221         to ix86_output_indirect_jmp.
18222         (*tablejump_1): Likewise.
18223         (simple_return_pop_internal): Change it to define_insn_and_split.
18224         Call ix86_split_simple_return_pop_internal to split it for
18225         -mfunction-return=.
18226         (simple_return_indirect_internal): Call
18227         ix86_output_indirect_function_return instead of
18228         ix86_output_indirect_jmp.
18230 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
18232         PR bootstrap/84405
18233         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
18234         memset and value initialization afterwards.
18236 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
18238         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
18240 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
18242         PR target/84521
18243         * common/config/aarch64/aarch64-common.c
18244         (aarch_option_optimization_table[]): Switch
18245         off fomit-frame-pointer
18247 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
18248             Chung-Ju Wu  <jasonwucj@gmail.com>
18250         * config/nds32/nds32-multiple.md (load_multiple): Disallow
18251         volatile memory.
18252         (store_multiple): Ditto.
18254 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
18256         * config.gcc: Add --with-cpu support for nds32 target.
18257         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
18258         * config/nds32/nds32.opt: Add -mcpu= option.
18260 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
18262         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
18263         isel=yes): Warn for these deprecated options.
18265 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
18267         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
18268         ISA_2_5_MASKS_EMBEDDED.
18270 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
18272         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
18273         p->max as pointers rather than using iterative_hash_expr.
18275 2018-02-23  Carl Love  <cel@us.ibm.com>
18277         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
18278         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
18279         BU_P8V_OVERLOAD_2.
18280         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
18281         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
18282         P8V_BUILTIN_VEC_VUNSIGNED2.
18284 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
18286         PR target/81572
18287         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
18288         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
18289         LRA_UNKNOWN_ALT.
18290         * lra-constraints.c (curr_insn_transform): Set up
18291         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
18292         LRA_UNKNOWN_ALT.
18293         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
18294         * lra-eliminations.c (spill_pseudos): Ditto.
18295         (process_insn_for_elimination): Ditto.
18296         * lra-lives.c (reg_early_clobber_p): Use the new macros.
18297         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
18298         LRA_NON_CLOBBERED_ALT.
18300 2018-02-22  Martin Sebor  <msebor@redhat.com>
18302         PR tree-optimization/84480
18303         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
18304         to maybe_diag_stxncpy_trunc.  Call it.
18305         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
18306         from gimple_fold_builtin_strcpy.  Print inlining stack.
18307         (handle_builtin_stxncpy): Print inlining stack.
18308         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
18310 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
18312         PR target/84176
18313         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
18314         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
18315         and -fcheck-pointer-bounds are used together.
18316         (indirect_thunk_prefix): New enum.
18317         (indirect_thunk_need_prefix): New function.
18318         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
18319         "_nt" instead of "_bnd" for NOTRACK prefix.
18320         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
18321         (output_indirect_thunk_function): Likewise.
18322         (): Likewise.
18323         (ix86_code_end): Update output_indirect_thunk_function calls.
18324         (ix86_output_indirect_branch_via_reg): Replace
18325         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
18326         (ix86_output_indirect_branch_via_push): Likewise.
18327         (ix86_output_function_return): Likewise.
18328         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
18329         incompatible with -fcf-protection=branch and
18330         -fcheck-pointer-bounds.
18332 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
18334         PR target/83335
18335         * config/aarch64/aarch64.c (aarch64_print_address_internal):
18336         Change gcc_assert call to output_operand_lossage.
18338 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
18340         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
18342 2018-02-22  DJ Delorie  <dj@redhat.com>
18343             Sebastian Perta  <sebastian.perta@renesas.com>
18344             Oleg Endo  <olegendo@gcc.gnu.org>
18346         * config/rx/rx.c (rx_rtx_costs): New function.
18347         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
18349 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
18351         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
18353 2018-02-22  Martin Liska  <mliska@suse.cz>
18355         PR driver/83193
18356         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
18357         Add "native" as a possible value.
18359 2018-02-22  Martin Liska  <mliska@suse.cz>
18361         PR driver/83193
18362         * config/i386/i386.c (ix86_option_override_internal):
18363         Add "native" as a possible value for -march and -mtune.
18365 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
18367         PR target/84502
18368         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
18369         to all type variants.
18371         PR tree-optimization/84503
18372         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
18373         width as info->bitpos + info->bitsize - start.
18374         (merged_store_group::merge_overlapping): Simplify width computation.
18375         (check_no_overlap): New function.
18376         (imm_store_chain_info::try_coalesce_bswap): Compute expected
18377         start + width and last_order of the group, fail if check_no_overlap
18378         fails.
18379         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
18380         to group if check_no_overlap fails.
18382 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
18384         * config/rs6000/altivec.md: Delete contraint arguments to
18385         define_expand, define_split, and define_peephole2, and in
18386         define_insn_and_split if always unused.
18387         * config/rs6000/darwin.md: Ditto.
18388         * config/rs6000/dfp.md: Ditto.
18389         * config/rs6000/rs6000.md: Ditto.
18390         * config/rs6000/sync.md: Ditto.
18391         * config/rs6000/vector.md: Ditto.
18392         * config/rs6000/vsx.md: Ditto.
18394 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
18396         * config/rs6000/altivec.md: Write output control strings as braced
18397         blocks instead of double-quoted strings.
18398         * config/rs6000/darwin.md: Ditto.
18399         * config/rs6000/rs6000.md: Ditto.
18400         * config/rs6000/vector.md: Ditto.
18401         * config/rs6000/vsx.md: Ditto.
18403 2018-02-21  Jason Merrill  <jason@redhat.com>
18405         PR c++/84314 - ICE with templates and fastcall attribute.
18406         * attribs.c (build_type_attribute_qual_variant): Remove assert.
18408 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
18410         * ipa-cp.c (determine_versionability): Fix comment typos.
18412 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
18414         PR c/84229
18415         * ipa-cp.c (determine_versionability): Do not version functions caling
18416         va_arg_pack.
18418 2018-02-21  Martin Liska  <mliska@suse.cz>
18420         PR driver/83193
18421         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
18422         Add "native" as a possible value.
18423         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
18424         the macro when native cpu detection is available.
18426 2018-02-21  Martin Liska  <mliska@suse.cz>
18428         PR driver/83193
18429         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
18430         Add "native" as a possible value.
18431         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
18432         when native cpu detection is available.
18434 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
18435             Martin Sebor  <msebor@redhat.com>
18437         PR tree-optimization/84478
18438         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
18439         false.
18440         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
18441         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
18442         support which is conservatively correct, for 2 only stay conservative
18443         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
18444         argument to the 2 argument get_range_strlen, adjust 6 arg
18445         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
18446         false.
18447         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18448         (gimple_fold_builtin_strlen): Pass true as last argument to
18449         get_range_strlen.
18451 2018-02-20  Martin Sebor  <msebor@redhat.com>
18453         PR middle-end/84095
18454         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
18455         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
18456         (builtin_memref::builtin_memref): Factor out parts into
18457         set_base_and_offset and call it.
18459 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
18461         PR middle-end/84406
18462         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
18463         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
18464         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
18465         search at the associated MODE_INT.
18467 2018-02-20  Jeff Law  <law@redhat.com>
18469         PR middle-end/82123
18470         PR tree-optimization/81592
18471         PR middle-end/79257
18472         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
18473         for range data rather than using global data.
18474         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
18475         range data rather than using global data.
18476         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
18477         pass it to children as needed.
18478         (struct directive::fmtresult): Similarly.
18479         (struct directive::set_width): Similarly.
18480         (struct directive::set_precision): Similarly.
18481         (format_integer, format_directive, parse_directive): Similarly.
18482         (format_none): Accept unnamed vr_values parameter.
18483         (format_percent, format_floating, format_character): Similarly.
18484         (format_string, format_plain): Similarly.
18485         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
18486         the EVRP range analyzer for range data rather than using global data.
18487         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
18488         gimple-ssa-evrp-analyze.h
18489         (class sprintf_dom_walker): Add after_dom_children member function.
18490         Add evrp_range_analyzer member.
18491         (sprintf_dom_walker::before_dom_children): Call into the EVRP
18492         range analyzer as needed.
18493         (sprintf_dom_walker::after_dom_children): New member function.
18494         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
18495         if not optimizing.
18496         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
18497         (evrp_range_analyzer::pop_to_marker): Likewise.
18499 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
18501         PR tree-optimization/84419
18502         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
18503         with the required type if its current type is compatible but
18504         different.
18506 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
18508         PR middle-end/82004
18509         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
18510         after vectorization.
18512 2018-02-20  Martin Liska  <mliska@suse.cz>
18514         PR driver/83193
18515         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
18516         possible values if we don't have a hint.
18518 2018-02-20  Martin Liska  <mliska@suse.cz>
18520         PR c/84310
18521         PR target/79747
18522         * final.c (shorten_branches): Build align_tab array with one
18523         more element.
18524         * opts.c (finish_options): Add alignment option limit check.
18525         (MAX_CODE_ALIGN): Likewise.
18526         (MAX_CODE_ALIGN_VALUE): Likewise.
18527         * doc/invoke.texi: Document maximum allowed option value for
18528         all -falign-* options.
18530 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
18532         PR target/84146
18533         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
18534         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
18535         * var-tracking.c (emit_note_insn_var_location): Remove all references
18536         to NOTE_INSN_CALL_ARG_LOCATION.
18537         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
18538         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
18539         Use copy_rtx_if_shared.
18540         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
18541         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
18542         (dwarf2out_var_location): Remove handling of
18543         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
18544         on call_insn.
18545         * final.c (final_scan_insn): Remove all references to
18546         NOTE_INSN_CALL_ARG_LOCATION.
18547         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
18548         before dumping final insns.
18549         * except.c (emit_note_eh_region_end): Remove all references to
18550         NOTE_INSN_CALL_ARG_LOCATION.
18551         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
18552         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
18553         * config/arc/arc.c (hwloop_optimize): Likewise.
18554         * config/arm/arm.c (create_fix_barrier): Likewise.
18555         * config/s390/s390.c (s390_chunkify_start): Likewise.
18556         * config/sh/sh.c (find_barrier): Likewise.
18557         * config/i386/i386.c (rest_of_insert_endbranch,
18558         ix86_seh_fixup_eh_fallthru): Likewise.
18559         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
18560         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
18561         * config/frv/frv.c (frv_function_prologue): Likewise.
18562         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
18563         reg note.
18564         (note_outside_basic_block_p): Remove all references to
18565         NOTE_INSN_CALL_ARG_LOCATION.
18566         * gengtype.c (adjust_field_rtx_def): Likewise.
18567         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
18568         Likewise.
18569         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18570         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18572         PR c++/84444
18573         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
18574         is ADDR_EXPR.
18576         PR tree-optimization/84452
18577         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
18578         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
18579         is NULL.
18581 2018-02-19  Martin Liska  <mliska@suse.cz>
18583         PR sanitizer/82183
18584         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
18586 2018-02-19  Martin Liska  <mliska@suse.cz>
18587             Richard Sandiford  <richard.sandiford@linaro.org>
18589         PR tree-optimization/82491
18590         * gimple-fold.c (get_base_constructor): Make earlier bail out
18591         to prevent ubsan.
18593 2018-02-19  Carl Love  <cel@us.ibm.com>
18595         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
18596         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
18597         BU_P8V_OVERLOAD_1.
18598         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
18599         P8V_BUILTIN_VEC_NEG.
18601 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
18603         * config/rl78/rl78.md (movdf): New define expand.
18605 2018-02-19  Martin Liska  <mliska@suse.cz>
18607         PR other/80589
18608         * doc/invoke.texi: Fix typo.
18609         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
18611 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
18613         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
18614         handle rs6000_single_float and rs6000_double_float specially for
18615         e500 family CPUs.
18617 2018-02-16  Jeff Law  <law@redhat.com>
18619         * config/rx/rx.c (add_pop_cfi_notes): New function.;
18620         (pop_regs): Use it.
18622 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
18624         PR ipa/84425
18625         * ipa-inline.c (inline_small_functions): Fix a typo.
18627 2018-02-16  Nathan Sidwell  <nathan@acm.org>
18629         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
18631 2018-02-16  Carl Love  <cel@us.ibm.com>
18633         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
18634         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
18635         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
18636         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
18637         expansion to P8V_BUILTIN_VEC_FLOAT2.
18639 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
18641         PR rtl-optimization/70023
18642         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
18643         src_regno into account.
18645 2018-02-16  Carl Love  <cel@us.ibm.com>
18647         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
18648         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
18649         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
18650         * config/rs6000/rs6000.c: Remove case statements for
18651         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
18652         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
18653         and P9V_BUILTIN_VEC_VINSERT4B.
18654         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
18655         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
18656         * config/rs6000/vsx.md:
18657         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
18658         vec_insert4b.
18660 2018-02-16  Carl Love  <cel@us.ibm.com>
18662         * config/rs6000/altivec.h: Add builtin names vec_extract4b
18663         vec_insert4b.
18664         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
18665         definitions.
18666         * config/rs6000/rs6000-c.c: Add the definitions for
18667         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
18668         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
18669         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
18670         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
18671         definition for insert4b and define insn *insert3b_internal.
18672         * doc/extend.texi: Add documentation for vec_extract4b.
18674 2018-02-16  Nathan Sidwell  <nathan@acm.org>
18676         * doc/extend.texi (Backwards Compatibility): Mention friend
18677         injection.  Note for-scope is deprecated.
18678         * doc/invoke.texi (-ffriend-injection): Deprecate.
18680 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
18682         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
18683         that moved to I2, also allow destinations that are a paradoxical
18684         subreg (instead of a normal reg).
18686 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
18688         PR target/83831
18689         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18690         to QImode.
18692 2018-02-16  Richard Biener  <rguenther@suse.de>
18694         PR tree-optimization/84037
18695         PR tree-optimization/84016
18696         PR target/82862
18697         * config/i386/i386.c (ix86_builtin_vectorization_cost):
18698         Adjust vec_construct for the fact we need additional higher latency
18699         128bit inserts for AVX256 and AVX512 vector builds.
18700         (ix86_add_stmt_cost): Scale vector construction cost for
18701         elementwise loads.
18703 2018-02-16  Richard Biener  <rguenther@suse.de>
18705         PR tree-optimization/84417
18706         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
18707         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
18708         (non_rewritable_lvalue_p): Likewise, use poly-ints.
18710 2018-02-16  Martin Liska  <mliska@suse.cz>
18712         PR sanitizer/84307
18713         * internal-fn.def (ASAN_CHECK): Set proper flags.
18714         (ASAN_MARK): Likewise.
18716 2018-02-16  Julia Koval  <julia.koval@intel.com>
18718         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
18719         from PTA_CANNONLAKE.
18721 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
18723         PR target/84272
18724         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
18725         Use ++iter rather than iter++ for std::list iterators.
18726         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
18727         defer deleting them until all nodes in the forest are processed.  Do
18728         free even leaf nodes.  Change to_process into auto_vec.
18730         PR bootstrap/84405
18731         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
18732         * vec.h (vec_default_construct): Use memset instead of placement new
18733         if BROKEN_VALUE_INITIALIZATION is defined.
18734         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
18735         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
18736         is defined.
18738         PR rtl-optimization/83723
18739         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
18740         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
18741         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
18742         recursive calls.
18743         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
18744         callers.
18745         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
18747 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
18749         PR rtl-optimization/81443
18750         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
18751         from inner REGs to paradoxical SUBREGs.
18753 2018-02-16  Richard Biener  <rguenther@suse.de>
18755         PR tree-optimization/84399
18756         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
18757         For operands we can analyze at their definition make sure we can
18758         analyze them at each use as well.
18760 2018-02-16  Richard Biener  <rguenther@suse.de>
18762         PR tree-optimization/84190
18763         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
18764         volatile accesses if the decl isn't volatile.
18766 2018-02-15  Jason Merrill  <jason@redhat.com>
18768         PR c++/84314 - ICE with templates and fastcall attribute.
18769         * attribs.c (build_type_attribute_qual_variant): Don't clobber
18770         TYPE_CANONICAL on an existing type.
18772 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
18774         PR tree-optimization/84383
18775         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
18776         dstoff nor call operand_equal_p if dstbase is NULL.
18778         PR tree-optimization/84334
18779         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
18780         also a CONSTANT_CLASS_P, punt.
18782 2018-02-14  Jim Wilson  <jimw@sifive.com>
18784         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
18785         first SMALL_OPERAND check.  New local min_second_step.  Move assert
18786         to where locals are set.  Add TARGET_RVC support.
18787         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
18789 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
18791         * doc/invoke.texi: Correct -Wformat-overflow code sample.
18793 2018-02-14  Martin Sebor  <msebor@redhat.com>
18795         PR tree-optimization/83698
18796         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
18797         arrays constrain the offset range to their bounds.
18798         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
18799         (builtin_access::overlap): Avoid setting the size of overlap if it's
18800         already been set.
18801         (maybe_diag_overlap): Also consider arrays when deciding what values
18802         of offsets to include in diagnostics.
18804 2018-02-14  Martin Sebor  <msebor@redhat.com>
18806         PR c/84108
18807         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
18808         that correspond to the kind of a declaration.
18810 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
18812         PR target/83984
18813         * config/pa/pa.md: Load address of PIC label using the linkage table
18814         if the label is nonlocal.
18816 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
18818         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
18819         warning message if user requests -maltivec=be.
18820         * doc/invoke.texi: Document deprecation of -maltivec=be.
18822 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
18824         PR target/84220
18825         * config/rs6000/rs6000-c.c: Update definitions for
18826         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
18827         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
18829 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
18831         PR target/84239
18832         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
18833         add _get_ssp intrinsics. Remove argument from
18834         __builtin_ia32_rdssp[d|q].
18835         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
18836         * config/i386/i386-builtin.def: Remove argument from
18837         __builtin_ia32_rdssp[d|q].
18838         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
18839         ix86_expand_special_args_builtin for _rdssp[d|q].
18840         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
18841         Clear register before usage.
18842         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
18843         Add documentation for new _get_ssp and _inc_ssp intrinsics.
18845 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
18847         PR tree-optimization/84357
18848         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
18849         operand 1 of an ARRAY_REF too.
18851 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
18853         PR target/83831
18854         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
18855         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18856         declarations.
18857         (set_of_reg): New struct.
18858         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
18859         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
18860         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
18861         functions.
18862         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
18863         Split into bitclr, bitset, bitinvert patterns if appropriate.
18864         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
18865         use rx_fuse_in_memory_bitop.
18866         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
18867         to named insn, correct maximum insn length.
18869 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
18871         PR target/79242
18872         * machmode.def: Define a complex mode for PARTIAL_INT.
18873         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18874         MODE_PARTIAL_INT.
18875         * doc/rtl.texi: Document CSPImode.
18876         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18877         handling.
18878         (msp430_hard_regno_nregs_with_padding): Likewise.
18880 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
18882         PR target/84279
18883         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
18885 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
18887         PR rtl-optimization/84169
18888         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
18889         we generated a parallel as new i3 and we split that to new i2 and i3
18890         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
18891         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
18892         those to i2, not i1.  Partially rewrite this scan code.
18894 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
18896         PR c/82210
18897         * stor-layout.c (place_field): For variable length fields, adjust
18898         offset_align afterwards not just based on the field's alignment,
18899         but also on the size.
18901         PR middle-end/84309
18902         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
18903         of exps and logs in the use_exp2 case.
18905 2018-02-13  Jeff Law  <law@redhat.com>
18907         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
18908         entry for "vector".
18910         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
18911         ARGS as unused.
18913 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
18915         PR debug/84342
18916         PR debug/84319
18917         * common.opt (gas-loc-support, gas-locview-support): New.
18918         (ginline-points, ginternal-reset-location-views): New.
18919         * doc/invoke.texi: Document them.  Use @itemx where intended.
18920         (gvariable-location-views): Adjust.
18921         * target.def (reset_location_view): New.
18922         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
18923         (TARGET_RESET_LOCATION_VIEW): New.
18924         * doc/tm.texi: Rebuilt.
18925         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
18926         (dwarf2out_default_as_locview_support): New.
18927         (output_asm_line_debug_info): Use option variables.
18928         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
18929         (output_loc_list): Likewise.
18930         (add_high_low_attributes): Check option variables.
18931         Don't output entry view attribute in strict mode.
18932         (gen_inlined_subroutine_die): Check option variables.
18933         (dwarf2out_inline_entry): Likewise.
18934         (init_sections_and_labels): Likewise.
18935         (dwarf2out_early_finish): Likewise.
18936         (maybe_reset_location_view): New, from...
18937         (dwarf2out_var_location): ... here.  Call it.
18938         * debug.h (dwarf2out_default_as_loc_support): Declare.
18939         (dwarf2out_default_as_locview_support): Declare.
18940         * hooks.c (hook_int_rtx_insn_0): New.
18941         * hooks.h (hook_int_rtx_insn_0): Declare.
18942         * toplev.c (process_options): Take -gas-loc-support and
18943         -gas-locview-support from dwarf2out.  Enable
18944         -gvariable-location-views by default only with locview
18945         assembler support.  Enable -ginternal-reset-location-views by
18946         default only if the target defines the corresponding hook.
18947         Enable -ginline-points by default if location views are
18948         enabled; force it disabled if statement frontiers are
18949         disabled.
18950         * tree-inline.c (expand_call_inline): Check option variables.
18951         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
18953 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
18955         PR tree-optimization/84321
18956         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
18957         handling.  Also check whether the anti-range contains any values
18958         that satisfy the mask; switch to a VR_RANGE if not.
18960 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
18962         PR sanitizer/84340
18963         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18965 2018-02-13  Martin Jambor  <mjambor@suse.cz>
18967         PR c++/83990
18968         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
18969         of call statements, also set location of a load to a temporary.
18971 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
18973         * config/rl78/rl78.c (add_vector_labels): New function.
18974         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
18975         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
18976         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
18977         which checks that no arguments are passed.
18978         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
18979         * doc/extend.texi: Documentation for the new attribute.
18981 2018-02-13  Andreas Schwab  <schwab@suse.de>
18983         * config/riscv/linux.h (CPP_SPEC): Define.
18985 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
18987         PR target/84335
18988         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
18989         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
18990         OPTION_MASK_ISA_AES as first argument to def_builtin_const
18991         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
18992         instead of OPTION_MASK_ISA_PCLMUL as first argument to
18993         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
18994         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
18995         temporarily for AES and PCLMUL builtins.
18997         PR tree-optimization/84339
18998         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
18999         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
19000         Formatting fixes.
19002         PR middle-end/84309
19003         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
19004         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
19005         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
19006         inline function.
19007         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
19008         inline function.
19009         * omp-simd-clone.h: New file.
19010         * omp-simd-clone.c: Include omp-simd-clone.h.
19011         (expand_simd_clones): No longer static.
19012         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
19013         cgraph.h and omp-simd-clone.h.
19014         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
19015         (vect_recog_widen_shift_pattern): Formatting fix.
19016         (vect_pattern_recog_1): Don't check optab for calls.
19018         PR target/84336
19019         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
19020         operands[2] into a REG before using gen_lowpart on it.
19022 2018-02-12  Jeff Law  <law@redhat.com>
19024         PR target/83760
19025         * config/sh/sh.c (find_barrier): Consider a sibling call
19026         a barrier as well.
19028         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
19029         successfully back substituting a reg.
19031 2018-02-12  Richard Biener  <rguenther@suse.de>
19033         PR tree-optimization/84037
19034         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
19035         parameter, move visited init to caller.
19036         (vect_slp_analyze_operations): Separate cost from validity
19037         check, initialize visited once for all instances.
19038         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
19039         for all instances.
19040         * tree-vect-stmts.c (vect_model_simple_cost): Make early
19041         out an assert.
19042         (vect_model_promotion_demotion_cost): Likewise.
19043         (vectorizable_bswap): Guard cost modeling with !slp_node
19044         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
19045         SLP stmts.
19046         (vectorizable_call): Likewise.
19047         (vectorizable_conversion): Likewise.
19048         (vectorizable_assignment): Likewise.
19049         (vectorizable_shift): Likewise.
19050         (vectorizable_operation): Likewise.
19051         (vectorizable_store): Likewise.
19052         (vectorizable_load): Likewise.
19053         (vectorizable_condition): Likewise.
19054         (vectorizable_comparison): Likewise.
19056 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
19058         PR sanitizer/84307
19059         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
19060         (ASAN_MARK): Fix fnspec to account for return value, change pointer
19061         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
19063 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
19065         PR middle-end/83665
19066         * params.def (inline-min-speedup): Increase from 8 to 15.
19067         (max-inline-insns-auto): Decrease from 40 to 30.
19068         * ipa-split.c (consider_split): Add some buffer for function to
19069         be considered inlining candidate.
19070         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
19071         default values.
19073 2018-02-12  Richard Biener  <rguenther@suse.de>
19075         PR tree-optimization/84037
19076         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
19077         matched stmts if we cannot swap the non-matched ones.
19079 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
19081         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
19082         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
19083         _mm_maskz_scalef_round_ss): New intrinsics.
19084         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
19085         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
19086         __builtin_ia32_scalefss_round): Remove.
19087         (__builtin_ia32_scalefsd_mask_round,
19088         __builtin_ia32_scalefss_mask_round): New intrinsics.
19089         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
19090         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
19091         ((match_operand:VF_128 2 "<round_nimm_predicate>"
19092         "<round_constraint>")): Changed to ...
19093         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
19094         "<round_scalar_constraint>")): ... this.
19095         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
19096         %0, %1, %2<round_op3>}"): Changed to ...
19097         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
19098         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
19099         %2<round_scalar_mask_op3>}"): ... this.
19100         * config/i386/subst.md (round_scalar_nimm_predicate): New.
19102 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
19104         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
19105         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
19106         (_mm_maskz_sqrt_round_ss): New intrinsics.
19107         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
19108         (__builtin_ia32_sqrtsd_mask_round)
19109         (__builtin_ia32_sqrtss_mask_round): New builtins.
19110         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
19111         (__builtin_ia32_sqrtss_round): Remove.
19112         (__builtin_ia32_sqrtsd_mask_round)
19113         (__builtin_ia32_sqrtss_mask_round): New builtins.
19114         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
19115         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
19116         ((match_operand:VF_128 1 "vector_operand"
19117         "xBm,<round_constraint>")): Changed to ...
19118         ((match_operand:VF_128 1 "vector_operand"
19119         "xBm,<round_scalar_constraint>")): ... this.
19120         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
19121         %0, %2, %<iptr>1<round_op3>}): Changed to ...
19122         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
19123         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
19124         %<iptr>1<round_scalar_mask_op3>}): ... this.
19125         ((set_attr "prefix" "<round_prefix>")): Changed to ...
19126         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
19128 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
19130         PR target/84266
19131         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
19132         Cast vec_cmpeq result to correct type.
19133         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
19134         Cast vec_cmpgt result to correct type.
19136 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
19138         * final.c (final_scan_insn_1): Renamed from...
19139         (final_scan_insn): ... this.  New wrapper, to recover
19140         seen from the outermost call in recursive ones.
19141         * config/sparc/sparc.c (output_return): Drop seen from call.
19142         (output_sibcall): Likewise.
19143         * config/visium/visium.c (output_branch): Likewise.
19145 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
19147         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
19148         function label.
19150 2018-02-10  Alan Modra  <amodra@gmail.com>
19152         PR target/84300
19153         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
19154         Specify LR as an input.
19156 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
19158         PR sanitizer/83987
19159         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
19160         remove_member_access_dummy_vars): New functions.
19161         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
19162         lower_omp_1, execute_lower_omp): Use them.
19164         PR rtl-optimization/84308
19165         * shrink-wrap.c (spread_components): Release todo vector.
19167 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
19169         PR rtl-optimization/57193
19170         * ira-color.c (struct allocno_color_data): Add member
19171         conflict_allocno_hard_prefs.
19172         (update_conflict_allocno_hard_prefs): New.
19173         (bucket_allocno_compare_func): Add a preference based on
19174         conflict_allocno_hard_prefs.
19175         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
19176         (color_allocnos): Remove a dead code.  Initiate
19177         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
19179 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
19181         PR target/84226
19182         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
19183         constraint from =wa to wa.  Avoid a subreg on the output operand,
19184         instead use a pseudo and subreg it in a move.
19185         (p9_xxbrd_<mode>): Changed to ...
19186         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
19187         (p9_xxbrd_v2df): New expander.
19188         (p9_xxbrw_<mode>): Changed to ...
19189         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
19190         (p9_xxbrw_v4sf): New expander.
19192 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
19194         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
19196 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
19198         PR target/83926
19199         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
19200         multiply in 32-bit mode.
19201         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
19202         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
19203         mode.
19205 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
19207         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
19208         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
19209         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
19210         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
19212 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
19214         PR lto/84213
19215         * dwarf2out.c (is_trivial_indirect_ref): New function.
19216         (dwarf2out_late_global_decl): Do not generate a location
19217         attribute for variables that have a non-trivial DECL_VALUE_EXPR
19218         and that are not defined in the current unit.
19220 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
19222         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
19223         instead of a libcall for UNORDERED.
19225 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
19227         PR target/82641
19228         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
19229         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
19231 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19233         PR target/PR84295
19234         * config/s390/s390.c (s390_set_current_function): Invoke
19235         s390_indirect_branch_settings also if fndecl didn't change.
19237 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
19239         * config/rs6000/rs6000.md (blockage): Set length to zero.
19241 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
19243         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
19245 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
19247         PR sanitizer/84285
19248         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19249         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
19250         -static-lib*san.
19252         PR debug/84252
19253         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
19254         PARALLEL incoming that failed vt_get_decl_and_offset check.
19256         PR middle-end/84237
19257         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
19258         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
19259         TREE_READONLY bit.
19260         (get_variable_section): For decls in named .bss* sections pass true as
19261         second argument to bss_initializer_p.
19263 2018-02-09  Marek Polacek  <polacek@redhat.com>
19264             Jakub Jelinek  <jakub@redhat.com>
19266         PR c++/83659
19267         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
19268         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
19269         Sync some changes from cxx_fold_indirect_ref.
19271 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
19273         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
19274         markers.
19275         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
19276         (BLOCK_INLINE_ENTRY_LABEL): New.
19277         (dwarf2out_var_location): Disregard inline entry markers.
19278         (inline_entry_data): New struct.
19279         (inline_entry_data_hasher): New hashtable type.
19280         (inline_entry_data_hasher::hash): New.
19281         (inline_entry_data_hasher::equal): New.
19282         (inline_entry_data_table): New variable.
19283         (add_high_low_attributes): Add DW_AT_entry_pc and
19284         DW_AT_GNU_entry_view attributes if a pending entry is found
19285         in inline_entry_data_table.  Add old entry_pc attribute only
19286         if debug nonbinding markers are disabled.
19287         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
19288         markers are enabled.
19289         (block_within_block_p, dwarf2out_inline_entry): New.
19290         (dwarf2out_finish): Check that no entries remained in
19291         inline_entry_data_table.
19292         * final.c (reemit_insn_block_notes): Handle inline entry notes.
19293         (final_scan_insn, notice_source_line): Likewise.
19294         (rest_of_clean_state): Skip inline entry markers.
19295         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
19296         markers.
19297         * gimple.c (gimple_build_debug_inline_entry): New.
19298         * gimple.h (enum gimple_debug_subcode): Add
19299         GIMPLE_DEBUG_INLINE_ENTRY.
19300         (gimple_build_debug_inline_entry): Declare.
19301         (gimple_debug_inline_entry_p): New.
19302         (gimple_debug_nonbind_marker_p): Adjust.
19303         * insn-notes.def (INLINE_ENTRY): New.
19304         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
19305         inline entry marker notes.
19306         (print_insn): Likewise.
19307         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
19308         (INSN_DEBUG_MARKER_KIND): Likewise.
19309         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
19310         * tree-inline.c (expand_call_inline): Build and insert
19311         debug_inline_entry stmt.
19312         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
19313         inline entry blocks early, if nonbind markers are enabled.
19314         (dump_scope_block): Dump fragment info.
19315         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
19316         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
19317         (gimple_build_debug_inline_entry): New.
19318         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
19319         Enable/disable inline entry points too.
19320         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
19321         (DEBUG_INSN): Describe inline entry markers.
19323         * common.opt (gvariable-location-views): New.
19324         (gvariable-location-views=incompat5): New.
19325         * config.in: Rebuilt.
19326         * configure: Rebuilt.
19327         * configure.ac: Test assembler for view support.
19328         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
19329         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
19330         * dwarf2out.c (var_loc_view): New typedef.
19331         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
19332         (dwarf2out_locviews_in_attribute): New.
19333         (dwarf2out_locviews_in_loclist): New.
19334         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
19335         (enum dw_line_info_opcode): Add LI_adv_address.
19336         (struct dw_line_info_table): Add view.
19337         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
19338         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
19339         (zero_view_p): New variable.
19340         (ZERO_VIEW_P): New macro.
19341         (output_asm_line_debug_info): New.
19342         (struct var_loc_node): Add view.
19343         (add_AT_view_list, AT_loc_list): New.
19344         (add_var_loc_to_decl): Add view param.  Test it against last.
19345         (new_loc_list): Add view params.  Record them.
19346         (AT_loc_list_ptr): Handle loc and view lists.
19347         (view_list_to_loc_list_val_node): New.
19348         (print_dw_val): Handle dw_val_class_view_list.
19349         (size_of_die): Likewise.
19350         (value_format): Likewise.
19351         (loc_list_has_views): New.
19352         (gen_llsym): Set vl_symbol too.
19353         (maybe_gen_llsym, skip_loc_list_entry): New.
19354         (dwarf2out_maybe_output_loclist_view_pair): New.
19355         (output_loc_list): Output view list or entries too.
19356         (output_view_list_offset): New.
19357         (output_die): Handle dw_val_class_view_list.
19358         (output_dwarf_version): New.
19359         (output_compilation_unit_header): Use it.
19360         (output_skeleton_debug_sections): Likewise.
19361         (output_rnglists, output_line_info): Likewise.
19362         (output_pubnames, output_aranges): Update version comments.
19363         (output_one_line_info_table): Output view numbers in asm comments.
19364         (dw_loc_list): Determine current endview, pass it to new_loc_list.
19365         Call maybe_gen_llsym.
19366         (loc_list_from_tree_1): Adjust.
19367         (add_AT_location_description): Create view list attribute if
19368         needed, check it's absent otherwise.
19369         (convert_cfa_to_fb_loc_list): Adjust.
19370         (maybe_emit_file): Call output_asm_line_debug_info for test.
19371         (dwarf2out_var_location): Reset views as needed.  Precompute
19372         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
19373         attribute.  Set view.
19374         (new_line_info_table): Reset next view.
19375         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
19376         (dwarf2out_source_line): Likewise.  Output view resets and labels to
19377         the assembler, or select appropriate line info opcodes.
19378         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
19379         (optimize_string_length): Catch it.  Adjust.
19380         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
19381         dw_val_class_view_list, and remove it if no longer needed.
19382         (hash_loc_list): Hash view numbers.
19383         (loc_list_hasher::equal): Compare them.
19384         (optimize_location_lists): Check whether a view list symbol is
19385         needed, and whether the locview attribute is present, and
19386         whether they match.  Remove the locview attribute if no longer
19387         needed.
19388         (index_location_lists): Call skip_loc_list_entry for test.
19389         (dwarf2out_finish): Call output_asm_line_debug_info for test.
19390         Use output_dwarf_version.
19391         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
19392         (struct dw_val_node): Add val_view_list.
19393         * final.c (SEEN_NEXT_VIEW): New.
19394         (set_next_view_needed): New.
19395         (clear_next_view_needed): New.
19396         (maybe_output_next_view): New.
19397         (final_start_function): Rename to...
19398         (final_start_function_1): ... this.  Take pointer to FIRST,
19399         add SEEN parameter.  Emit param bindings in the initial view.
19400         (final_start_function): Reintroduce SEEN-less interface.
19401         (final): Rename to...
19402         (final_1): ... this.  Take SEEN parameter.  Output final pending
19403         next view at the end.
19404         (final): Reintroduce seen-less interface.
19405         (final_scan_insn): Output pending next view before switching
19406         sections or ending a block.  Mark the next view as needed when
19407         outputting variable locations.  Notify debug backend of section
19408         changes, and of location view changes.
19409         (rest_of_handle_final): Adjust.
19410         * toplev.c (process_options): Autodetect value for debug variable
19411         location views option.  Warn on incompat5 without -gdwarf-5.
19412         * doc/invoke.texi (gvariable-location-views): New.
19413         (gvariable-location-views=incompat5): New.
19414         (gno-variable-location-views): New.
19416 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
19418         PR tree-optimization/84136
19419         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
19420         that the result of find_edge is non-NULL.
19422 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
19424         PR target/83008
19425         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
19426         storing integer register in SImode.  Fix cost of 256 and 512
19427         byte aligned SSE register store.
19429 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
19431         * config/i386/i386.c (ix86_multiplication_cost): Fix
19432         multiplication cost for TARGET_AVX512DQ.
19434 2018-02-08  Marek Polacek  <polacek@redhat.com>
19436         PR tree-optimization/84238
19437         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
19438         get_range_strlen.
19440 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19442         PR tree-optimization/84265
19443         * tree-vect-stmts.c (vectorizable_store): Don't treat
19444         VMAT_CONTIGUOUS accesses as grouped.
19445         (vectorizable_load): Likewise.
19447 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19449         PR tree-optimization/81635
19450         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
19451         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
19452         (test_round_for_mask): New functions.
19453         (wide_int_cc_tests): Call test_round_for_mask.
19454         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
19455         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
19456         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
19457         range returned by get_range_info.
19459 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
19461         PR ipa/81360
19462         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
19463         * symtab.c: Include builtins.h
19464         (symtab_node::output_to_lto_symbol_table_p): Move here
19465         from lto-streamer-out.c:output_symbol_p.
19466         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
19467         (output_symbol_p): Move all logic to symtab.c
19468         (produce_symtab): Update.
19470 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
19472         * config/s390/s390-opts.h (enum indirect_branch): Define.
19473         * config/s390/s390-protos.h (s390_return_addr_from_memory)
19474         (s390_indirect_branch_via_thunk)
19475         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
19476         (enum s390_indirect_branch_type): Define.
19477         * config/s390/s390.c (struct s390_frame_layout, struct
19478         machine_function): Remove.
19479         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
19480         (indirect_branch_table_label_no, indirect_branch_table_name):
19481         Define variables.
19482         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
19483         (enum s390_indirect_branch_option): Define.
19484         (s390_return_addr_from_memory): New function.
19485         (s390_handle_string_attribute): New function.
19486         (s390_attribute_table): Add new attribute handler.
19487         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
19488         (s390_indirect_branch_via_thunk): New function.
19489         (s390_indirect_branch_via_inline_thunk): New function.
19490         (s390_function_ok_for_sibcall): When jumping via thunk disallow
19491         sibling call optimization for non z10 compiles.
19492         (s390_emit_call): Force indirect branch target to be a single
19493         register.  Add r1 clobber for non-z10 compiles.
19494         (s390_emit_epilogue): Emit return jump via return_use expander.
19495         (s390_reorg): Handle JUMP_INSNs as execute targets.
19496         (s390_option_override_internal): Perform validity checks for the
19497         new command line options.
19498         (s390_indirect_branch_attrvalue): New function.
19499         (s390_indirect_branch_settings): New function.
19500         (s390_set_current_function): Invoke s390_indirect_branch_settings.
19501         (s390_output_indirect_thunk_function):  New function.
19502         (s390_code_end): Implement target hook.
19503         (s390_case_values_threshold): Implement target hook.
19504         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
19505         macros.
19506         * config/s390/s390.h (struct s390_frame_layout)
19507         (struct machine_function): Move here from s390.c.
19508         (TARGET_INDIRECT_BRANCH_NOBP_RET)
19509         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
19510         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
19511         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
19512         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
19513         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
19514         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
19515         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
19516         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
19517         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
19518         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
19519         (mnemonic attribute): Add values which aren't recognized
19520         automatically.
19521         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
19522         pattern for branch conversion.  Fix mnemonic attribute.
19523         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
19524         indirect branch via thunk if requested.
19525         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
19526         ("*indirect_jump"): Disable for branch conversion using out of
19527         line thunks.
19528         ("indirect_jump_via_thunk<mode>_z10")
19529         ("indirect_jump_via_thunk<mode>")
19530         ("indirect_jump_via_inlinethunk<mode>_z10")
19531         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
19532         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
19533         ("casesi_jump_via_inlinethunk<mode>_z10")
19534         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
19535         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
19536         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
19537         ("*indirect2_jump"): Disable for branch conversion.
19538         ("casesi_jump"): Turn into expander and expand patterns for branch
19539         conversion.
19540         ("return_use"): New expander.
19541         ("*return"): Emit return via thunk and rename it to ...
19542         ("*return<mode>"): ... this one.
19543         * config/s390/s390.opt: Add new options and and enum for the
19544         option values.
19546 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
19548         * lra-constraints.c (match_reload): Unconditionally use
19549         gen_lowpart_SUBREG, rather than selecting between that
19550         and equivalent gen_rtx_SUBREG code.
19552 2018-02-08  Richard Biener  <rguenther@suse.de>
19554         PR tree-optimization/84233
19555         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
19556         changed flag instead of boguously re-using phi_inserted.
19558 2018-02-08  Martin Jambor  <mjambor@suse.cz>
19560         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
19561         static local variables.
19563 2018-02-08  Richard Biener  <rguenther@suse.de>
19565         PR tree-optimization/84278
19566         * tree-vect-stmts.c (vectorizable_store): When looking for
19567         smaller vector types to perform grouped strided loads/stores
19568         make sure the mode is supported by the target.
19569         (vectorizable_load): Likewise.
19571 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
19573         * config/aarch64/aarch64.c (aarch64_components_for_bb):
19574         Increase LDP/STP opportunities by adding adjacent callee-saves.
19576 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
19578         PR rtl-optimization/84068
19579         PR rtl-optimization/83459
19580         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
19582 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
19584         PR tree-optimization/84224
19585         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
19586         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
19587         non-zero arguments.
19589 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
19591         PR target/84113
19592         * config/rs6000/altivec.md (*restore_world): Remove LR use.
19593         * config/rs6000/predicates.md (restore_world_operation): Adjust op
19594         count, remove one USE.
19596 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
19598         * doc/install.texi (Configuration): Document the
19599         --with-long-double-format={ibm,ieee} PowerPC configuration
19600         options.
19602         PR target/84154
19603         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
19604         Convert from define_expand to be define_insn_and_split.  Rework
19605         float/double/_Float128 conversions to QI/HI/SImode to work with
19606         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
19607         conversions to QI/HImode types did a store and then a load to
19608         truncate the value.  For conversions to VSX registers, don't split
19609         the insn, instead emit the code directly.  Use the code iterator
19610         any_fix to combine signed and unsigned conversions.
19611         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
19612         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
19613         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19614         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
19615         (fix_<mode>di2_hw): Likewise.
19616         (fixuns_<mode>di2_hw): Likewise.
19617         (fix_<mode>si2_hw): Likewise.
19618         (fixuns_<mode>si2_hw): Likewise.
19619         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
19620         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
19621         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
19622         fix<uns>_trunc<SFDF:mode>si2_p8.
19623         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
19624         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
19625         (fix<uns>_<mode>_mem): Likewise.
19626         (fctiw<u>z_<mode>_mem): Likewise.
19627         (fix<uns>_<mode>_mem): Likewise.
19628         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
19629         the register allocator from doing a direct move to the GPRs to do
19630         a store, and instead use the ISA 3.0 store byte/half-word from
19631         vector register instruction.  For IEEE 128-bit floating point,
19632         also optimize stores of 32-bit ints.
19633         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
19635 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
19637         * genextract.c (push_pathstr_operand): New function to support
19638         [a-zA-Z].
19639         (walk_rtx): Call push_pathstr_operand.
19640         (print_path): Support [a-zA-Z].
19642 2018-02-07  Richard Biener  <rguenther@suse.de>
19644         PR tree-optimization/84037
19645         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
19646         (cse_and_gimplify_to_preheader): Declare.
19647         (vect_get_place_in_interleaving_chain): Likewise.
19648         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
19649         ivexpr_map.
19650         (_loop_vec_info::~_loop_vec_info): Delete it.
19651         (cse_and_gimplify_to_preheader): New function.
19652         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
19653         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
19654         (vectorizable_load): Likewise.  For grouped stores always base
19655         the IV on the first element.
19656         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
19657         condition before gimplifying.
19659 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
19661         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
19662         *DIV_EXPR and *MOD_EXPR.
19664 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
19666         PR target/84248
19667         * config/i386/i386.c (ix86_option_override_internal): Mask out
19668         the CF_SET bit when checking -fcf-protection.
19670 2018-02-07  Tom de Vries  <tom@codesourcery.com>
19672         PR libgomp/84217
19673         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
19674         enough.
19676 2018-02-07  Richard Biener  <rguenther@suse.de>
19678         PR tree-optimization/84204
19679         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
19680         this place.
19682         PR tree-optimization/84205
19683         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
19684         special-case isl_ast_op_zdiv_r.
19686         PR tree-optimization/84223
19687         * graphite-scop-detection.c (gather_bbs::before_dom_children):
19688         Only add conditions from within the region.
19689         (gather_bbs::after_dom_children): Adjust.
19691 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
19693         PR target/84209
19694         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
19695         * config/avr/avr.md: Only post-reload split REG-REG moves if
19696         either register is GENERAL_REG_P.
19698 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
19700         PR tree-optimization/84235
19701         * tree-ssa-scopedtables.c
19702         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
19703         if the subtraction is performed in floating point type where NaNs are
19704         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
19705         build 1.  Formatting fix.
19707 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
19709         PR target/84146
19710         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
19711         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
19712         and skip it regardless of bb boundaries.  Use CALL_P macro,
19713         don't test INSN_P (insn) together with CALL_P or JUMP_P check
19714         unnecessarily, formatting fix.
19716 2018-02-06  Michael Collison  <michael.collison@arm.com>
19718         * config/arm/thumb2.md:
19719         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
19720         (*thumb_mov_notscc): Ditto.
19722 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
19724         PR target/84154
19725         * config/rs6000/rs6000.md (su code attribute): Use "u" for
19726         unsigned_fix, not "s".
19728 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
19730         * configure.ac (gcc_fn_eh_frame_ro): New function.
19731         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
19732         correct .eh_frame permissions.
19733         * configure: Regenerate.
19735 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
19737         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
19738         irrelevant options.
19740 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
19742         * config/rs6000/rs6000.c (rs6000_option_override_internal):
19743         Display warning message for -mno-speculate-indirect-jumps.
19745 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
19747         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
19748         Undocumented.
19749         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
19751 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
19753         PR tree-optimization/84225
19754         * tree-eh.c (find_trapping_overflow): Only call
19755         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
19757 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
19759         PR target/84145
19760         * config/i386/i386.c: Reimplement the check of possible options
19761         -mibt/-mshstk conbination. Change error messages.
19762         * doc/invoke.texi: Fix a typo: remove extra '='.
19764 2018-02-06  Marek Polacek  <polacek@redhat.com>
19766         PR tree-optimization/84228
19767         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
19769 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
19771         PR target/82641
19772         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
19773         emitted arch directives.
19774         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
19775         __ARM_FEATURE_COPROC before changing architectures.
19777 2018-02-06  Richard Biener  <rguenther@suse.de>
19779         * config/i386/i386.c (print_reg): Fix typo.
19780         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
19782 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
19784         * configure: Regenerate.
19786 2018-02-05  Martin Sebor  <msebor@redhat.com>
19788         PR tree-optimization/83369
19789         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
19790         inlining context.
19792 2018-02-05  Martin Liska  <mliska@suse.cz>
19794         * doc/invoke.texi: Cherry-pick upstream r323995.
19796 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
19798         * ira.c (ira_init_register_move_cost): Adjust comment.
19800 2018-02-05  Martin Liska  <mliska@suse.cz>
19802         PR gcov-profile/84137
19803         * doc/gcov.texi: Fix typo in documentation.
19805 2018-02-05  Martin Liska  <mliska@suse.cz>
19807         PR gcov-profile/83879
19808         * doc/gcov.texi: Document necessity of --dynamic-list-data when
19809         using dlopen functionality.
19811 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
19813         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
19814         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
19815         _mm_maskz_range_ss, _mm_mask_range_round_ss,
19816         _mm_maskz_range_round_ss): New intrinsics.
19817         (__builtin_ia32_rangesd128_round)
19818         (__builtin_ia32_rangess128_round): Remove.
19819         (__builtin_ia32_rangesd128_mask_round,
19820         __builtin_ia32_rangess128_mask_round): New builtins.
19821         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
19822         __builtin_ia32_rangess128_round): Remove.
19823         (__builtin_ia32_rangesd128_mask_round,
19824         __builtin_ia32_rangess128_mask_round): New builtins.
19825         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
19826         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
19827         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
19828         "<round_saeonly_constraint>")): Changed to ...
19829         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
19830         "<round_saeonly_scalar_constraint>")): ... this.
19831         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
19832         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
19833         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
19834         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
19835         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
19837 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
19839         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
19840         options.
19841         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
19842         Remove all values except native, 8540 and 8548.
19844 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
19846         * config/i386/i386.c (ix86_output_function_return): Pass
19847         INVALID_REGNUM, instead of -1, as invalid register number to
19848         indirect_thunk_name and output_indirect_thunk.
19850 2018-02-02  Julia Koval  <julia.koval@intel.com>
19852         * config.gcc: Add -march=icelake.
19853         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
19854         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
19855         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
19856         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
19857         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
19858         (processor_target_table): Add icelake.
19859         (ix86_option_override_internal): Handle new PTAs.
19860         (get_builtin_code_for_version): Handle icelake.
19861         (M_INTEL_COREI7_ICELAKE): New.
19862         (fold_builtin_cpu): Handle icelake.
19863         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
19864         * doc/invoke.texi: Add -march=icelake.
19866 2018-02-02  Julia Koval  <julia.koval@intel.com>
19868         * config/i386/i386.c (ix86_option_override_internal): Change flags type
19869         to wide_int_bitmask.
19870         * wide-int-bitmask.h: New.
19872 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
19874         PR target/84066
19875         * config/i386/i386.md: Replace Pmode with word_mode in
19876         builtin_setjmp_setup and builtin_longjmp to support x32.
19878 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
19880         PR target/56010
19881         PR target/83743
19882         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
19883         #include "opts.h".
19884         (rs6000_supported_cpu_names): New static variable.
19885         (linux_cpu_translation_table): Likewise.
19886         (elf_platform) <cpu>: Define new static variable and use it.
19887         Translate kernel AT_PLATFORM name to canonical name if needed.
19888         Error if platform name is unknown.
19890 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
19892         PR target/84089
19893         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19895 2018-02-01  Jeff Law  <law@redhat.com>
19897         PR target/84128
19898         * config/i386/i386.c (release_scratch_register_on_entry): Add new
19899         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
19900         the scratch if RELEASE_VIA_POP is false.
19901         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
19902         If we have to save a temporary register, decrement SIZE appropriately.
19903         Pass new arguments to release_scratch_register_on_entry.
19904         (ix86_adjust_stack_and_probe): Likewise.
19905         (ix86_emit_probe_stack_range): Pass new arguments to
19906         release_scratch_register_on_entry.
19908 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
19910         PR rtl-optimization/84157
19911         * combine.c (change_zero_ext): Use REG_P predicate in
19912         front of HARD_REGISTER_P predicate.
19914 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
19916         * config/avr/avr.c (avr_option_override): Move disabling of
19917         -fdelete-null-pointer-checks to...
19918         * common/config/avr/avr-common.c (avr_option_optimization_table):
19919         ...here.
19921 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
19923         PR tree-optimization/81635
19924         * tree-data-ref.c (split_constant_offset_1): For types that
19925         wrap on overflow, try to use range info to prove that wrapping
19926         cannot occur.
19928 2018-02-01  Renlin Li  <renlin.li@arm.com>
19930         PR target/83370
19931         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
19932         TAILCALL_ADDR_REGS.
19933         (aarch64_register_move_cost): Likewise.
19934         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
19935         TAILCALL_ADDR_REGS.
19936         (REG_CLASS_NAMES): Likewise.
19937         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
19938         TAILCALL_ADDR_REGS. Remove IP registers.
19939         * config/aarch64/aarch64.md (Ucs): Update register constraint.
19941 2018-02-01  Richard Biener  <rguenther@suse.de>
19943         * domwalk.h (dom_walker::dom_walker): Add additional constructor
19944         for specifying RPO order and allow NULL for that.
19945         * domwalk.c (dom_walker::dom_walker): Likewise.
19946         (dom_walker::walk): Handle NULL RPO order.
19947         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
19948         in RPO order.
19949         (rewrite_update_dom_walker): Likewise.
19950         (mark_def_dom_walker): Likewise.
19952 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
19954         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
19955         (aarch64_maybe_expand_sve_subreg_move): Declare.
19956         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
19957         * config/aarch64/predicates.md (aarch64_any_register_operand): New
19958         predicate.
19959         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
19960         that are semantically a reverse operation.
19961         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
19962         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
19963         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
19964         functions.
19965         (aarch64_can_change_mode_class): For big-endian, forbid changes
19966         between two SVE modes if they have different element sizes.
19968 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
19970         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
19971         the TImode handling for big-endian targets.
19973 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
19975         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
19976         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
19977         not just bytes.
19978         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
19979         Remove BSWAP handing for big-endian targets and use the form of
19980         LD1RQ appropariate for the mode.
19982 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
19984         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
19985         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
19986         duplicated element.
19988 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
19990         PR tearget/83845
19991         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
19992         check for operands that need to go through aarch64_sve_reload_be.
19994 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
19996         PR tree-optimization/81661
19997         PR tree-optimization/84117
19998         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
19999         * tree-eh.c: Include gimplify.h.
20000         (find_trapping_overflow, replace_trapping_overflow,
20001         rewrite_to_non_trapping_overflow): New functions.
20002         * tree-vect-loop.c: Include tree-eh.h.
20003         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
20004         * tree-data-ref.c: Include tree-eh.h.
20005         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
20007 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
20009         PR rtl-optimization/84123
20010         * combine.c (change_zero_ext): Check if hard register satisfies
20011         can_change_dest_mode before calling gen_lowpart_SUBREG.
20013 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
20015         PR target/82444
20016         * ira.c (ira_init_register_move_cost): Remove assert.
20018 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20020         PR rtl-optimization/84071
20021         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
20022         * doc/tm.texi: Regenerate.
20024 2018-01-31  Richard Biener  <rguenther@suse.de>
20026         PR tree-optimization/84132
20027         * tree-data-ref.c (analyze_miv_subscript): Properly
20028         check whether evolution_function_is_affine_multivariate_p
20029         before calling gcd_of_steps_may_divide_p.
20031 2018-01-31  Julia Koval  <julia.koval@intel.com>
20033         PR target/83618
20034         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
20035         * config/i386/i386.md (rdpid_rex64) New.
20036         (rdpid): Make 32bit only.
20038 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
20040         PR lto/84105
20041         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
20042         an IDENTIFIER_NODE for FUNCTION_TYPE's.
20044 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20046         Revert
20047         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
20049         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
20051 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
20053         PR rtl-optimization/84071
20054         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
20055         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
20057 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
20059         * config/arc/arc.c (arc_handle_aux_attribute): New function.
20060         (arc_attribute_table): Add 'aux' attribute.
20061         (arc_in_small_data_p): Consider aux like variables.
20062         (arc_is_aux_reg_p): New function.
20063         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
20064         (arc_get_aux_arg): New function.
20065         (prepare_move_operands): Handle aux-register access.
20066         (arc_handle_aux_attribute): New function.
20067         * doc/extend.texi (ARC Variable attributes): Add subsection.
20069 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
20071         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
20072         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
20073         (arc_attribute_table): Add 'uncached' attribute.
20074         (arc_print_operand): Print '.di' flag for uncached memory
20075         accesses.
20076         (arc_in_small_data_p): Do not consider for small data the uncached
20077         types.
20078         (arc_is_uncached_mem_p): New function.
20079         * config/arc/predicates.md (compact_store_memory_operand): Check
20080         for uncached memory accesses.
20081         (nonvol_nonimm_operand): Likewise.
20082         * doc/extend.texi (ARC Type Attribute): New subsection.
20084 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
20086         PR c/84100
20087         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20088         falign-loops=): Add Optimization flag.
20090 2018-01-30  Jeff Law  <law@redhat.com>
20092         PR target/84064
20093         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
20094         INT_REGISTERS_SAVED.  Check it prior to calling
20095         get_scratch_register_on_entry.
20096         (ix86_adjust_stack_and_probe): Similarly.
20097         (ix86_emit_probe_stack_range): Similarly.
20098         (ix86_expand_prologue): Corresponding changes.
20100 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20102         PR target/40411
20103         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
20104         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
20106 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
20108         PR target/84112
20109         * lra-constraints.c (curr_insn_transform): Process AND in the
20110         address.
20112 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
20114         PR rtl-optimization/83986
20115         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
20116         dependence against last_pending_memory_flush in addition to
20117         pending_jump_insns.
20119 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
20121         PR tree-optimization/81611
20122         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
20123         copies.
20125 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
20127         PR target/83758
20128         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
20129         a reg rtx.
20131 2018-01-30  Richard Biener  <rguenther@suse.de>
20132             Jakub Jelinek  <jakub@redhat.com>
20134         PR tree-optimization/84111
20135         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
20136         inner loops added during recursion, as they don't have up-to-date
20137         SSA form.
20139 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
20141         PR ipa/81360
20142         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
20143         (can_inline_edge_by_limits_p): ... here.
20144         (can_early_inline_edge_p, check_callers,
20145         update_caller_keys, update_callee_keys, recursive_inlining,
20146         add_new_edges_to_heap, speculation_useful_p,
20147         inline_small_functions,
20148         inline_small_functions, flatten_function,
20149         inline_to_all_callers_1): Update.
20151 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
20153         * profile-count.c (profile_count::combine_with_ipa_count): Handle
20154         zeros correctly.
20156 2018-01-30  Richard Biener  <rguenther@suse.de>
20158         PR tree-optimization/83008
20159         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
20160         invariant and constant vector uses in stmts when they need
20161         more than one stmt.
20163 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20165         PR bootstrap/84017
20166         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
20167         * configure: Regenerate.
20169 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20171         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
20172         pattern.
20173         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
20174         Use gen_rtx_REG rather than gen_lowpart.
20176 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20178         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
20179         rather than 0 when creating partial subregs.
20181 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
20183         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
20184         of usage.
20186 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20188         PR target/81550
20189         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
20190         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
20191         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
20192         flags.  This restores the settings used before the 2017-07-24.
20193         Turning off pre increment/decrement/modify allows IVOPTS to
20194         optimize DF/SF loops where the index is an int.
20196 2018-01-29  Richard Biener  <rguenther@suse.de>
20197             Kelvin Nilsen  <kelvin@gcc.gnu.org>
20199         PR bootstrap/80867
20200         * tree-vect-stmts.c (vectorizable_call): Don't call
20201         targetm.vectorize_builtin_md_vectorized_function if callee is
20202         NULL.
20204 2018-01-22  Carl Love  <cel@us.ibm.com>
20206         * doc/extend.tex: Fix typo in second arg in
20207         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
20209 2018-01-29  Richard Biener  <rguenther@suse.de>
20211         PR tree-optimization/84086
20212         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
20213         (flush_ssaname_freelist): When SSA names were released reset
20214         the SCEV hash table.
20216 2018-01-29  Richard Biener  <rguenther@suse.de>
20218         PR tree-optimization/84057
20219         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
20220         removed paths when removing edges.
20222 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
20224         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
20225         -mfunction-return=@var{choice}.
20227 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
20229         PR diagnostic/84034
20230         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
20231         Handle CR like TAB.
20232         (layout::print_source_line): Likewise.
20233         (test_get_line_width_without_trailing_whitespace): Add test cases.
20235 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
20237         PR middle-end/84040
20238         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
20239         debug insns.
20241 2018-01-26  Jim Wilson  <jimw@sifive.com>
20243         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
20245         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
20246         specified.
20248 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20250         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
20251         and CMP + SUB-immediate -> SUBS.
20253 2018-01-26  Martin Sebor  <msebor@redhat.com>
20255         PR tree-optimization/83896
20256         * tree-ssa-strlen.c (get_string_len): Rename...
20257         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
20258         Avoid assuming length is constant.
20259         (handle_char_store): Use HOST_WIDE_INT for string length.
20261 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
20263         PR target/81763
20264         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
20265         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
20267 2018-01-26  Richard Biener  <rguenther@suse.de>
20269         PR rtl-optimization/84003
20270         * dse.c (record_store): Only record redundant stores when
20271         the earlier store aliases at least all accesses the later one does.
20273 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
20275         PR rtl-optimization/83985
20276         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
20277         REG_CFA_RESTORE insns.
20278         (delete_unmarked_insns): Don't ignore separate shrink wrapping
20279         REG_CFA_RESTORE insns here.
20281         PR c/83989
20282         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
20283         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
20285 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20287         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
20288         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
20289         (arc_init): Likewise.
20290         (arc_override_options): Likewise.
20291         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
20292         value.
20293         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
20294         support.
20295         * config/arc/arc.h (TARGET_DBNZ): Define.
20296         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
20297         properly set the tune attribute.
20298         (dbnz): Use TARGET_DBNZ guard.
20299         * config/arc/arc.opt (mtune): Add core3 option.
20301 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20303         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
20304         recognize new pic like addresses.
20305         (arc_delegitimize_address): Clean up.
20307 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20309         * config/arc/arc-arches.def: Option mrf16 valid for all
20310         architectures.
20311         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
20312         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
20313         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
20314         * config/arc/arc-tables.opt: Regenerate.
20315         * config/arc/arc.c (arc_conditional_register_usage): Handle
20316         reduced register file case.
20317         (arc_file_start): Set must have build attributes.
20318         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
20319         mrf16 option value.
20320         * config/arc/arc.opt (mrf16): Add new option.
20321         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
20322         * config/arc/genmultilib.awk: Handle new mrf16 option.
20323         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
20324         * config/arc/t-multilib: Regenerate.
20325         * doc/invoke.texi (ARC Options): Document mrf16 option.
20327 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20329         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
20330         * config/arc/arc.c (arc_handle_secure_attribute): New function.
20331         (arc_attribute_table): Add 'secure_call' attribute.
20332         (arc_print_operand): Print secure call operand.
20333         (arc_function_ok_for_sibcall): Don't optimize tail calls when
20334         secure.
20335         (arc_is_secure_call_p): New function.  * config/arc/arc.md
20336         (call_i): Add support for sjli instruction.
20337         (call_value_i): Likewise.
20338         * config/arc/constraints.md (Csc): New constraint.
20340 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
20341             John Eric Martin  <John.Martin@emmicro-us.com>
20343         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
20344         * config/arc/arc.c (_arc_jli_section): New struct.
20345         (arc_jli_section): New type.
20346         (rc_jli_sections): New static variable.
20347         (arc_handle_jli_attribute): New function.
20348         (arc_attribute_table): Add jli_always and jli_fixed attribute.
20349         (arc_file_end): New function.
20350         (TARGET_ASM_FILE_END): Define.
20351         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
20352         (arc_add_jli_section): New function.
20353         (jli_call_scan): Likewise.
20354         (arc_reorg): Call jli_call_scan.
20355         (arc_output_addsi): Remove 'S' from printing asm operand.
20356         (arc_is_jli_call_p): New function.
20357         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
20358         operand.
20359         (movhi_insn): Likewise.
20360         (movsi_insn): Likewise.
20361         (movsi_set_cc_insn): Likewise.
20362         (loadqi_update): Likewise.
20363         (load_zeroextendqisi_update): Likewise.
20364         (load_signextendqisi_update): Likewise.
20365         (loadhi_update): Likewise.
20366         (load_zeroextendhisi_update): Likewise.
20367         (load_signextendhisi_update): Likewise.
20368         (loadsi_update): Likewise.
20369         (loadsf_update): Likewise.
20370         (movsicc_insn): Likewise.
20371         (bset_insn): Likewise.
20372         (bxor_insn): Likewise.
20373         (bclr_insn): Likewise.
20374         (bmsk_insn): Likewise.
20375         (bicsi3_insn): Likewise.
20376         (cmpsi_cc_c_insn): Likewise.
20377         (movsi_ne): Likewise.
20378         (movsi_cond_exec): Likewise.
20379         (clrsbsi2): Likewise.
20380         (norm_f): Likewise.
20381         (normw): Likewise.
20382         (swap): Likewise.
20383         (divaw): Likewise.
20384         (flag): Likewise.
20385         (sr): Likewise.
20386         (kflag): Likewise.
20387         (ffs): Likewise.
20388         (ffs_f): Likewise.
20389         (fls): Likewise.
20390         (call_i): Remove 'S' asm letter, add jli instruction.
20391         (call_value_i): Likewise.
20392         * config/arc/arc.op (mjli-always): New option.
20393         * config/arc/constraints.md (Cji): New constraint.
20394         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
20395         operand.
20396         (subsf3_fpx): Likewise.
20397         (mulsf3_fpx): Likewise.
20398         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
20399         asm operand.
20400         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
20401         function attrbutes.
20402         * doc/invoke.texi (ARC): Document mjli-always option.
20404 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
20406         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
20407         avoid addition with 0 and use incw and decw where possible.
20409 2018-01-26  Richard Biener  <rguenther@suse.de>
20411         PR tree-optimization/81082
20412         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
20413         association if it requires casting to unsigned.
20414         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
20415         from fold_plusminus_mult_expr to catch important cases late when
20416         range info is available.
20418 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
20420         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
20421         * configure.ac (hidden_linkonce): New test.
20422         * configure: Regenerate.
20423         * config.in: Regenerate.
20425 2018-01-26  Julia Koval  <julia.koval@intel.com>
20427         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
20428         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
20429         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
20430         _mm_mask_bitshuffle_epi64_mask): Fix type.
20431         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
20432         USI_FTYPE_V4DI_V4DI_USI): Remove.
20433         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
20434         __builtin_ia32_vpshufbitqmb256_mask,
20435         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
20436         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
20437         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
20439 2018-01-26  Alan Modra  <amodra@gmail.com>
20441         PR target/84033
20442         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
20443         UNSPEC_VBPERMQ.  Sort other unspecs.
20445 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
20447         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
20449 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
20451         PR middle-end/83055
20452         * predict.c (drop_profile): Do not push/pop cfun; update also
20453         node->count.
20454         (handle_missing_profiles): Fix logic looking for zero profiles.
20456 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
20458         PR middle-end/83977
20459         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
20460         on functions with #pragma omp declare simd or functions with simd
20461         attribute.
20462         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
20463         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
20464         Remove trailing \n from warning_at calls.
20466 2018-01-25  Tom de Vries  <tom@codesourcery.com>
20468         PR target/84028
20469         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20470         for neutered workers.
20472 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
20474         PR target/68467
20475         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
20476         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
20478 2018-01-24  Jeff Law  <law@redhat.com>
20480         PR target/83994
20481         * i386.c (get_probe_interval): Move to earlier point.
20482         (ix86_compute_frame_layout): If -fstack-clash-protection and
20483         the frame is larger than the probe interval, then use pushes
20484         to save registers rather than reg->mem moves.
20485         (ix86_expand_prologue): Remove conditional for int_registers_saved
20486         assertion.
20488 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
20490         PR target/84014
20491         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
20492         min/max for never referenced object.
20494 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
20496         PR middle-end/83977
20497         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
20498         here.
20499         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
20500         attributes from DECL_ATTRIBUTES (decl) without affecting
20501         DECL_ATTRIBUTES (current_function_decl).
20502         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
20503         functions with non-NULL DECL_ABSTRACT_ORIGIN.
20505 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
20507         PR tree-optimization/83979
20508         * fold-const.c (fold_comparison): Use constant_boolean_node
20509         instead of boolean_{true,false}_node.
20511 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
20513         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
20514         with zero counts.
20516 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20518         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20519         Simplify the clause that sets the length attribute.
20520         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20521         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
20522         clause that sets the length attribute.
20523         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20525 2018-01-24  Tom de Vries  <tom@codesourcery.com>
20527         PR target/83589
20528         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
20529         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
20530         Add strict parameter.
20531         (prevent_branch_around_nothing): Insert dummy insn between branch to
20532         label and label with no ptx insn inbetween.
20533         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
20535 2018-01-24  Tom de Vries  <tom@codesourcery.com>
20537         PR target/81352
20538         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
20539         for neutered threads in warp.
20540         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
20542 2018-01-24  Richard Biener  <rguenther@suse.de>
20544         PR tree-optimization/83176
20545         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
20546         operands.
20548 2018-01-24  Richard Biener  <rguenther@suse.de>
20550         PR tree-optimization/82819
20551         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
20552         code generating pluses that are no-ops in the target precision.
20554 2018-01-24  Richard Biener  <rguenther@suse.de>
20556         PR middle-end/84000
20557         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
20559 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
20561         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
20562         to merge probabilities.
20563         * predict.c (probably_never_executed): Also mark as cold functions
20564         with global 0 profile and guessed local profile.
20565         * profile-count.c (profile_probability::combine_with_count): New
20566         member function.
20567         * profile-count.h (profile_probability::operator*,
20568         profile_probability::operator*=, profile_probability::operator/,
20569         profile_probability::operator/=): Reduce precision to adjusted
20570         and set value to guessed on contradictory divisions.
20571         (profile_probability::combine_with_freq): Remove.
20572         (profile_probability::combine_wiht_count): Declare.
20573         (profile_count::force_nonzero):: Set to adjusted.
20574         (profile_count::probability_in):: Set quality to adjusted.
20575         * tree-ssa-tail-merge.c (replace_block_by): Use
20576         combine_with_count.
20578 2018-01-23  Andrew Waterman  <andrew@sifive.com>
20579             Jim Wilson  <jimw@sifive.com>
20581         * config/riscv/riscv.c (riscv_stack_boundary): New.
20582         (riscv_option_override): Set riscv_stack_boundary.  Handle
20583         riscv_preferred_stack_boundary_arg.
20584         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
20585         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
20586         (STACK_BOUNDARY): Set to riscv_stack_boundary.
20587         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
20588         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
20589         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
20591 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
20593         PR target/83905
20594         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
20595         of struct ix86_frame.
20596         (ix86_expand_epilogue): Likewise.  Add a local variable for
20597         the reg_save_offset field in struct ix86_frame.
20599 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
20601         PR tree-optimization/82604
20602         * tree-loop-distribution.c (enum partition_kind): New enum item
20603         PKIND_PARTIAL_MEMSET.
20604         (partition_builtin_p): Support above new enum item.
20605         (generate_code_for_partition): Ditto.
20606         (compute_access_range): Differentiate cases that equality can be
20607         proven at all loops, the innermost loops or no loops.
20608         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
20609         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
20610         (finalize_partitions, distribute_loop): Don't fuse partition of
20611         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
20612         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
20613         parloop is enabled.
20615 2018-01-23  Martin Liska  <mliska@suse.cz>
20617         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
20618         order to ignore the predictor.
20619         (PRED_POLYMORPHIC_CALL): Likewise.
20620         (PRED_RECURSIVE_CALL): Likewise.
20622 2018-01-23  Martin Liska  <mliska@suse.cz>
20624         * tree-profile.c (tree_profiling): Print function header to
20625         aware reader which function we are working on.
20626         * value-prof.c (gimple_find_values_to_profile): Do not print
20627         not interesting value histograms.
20629 2018-01-23  Martin Liska  <mliska@suse.cz>
20631         * profile-count.h (enum profile_quality): Add
20632         profile_uninitialized as the first value. Do not number values
20633         as they are zero based.
20634         (profile_count::verify): Update sanity check.
20635         (profile_probability::verify): Likewise.
20637 2018-01-23  Nathan Sidwell  <nathan@acm.org>
20639         * doc/invoke.texi (ffor-scope): Deprecate.
20641 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
20643         PR tree-optimization/83510
20644         * domwalk.c (set_all_edges_as_executable): New function.
20645         (dom_walker::dom_walker): Convert bool param
20646         "skip_unreachable_blocks" to enum reachability.  Move setup of
20647         edge flags to set_all_edges_as_executable and only do it when
20648         reachability is REACHABLE_BLOCKS.
20649         * domwalk.h (enum dom_walker::reachability): New enum.
20650         (dom_walker::dom_walker): Convert bool param
20651         "skip_unreachable_blocks" to enum reachability.
20652         (set_all_edges_as_executable): New decl.
20653         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
20654         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
20655         "reachability".
20656         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
20657         but converting true to REACHABLE_BLOCKS.
20658         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
20659         * tree-vrp.c
20660         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
20661         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
20662         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
20663         REACHABLE_BLOCKS.
20664         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
20665         if check_all_array_refs will be called.
20667 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
20669         * tree.c (selftest::test_location_wrappers): Add more test
20670         coverage.
20672 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
20674         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
20675         (selftest::test_bit_in_range): Likewise.
20677 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
20679         PR testsuite/83888
20680         * doc/sourcebuild.texi (vect_float): Say that the selector
20681         only describes the situation when -funsafe-math-optimizations is on.
20682         (vect_float_strict): Document.
20684 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
20686         PR tree-optimization/83965
20687         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
20688         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
20689         instead of checking only for a reduction.
20690         (vect_recog_widen_sum_pattern): Likewise.
20692 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
20694         * predict.c (probably_never_executed): Only use precise profile info.
20695         (compute_function_frequency): Skip after inlining hack since we now
20696         have quality checking.
20698 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
20700         * profile-count.h (profile_probability::very_unlikely,
20701         profile_probability::unlikely, profile_probability::even): Set
20702         precision to guessed.
20704 2018-01-23  Richard Biener  <rguenther@suse.de>
20706         PR tree-optimization/83963
20707         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
20708         Properly terminate dominator walk when crossing the exit edge not
20709         when visiting its source block.
20711 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
20713         PR c++/83918
20714         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
20715         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
20717 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
20719         PR tree-optimization/83957
20720         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
20721         semicolon after for body surrounded by braces.
20723         PR tree-optimization/83081
20724         * profile-count.h (profile_probability::split): New method.
20725         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
20726         Use profile_probability::split.
20727         (do_compare_rtx_and_jump): Fix adjustment of probabilities
20728         when splitting a single conditional jump into 2.
20730 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
20732         PR tree-optimization/69452
20733         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
20734         decl.
20736 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20738         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
20739         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
20740         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
20742 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20744         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
20745         declaration.
20746         * config/rl78/rl78.md (movdi): New define_expand.
20747         * config/rl78/rl78.c (rl78_split_movdi): New function.
20749 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
20751         PR target/83862
20752         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
20753         no longer used.
20754         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
20755         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
20756         128-bit to produce an UNSPEC move to get the double word with the
20757         signbit and then a shift directly to do signbit.
20758         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
20759         implementation with a new version that just does either a direct
20760         move or a regular move.  Move memory interface to separate insns.
20761         Move insns so they are next to the expander.
20762         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
20763         with signbit move.  Split big and little endian case.
20764         (signbit<mode>2_dm_mem_le): Likewise.
20765         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
20766         (signbit<mode>2_dm2): Likewise.
20768 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20770         * config/rl78/rl78.md (anddi3): New define_expand.
20772 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20774         * config/rl78/rl78.md (umindi3): New define_expand.
20776 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20778         * config/rl78/rl78.md (smindi3): New define_expand.
20780 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20782         * config/rl78/rl78.md (smaxdi3): New define_expand.
20784 2018-01-22  Carl Love  <cel@us.ibm.com>
20786         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
20787         LVX_V1TI): Add macro expansion.
20788         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
20789         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
20790         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
20791         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
20792         Change check to determine if the instruction is a byte reversing
20793         entry.  Fix typo in comment.
20794         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
20795         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
20796         Add def_builtin calls for new builtins.
20797         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
20798         Add define_insn expansion.
20800 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20802         * config/rl78/rl78.md (umaxdi3): New define_expand.
20804 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
20806         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
20807         for non-QImode registers.
20809 2018-01-22  Richard Biener  <rguenther@suse.de>
20811         PR tree-optimization/83963
20812         * graphite-scop-detection.c (scop_detection::get_sese): Delay
20813         including the loop exit block.
20814         (scop_detection::merge_sese): Likewise.
20815         (scop_detection::add_scop): Do it here instead.
20817 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20819         * doc/sourcebuild.texi (arm_softfloat): Document.
20821 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
20823         PR gcc/77734
20824         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
20825         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
20826         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
20828 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20829             David Edelsohn  <dje.gcc@gmail.com>
20831         PR target/83946
20832         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
20833         Change "crset eq" to "crset 2".
20834         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
20835         (*call_indirect_aix<mode>_nospec): Likewise.
20836         (*call_value_indirect_aix<mode>_nospec): Likewise.
20837         (*call_indirect_elfv2<mode>_nospec): Likewise.
20838         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
20839         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
20840         change assembly output from . to $.
20841         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
20842         (indirect_jump<mode>_nospec): Change assembly output from . to $.
20843         (*tablejump<mode>_internal1_nospec): Likewise.
20845 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
20847         PR target/80870
20848         * config/sh/sh_optimize_sett_clrt.cc:
20849         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
20851 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
20853         PR tree-optimization/83940
20854         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
20855         offset_dt to vect_constant_def rather than vect_unknown_def_type.
20856         (vect_check_load_store_mask): Add a mask_dt_out parameter and
20857         use it to pass back the definition type.
20858         (vect_check_store_rhs): Likewise rhs_dt_out.
20859         (vect_build_gather_load_calls): Add a mask_dt argument and use
20860         it instead of a call to vect_is_simple_use.
20861         (vectorizable_store): Update calls to vect_check_load_store_mask
20862         and vect_check_store_rhs.  Use the dt returned by the latter instead
20863         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
20864         instead of calls to vect_is_simple_use.  Pass the scalar rather
20865         than the vector operand to vect_is_simple_use when handling
20866         second and subsequent copies of an rhs value.
20867         (vectorizable_load): Update calls to vect_check_load_store_mask
20868         and vect_build_gather_load_calls.  Use the cached mask_dt and
20869         gs_info.offset_dt instead of calls to vect_is_simple_use.
20871 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
20873         PR middle-end/83945
20874         * tree-emutls.c: Include gimplify.h.
20875         (lower_emutls_2): New function.
20876         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
20877         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
20878         it before further processing.
20880         PR target/83930
20881         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
20882         UINTVAL (trueop1) instead of INTVAL (op1).
20884 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
20886         PR debug/81570
20887         PR debug/83728
20888         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
20889         INCOMING_FRAME_SP_OFFSET if not defined.
20890         (scan_trace): Add ENTRY argument.  If true and
20891         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
20892         emit a note to adjust the CFA offset.
20893         (create_cfi_notes): Adjust scan_trace callers.
20894         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
20895         INCOMING_FRAME_SP_OFFSET in the CIE.
20896         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
20897         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
20898         Likewise.
20899         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
20900         * doc/tm.texi: Regenerated.
20902 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20904         PR rtl-optimization/83147
20905         * lra-constraints.c (remove_inheritance_pseudos): Use
20906         lra_substitute_pseudo_within_insn.
20908 2018-01-19  Tom de Vries  <tom@codesourcery.com>
20909             Cesar Philippidis  <cesar@codesourcery.com>
20911         PR target/83920
20912         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20914 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
20916         PR target/83790
20917         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
20918         spaces for function labels.
20920 2018-01-19  Martin Liska  <mliska@suse.cz>
20922         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
20923         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
20924         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
20925         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
20926         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
20927         (PRED_CONST_RETURN): Change from 69 to 65.
20928         (PRED_NULL_RETURN): Change from 91 to 71.
20929         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
20930         (PRED_LOOP_GUARD): Change from 66 to 73.
20932 2018-01-19  Martin Liska  <mliska@suse.cz>
20934         * predict.c (predict_insn_def): Add new assert.
20935         (struct branch_predictor): Change type to signed integer.
20936         (test_prediction_value_range): Amend test to cover
20937         PROB_UNINITIALIZED.
20938         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
20939         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
20940         (PRED_LOOP_ITERATIONS_MAX): Likewise.
20941         (PRED_LOOP_IV_COMPARE): Likewise.
20942         * predict.h (PROB_UNINITIALIZED): Define new constant.
20944 2018-01-19  Martin Liska  <mliska@suse.cz>
20946         * predict.c (dump_prediction): Add new format for
20947         analyze_brprob.py script which is enabled with -details
20948         suboption.
20949         * profile-count.h (precise_p): New function.
20951 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
20953         PR tree-optimization/83922
20954         * tree-vect-loop.c (vect_verify_full_masking): Return false if
20955         there are no statements that need masking.
20956         (vect_active_double_reduction_p): New function.
20957         (vect_analyze_loop_operations): Use it when handling phis that
20958         are not in the loop header.
20960 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
20962         PR tree-optimization/83914
20963         * tree-vect-loop.c (vectorizable_induction): Don't convert
20964         init_expr or apply the peeling adjustment for inductions
20965         that are nested within the vectorized loop.
20967 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20969         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
20970         instead of NEG.
20972 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
20974         PR sanitizer/81715
20975         PR testsuite/83882
20976         * function.h (gimplify_parameters): Add gimple_seq * argument.
20977         * function.c: Include gimple.h and options.h.
20978         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
20979         for the added local temporaries if needed.
20980         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
20981         if there are any parameter cleanups, wrap whole body into a
20982         try/finally with the cleanups.
20984 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
20986         PR target/82964
20987         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
20988         Use GET_MODE_CLASS for scalar floating point.
20990 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
20992         PR ipa/82256
20993         patch by PaX Team
20994         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
20995         Fix call of call_cgraph_insertion_hooks.
20997 2018-01-18  Martin Sebor  <msebor@redhat.com>
20999         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
21001 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
21003         PR ipa/83619
21004         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21005         frequencies.
21007 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
21009         PR other/70268
21010         * common.opt: (-ffile-prefix-map): New option.
21011         * opts.c (common_handle_option): Defer it.
21012         * opts-global.c (handle_common_deferred_options): Handle it.
21013         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
21014         * file-prefix-map.h: New file.
21015         (remap_debug_filename, add_debug_prefix_map): ...here.
21016         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
21017         * final.c (debug_prefix_map, add_debug_prefix_map
21018         remap_debug_filename): Move to...
21019         * file-prefix-map.c: New file.
21020         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
21021         generalize, get rid of alloca(), use strrchr() instead of strchr().
21022         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
21023         Implement in terms of add_prefix_map().
21024         (remap_macro_filename, remap_debug_filename): Implement in term of
21025         remap_filename().
21026         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
21027         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
21028         * dbxout.c: Include file-prefix-map.h.
21029         * varasm.c: Likewise.
21030         * vmsdbgout.c: Likewise.
21031         * xcoffout.c: Likewise.
21032         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
21033         * doc/cppopts.texi (-fmacro-prefix-map): Document.
21034         * doc/invoke.texi (-ffile-prefix-map): Document.
21035         (-fdebug-prefix-map): Update description.
21037 2018-01-18  Martin Liska  <mliska@suse.cz>
21039         * config/i386/i386.c (indirect_thunk_name): Document that also
21040         lfence is emitted.
21041         (output_indirect_thunk): Document why both instructions
21042         (pause and lfence) are generated.
21044 2018-01-18  Richard Biener  <rguenther@suse.de>
21046         PR tree-optimization/83887
21047         * graphite-scop-detection.c
21048         (scop_detection::get_nearest_dom_with_single_entry): Remove.
21049         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
21050         (scop_detection::merge_sese): Re-implement with a flood-fill
21051         algorithm that properly finds a SESE region if it exists.
21053 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
21055         PR c/61240
21056         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
21057         pointer_diff optimizations use view_convert instead of convert.
21059 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21061         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21062         Generate different code for -mno-speculate-indirect-jumps.
21063         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21064         (*call_indirect_aix<mode>): Disable for
21065         -mno-speculate-indirect-jumps.
21066         (*call_indirect_aix<mode>_nospec): New define_insn.
21067         (*call_value_indirect_aix<mode>): Disable for
21068         -mno-speculate-indirect-jumps.
21069         (*call_value_indirect_aix<mode>_nospec): New define_insn.
21070         (*sibcall_nonlocal_sysv<mode>): Generate different code for
21071         -mno-speculate-indirect-jumps.
21072         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21074 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21076         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
21077         long double type, set the flags for noting the default long double
21078         type, even if we don't pass or return a long double type.
21080 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
21082         PR ipa/83051
21083         * ipa-inline.c (flatten_function): Do not overwrite final inlining
21084         failure.
21086 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
21088         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
21089         support for merge[hl].
21090         (fold_mergehl_helper): New helper function.
21091         (tree-vector-builder.h): New #include for tree_vector_builder usage.
21092         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
21093         (altivec_vmrglw_direct): Add xxmrglw insn.
21095 2018-01-17  Andrew Waterman  <andrew@sifive.com>
21097         * config/riscv/riscv.c (riscv_conditional_register_usage): If
21098         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
21100 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
21102         PR lto/83121
21103         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
21104         call the lto_location_cache before reading the
21105         DECL_SOURCE_LOCATION of the types.
21107 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
21108             Richard Sandiford  <richard.sandiford@linaro.org>
21110         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
21111         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
21112         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
21113         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
21114         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
21115         Add declaration.
21116         * config/aarch64/constraints.md (aarch64_movti_operand):
21117         Limit immediates.
21118         * config/aarch64/predicates.md (Uti): Add new constraint.
21120 2018-01-17  Carl Love  <cel@us.ibm.com>
21122         * config/rs6000/vsx.md (define_expand xl_len_r,
21123         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
21124         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
21125         lxvll.
21126         (define_expand, define_insn): Move the shift left from  the
21127         define_insn to the define_expand for lxvl and stxvl instructions.
21128         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
21129         and XL_LEN_R definitions to PURE.
21131 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
21133         * config/i386/i386.c (indirect_thunk_name): Declare regno
21134         as unsigned int.  Compare regno with INVALID_REGNUM.
21135         (output_indirect_thunk): Ditto.
21136         (output_indirect_thunk_function): Ditto.
21137         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
21138         in the call to output_indirect_thunk_function.
21140 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
21142         PR middle-end/83884
21143         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
21144         rather than the size of inner_type to determine the stack slot size
21145         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
21147 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
21149         PR target/83546
21150         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
21151         to PTA_SILVERMONT.
21153 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
21155         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
21156         endian Linux systems to optionally enable multilibs for selecting
21157         the long double type if the user configured an explicit type.
21158         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
21159         have no long double multilibs if not defined.
21160         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
21161         warn if the user used -mabi={ieee,ibm}longdouble and we built
21162         multilibs for long double.
21163         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
21164         appropriate multilib option.
21165         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
21166         multilib options.
21167         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
21168         for building long double multilibs.
21169         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
21171 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
21173         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
21174         copies.
21176         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21177         64 bits.
21178         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21179         128 bits.
21181         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21182         variables.
21184         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
21185         return value.
21187 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
21189         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
21190         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
21192 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21194         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
21195         different rtl trees depending on TARGET_64BIT.
21196         (rs6000_gen_lvx): Likewise.
21198 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
21200         * config/visium/visium.md (nop): Tweak comment.
21201         (hazard_nop): Likewise.
21203 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21205         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
21206         -mspeculate-indirect-jumps.
21207         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
21208         for -mno-speculate-indirect-jumps.
21209         (*call_indirect_elfv2<mode>_nospec): New define_insn.
21210         (*call_value_indirect_elfv2<mode>): Disable for
21211         -mno-speculate-indirect-jumps.
21212         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
21213         (indirect_jump): Emit different RTL for
21214         -mno-speculate-indirect-jumps.
21215         (*indirect_jump<mode>): Disable for
21216         -mno-speculate-indirect-jumps.
21217         (*indirect_jump<mode>_nospec): New define_insn.
21218         (tablejump): Emit different RTL for
21219         -mno-speculate-indirect-jumps.
21220         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
21221         (tablejumpsi_nospec): New define_expand.
21222         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
21223         (tablejumpdi_nospec): New define_expand.
21224         (*tablejump<mode>_internal1): Disable for
21225         -mno-speculate-indirect-jumps.
21226         (*tablejump<mode>_internal1_nospec): New define_insn.
21227         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
21228         option.
21230 2018-01-16  Artyom Skrobov tyomitch@gmail.com
21232         * caller-save.c (insert_save): Drop unnecessary parameter.  All
21233         callers updated.
21235 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21236             Richard Biener  <rguenth@suse.de>
21238         PR libgomp/83590
21239         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
21240         return early, inline manually is_gimple_sizepos.  Make sure if we
21241         call gimplify_expr we don't end up with a gimple constant.
21242         * tree.c (variably_modified_type_p): Don't return true for
21243         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
21244         * gimplify.h (is_gimple_sizepos): Remove.
21246 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21248         PR tree-optimization/83857
21249         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
21250         vectorizable_live_operation for pure SLP statements.
21251         (vectorizable_live_operation): Handle PHIs.
21253 2018-01-16  Richard Biener  <rguenther@suse.de>
21255         PR tree-optimization/83867
21256         * tree-vect-stmts.c (vect_transform_stmt): Precompute
21257         nested_in_vect_loop_p since the scalar stmt may get invalidated.
21259 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21261         PR c/83844
21262         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
21263         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
21264         If off is not INTEGER_CST, issue a may not be aligned warning
21265         rather than isn't aligned.  Use isn%'t rather than isn't.
21266         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
21267         into MULT_EXPR.
21268         <case MULT_EXPR>: Improve the case when bottom and one of the
21269         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
21270         operand, in that case check if the other operand is multiple of
21271         bottom divided by the INTEGER_CST operand.
21273 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21275         PR target/83858
21276         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
21277         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
21278         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
21279         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
21280         * config/pa/pa.c (pa_function_arg_advance): Likewise.
21281         (pa_function_arg, pa_arg_partial_bytes): Likewise.
21282         (pa_function_arg_size): New function.
21284 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21286         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
21287         in a separate statement.
21289 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
21291         PR tree-optimization/83847
21292         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
21293         group gathers and scatters.
21295 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
21297         PR rtl-optimization/86620
21298         * params.def (max-sched-ready-insns): Bump minimum value to 1.
21300         PR rtl-optimization/83213
21301         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
21302         to last if both are JUMP_INSNs.
21304         PR tree-optimization/83843
21305         * gimple-ssa-store-merging.c
21306         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
21307         store_immediate_info for bswap/nop orig_stores.
21309 2018-01-15  Andrew Waterman  <andrew@sifive.com>
21311         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
21312         !TARGET_MUL.
21313         <UDIV>: Increase cost if !TARGET_DIV.
21315 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
21317         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
21318         (define_attr "cr_logical_3op"): New.
21319         (cceq_ior_compare): Adjust.
21320         (cceq_ior_compare_complement): Adjust.
21321         (*cceq_rev_compare): Adjust.
21322         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
21323         (is_cracked_insn): Adjust.
21324         (insn_must_be_first_in_group): Adjust.
21325         * config/rs6000/40x.md: Adjust.
21326         * config/rs6000/440.md: Adjust.
21327         * config/rs6000/476.md: Adjust.
21328         * config/rs6000/601.md: Adjust.
21329         * config/rs6000/603.md: Adjust.
21330         * config/rs6000/6xx.md: Adjust.
21331         * config/rs6000/7450.md: Adjust.
21332         * config/rs6000/7xx.md: Adjust.
21333         * config/rs6000/8540.md: Adjust.
21334         * config/rs6000/cell.md: Adjust.
21335         * config/rs6000/e300c2c3.md: Adjust.
21336         * config/rs6000/e500mc.md: Adjust.
21337         * config/rs6000/e500mc64.md: Adjust.
21338         * config/rs6000/e5500.md: Adjust.
21339         * config/rs6000/e6500.md: Adjust.
21340         * config/rs6000/mpc.md: Adjust.
21341         * config/rs6000/power4.md: Adjust.
21342         * config/rs6000/power5.md: Adjust.
21343         * config/rs6000/power6.md: Adjust.
21344         * config/rs6000/power7.md: Adjust.
21345         * config/rs6000/power8.md: Adjust.
21346         * config/rs6000/power9.md: Adjust.
21347         * config/rs6000/rs64.md: Adjust.
21348         * config/rs6000/titan.md: Adjust.
21350 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21352         * config/i386/predicates.md (indirect_branch_operand): Rewrite
21353         ix86_indirect_branch_register logic.
21355 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21357         * config/i386/constraints.md (Bs): Update
21358         ix86_indirect_branch_register check.  Don't check
21359         ix86_indirect_branch_register with GOT_memory_operand.
21360         (Bw): Likewise.
21361         * config/i386/predicates.md (GOT_memory_operand): Don't check
21362         ix86_indirect_branch_register here.
21363         (GOT32_symbol_operand): Likewise.
21365 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21367         * config/i386/predicates.md (constant_call_address_operand):
21368         Rewrite ix86_indirect_branch_register logic.
21369         (sibcall_insn_operand): Likewise.
21371 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21373         * config/i386/constraints.md (Bs): Replace
21374         ix86_indirect_branch_thunk_register with
21375         ix86_indirect_branch_register.
21376         (Bw): Likewise.
21377         * config/i386/i386.md (indirect_jump): Likewise.
21378         (tablejump): Likewise.
21379         (*sibcall_memory): Likewise.
21380         (*sibcall_value_memory): Likewise.
21381         Peepholes of indirect call and jump via memory: Likewise.
21382         * config/i386/i386.opt: Likewise.
21383         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21384         (GOT_memory_operand): Likewise.
21385         (call_insn_operand): Likewise.
21386         (sibcall_insn_operand): Likewise.
21387         (GOT32_symbol_operand): Likewise.
21389 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
21391         PR middle-end/83837
21392         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
21393         type rather than type addr's type points to.
21394         (expand_omp_atomic_mutex): Likewise.
21395         (expand_omp_atomic): Likewise.
21397 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
21399         PR target/83839
21400         * config/i386/i386.c (output_indirect_thunk_function): Use
21401         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
21402         for  __x86_return_thunk.
21404 2018-01-15  Richard Biener  <rguenther@suse.de>
21406         PR middle-end/83850
21407         * expmed.c (extract_bit_field_1): Fix typo.
21409 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21411         PR target/83687
21412         * config/arm/iterators.md (VF): New mode iterator.
21413         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
21414         Remove integer-related logic from pattern.
21415         (neon_vabd<mode>_3): Likewise.
21417 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
21419         PR middle-end/82694
21420         * common.opt (fstrict-overflow): No longer an alias.
21421         (fwrapv-pointer): New option.
21422         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
21423         also for pointer types based on flag_wrapv_pointer.
21424         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
21425         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
21426         opts->x_flag_wrapv got set.
21427         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
21428         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
21429         POINTER_TYPE_OVERFLOW_UNDEFINED.
21430         * match.pd: Likewise in address comparison pattern.
21431         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
21433 2018-01-15  Richard Biener  <rguenther@suse.de>
21435         PR lto/83804
21436         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
21437         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
21438         Reset type names to their identifier if their TYPE_DECL doesn't
21439         have linkage (and thus is used for ODR and devirt).
21440         (save_debug_info_for_decl): Remove.
21441         (save_debug_info_for_type): Likewise.
21442         (add_tree_to_fld_list): Adjust.
21443         * tree-pretty-print.c (dump_generic_node): Make dumping of
21444         type names more robust.
21446 2018-01-15  Richard Biener  <rguenther@suse.de>
21448         * BASE-VER: Bump to 8.0.1.
21450 2018-01-14  Martin Sebor  <msebor@redhat.com>
21452         PR other/83508
21453         * builtins.c (check_access): Avoid warning when the no-warning bit
21454         is set.
21456 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
21458         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
21459         * ira-color (allocno_hard_regs_compare): Likewise.
21461 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
21463         PR target/83013
21464         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21465         Use .pushsection/.popsection.
21467 2018-01-14  Martin Sebor  <msebor@redhat.com>
21469         PR c++/81327
21470         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
21472 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
21474         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
21475         entry from extra_headers.
21476         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
21477         extra_headers, make the list bitwise identical to the i?86-*-* one.
21479 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21481         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
21482         -mcmodel=large with -mindirect-branch=thunk,
21483         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
21484         -mfunction-return=thunk-extern.
21485         * doc/invoke.texi: Document -mcmodel=large is incompatible with
21486         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
21487         -mfunction-return=thunk and -mfunction-return=thunk-extern.
21489 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21491         * config/i386/i386.c (print_reg): Print the name of the full
21492         integer register without '%'.
21493         (ix86_print_operand): Handle 'V'.
21494         * doc/extend.texi: Document 'V' modifier.
21496 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21498         * config/i386/constraints.md (Bs): Disallow memory operand for
21499         -mindirect-branch-register.
21500         (Bw): Likewise.
21501         * config/i386/predicates.md (indirect_branch_operand): Likewise.
21502         (GOT_memory_operand): Likewise.
21503         (call_insn_operand): Likewise.
21504         (sibcall_insn_operand): Likewise.
21505         (GOT32_symbol_operand): Likewise.
21506         * config/i386/i386.md (indirect_jump): Call convert_memory_address
21507         for -mindirect-branch-register.
21508         (tablejump): Likewise.
21509         (*sibcall_memory): Likewise.
21510         (*sibcall_value_memory): Likewise.
21511         Disallow peepholes of indirect call and jump via memory for
21512         -mindirect-branch-register.
21513         (*call_pop): Replace m with Bw.
21514         (*call_value_pop): Likewise.
21515         (*sibcall_pop_memory): Replace m with Bs.
21516         * config/i386/i386.opt (mindirect-branch-register): New option.
21517         * doc/invoke.texi: Document -mindirect-branch-register option.
21519 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21521         * config/i386/i386-protos.h (ix86_output_function_return): New.
21522         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
21523         set function_return_type.
21524         (indirect_thunk_name): Add ret_p to indicate thunk for function
21525         return.
21526         (output_indirect_thunk_function): Pass false to
21527         indirect_thunk_name.
21528         (ix86_output_indirect_branch_via_reg): Likewise.
21529         (ix86_output_indirect_branch_via_push): Likewise.
21530         (output_indirect_thunk_function): Create alias for function
21531         return thunk if regno < 0.
21532         (ix86_output_function_return): New function.
21533         (ix86_handle_fndecl_attribute): Handle function_return.
21534         (ix86_attribute_table): Add function_return.
21535         * config/i386/i386.h (machine_function): Add
21536         function_return_type.
21537         * config/i386/i386.md (simple_return_internal): Use
21538         ix86_output_function_return.
21539         (simple_return_internal_long): Likewise.
21540         * config/i386/i386.opt (mfunction-return=): New option.
21541         (indirect_branch): Mention -mfunction-return=.
21542         * doc/extend.texi: Document function_return function attribute.
21543         * doc/invoke.texi: Document -mfunction-return= option.
21545 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
21547         * config/i386/i386-opts.h (indirect_branch): New.
21548         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
21549         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
21550         with local indirect jump when converting indirect call and jump.
21551         (ix86_set_indirect_branch_type): New.
21552         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
21553         (indirectlabelno): New.
21554         (indirect_thunk_needed): Likewise.
21555         (indirect_thunk_bnd_needed): Likewise.
21556         (indirect_thunks_used): Likewise.
21557         (indirect_thunks_bnd_used): Likewise.
21558         (INDIRECT_LABEL): Likewise.
21559         (indirect_thunk_name): Likewise.
21560         (output_indirect_thunk): Likewise.
21561         (output_indirect_thunk_function): Likewise.
21562         (ix86_output_indirect_branch_via_reg): Likewise.
21563         (ix86_output_indirect_branch_via_push): Likewise.
21564         (ix86_output_indirect_branch): Likewise.
21565         (ix86_output_indirect_jmp): Likewise.
21566         (ix86_code_end): Call output_indirect_thunk_function if needed.
21567         (ix86_output_call_insn): Call ix86_output_indirect_branch if
21568         needed.
21569         (ix86_handle_fndecl_attribute): Handle indirect_branch.
21570         (ix86_attribute_table): Add indirect_branch.
21571         * config/i386/i386.h (machine_function): Add indirect_branch_type
21572         and has_local_indirect_jump.
21573         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
21574         to true.
21575         (tablejump): Likewise.
21576         (*indirect_jump): Use ix86_output_indirect_jmp.
21577         (*tablejump_1): Likewise.
21578         (simple_return_indirect_internal): Likewise.
21579         * config/i386/i386.opt (mindirect-branch=): New option.
21580         (indirect_branch): New.
21581         (keep): Likewise.
21582         (thunk): Likewise.
21583         (thunk-inline): Likewise.
21584         (thunk-extern): Likewise.
21585         * doc/extend.texi: Document indirect_branch function attribute.
21586         * doc/invoke.texi: Document -mindirect-branch= option.
21588 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
21590         PR ipa/83051
21591         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
21593 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
21595         * ipa-inline.c (want_inline_small_function_p): Return false if
21596         inlining has already failed with CIF_FINAL_ERROR.
21597         (update_caller_keys): Call want_inline_small_function_p before
21598         can_inline_edge_p.
21599         (update_callee_keys): Likewise.
21601 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21603         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
21604         New function.
21605         (rs6000_quadword_masked_address_p): Likewise.
21606         (quad_aligned_load_p): Likewise.
21607         (quad_aligned_store_p): Likewise.
21608         (const_load_sequence_p): Add comment to describe the outer-most loop.
21609         (mimic_memory_attributes_and_flags): New function.
21610         (rs6000_gen_stvx): Likewise.
21611         (replace_swapped_aligned_store): Likewise.
21612         (rs6000_gen_lvx): Likewise.
21613         (replace_swapped_aligned_load): Likewise.
21614         (replace_swapped_load_constant): Capitalize argument name in
21615         comment describing this function.
21616         (rs6000_analyze_swaps): Add a third pass to search for vector loads
21617         and stores that access quad-word aligned addresses and replace
21618         with stvx or lvx instructions when appropriate.
21619         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
21620         New function prototype.
21621         (rs6000_quadword_masked_address_p): Likewise.
21622         (rs6000_gen_lvx): Likewise.
21623         (rs6000_gen_stvx): Likewise.
21624         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
21625         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
21626         when memory address is aligned.
21627         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
21628         this split to select lvx instruction when memory address is aligned.
21629         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
21630         instruction when memory address is aligned.
21631         (*vsx_le_perm_load_v16qi): Likewise.
21632         (four unnamed splitters): Modify to select the stvx instruction
21633         when memory is aligned.
21635 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
21637         * predict.c (determine_unlikely_bbs): Handle correctly BBs
21638         which appears in the queue multiple times.
21640 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21641             Alan Hayward  <alan.hayward@arm.com>
21642             David Sherwood  <david.sherwood@arm.com>
21644         * tree-vectorizer.h (vec_lower_bound): New structure.
21645         (_loop_vec_info): Add check_nonzero and lower_bounds.
21646         (LOOP_VINFO_CHECK_NONZERO): New macro.
21647         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
21648         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
21649         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
21650         fields.  Make seg_len the distance travelled, not including the
21651         access size.
21652         (dr_direction_indicator): Declare.
21653         (dr_zero_step_indicator): Likewise.
21654         (dr_known_forward_stride_p): Likewise.
21655         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
21656         tree-ssanames.h.
21657         (runtime_alias_check_p): Allow runtime alias checks with
21658         variable strides.
21659         (operator ==): Compare access_size and align.
21660         (prune_runtime_alias_test_list): Rework for new distinction between
21661         the access_size and seg_len.
21662         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
21663         segment lengths.
21664         (get_segment_min_max): New function.
21665         (create_intersect_range_checks): Use it.
21666         (dr_step_indicator): New function.
21667         (dr_direction_indicator): Likewise.
21668         (dr_zero_step_indicator): Likewise.
21669         (dr_known_forward_stride_p): Likewise.
21670         * tree-loop-distribution.c (data_ref_segment_size): Return
21671         DR_STEP * (niters - 1).
21672         (compute_alias_check_pairs): Update call to the dr_with_seg_len
21673         constructor.
21674         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
21675         (vect_preserves_scalar_order_p): New function, split out from...
21676         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
21677         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
21678         (vect_vfa_access_size): New function.
21679         (vect_vfa_align): Likewise.
21680         (vect_compile_time_alias): Take access_size_a and access_b arguments.
21681         (dump_lower_bound): New function.
21682         (vect_check_lower_bound): Likewise.
21683         (vect_small_gap_p): Likewise.
21684         (vectorizable_with_step_bound_p): Likewise.
21685         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
21686         depencies if the vectorization factor is 1.  Convert the checks
21687         for nonzero steps into checks on the bounds of DR_STEP.  Try using
21688         a bunds check for variable steps if the minimum required step is
21689         relatively small. Update calls to the dr_with_seg_len
21690         constructor and to vect_compile_time_alias.
21691         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
21692         function.
21693         (vect_loop_versioning): Call it.
21694         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
21695         when retrying.
21696         (vect_estimate_min_profitable_iters): Account for any bounds checks.
21698 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21699             Alan Hayward  <alan.hayward@arm.com>
21700             David Sherwood  <david.sherwood@arm.com>
21702         * doc/sourcebuild.texi (vect_scatter_store): Document.
21703         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
21704         optabs.
21705         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
21706         Document.
21707         * genopinit.c (main): Add supports_vec_scatter_store and
21708         supports_vec_scatter_store_cached to target_optabs.
21709         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
21710         IFN_MASK_SCATTER_STORE.
21711         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
21712         functions.
21713         * internal-fn.h (internal_store_fn_p): Declare.
21714         (internal_fn_stored_value_index): Likewise.
21715         * internal-fn.c (scatter_store_direct): New macro.
21716         (expand_scatter_store_optab_fn): New function.
21717         (direct_scatter_store_optab_supported_p): New macro.
21718         (internal_store_fn_p): New function.
21719         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
21720         IFN_MASK_SCATTER_STORE.
21721         (internal_fn_mask_index): Likewise.
21722         (internal_fn_stored_value_index): New function.
21723         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
21724         for scatter stores.
21725         * optabs-query.h (supports_vec_scatter_store_p): Declare.
21726         * optabs-query.c (supports_vec_scatter_store_p): New function.
21727         * tree-vectorizer.h (vect_get_store_rhs): Declare.
21728         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
21729         true for scatter stores.
21730         (vect_gather_scatter_fn_p): Handle scatter stores too.
21731         (vect_check_gather_scatter): Consider using scatter stores if
21732         supports_vec_scatter_store_p.
21733         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
21734         scatter stores too.
21735         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21736         internal_fn_stored_value_index.
21737         (check_load_store_masking): Handle scatter stores too.
21738         (vect_get_store_rhs): Make public.
21739         (vectorizable_call): Use internal_store_fn_p.
21740         (vectorizable_store): Handle scatter store internal functions.
21741         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
21742         when deciding whether the end of the group has been reached.
21743         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
21744         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
21745         (mask_scatter_store<mode>): New insns.
21747 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21748             Alan Hayward  <alan.hayward@arm.com>
21749             David Sherwood  <david.sherwood@arm.com>
21751         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
21752         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
21753         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
21754         function.
21755         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
21756         Use vect_truncate_gather_scatter_offset if we can't treat the
21757         operation as a normal gather load or scatter store.
21758         (get_group_load_store_type): Take the gather_scatter_info
21759         as argument.  Try using a gather load or scatter store for
21760         single-element groups.
21761         (get_load_store_type): Update calls to get_group_load_store_type
21762         and vect_use_strided_gather_scatters_p.
21764 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21765             Alan Hayward  <alan.hayward@arm.com>
21766             David Sherwood  <david.sherwood@arm.com>
21768         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
21769         optional tree argument.
21770         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
21771         null target hooks.
21772         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
21773         but continue to use the current value as a fallback.
21774         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
21775         to compare the updates.
21776         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
21777         (get_load_store_type): Use it when handling a strided access.
21778         (vect_get_strided_load_store_ops): New function.
21779         (vect_get_data_ptr_increment): Likewise.
21780         (vectorizable_load): Handle strided gather loads.  Always pass
21781         a step to vect_create_data_ref_ptr and bump_vector_ptr.
21783 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21784             Alan Hayward  <alan.hayward@arm.com>
21785             David Sherwood  <david.sherwood@arm.com>
21787         * doc/md.texi (gather_load@var{m}): Document.
21788         (mask_gather_load@var{m}): Likewise.
21789         * genopinit.c (main): Add supports_vec_gather_load and
21790         supports_vec_gather_load_cached to target_optabs.
21791         * optabs-tree.c (init_tree_optimization_optabs): Use
21792         ggc_cleared_alloc to allocate target_optabs.
21793         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
21794         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
21795         functions.
21796         * internal-fn.h (internal_load_fn_p): Declare.
21797         (internal_gather_scatter_fn_p): Likewise.
21798         (internal_fn_mask_index): Likewise.
21799         (internal_gather_scatter_fn_supported_p): Likewise.
21800         * internal-fn.c (gather_load_direct): New macro.
21801         (expand_gather_load_optab_fn): New function.
21802         (direct_gather_load_optab_supported_p): New macro.
21803         (direct_internal_fn_optab): New function.
21804         (internal_load_fn_p): Likewise.
21805         (internal_gather_scatter_fn_p): Likewise.
21806         (internal_fn_mask_index): Likewise.
21807         (internal_gather_scatter_fn_supported_p): Likewise.
21808         * optabs-query.c (supports_at_least_one_mode_p): New function.
21809         (supports_vec_gather_load_p): Likewise.
21810         * optabs-query.h (supports_vec_gather_load_p): Declare.
21811         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
21812         and memory_type field.
21813         (NUM_PATTERNS): Bump to 15.
21814         * tree-vect-data-refs.c: Include internal-fn.h.
21815         (vect_gather_scatter_fn_p): New function.
21816         (vect_describe_gather_scatter_call): Likewise.
21817         (vect_check_gather_scatter): Try using internal functions for
21818         gather loads.  Recognize existing calls to a gather load function.
21819         (vect_analyze_data_refs): Consider using gather loads if
21820         supports_vec_gather_load_p.
21821         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
21822         (vect_get_gather_scatter_offset_type): Likewise.
21823         (vect_convert_mask_for_vectype): Likewise.
21824         (vect_add_conversion_to_patterm): Likewise.
21825         (vect_try_gather_scatter_pattern): Likewise.
21826         (vect_recog_gather_scatter_pattern): New pattern recognizer.
21827         (vect_vect_recog_func_ptrs): Add it.
21828         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
21829         internal_fn_mask_index and internal_gather_scatter_fn_p.
21830         (check_load_store_masking): Take the gather_scatter_info as an
21831         argument and handle gather loads.
21832         (vect_get_gather_scatter_ops): New function.
21833         (vectorizable_call): Check internal_load_fn_p.
21834         (vectorizable_load): Likewise.  Handle gather load internal
21835         functions.
21836         (vectorizable_store): Update call to check_load_store_masking.
21837         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
21838         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
21839         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
21840         (aarch64_gather_scale_operand_d): New predicates.
21841         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
21842         (mask_gather_load<mode>): New insns.
21844 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21845             Alan Hayward  <alan.hayward@arm.com>
21846             David Sherwood  <david.sherwood@arm.com>
21848         * optabs.def (fold_left_plus_optab): New optab.
21849         * doc/md.texi (fold_left_plus_@var{m}): Document.
21850         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
21851         * internal-fn.c (fold_left_direct): Define.
21852         (expand_fold_left_optab_fn): Likewise.
21853         (direct_fold_left_optab_supported_p): Likewise.
21854         * fold-const-call.c (fold_const_fold_left): New function.
21855         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
21856         * tree-parloops.c (valid_reduction_p): New function.
21857         (gather_scalar_reductions): Use it.
21858         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
21859         (vect_finish_replace_stmt): Declare.
21860         * tree-vect-loop.c (fold_left_reduction_fn): New function.
21861         (needs_fold_left_reduction_p): New function, split out from...
21862         (vect_is_simple_reduction): ...here.  Accept reductions that
21863         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
21864         (vect_force_simple_reduction): Also store the reduction type in
21865         the assignment's STMT_VINFO_REDUC_TYPE.
21866         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
21867         (merge_with_identity): New function.
21868         (vect_expand_fold_left): Likewise.
21869         (vectorize_fold_left_reduction): Likewise.
21870         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
21871         scalar phi in place for it.  Check for target support and reject
21872         cases that would reassociate the operation.  Defer the transform
21873         phase to vectorize_fold_left_reduction.
21874         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
21875         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
21876         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
21878 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21880         * tree-if-conv.c (predicate_mem_writes): Remove redundant
21881         call to ifc_temp_var.
21883 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21884             Alan Hayward  <alan.hayward@arm.com>
21885             David Sherwood  <david.sherwood@arm.com>
21887         * target.def (legitimize_address_displacement): Take the original
21888         offset as a poly_int.
21889         * targhooks.h (default_legitimize_address_displacement): Update
21890         accordingly.
21891         * targhooks.c (default_legitimize_address_displacement): Likewise.
21892         * doc/tm.texi: Regenerate.
21893         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
21894         as an argument, moving assert of ad->disp == ad->disp_term to...
21895         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
21896         Try calling targetm.legitimize_address_displacement before expanding
21897         the address rather than afterwards, and adjust for the new interface.
21898         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
21899         Match the new hook interface.  Handle SVE addresses.
21900         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
21901         new hook interface.
21903 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21905         * Makefile.in (OBJS): Add early-remat.o.
21906         * target.def (select_early_remat_modes): New hook.
21907         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
21908         * doc/tm.texi: Regenerate.
21909         * targhooks.h (default_select_early_remat_modes): Declare.
21910         * targhooks.c (default_select_early_remat_modes): New function.
21911         * timevar.def (TV_EARLY_REMAT): New timevar.
21912         * passes.def (pass_early_remat): New pass.
21913         * tree-pass.h (make_pass_early_remat): Declare.
21914         * early-remat.c: New file.
21915         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
21916         function.
21917         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
21919 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21920             Alan Hayward  <alan.hayward@arm.com>
21921             David Sherwood  <david.sherwood@arm.com>
21923         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
21924         vfm1 with a bound_epilog parameter.
21925         (vect_do_peeling): Update calls accordingly, and move the prologue
21926         call earlier in the function.  Treat the base bound_epilog as 0 for
21927         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
21928         this base when peeling for gaps.
21929         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
21930         with fully-masked loops.
21931         (vect_estimate_min_profitable_iters): Handle the single peeled
21932         iteration in that case.
21934 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21935             Alan Hayward  <alan.hayward@arm.com>
21936             David Sherwood  <david.sherwood@arm.com>
21938         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
21939         single-element interleaving even if the size is not a power of 2.
21940         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
21941         accesses for single-element interleaving if the group size is
21942         not a power of 2.
21944 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21945             Alan Hayward  <alan.hayward@arm.com>
21946             David Sherwood  <david.sherwood@arm.com>
21948         * doc/md.texi (fold_extract_last_@var{m}): Document.
21949         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
21950         * optabs.def (fold_extract_last_optab): New optab.
21951         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
21952         * internal-fn.c (fold_extract_direct): New macro.
21953         (expand_fold_extract_optab_fn): Likewise.
21954         (direct_fold_extract_optab_supported_p): Likewise.
21955         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
21956         * tree-vect-loop.c (vect_model_reduction_cost): Handle
21957         EXTRACT_LAST_REDUCTION.
21958         (get_initial_def_for_reduction): Do not create an initial vector
21959         for EXTRACT_LAST_REDUCTION reductions.
21960         (vectorizable_reduction): Leave the scalar phi in place for
21961         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
21962         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
21963         epilogue code for EXTRACT_LAST_REDUCTION and defer the
21964         transform phase to vectorizable_condition.
21965         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
21966         split out from...
21967         (vect_finish_stmt_generation): ...here.
21968         (vect_finish_replace_stmt): New function.
21969         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
21970         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
21971         pattern.
21972         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
21974 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21975             Alan Hayward  <alan.hayward@arm.com>
21976             David Sherwood  <david.sherwood@arm.com>
21978         * doc/md.texi (extract_last_@var{m}): Document.
21979         * optabs.def (extract_last_optab): New optab.
21980         * internal-fn.def (EXTRACT_LAST): New internal function.
21981         * internal-fn.c (cond_unary_direct): New macro.
21982         (expand_cond_unary_optab_fn): Likewise.
21983         (direct_cond_unary_optab_supported_p): Likewise.
21984         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
21985         loops using EXTRACT_LAST.
21986         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
21987         (extract_last_<mode>): ...this optab.
21988         (vec_extract<mode><Vel>): Update accordingly.
21990 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
21991             Alan Hayward  <alan.hayward@arm.com>
21992             David Sherwood  <david.sherwood@arm.com>
21994         * target.def (empty_mask_is_expensive): New hook.
21995         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
21996         * doc/tm.texi: Regenerate.
21997         * targhooks.h (default_empty_mask_is_expensive): Declare.
21998         * targhooks.c (default_empty_mask_is_expensive): New function.
21999         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
22000         if the target says that empty masks are expensive.
22001         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
22002         New function.
22003         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
22005 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22006             Alan Hayward  <alan.hayward@arm.com>
22007             David Sherwood  <david.sherwood@arm.com>
22009         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
22010         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
22011         (vect_use_loop_mask_for_alignment_p): New function.
22012         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
22013         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
22014         niters_skip argument.  Make sure that the first niters_skip elements
22015         of the first iteration are inactive.
22016         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
22017         Update call to vect_set_loop_masks_directly.
22018         (get_misalign_in_elems): New function, split out from...
22019         (vect_gen_prolog_loop_niters): ...here.
22020         (vect_update_init_of_dr): Take a code argument that specifies whether
22021         the adjustment should be added or subtracted.
22022         (vect_update_init_of_drs): Likewise.
22023         (vect_prepare_for_masked_peels): New function.
22024         (vect_do_peeling): Skip prologue peeling if we're using a mask
22025         instead.  Update call to vect_update_inits_of_drs.
22026         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22027         mask_skip_niters.
22028         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
22029         alignment.  Do not include the number of peeled iterations in
22030         the minimum threshold in that case.
22031         (vectorizable_induction): Adjust the start value down by
22032         LOOP_VINFO_MASK_SKIP_NITERS iterations.
22033         (vect_transform_loop): Call vect_prepare_for_masked_peels.
22034         Take the number of skipped iterations into account when calculating
22035         the loop bounds.
22036         * tree-vect-stmts.c (vect_gen_while_not): New function.
22038 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22039             Alan Hayward  <alan.hayward@arm.com>
22040             David Sherwood  <david.sherwood@arm.com>
22042         * doc/sourcebuild.texi (vect_fully_masked): Document.
22043         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
22044         default value to 0.
22045         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
22046         split out from...
22047         (vect_analyze_loop_2): ...here. Don't check the vectorization
22048         factor against the number of loop iterations if the loop is
22049         fully-masked.
22051 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22052             Alan Hayward  <alan.hayward@arm.com>
22053             David Sherwood  <david.sherwood@arm.com>
22055         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
22056         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
22057         (dump_groups): Update accordingly.
22058         (iv_use::mem_type): New member variable.
22059         (address_p): New function.
22060         (record_use): Add a mem_type argument and initialize the new
22061         mem_type field.
22062         (record_group_use): Add a mem_type argument.  Use address_p.
22063         Remove obsolete null checks of base_object.  Update call to record_use.
22064         (find_interesting_uses_op): Update call to record_group_use.
22065         (find_interesting_uses_cond): Likewise.
22066         (find_interesting_uses_address): Likewise.
22067         (get_mem_type_for_internal_fn): New function.
22068         (find_address_like_use): Likewise.
22069         (find_interesting_uses_stmt): Try find_address_like_use before
22070         calling find_interesting_uses_op.
22071         (addr_offset_valid_p): Use the iv mem_type field as the type
22072         of the addressed memory.
22073         (add_autoinc_candidates): Likewise.
22074         (get_address_cost): Likewise.
22075         (split_small_address_groups_p): Use address_p.
22076         (split_address_groups): Likewise.
22077         (add_iv_candidate_for_use): Likewise.
22078         (autoinc_possible_for_pair): Likewise.
22079         (rewrite_groups): Likewise.
22080         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
22081         (determine_group_iv_cost): Update after split of USE_ADDRESS.
22082         (get_alias_ptr_type_for_ptr_address): New function.
22083         (rewrite_use_address): Rewrite address uses in calls that were
22084         identified by find_address_like_use.
22086 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22087             Alan Hayward  <alan.hayward@arm.com>
22088             David Sherwood  <david.sherwood@arm.com>
22090         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
22091         TARGET_MEM_REFs.
22092         * gimple-expr.h (is_gimple_addressable: Likewise.
22093         * gimple-expr.c (is_gimple_address): Likewise.
22094         * internal-fn.c (expand_call_mem_ref): New function.
22095         (expand_mask_load_optab_fn): Use it.
22096         (expand_mask_store_optab_fn): Likewise.
22098 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22099             Alan Hayward  <alan.hayward@arm.com>
22100             David Sherwood  <david.sherwood@arm.com>
22102         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
22103         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
22104         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
22105         (cond_umax@var{mode}): Document.
22106         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
22107         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
22108         (cond_umin_optab, cond_umax_optab): New optabs.
22109         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
22110         (COND_IOR, COND_XOR): New internal functions.
22111         * internal-fn.h (get_conditional_internal_fn): Declare.
22112         * internal-fn.c (cond_binary_direct): New macro.
22113         (expand_cond_binary_optab_fn): Likewise.
22114         (direct_cond_binary_optab_supported_p): Likewise.
22115         (get_conditional_internal_fn): New function.
22116         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
22117         Cope with reduction statements that are vectorized as calls rather
22118         than assignments.
22119         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
22120         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
22121         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
22122         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
22123         (UNSPEC_COND_EOR): New unspecs.
22124         (optab): Add mappings for them.
22125         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
22126         (sve_int_op, sve_fp_op): New int attributes.
22128 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22129             Alan Hayward  <alan.hayward@arm.com>
22130             David Sherwood  <david.sherwood@arm.com>
22132         * optabs.def (while_ult_optab): New optab.
22133         * doc/md.texi (while_ult@var{m}@var{n}): Document.
22134         * internal-fn.def (WHILE_ULT): New internal function.
22135         * internal-fn.h (direct_internal_fn_supported_p): New override
22136         that takes two types as argument.
22137         * internal-fn.c (while_direct): New macro.
22138         (expand_while_optab_fn): New function.
22139         (convert_optab_supported_p): Likewise.
22140         (direct_while_optab_supported_p): New macro.
22141         * wide-int.h (wi::udiv_ceil): New function.
22142         * tree-vectorizer.h (rgroup_masks): New structure.
22143         (vec_loop_masks): New typedef.
22144         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
22145         and fully_masked_p.
22146         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
22147         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
22148         (vect_max_vf): New function.
22149         (slpeel_make_loop_iterate_ntimes): Delete.
22150         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
22151         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
22152         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
22153         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
22154         internal-fn.h, stor-layout.h and optabs-query.h.
22155         (vect_set_loop_mask): New function.
22156         (add_preheader_seq): Likewise.
22157         (add_header_seq): Likewise.
22158         (interleave_supported_p): Likewise.
22159         (vect_maybe_permute_loop_masks): Likewise.
22160         (vect_set_loop_masks_directly): Likewise.
22161         (vect_set_loop_condition_masked): Likewise.
22162         (vect_set_loop_condition_unmasked): New function, split out from
22163         slpeel_make_loop_iterate_ntimes.
22164         (slpeel_make_loop_iterate_ntimes): Rename to..
22165         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
22166         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
22167         (vect_do_peeling): Update call accordingly.
22168         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
22169         loops.
22170         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
22171         mask_compare_type, can_fully_mask_p and fully_masked_p.
22172         (release_vec_loop_masks): New function.
22173         (_loop_vec_info): Use it to free the loop masks.
22174         (can_produce_all_loop_masks_p): New function.
22175         (vect_get_max_nscalars_per_iter): Likewise.
22176         (vect_verify_full_masking): Likewise.
22177         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
22178         retries, and free the mask rgroups before retrying.  Check loop-wide
22179         reasons for disallowing fully-masked loops.  Make the final decision
22180         about whether use a fully-masked loop or not.
22181         (vect_estimate_min_profitable_iters): Do not assume that peeling
22182         for the number of iterations will be needed for fully-masked loops.
22183         (vectorizable_reduction): Disable fully-masked loops.
22184         (vectorizable_live_operation): Likewise.
22185         (vect_halve_mask_nunits): New function.
22186         (vect_double_mask_nunits): Likewise.
22187         (vect_record_loop_mask): Likewise.
22188         (vect_get_loop_mask): Likewise.
22189         (vect_transform_loop): Handle the case in which the final loop
22190         iteration might handle a partial vector.  Call vect_set_loop_condition
22191         instead of slpeel_make_loop_iterate_ntimes.
22192         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
22193         (check_load_store_masking): New function.
22194         (prepare_load_store_mask): Likewise.
22195         (vectorizable_store): Handle fully-masked loops.
22196         (vectorizable_load): Likewise.
22197         (supportable_widening_operation): Use vect_halve_mask_nunits for
22198         booleans.
22199         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
22200         (vect_gen_while): New function.
22201         * config/aarch64/aarch64.md (umax<mode>3): New expander.
22202         (aarch64_uqdec<mode>): New insn.
22204 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22205             Alan Hayward  <alan.hayward@arm.com>
22206             David Sherwood  <david.sherwood@arm.com>
22208         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
22209         (reduc_xor_scal_optab): New optabs.
22210         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
22211         (reduc_xor_scal_@var{m}): Document.
22212         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
22213         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
22214         internal functions.
22215         * fold-const-call.c (fold_const_call): Handle them.
22216         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
22217         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
22218         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
22219         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
22220         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
22221         (UNSPEC_XORV): New unspecs.
22222         (optab): Add entries for them.
22223         (BITWISEV): New int iterator.
22224         (bit_reduc_op): New int attributes.
22226 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22227             Alan Hayward  <alan.hayward@arm.com>
22228             David Sherwood  <david.sherwood@arm.com>
22230         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
22231         * internal-fn.def (VEC_SHL_INSERT): New internal function.
22232         * optabs.def (vec_shl_insert_optab): New optab.
22233         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
22234         (duplicate_and_interleave): Likewise.
22235         * tree-vect-loop.c: Include internal-fn.h.
22236         (neutral_op_for_slp_reduction): New function, split out from
22237         get_initial_defs_for_reduction.
22238         (get_initial_def_for_reduction): Handle option 2 for variable-length
22239         vectors by loading the neutral value into a vector and then shifting
22240         the initial value into element 0.
22241         (get_initial_defs_for_reduction): Replace the code argument with
22242         the neutral value calculated by neutral_op_for_slp_reduction.
22243         Use gimple_build_vector for constant-length vectors.
22244         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
22245         but the first group_size elements have a neutral value.
22246         Use duplicate_and_interleave otherwise.
22247         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
22248         Update call to get_initial_defs_for_reduction.  Handle SLP
22249         reductions for variable-length vectors by creating one vector
22250         result for each scalar result, with the elements associated
22251         with other scalar results stubbed out with the neutral value.
22252         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
22253         Require IFN_VEC_SHL_INSERT for double reductions on
22254         variable-length vectors, or SLP reductions that have
22255         a neutral value.  Require can_duplicate_and_interleave_p
22256         support for variable-length unchained SLP reductions if there
22257         is no neutral value, such as for MIN/MAX reductions.  Also require
22258         the number of vector elements to be a multiple of the number of
22259         SLP statements when doing variable-length unchained SLP reductions.
22260         Update call to vect_create_epilog_for_reduction.
22261         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
22262         and remove initial values.
22263         (duplicate_and_interleave): Make public.
22264         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
22265         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
22267 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22268             Alan Hayward  <alan.hayward@arm.com>
22269             David Sherwood  <david.sherwood@arm.com>
22271         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
22272         (can_duplicate_and_interleave_p): New function.
22273         (vect_get_and_check_slp_defs): Take the vector of statements
22274         rather than just the current one.  Remove excess parentheses.
22275         Restriction rejectinon of vect_constant_def and vect_external_def
22276         for variable-length vectors to boolean types, or types for which
22277         can_duplicate_and_interleave_p is false.
22278         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
22279         (duplicate_and_interleave): New function.
22280         (vect_get_constant_vectors): Use gimple_build_vector for
22281         constant-length vectors and suitable variable-length constant
22282         vectors.  Use duplicate_and_interleave for other variable-length
22283         vectors.  Don't defer the update when inserting new statements.
22285 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22286             Alan Hayward  <alan.hayward@arm.com>
22287             David Sherwood  <david.sherwood@arm.com>
22289         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
22290         min_profitable_iters doesn't go negative.
22292 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22293             Alan Hayward  <alan.hayward@arm.com>
22294             David Sherwood  <david.sherwood@arm.com>
22296         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
22297         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
22298         * optabs.def (vec_mask_load_lanes_optab): New optab.
22299         (vec_mask_store_lanes_optab): Likewise.
22300         * internal-fn.def (MASK_LOAD_LANES): New internal function.
22301         (MASK_STORE_LANES): Likewise.
22302         * internal-fn.c (mask_load_lanes_direct): New macro.
22303         (mask_store_lanes_direct): Likewise.
22304         (expand_mask_load_optab_fn): Handle masked operations.
22305         (expand_mask_load_lanes_optab_fn): New macro.
22306         (expand_mask_store_optab_fn): Handle masked operations.
22307         (expand_mask_store_lanes_optab_fn): New macro.
22308         (direct_mask_load_lanes_optab_supported_p): Likewise.
22309         (direct_mask_store_lanes_optab_supported_p): Likewise.
22310         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
22311         parameter.
22312         (vect_load_lanes_supported): Likewise.
22313         * tree-vect-data-refs.c (strip_conversion): New function.
22314         (can_group_stmts_p): Likewise.
22315         (vect_analyze_data_ref_accesses): Use it instead of checking
22316         for a pair of assignments.
22317         (vect_store_lanes_supported): Take a masked_p parameter.
22318         (vect_load_lanes_supported): Likewise.
22319         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
22320         vect_store_lanes_supported and vect_load_lanes_supported.
22321         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
22322         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
22323         parameter.  Don't allow gaps for masked accesses.
22324         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
22325         and vect_load_lanes_supported.
22326         (get_load_store_type): Take a masked_p parameter and update
22327         call to get_group_load_store_type.
22328         (vectorizable_store): Update call to get_load_store_type.
22329         Handle IFN_MASK_STORE_LANES.
22330         (vectorizable_load): Update call to get_load_store_type.
22331         Handle IFN_MASK_LOAD_LANES.
22333 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22334             Alan Hayward  <alan.hayward@arm.com>
22335             David Sherwood  <david.sherwood@arm.com>
22337         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
22338         modes for SVE.
22339         * config/aarch64/aarch64-protos.h
22340         (aarch64_sve_struct_memory_operand_p): Declare.
22341         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
22342         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
22343         (VPRED, vpred): Handle SVE structure modes.
22344         * config/aarch64/constraints.md (Utx): New constraint.
22345         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
22346         (aarch64_sve_struct_nonimmediate_operand): New predicates.
22347         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
22348         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
22349         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
22350         structure modes.  Split into pieces after RA.
22351         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
22352         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
22353         New patterns.
22354         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
22355         SVE structure modes.
22356         (aarch64_classify_address): Likewise.
22357         (sizetochar): Move earlier in file.
22358         (aarch64_print_operand): Handle SVE register lists.
22359         (aarch64_array_mode): New function.
22360         (aarch64_sve_struct_memory_operand_p): Likewise.
22361         (TARGET_ARRAY_MODE): Redefine.
22363 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22364             Alan Hayward  <alan.hayward@arm.com>
22365             David Sherwood  <david.sherwood@arm.com>
22367         * target.def (array_mode): New target hook.
22368         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
22369         * doc/tm.texi: Regenerate.
22370         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
22371         * hooks.c (hook_optmode_mode_uhwi_none): New function.
22372         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
22373         targetm.array_mode.
22374         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
22375         type sizes.
22377 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22378             Alan Hayward  <alan.hayward@arm.com>
22379             David Sherwood  <david.sherwood@arm.com>
22381         * fold-const.c (fold_binary_loc): Check the argument types
22382         rather than the result type when testing for a vector operation.
22384 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22386         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
22387         * doc/tm.texi: Regenerate.
22389 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22390             Alan Hayward  <alan.hayward@arm.com>
22391             David Sherwood  <david.sherwood@arm.com>
22393         * doc/invoke.texi (-msve-vector-bits=): Document new option.
22394         (sve): Document new AArch64 extension.
22395         * doc/md.texi (w): Extend the description of the AArch64
22396         constraint to include SVE vectors.
22397         (Upl, Upa): Document new AArch64 predicate constraints.
22398         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
22399         enum.
22400         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
22401         (msve-vector-bits=): New option.
22402         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
22403         SVE when these are disabled.
22404         (sve): New extension.
22405         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
22406         modes.  Adjust their number of units based on aarch64_sve_vg.
22407         (MAX_BITSIZE_MODE_ANY_MODE): Define.
22408         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
22409         aarch64_addr_query_type.
22410         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
22411         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
22412         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
22413         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
22414         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
22415         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
22416         (aarch64_simd_imm_zero_p): Delete.
22417         (aarch64_check_zero_based_sve_index_immediate): Declare.
22418         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22419         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22420         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22421         (aarch64_sve_float_mul_immediate_p): Likewise.
22422         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22423         rather than an rtx.
22424         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
22425         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
22426         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
22427         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
22428         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
22429         (aarch64_regmode_natural_size): Likewise.
22430         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
22431         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
22432         left one place.
22433         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
22434         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
22435         for VG and the SVE predicate registers.
22436         (V_ALIASES): Add a "z"-prefixed alias.
22437         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
22438         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
22439         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
22440         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
22441         (REG_CLASS_NAMES): Add entries for them.
22442         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
22443         and the predicate registers.
22444         (aarch64_sve_vg): Declare.
22445         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
22446         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
22447         (REGMODE_NATURAL_SIZE): Define.
22448         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
22449         SVE macros.
22450         * config/aarch64/aarch64.c: Include cfgrtl.h.
22451         (simd_immediate_info): Add a constructor for series vectors,
22452         and an associated step field.
22453         (aarch64_sve_vg): New variable.
22454         (aarch64_dbx_register_number): Handle VG and the predicate registers.
22455         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
22456         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
22457         (VEC_ANY_DATA, VEC_STRUCT): New constants.
22458         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
22459         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
22460         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
22461         (aarch64_get_mask_mode): New functions.
22462         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
22463         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
22464         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
22465         predicate modes and predicate registers.  Explicitly restrict
22466         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
22467         to store a vector mode if it is recognized by
22468         aarch64_classify_vector_mode.
22469         (aarch64_regmode_natural_size): New function.
22470         (aarch64_hard_regno_caller_save_mode): Return the original mode
22471         for predicates.
22472         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
22473         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
22474         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
22475         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
22476         functions.
22477         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
22478         does not overlap dest if the function is frame-related.  Handle
22479         SVE constants.
22480         (aarch64_split_add_offset): New function.
22481         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
22482         them aarch64_add_offset.
22483         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
22484         and update call to aarch64_sub_sp.
22485         (aarch64_add_cfa_expression): New function.
22486         (aarch64_expand_prologue): Pass extra temporary registers to the
22487         functions above.  Handle the case in which we need to emit new
22488         DW_CFA_expressions for registers that were originally saved
22489         relative to the stack pointer, but now have to be expressed
22490         relative to the frame pointer.
22491         (aarch64_output_mi_thunk): Pass extra temporary registers to the
22492         functions above.
22493         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
22494         IP0 and IP1 values for SVE frames.
22495         (aarch64_expand_vec_series): New function.
22496         (aarch64_expand_sve_widened_duplicate): Likewise.
22497         (aarch64_expand_sve_const_vector): Likewise.
22498         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
22499         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
22500         into the register, rather than emitting a SET directly.
22501         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
22502         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
22503         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
22504         (offset_9bit_signed_scaled_p): New functions.
22505         (aarch64_replicate_bitmask_imm): New function.
22506         (aarch64_bitmask_imm): Use it.
22507         (aarch64_cannot_force_const_mem): Reject expressions involving
22508         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
22509         (aarch64_classify_index): Handle SVE indices, by requiring
22510         a plain register index with a scale that matches the element size.
22511         (aarch64_classify_address): Handle SVE addresses.  Assert that
22512         the mode of the address is VOIDmode or an integer mode.
22513         Update call to aarch64_classify_symbol.
22514         (aarch64_classify_symbolic_expression): Update call to
22515         aarch64_classify_symbol.
22516         (aarch64_const_vec_all_in_range_p): New function.
22517         (aarch64_print_vector_float_operand): Likewise.
22518         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
22519         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
22520         and the FP immediates 1.0 and 0.5.
22521         (aarch64_print_address_internal): Handle SVE addresses.
22522         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
22523         (aarch64_regno_regclass): Handle predicate registers.
22524         (aarch64_secondary_reload): Handle big-endian reloads of SVE
22525         data modes.
22526         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
22527         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
22528         (aarch64_convert_sve_vector_bits): New function.
22529         (aarch64_override_options): Use it to handle -msve-vector-bits=.
22530         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
22531         rather than an rtx.
22532         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
22533         Handle SVE vector and predicate modes.  Accept VL-based constants
22534         that need only one temporary register, and VL offsets that require
22535         no temporary registers.
22536         (aarch64_conditional_register_usage): Mark the predicate registers
22537         as fixed if SVE isn't available.
22538         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
22539         Return true for SVE vector and predicate modes.
22540         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
22541         rather than an unsigned int.  Handle SVE modes.
22542         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
22543         SVE modes.
22544         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
22545         if SVE is enabled.
22546         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
22547         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
22548         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
22549         (aarch64_sve_float_mul_immediate_p): New functions.
22550         (aarch64_sve_valid_immediate): New function.
22551         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
22552         Explicitly reject structure modes.  Check for INDEX constants.
22553         Handle PTRUE and PFALSE constants.
22554         (aarch64_check_zero_based_sve_index_immediate): New function.
22555         (aarch64_simd_imm_zero_p): Delete.
22556         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
22557         vector modes.  Accept constants in the range of CNT[BHWD].
22558         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
22559         ask for an Advanced SIMD mode.
22560         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
22561         (aarch64_simd_vector_alignment): Handle SVE predicates.
22562         (aarch64_vectorize_preferred_vector_alignment): New function.
22563         (aarch64_simd_vector_alignment_reachable): Use it instead of
22564         the vector size.
22565         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22566         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
22567         functions.
22568         (MAX_VECT_LEN): Delete.
22569         (expand_vec_perm_d): Add a vec_flags field.
22570         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
22571         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
22572         (aarch64_evpc_ext): Don't apply a big-endian lane correction
22573         for SVE modes.
22574         (aarch64_evpc_rev): Rename to...
22575         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
22576         (aarch64_evpc_rev_global): New function.
22577         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
22578         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
22579         MAX_VECT_LEN.
22580         (aarch64_evpc_sve_tbl): New function.
22581         (aarch64_expand_vec_perm_const_1): Update after rename of
22582         aarch64_evpc_rev.  Handle SVE permutes too, trying
22583         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
22584         than aarch64_evpc_tbl.
22585         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
22586         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
22587         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
22588         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
22589         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
22590         (aarch64_expand_sve_vcond): New functions.
22591         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
22592         of aarch64_vector_mode_p.
22593         (aarch64_dwarf_poly_indeterminate_value): New function.
22594         (aarch64_compute_pressure_classes): Likewise.
22595         (aarch64_can_change_mode_class): Likewise.
22596         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
22597         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
22598         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
22599         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
22600         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
22601         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
22602         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
22603         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
22604         constraints.
22605         (Dn, Dl, Dr): Accept const as well as const_vector.
22606         (Dz): Likewise.  Compare against CONST0_RTX.
22607         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
22608         of "vector" where appropriate.
22609         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
22610         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
22611         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
22612         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
22613         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
22614         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
22615         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
22616         (v_int_equiv): Extend to SVE modes.
22617         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
22618         mode attributes.
22619         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
22620         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
22621         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
22622         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
22623         (SVE_COND_FP_CMP): New int iterators.
22624         (perm_hilo): Handle the new unpack unspecs.
22625         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
22626         attributes.
22627         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
22628         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
22629         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
22630         (aarch64_equality_operator, aarch64_constant_vector_operand)
22631         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
22632         (aarch64_sve_nonimmediate_operand): Likewise.
22633         (aarch64_sve_general_operand): Likewise.
22634         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
22635         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
22636         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
22637         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
22638         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
22639         (aarch64_sve_float_arith_immediate): Likewise.
22640         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
22641         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
22642         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
22643         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
22644         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
22645         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
22646         (aarch64_sve_float_arith_operand): Likewise.
22647         (aarch64_sve_float_arith_with_sub_operand): Likewise.
22648         (aarch64_sve_float_mul_operand): Likewise.
22649         (aarch64_sve_vec_perm_operand): Likewise.
22650         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
22651         (aarch64_mov_operand): Accept const_poly_int and const_vector.
22652         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
22653         as well as const_vector.
22654         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
22655         in file.  Use CONST0_RTX and CONSTM1_RTX.
22656         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
22657         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
22658         Use aarch64_simd_imm_zero.
22659         * config/aarch64/aarch64-sve.md: New file.
22660         * config/aarch64/aarch64.md: Include it.
22661         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
22662         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
22663         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
22664         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
22665         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
22666         (sve): New attribute.
22667         (enabled): Disable instructions with the sve attribute unless
22668         TARGET_SVE.
22669         (movqi, movhi): Pass CONST_POLY_INT operaneds through
22670         aarch64_expand_mov_immediate.
22671         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
22672         CNT[BHSD] immediates.
22673         (movti): Split CONST_POLY_INT moves into two halves.
22674         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
22675         Split additions that need a temporary here if the destination
22676         is the stack pointer.
22677         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
22678         (*add<mode>3_poly_1): New instruction.
22679         (set_clobber_cc): New expander.
22681 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22683         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
22684         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
22685         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
22686         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
22687         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
22688         Change innermode from fixed_mode_size to machine_mode.
22689         (simplify_subreg): Update call accordingly.  Handle a constant-sized
22690         subreg of a variable-length CONST_VECTOR.
22692 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22693             Alan Hayward  <alan.hayward@arm.com>
22694             David Sherwood  <david.sherwood@arm.com>
22696         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
22697         (add_offset_to_base): New function, split out from...
22698         (create_mem_ref): ...here.  When handling a scale other than 1,
22699         check first whether the address is valid without the offset.
22700         Add it into the base if so, leaving the index and scale as-is.
22702 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
22704         PR c++/83778
22705         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
22706         fold_for_warn before checking if arg2 is INTEGER_CST.
22708 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
22710         * config/rs6000/predicates.md (load_multiple_operation): Delete.
22711         (store_multiple_operation): Delete.
22712         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
22713         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
22714         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
22715         guarded by TARGET_STRING.
22716         (rs6000_output_load_multiple): Delete.
22717         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
22718         OPTION_MASK_STRING / TARGET_STRING handling.
22719         (print_operand) <'N', 'O'>: Add comment that these are unused now.
22720         (const rs6000_opt_masks) <"string">: Change mask to 0.
22721         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
22722         (MASK_STRING): Delete.
22723         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
22724         parts.  Simplify.
22725         (load_multiple): Delete.
22726         (*ldmsi8): Delete.
22727         (*ldmsi7): Delete.
22728         (*ldmsi6): Delete.
22729         (*ldmsi5): Delete.
22730         (*ldmsi4): Delete.
22731         (*ldmsi3): Delete.
22732         (store_multiple): Delete.
22733         (*stmsi8): Delete.
22734         (*stmsi7): Delete.
22735         (*stmsi6): Delete.
22736         (*stmsi5): Delete.
22737         (*stmsi4): Delete.
22738         (*stmsi3): Delete.
22739         (movmemsi_8reg): Delete.
22740         (corresponding unnamed define_insn): Delete.
22741         (movmemsi_6reg): Delete.
22742         (corresponding unnamed define_insn): Delete.
22743         (movmemsi_4reg): Delete.
22744         (corresponding unnamed define_insn): Delete.
22745         (movmemsi_2reg): Delete.
22746         (corresponding unnamed define_insn): Delete.
22747         (movmemsi_1reg): Delete.
22748         (corresponding unnamed define_insn): Delete.
22749         * config/rs6000/rs6000.opt (mno-string): New.
22750         (mstring): Replace by deprecation warning stub.
22751         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
22753 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
22755         * regrename.c (regrename_do_replace): If replacing the same
22756         reg multiple times, try to reuse last created gen_raw_REG.
22758         PR debug/81155
22759         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
22760         main to workaround a bug in GDB.
22762 2018-01-12  Tom de Vries  <tom@codesourcery.com>
22764         PR target/83737
22765         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
22767 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
22769         PR rtl-optimization/80481
22770         * ira-color.c (get_cap_member): New function.
22771         (allocnos_conflict_by_live_ranges_p): Use it.
22772         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
22773         (setup_slot_coalesced_allocno_live_ranges): Ditto.
22775 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
22777         PR target/83628
22778         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
22779         (*saddl_se_1): Ditto.
22780         (*ssubsi_1): Ditto.
22781         (*ssubl_se_1): Ditto.
22783 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
22785         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
22786         rather than wi::to_widest for DR_INITs.
22787         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
22788         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
22789         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
22790         INTEGER_CSTs.
22791         (vect_analyze_group_access_1): Note that here.
22793 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
22795         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
22796         polynomial type sizes.
22798 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
22800         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
22801         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
22802         (gimple_add_tmp_var): Likewise.
22804 2018-01-12  Martin Liska  <mliska@suse.cz>
22806         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
22807         (gimple_alloc_sizes): Likewise.
22808         (dump_gimple_statistics): Use PRIu64 in printf format.
22809         * gimple.h: Change uint64_t to int.
22811 2018-01-12  Martin Liska  <mliska@suse.cz>
22813         * tree-core.h: Use uint64_t instead of int.
22814         * tree.c (tree_node_counts): Likewise.
22815         (tree_node_sizes): Likewise.
22816         (dump_tree_statistics): Use PRIu64 in printf format.
22818 2018-01-12  Martin Liska  <mliska@suse.cz>
22820         * Makefile.in: As qsort_chk is implemented in vec.c, add
22821         vec.o to linkage of gencfn-macros.
22822         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
22823         passing the info to record_node_allocation_statistics.
22824         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
22825         and pass the info.
22826         * ggc-common.c (struct ggc_usage): Add operator== and use
22827         it in operator< and compare function.
22828         * mem-stats.h (struct mem_usage): Likewise.
22829         * vec.c (struct vec_usage): Remove operator< and compare
22830         function. Can be simply inherited.
22832 2018-01-12  Martin Jambor  <mjambor@suse.cz>
22834         PR target/81616
22835         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
22836         * tree-ssa-math-opts.c: Include domwalk.h.
22837         (convert_mult_to_fma_1): New function.
22838         (fma_transformation_info): New type.
22839         (fma_deferring_state): Likewise.
22840         (cancel_fma_deferring): New function.
22841         (result_of_phi): Likewise.
22842         (last_fma_candidate_feeds_initial_phi): Likewise.
22843         (convert_mult_to_fma): Added deferring logic, split actual
22844         transformation to convert_mult_to_fma_1.
22845         (math_opts_dom_walker): New type.
22846         (math_opts_dom_walker::after_dom_children): New method, body moved
22847         here from pass_optimize_widening_mul::execute, added deferring logic
22848         bits.
22849         (pass_optimize_widening_mul::execute): Moved most of code to
22850         math_opts_dom_walker::after_dom_children.
22851         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
22852         * config/i386/i386.c (ix86_option_override_internal): Added
22853         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
22855 2018-01-12  Richard Biener  <rguenther@suse.de>
22857         PR debug/83157
22858         * dwarf2out.c (gen_variable_die): Do not reset old_die for
22859         inline instance vars.
22861 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
22863         PR target/81819
22864         * config/rx/rx.c (rx_is_restricted_memory_address):
22865         Handle SUBREG case.
22867 2018-01-12  Richard Biener  <rguenther@suse.de>
22869         PR tree-optimization/80846
22870         * target.def (split_reduction): New target hook.
22871         * targhooks.c (default_split_reduction): New function.
22872         * targhooks.h (default_split_reduction): Declare.
22873         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
22874         target requests first reduce vectors by combining low and high
22875         parts.
22876         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
22877         (get_vectype_for_scalar_type_and_size): Export.
22878         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
22879         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
22880         * doc/tm.texi: Regenerate.
22881         * config/i386/i386.c (ix86_split_reduction): Implement
22882         TARGET_VECTORIZE_SPLIT_REDUCTION.
22884 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
22886         PR target/83368
22887         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
22888         in PIC mode except for TARGET_VXWORKS_RTP.
22889         * config/sparc/sparc.c: Include cfgrtl.h.
22890         (TARGET_INIT_PIC_REG): Define.
22891         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
22892         (sparc_pic_register_p): New predicate.
22893         (sparc_legitimate_address_p): Use it.
22894         (sparc_legitimize_pic_address): Likewise.
22895         (sparc_delegitimize_address): Likewise.
22896         (sparc_mode_dependent_address_p): Likewise.
22897         (gen_load_pcrel_sym): Remove 4th parameter.
22898         (load_got_register): Adjust call to above.  Remove obsolete stuff.
22899         (sparc_expand_prologue): Do not call load_got_register here.
22900         (sparc_flat_expand_prologue): Likewise.
22901         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
22902         (sparc_use_pseudo_pic_reg): New function.
22903         (sparc_init_pic_reg): Likewise.
22904         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22905         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
22907 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
22909         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
22910         Add item for branch_cost.
22912 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
22914         PR rtl-optimization/83565
22915         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
22916         not extend the result to a larger mode for rotate operations.
22917         (num_sign_bit_copies1): Likewise.
22919 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
22921         PR target/40411
22922         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
22923         -symbolic.
22924         Use values-Xc.o for -pedantic.
22925         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
22927 2018-01-12  Martin Liska  <mliska@suse.cz>
22929         PR ipa/83054
22930         * ipa-devirt.c (final_warning_record::grow_type_warnings):
22931         New function.
22932         (possible_polymorphic_call_targets): Use it.
22933         (ipa_devirt): Likewise.
22935 2018-01-12  Martin Liska  <mliska@suse.cz>
22937         * profile-count.h (enum profile_quality): Use 0 as invalid
22938         enum value of profile_quality.
22940 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
22942         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
22943         -mext-string options.
22945 2018-01-12  Richard Biener  <rguenther@suse.de>
22947         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
22948         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
22949         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
22950         Likewise.
22951         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
22953 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
22955         * configure.ac (--with-long-double-format): Add support for the
22956         configuration option to change the default long double format on
22957         PowerPC systems.
22958         * config.gcc (powerpc*-linux*-*): Likewise.
22959         * configure: Regenerate.
22960         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
22961         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
22962         used without modification.
22964 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22966         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
22967         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
22968         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
22969         MISC_BUILTIN_SPEC_BARRIER.
22970         (rs6000_init_builtins): Likewise.
22971         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
22972         enum value.
22973         (speculation_barrier): New define_insn.
22974         * doc/extend.texi: Document __builtin_speculation_barrier.
22976 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
22978         PR target/83203
22979         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
22980         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
22981         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
22982         iterators.
22983         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
22984         integral modes instead of "ss" and "sd".
22985         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
22986         vectors with 32-bit and 64-bit elements.
22987         (vecdupssescalarmodesuffix): New mode attribute.
22988         (vec_dup<mode>): Use it.
22990 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
22992         PR target/83330
22993         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
22994         frame if argument is passed on stack.
22996 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
22998         PR target/82682
22999         * ree.c (combine_reaching_defs): Optimize also
23000         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
23001         reg2=any_extend(exp); reg1=reg2;, formatting fix.
23003 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
23005         PR middle-end/83189
23006         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
23008 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
23010         PR middle-end/83718
23011         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
23012         after they are computed.
23014 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
23016         PR tree-optimization/83695
23017         * gimple-loop-linterchange.cc
23018         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
23019         reset cached scev information after interchange.
23020         (pass_linterchange::execute): Remove call to scev_reset_htab.
23022 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23024         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
23025         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
23026         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
23027         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
23028         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
23029         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
23030         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
23031         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
23032         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
23033         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
23034         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
23035         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
23036         (V_lane_reg): Likewise.
23037         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
23038         New define_expand.
23039         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
23040         (vfmal_lane_low<mode>_intrinsic,
23041         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
23042         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
23043         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
23044         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
23045         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
23046         vfmsl_lane_high<mode>_intrinsic): New define_insns.
23048 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23050         * config/arm/arm-cpus.in (fp16fml): New feature.
23051         (ALL_SIMD): Add fp16fml.
23052         (armv8.2-a): Add fp16fml as an option.
23053         (armv8.3-a): Likewise.
23054         (armv8.4-a): Add fp16fml as part of fp16.
23055         * config/arm/arm.h (TARGET_FP16FML): Define.
23056         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
23057         when appropriate.
23058         * config/arm/arm-modes.def (V2HF): Define.
23059         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
23060         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
23061         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
23062         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
23063         vfmsl_low, vfmsl_high): New set of builtins.
23064         * config/arm/iterators.md (PLUSMINUS): New code iterator.
23065         (vfml_op): New code attribute.
23066         (VFMLHALVES): New int iterator.
23067         (VFML, VFMLSEL): New mode attributes.
23068         (V_reg): Define mapping for V2HF.
23069         (V_hi, V_lo): New mode attributes.
23070         (VF_constraint): Likewise.
23071         (vfml_half, vfml_half_selector): New int attributes.
23072         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
23073         define_expand.
23074         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23075         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
23076         New define_insn.
23077         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
23078         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
23079         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
23080         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
23081         documentation.
23082         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
23083         Document new effective target and option set.
23085 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23087         * config/arm/arm-cpus.in (armv8_4): New feature.
23088         (ARMv8_4a): New fgroup.
23089         (armv8.4-a): New arch.
23090         * config/arm/arm-tables.opt: Regenerate.
23091         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
23092         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
23093         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
23094         Add matching rules for -march=armv8.4-a and extensions.
23095         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
23097 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
23099         PR target/81821
23100         * config/rx/rx.md (BW): New mode attribute.
23101         (sync_lock_test_and_setsi): Add mode suffix to insn output.
23103 2018-01-11  Richard Biener  <rguenther@suse.de>
23105         PR tree-optimization/83435
23106         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
23107         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
23108         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
23110 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23111             Alan Hayward  <alan.hayward@arm.com>
23112             David Sherwood  <david.sherwood@arm.com>
23114         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
23115         field.
23116         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
23117         (aarch64_print_address_internal): Use it to check for a zero offset.
23119 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23120             Alan Hayward  <alan.hayward@arm.com>
23121             David Sherwood  <david.sherwood@arm.com>
23123         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
23124         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
23125         Return a poly_int64 rather than a HOST_WIDE_INT.
23126         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
23127         rather than a HOST_WIDE_INT.
23128         * config/aarch64/aarch64.h (aarch64_frame): Protect with
23129         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
23130         hard_fp_offset, frame_size, initial_adjust, callee_offset and
23131         final_offset from HOST_WIDE_INT to poly_int64.
23132         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
23133         to_constant when getting the number of units in an Advanced SIMD
23134         mode.
23135         (aarch64_builtin_vectorized_function): Check for a constant number
23136         of units.
23137         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
23138         GET_MODE_SIZE.
23139         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
23140         attribute instead of GET_MODE_NUNITS.
23141         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
23142         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
23143         GET_MODE_SIZE for fixed-size registers.
23144         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
23145         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
23146         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
23147         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
23148         (aarch64_print_operand, aarch64_print_address_internal)
23149         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
23150         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
23151         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
23152         Handle polynomial GET_MODE_SIZE.
23153         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
23154         wider than SImode without modification.
23155         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
23156         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
23157         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
23158         passing and returning SVE modes.
23159         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
23160         rather than GEN_INT.
23161         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
23162         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
23163         (aarch64_allocate_and_probe_stack_space): Likewise.
23164         (aarch64_layout_frame): Cope with polynomial offsets.
23165         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
23166         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
23167         polynomial offsets.
23168         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
23169         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
23170         poly_int64 rather than a HOST_WIDE_INT.
23171         (aarch64_get_separate_components, aarch64_process_components)
23172         (aarch64_expand_prologue, aarch64_expand_epilogue)
23173         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
23174         (aarch64_anchor_offset): New function, split out from...
23175         (aarch64_legitimize_address): ...here.
23176         (aarch64_builtin_vectorization_cost): Handle polynomial
23177         TYPE_VECTOR_SUBPARTS.
23178         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
23179         GET_MODE_NUNITS.
23180         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
23181         number of elements from the PARALLEL rather than the mode.
23182         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
23183         rather than GET_MODE_BITSIZE.
23184         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
23185         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
23186         (aarch64_expand_vec_perm_const_1): Handle polynomial
23187         d->perm.length () and d->perm elements.
23188         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
23189         Apply to_constant to d->perm elements.
23190         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
23191         polynomial CONST_VECTOR_NUNITS.
23192         (aarch64_move_pointer): Take amount as a poly_int64 rather
23193         than an int.
23194         (aarch64_progress_pointer): Avoid temporary variable.
23195         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
23196         the mode attribute instead of GET_MODE.
23198 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23199             Alan Hayward  <alan.hayward@arm.com>
23200             David Sherwood  <david.sherwood@arm.com>
23202         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
23203         x exists before using it.
23204         (aarch64_add_constant_internal): Rename to...
23205         (aarch64_add_offset_1): ...this.  Replace regnum with separate
23206         src and dest rtxes.  Handle the case in which they're different,
23207         including when the offset is zero.  Replace scratchreg with an rtx.
23208         Use 2 additions if there is no spare register into which we can
23209         move a 16-bit constant.
23210         (aarch64_add_constant): Delete.
23211         (aarch64_add_offset): Replace reg with separate src and dest
23212         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
23213         Use aarch64_add_offset_1.
23214         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
23215         an rtx rather than an int.  Take the delta as a poly_int64
23216         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
23217         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
23218         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
23219         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
23220         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
23221         and aarch64_add_sp.
23222         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
23223         aarch64_add_constant.
23225 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23227         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
23228         Use scalar_float_mode.
23230 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23232         * config/aarch64/aarch64-simd.md
23233         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
23234         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
23235         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
23236         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
23237         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
23238         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
23239         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
23240         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
23241         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
23242         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
23244 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23246         PR target/83514
23247         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
23248         targ_options->x_arm_arch_string is non NULL.
23250 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
23252         * config/aarch64/aarch64.h
23253         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
23255 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
23257         PR target/82096
23258         * expmed.c (emit_store_flag_force): Swap if const op0
23259         and change VOIDmode to mode of op0.
23261 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
23263         PR rtl-optimization/83761
23264         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
23265         than bytes to mode_for_size.
23267 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23269         PR middle-end/83189
23270         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
23271         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
23272         profile.
23274 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23276         PR middle-end/83575
23277         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
23278         when in layout mode.
23279         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
23280         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
23281         partition fixup.
23283 2018-01-10  Michael Collison  <michael.collison@arm.com>
23285         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
23286         * config/aarch64/aarch64-option-extension.def: Add
23287         AARCH64_OPT_EXTENSION of 'fp16fml'.
23288         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23289         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
23290         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
23291         * config/aarch64/constraints.md (Ui7): New constraint.
23292         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
23293         (VFMLA_SEL_W): Ditto.
23294         (f16quad): Ditto.
23295         (f16mac1): Ditto.
23296         (VFMLA16_LOW): New int iterator.
23297         (VFMLA16_HIGH): Ditto.
23298         (UNSPEC_FMLAL): New unspec.
23299         (UNSPEC_FMLSL): Ditto.
23300         (UNSPEC_FMLAL2): Ditto.
23301         (UNSPEC_FMLSL2): Ditto.
23302         (f16mac): New code attribute.
23303         * config/aarch64/aarch64-simd-builtins.def
23304         (aarch64_fmlal_lowv2sf): Ditto.
23305         (aarch64_fmlsl_lowv2sf): Ditto.
23306         (aarch64_fmlalq_lowv4sf): Ditto.
23307         (aarch64_fmlslq_lowv4sf): Ditto.
23308         (aarch64_fmlal_highv2sf): Ditto.
23309         (aarch64_fmlsl_highv2sf): Ditto.
23310         (aarch64_fmlalq_highv4sf): Ditto.
23311         (aarch64_fmlslq_highv4sf): Ditto.
23312         (aarch64_fmlal_lane_lowv2sf): Ditto.
23313         (aarch64_fmlsl_lane_lowv2sf): Ditto.
23314         (aarch64_fmlal_laneq_lowv2sf): Ditto.
23315         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
23316         (aarch64_fmlalq_lane_lowv4sf): Ditto.
23317         (aarch64_fmlsl_lane_lowv4sf): Ditto.
23318         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
23319         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
23320         (aarch64_fmlal_lane_highv2sf): Ditto.
23321         (aarch64_fmlsl_lane_highv2sf): Ditto.
23322         (aarch64_fmlal_laneq_highv2sf): Ditto.
23323         (aarch64_fmlsl_laneq_highv2sf): Ditto.
23324         (aarch64_fmlalq_lane_highv4sf): Ditto.
23325         (aarch64_fmlsl_lane_highv4sf): Ditto.
23326         (aarch64_fmlalq_laneq_highv4sf): Ditto.
23327         (aarch64_fmlsl_laneq_highv4sf): Ditto.
23328         * config/aarch64/aarch64-simd.md:
23329         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
23330         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23331         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
23332         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
23333         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
23334         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
23335         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
23336         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
23337         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
23338         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
23339         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
23340         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
23341         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
23342         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
23343         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
23344         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
23345         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
23346         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
23347         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
23348         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
23349         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
23350         (vfmlsl_low_u32): Ditto.
23351         (vfmlalq_low_u32): Ditto.
23352         (vfmlslq_low_u32): Ditto.
23353         (vfmlal_high_u32): Ditto.
23354         (vfmlsl_high_u32): Ditto.
23355         (vfmlalq_high_u32): Ditto.
23356         (vfmlslq_high_u32): Ditto.
23357         (vfmlal_lane_low_u32): Ditto.
23358         (vfmlsl_lane_low_u32): Ditto.
23359         (vfmlal_laneq_low_u32): Ditto.
23360         (vfmlsl_laneq_low_u32): Ditto.
23361         (vfmlalq_lane_low_u32): Ditto.
23362         (vfmlslq_lane_low_u32): Ditto.
23363         (vfmlalq_laneq_low_u32): Ditto.
23364         (vfmlslq_laneq_low_u32): Ditto.
23365         (vfmlal_lane_high_u32): Ditto.
23366         (vfmlsl_lane_high_u32): Ditto.
23367         (vfmlal_laneq_high_u32): Ditto.
23368         (vfmlsl_laneq_high_u32): Ditto.
23369         (vfmlalq_lane_high_u32): Ditto.
23370         (vfmlslq_lane_high_u32): Ditto.
23371         (vfmlalq_laneq_high_u32): Ditto.
23372         (vfmlslq_laneq_high_u32): Ditto.
23373         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
23374         (AARCH64_FL_FOR_ARCH8_4): New.
23375         (AARCH64_ISA_F16FML): New ISA flag.
23376         (TARGET_F16FML): New feature flag for fp16fml.
23377         (doc/invoke.texi): Document new fp16fml option.
23379 2018-01-10  Michael Collison  <michael.collison@arm.com>
23381         * config/aarch64/aarch64-builtins.c:
23382         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
23383         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23384         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
23385         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
23386         (AARCH64_ISA_SHA3): New ISA flag.
23387         (TARGET_SHA3): New feature flag for sha3.
23388         * config/aarch64/iterators.md (sha512_op): New int attribute.
23389         (CRYPTO_SHA512): New int iterator.
23390         (UNSPEC_SHA512H): New unspec.
23391         (UNSPEC_SHA512H2): Ditto.
23392         (UNSPEC_SHA512SU0): Ditto.
23393         (UNSPEC_SHA512SU1): Ditto.
23394         * config/aarch64/aarch64-simd-builtins.def
23395         (aarch64_crypto_sha512hqv2di): New builtin.
23396         (aarch64_crypto_sha512h2qv2di): Ditto.
23397         (aarch64_crypto_sha512su0qv2di): Ditto.
23398         (aarch64_crypto_sha512su1qv2di): Ditto.
23399         (aarch64_eor3qv8hi): Ditto.
23400         (aarch64_rax1qv2di): Ditto.
23401         (aarch64_xarqv2di): Ditto.
23402         (aarch64_bcaxqv8hi): Ditto.
23403         * config/aarch64/aarch64-simd.md:
23404         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
23405         (aarch64_crypto_sha512su0qv2di): Ditto.
23406         (aarch64_crypto_sha512su1qv2di): Ditto.
23407         (aarch64_eor3qv8hi): Ditto.
23408         (aarch64_rax1qv2di): Ditto.
23409         (aarch64_xarqv2di): Ditto.
23410         (aarch64_bcaxqv8hi): Ditto.
23411         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
23412         (vsha512h2q_u64): Ditto.
23413         (vsha512su0q_u64): Ditto.
23414         (vsha512su1q_u64): Ditto.
23415         (veor3q_u16): Ditto.
23416         (vrax1q_u64): Ditto.
23417         (vxarq_u64): Ditto.
23418         (vbcaxq_u16): Ditto.
23419         * config/arm/types.md (crypto_sha512): New type attribute.
23420         (crypto_sha3): Ditto.
23421         (doc/invoke.texi): Document new sha3 option.
23423 2018-01-10  Michael Collison  <michael.collison@arm.com>
23425         * config/aarch64/aarch64-builtins.c:
23426         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
23427         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23428         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
23429         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
23430         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
23431         (AARCH64_ISA_SM4): New ISA flag.
23432         (TARGET_SM4): New feature flag for sm4.
23433         * config/aarch64/aarch64-simd-builtins.def
23434         (aarch64_sm3ss1qv4si): Ditto.
23435         (aarch64_sm3tt1aq4si): Ditto.
23436         (aarch64_sm3tt1bq4si): Ditto.
23437         (aarch64_sm3tt2aq4si): Ditto.
23438         (aarch64_sm3tt2bq4si): Ditto.
23439         (aarch64_sm3partw1qv4si): Ditto.
23440         (aarch64_sm3partw2qv4si): Ditto.
23441         (aarch64_sm4eqv4si): Ditto.
23442         (aarch64_sm4ekeyqv4si): Ditto.
23443         * config/aarch64/aarch64-simd.md:
23444         (aarch64_sm3ss1qv4si): Ditto.
23445         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
23446         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
23447         (aarch64_sm4eqv4si): Ditto.
23448         (aarch64_sm4ekeyqv4si): Ditto.
23449         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
23450         (sm3part_op): Ditto.
23451         (CRYPTO_SM3TT): Ditto.
23452         (CRYPTO_SM3PART): Ditto.
23453         (UNSPEC_SM3SS1): New unspec.
23454         (UNSPEC_SM3TT1A): Ditto.
23455         (UNSPEC_SM3TT1B): Ditto.
23456         (UNSPEC_SM3TT2A): Ditto.
23457         (UNSPEC_SM3TT2B): Ditto.
23458         (UNSPEC_SM3PARTW1): Ditto.
23459         (UNSPEC_SM3PARTW2): Ditto.
23460         (UNSPEC_SM4E): Ditto.
23461         (UNSPEC_SM4EKEY): Ditto.
23462         * config/aarch64/constraints.md (Ui2): New constraint.
23463         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
23464         * config/arm/types.md (crypto_sm3): New type attribute.
23465         (crypto_sm4): Ditto.
23466         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
23467         (vsm3tt1aq_u32): Ditto.
23468         (vsm3tt1bq_u32): Ditto.
23469         (vsm3tt2aq_u32): Ditto.
23470         (vsm3tt2bq_u32): Ditto.
23471         (vsm3partw1q_u32): Ditto.
23472         (vsm3partw2q_u32): Ditto.
23473         (vsm4eq_u32): Ditto.
23474         (vsm4ekeyq_u32): Ditto.
23475         (doc/invoke.texi): Document new sm4 option.
23477 2018-01-10  Michael Collison  <michael.collison@arm.com>
23479         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
23480         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
23481         (AARCH64_FL_FOR_ARCH8_4): New.
23482         (AARCH64_FL_V8_4): New flag.
23483         (doc/invoke.texi): Document new armv8.4-a option.
23485 2018-01-10  Michael Collison  <michael.collison@arm.com>
23487         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
23488         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
23489         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
23490         * config/aarch64/aarch64-option-extension.def: Add
23491         AARCH64_OPT_EXTENSION of 'sha2'.
23492         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
23493         (crypto): Disable sha2 and aes if crypto disabled.
23494         (crypto): Enable aes and sha2 if enabled.
23495         (simd): Disable sha2 and aes if simd disabled.
23496         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
23497         New flags.
23498         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
23499         (TARGET_SHA2): New feature flag for sha2.
23500         (TARGET_AES): New feature flag for aes.
23501         * config/aarch64/aarch64-simd.md:
23502         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
23503         conditional on TARGET_AES.
23504         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
23505         (aarch64_crypto_sha1hsi): Make pattern conditional
23506         on TARGET_SHA2.
23507         (aarch64_crypto_sha1hv4si): Ditto.
23508         (aarch64_be_crypto_sha1hv4si): Ditto.
23509         (aarch64_crypto_sha1su1v4si): Ditto.
23510         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
23511         (aarch64_crypto_sha1su0v4si): Ditto.
23512         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
23513         (aarch64_crypto_sha256su0v4si): Ditto.
23514         (aarch64_crypto_sha256su1v4si): Ditto.
23515         (doc/invoke.texi): Document new aes and sha2 options.
23517 2018-01-10  Martin Sebor  <msebor@redhat.com>
23519         PR tree-optimization/83781
23520         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
23521         as string arrays.
23523 2018-01-11  Martin Sebor  <msebor@gmail.com>
23524             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
23526         PR tree-optimization/83501
23527         PR tree-optimization/81703
23529         * tree-ssa-strlen.c (get_string_cst): Rename...
23530         (get_string_len): ...to this.  Handle global constants.
23531         (handle_char_store): Adjust.
23533 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
23534             Jim Wilson  <jimw@sifive.com>
23536         * config/riscv/riscv-protos.h (riscv_output_return): New.
23537         * config/riscv/riscv.c (struct machine_function): New naked_p field.
23538         (riscv_attribute_table, riscv_output_return),
23539         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
23540         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
23541         (riscv_compute_frame_info): Only compute frame->mask if not a naked
23542         function.
23543         (riscv_expand_prologue): Add early return for naked function.
23544         (riscv_expand_epilogue): Likewise.
23545         (riscv_function_ok_for_sibcall): Return false for naked function.
23546         (riscv_set_current_function): New.
23547         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
23548         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
23549         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
23550         * doc/extend.texi (RISC-V Function Attributes): New.
23552 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
23554         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
23555         check for 128-bit long double before checking TCmode.
23556         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
23557         128-bit long doubles before checking TFmode or TCmode.
23558         (FLOAT128_IBM_P): Likewise.
23560 2018-01-10  Martin Sebor  <msebor@redhat.com>
23562         PR tree-optimization/83671
23563         * builtins.c (c_strlen): Unconditionally return zero for the empty
23564         string.
23565         Use -Warray-bounds for warnings.
23566         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
23567         for non-constant array indices with COMPONENT_REF, arrays of
23568         arrays, and pointers to arrays.
23569         (gimple_fold_builtin_strlen): Determine and set length range for
23570         non-constant character arrays.
23572 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
23574         PR middle-end/81897
23575         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
23576         empty blocks.
23578 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
23580         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
23582 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
23584         PR target/83399
23585         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
23586         VECTOR_MEM_ALTIVEC_OR_VSX_P.
23587         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
23588         indexed_or_indirect_operand predicate.
23589         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
23590         (*vsx_le_perm_load_v8hi): Likewise.
23591         (*vsx_le_perm_load_v16qi): Likewise.
23592         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
23593         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
23594         (*vsx_le_perm_store_v8hi): Likewise.
23595         (*vsx_le_perm_store_v16qi): Likewise.
23596         (eight unnamed splitters): Likewise.
23598 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
23600         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
23601         * config/rs6000/emmintrin.h: Likewise.
23602         * config/rs6000/mmintrin.h: Likewise.
23603         * config/rs6000/xmmintrin.h: Likewise.
23605 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
23607         PR c++/43486
23608         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23609         "public_flag".
23610         * tree.c (tree_nop_conversion): Return true for location wrapper
23611         nodes.
23612         (maybe_wrap_with_location): New function.
23613         (selftest::check_strip_nops): New function.
23614         (selftest::test_location_wrappers): New function.
23615         (selftest::tree_c_tests): Call it.
23616         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
23617         (maybe_wrap_with_location): New decl.
23618         (EXPR_LOCATION_WRAPPER_P): New macro.
23619         (location_wrapper_p): New inline function.
23620         (tree_strip_any_location_wrapper): New inline function.
23622 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
23624         PR target/83735
23625         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23626         stack_realign_offset for the largest alignment of stack slot
23627         actually used.
23628         (ix86_find_max_used_stack_alignment): New function.
23629         (ix86_finalize_stack_frame_flags): Use it.  Set
23630         max_used_stack_alignment if we don't realign stack.
23631         * config/i386/i386.h (machine_function): Add
23632         max_used_stack_alignment.
23634 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
23636         * config/arm/arm.opt (-mbranch-cost): New option.
23637         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
23638         account.
23640 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
23642         PR target/83629
23643         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
23644         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
23646 2018-01-10  Richard Biener  <rguenther@suse.de>
23648         PR debug/83765
23649         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
23650         early out so it also covers the case where we have a non-NULL
23651         origin.
23653 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
23655         PR tree-optimization/83753
23656         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
23657         for non-strided grouped accesses if the number of elements is 1.
23659 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
23661         PR target/81616
23662         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
23663         * i386.h (TARGET_USE_GATHER): Define.
23664         * x86-tune.def (X86_TUNE_USE_GATHER): New.
23666 2018-01-10  Martin Liska  <mliska@suse.cz>
23668         PR bootstrap/82831
23669         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23670         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
23671         partitioning.
23672         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
23673         CLEANUP_NO_PARTITIONING is not set.
23675 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
23677         * doc/rtl.texi: Remove documentation of (const ...) wrappers
23678         for vectors, as a partial revert of r254296.
23679         * rtl.h (const_vec_p): Delete.
23680         (const_vec_duplicate_p): Don't test for vector CONSTs.
23681         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
23682         * expmed.c (make_tree): Likewise.
23684         Revert:
23685         * common.md (E, F): Use CONSTANT_P instead of checking for
23686         CONST_VECTOR.
23687         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
23688         checking for CONST_VECTOR.
23690 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
23692         PR middle-end/83575
23693         * predict.c (force_edge_cold): Handle in more sane way edges
23694         with no prediction.
23696 2018-01-09  Carl Love  <cel@us.ibm.com>
23698         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
23699         V4SI, V4SF types.
23700         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
23701         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
23702         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
23703         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
23704         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
23705         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
23706         * config/rs6000/rs6000-protos.h: Add extern defition for
23707         rs6000_generate_float2_double_code.
23708         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
23709         function.
23710         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
23711         (float2_v2df): Add define_expand.
23713 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
23715         PR target/83628
23716         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
23717         op_mode in the force_to_mode call.
23719 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
23721         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
23722         instead of checking each element individually.
23723         (aarch64_evpc_uzp): Likewise.
23724         (aarch64_evpc_zip): Likewise.
23725         (aarch64_evpc_ext): Likewise.
23726         (aarch64_evpc_rev): Likewise.
23727         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
23728         instead of checking each element individually.  Return true without
23729         generating rtl if
23730         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
23731         whether all selected elements come from the same input, instead of
23732         checking each element individually.  Remove calls to gen_rtx_REG,
23733         start_sequence and end_sequence and instead assert that no rtl is
23734         generated.
23736 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
23738         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
23739         order of HIGH and CONST checks.
23741 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
23743         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
23744         if the destination isn't an SSA_NAME.
23746 2018-01-09  Richard Biener  <rguenther@suse.de>
23748         PR tree-optimization/83668
23749         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
23750         move prologue...
23751         (canonicalize_loop_form): ... here, renamed from ...
23752         (canonicalize_loop_closed_ssa_form): ... this and amended to
23753         swap successor edges for loop exit blocks to make us use
23754         the RPO order we need for initial schedule generation.
23756 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
23758         PR tree-optimization/64811
23759         * match.pd: When optimizing comparisons with Inf, avoid
23760         introducing or losing exceptions from comparisons with NaN.
23762 2018-01-09  Martin Liska  <mliska@suse.cz>
23764         PR sanitizer/82517
23765         * asan.c (shadow_mem_size): Add gcc_assert.
23767 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
23769         Don't save registers in main().
23771         PR target/83738
23772         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
23773         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
23774         * config/avr/avr.c (avr_set_current_function): Don't error if
23775         naked, OS_task or OS_main are specified at the same time.
23776         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
23777         OS_main.
23778         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
23779         attribute.
23780         * common/config/avr/avr-common.c (avr_option_optimization_table):
23781         Switch on -mmain-is-OS_task for optimizing compilations.
23783 2018-01-09  Richard Biener  <rguenther@suse.de>
23785         PR tree-optimization/83572
23786         * graphite.c: Include cfganal.h.
23787         (graphite_transform_loops): Connect infinite loops to exit
23788         and remove fake edges at the end.
23790 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
23792         * ipa-inline.c (edge_badness): Revert accidental checkin.
23794 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
23796         PR ipa/80763
23797         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
23798         symbols; not inline clones.
23800 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
23802         PR target/83507
23803         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
23804         hard registers.  Formatting fixes.
23806         PR preprocessor/83722
23807         * gcc.c (try_generate_repro): Pass
23808         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
23809         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
23810         do_report_bug.
23812 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
23813             Kito Cheng  <kito.cheng@gmail.com>
23815         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
23816         (riscv_leaf_function_p): Delete.
23817         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
23819 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23821         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
23822         function.
23823         (do_ifelse): New function.
23824         (do_isel): New function.
23825         (do_sub3): New function.
23826         (do_add3): New function.
23827         (do_load_mask_compare): New function.
23828         (do_overlap_load_compare): New function.
23829         (expand_compare_loop): New function.
23830         (expand_block_compare): Call expand_compare_loop() when appropriate.
23831         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
23832         option description.
23833         (-mblock-compare-inline-loop-limit): New option.
23835 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23837         PR target/83677
23838         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
23839         Reverse order of second and third operands in first alternative.
23840         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
23841         of first and second elements in UNSPEC_VPERMR vector.
23842         (altivec_expand_vec_perm_le): Likewise.
23844 2018-01-08  Jeff Law  <law@redhat.com>
23846         PR rtl-optimizatin/81308
23847         * tree-switch-conversion.c (cfg_altered): New file scoped static.
23848         (process_switch): If group_case_labels makes a change, then set
23849         cfg_altered.
23850         (pass_convert_switch::execute): If a switch is converted, then
23851         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
23853         PR rtl-optimization/81308
23854         * recog.c (split_all_insns): Conditionally cleanup the CFG after
23855         splitting insns.
23857 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
23859         PR target/83663 - Revert r255946
23860         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
23861         generation for cases where splatting a value is not useful.
23862         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
23863         across a vec_duplicate and a paradoxical subreg forming a vector
23864         mode to a vec_concat.
23866 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23868         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
23869         -march=armv8.3-a variants.
23870         * config/arm/t-multilib: Likewise.
23871         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
23873 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23875         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
23876         to generate rtl.
23877         (cceq_ior_compare_complement): Give it a name so I can use it, and
23878         change boolean_or_operator predicate to boolean_operator so it can
23879         be used to generate a crand.
23880         (eqne): New code iterator.
23881         (bd/bd_neg): New code_attrs.
23882         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
23883         a single define_insn.
23884         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
23885         decrement (bdnzt/bdnzf/bdzt/bdzf).
23886         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
23887         with the new names of the branch decrement patterns, and added the
23888         names of the branch decrement conditional patterns.
23890 2018-01-08  Richard Biener  <rguenther@suse.de>
23892         PR tree-optimization/83563
23893         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
23894         cache.
23896 2018-01-08  Richard Biener  <rguenther@suse.de>
23898         PR middle-end/83713
23899         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
23901 2018-01-08  Richard Biener  <rguenther@suse.de>
23903         PR tree-optimization/83685
23904         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
23905         references to abnormals.
23907 2018-01-08  Richard Biener  <rguenther@suse.de>
23909         PR lto/83719
23910         * dwarf2out.c (output_indirect_strings): Handle empty
23911         skeleton_debug_str_hash.
23912         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
23914 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
23916         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
23917         (emit_store_direct): Likewise.
23918         (arc_trampoline_adjust_address): Likewise.
23919         (arc_asm_trampoline_template): New function.
23920         (arc_initialize_trampoline): Use asm_trampoline_template.
23921         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
23922         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
23923         * config/arc/arc.md (flush_icache): Delete pattern.
23925 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
23927         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
23928         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
23929         munaligned-access.
23931 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23933         PR target/83681
23934         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
23935         by not USED_FOR_TARGET.
23936         (make_pass_resolve_sw_modes): Likewise.
23938 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23940         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
23941         USED_FOR_TARGET.
23943 2018-01-08  Richard Biener  <rguenther@suse.de>
23945         PR middle-end/83580
23946         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
23948 2018-01-08  Richard Biener  <rguenther@suse.de>
23950         PR middle-end/83517
23951         * match.pd ((t * 2) / 2) -> t): Add missing :c.
23953 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
23955         PR middle-end/81897
23956         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
23957         basic blocks with a small number of successors.
23958         (convert_control_dep_chain_into_preds): Improve handling of
23959         forwarder blocks.
23960         (dump_predicates): Split apart into...
23961         (dump_pred_chain): ...here...
23962         (dump_pred_info): ...and here.
23963         (can_one_predicate_be_invalidated_p): Add debugging printfs.
23964         (can_chain_union_be_invalidated_p): Improve check for invalidation
23965         of paths.
23966         (uninit_uses_cannot_happen): Avoid unnecessary if
23967         convert_control_dep_chain_into_preds yielded nothing.
23969 2018-01-06  Martin Sebor  <msebor@redhat.com>
23971         PR tree-optimization/83640
23972         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
23973         subtracting negative offset from size.
23974         (builtin_access::overlap): Adjust offset bounds of the access to fall
23975         within the size of the object if possible.
23977 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
23979         PR rtl-optimization/83699
23980         * expmed.c (extract_bit_field_1): Restrict the vector usage of
23981         extract_bit_field_as_subreg to cases in which the extracted
23982         value is also a vector.
23984         * lra-constraints.c (process_alt_operands): Test for the equivalence
23985         substitutions when detecting a possible reload cycle.
23987 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
23989         PR debug/83480
23990         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
23991         by default if flag_selective_schedling{,2}.  Formatting fixes.
23993         PR rtl-optimization/83682
23994         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
23995         if it has non-VECTOR_MODE element mode.
23996         (vec_duplicate_p): Likewise.
23998         PR middle-end/83694
23999         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
24000         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
24002 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
24004         PR target/83604
24005         * config/i386/i386-builtin.def
24006         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
24007         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
24008         Require also OPTION_MASK_ISA_AVX512F in addition to
24009         OPTION_MASK_ISA_GFNI.
24010         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
24011         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
24012         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
24013         to OPTION_MASK_ISA_GFNI.
24014         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
24015         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
24016         OPTION_MASK_ISA_AVX512BW.
24017         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
24018         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
24019         addition to OPTION_MASK_ISA_GFNI.
24020         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
24021         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
24022         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
24023         to OPTION_MASK_ISA_GFNI.
24024         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
24025         a requirement for all ISAs rather than any of them with a few
24026         exceptions.
24027         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
24028         processing.
24029         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
24030         bitmasks to be enabled with 3 exceptions, instead of requiring any
24031         enabled ISA with lots of exceptions.
24032         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
24033         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
24034         Change avx512bw in isa attribute to avx512f.
24035         * config/i386/sgxintrin.h: Add license boilerplate.
24036         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
24037         to __AVX512F__ and __AVX512VL to __AVX512VL__.
24038         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
24039         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
24040         defined.
24041         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
24042         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
24043         temporarily sse2 rather than sse if not enabled already.
24045         PR target/83604
24046         * config/i386/sse.md (VI248_VLBW): Rename to ...
24047         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
24048         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
24049         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
24050         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
24051         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
24052         mode iterator instead of VI248_VLBW.
24054 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
24056         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
24057         (record_modified): Skip clobbers; add debug output.
24058         (param_change_prob): Use sreal frequencies.
24060 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
24062         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
24063         punt for user-aligned variables.
24065 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
24067         * tree-chrec.c (chrec_contains_symbols): Return true for
24068         POLY_INT_CST.
24070 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
24072         PR target/82439
24073         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
24074         of (x|y) == x for BICS pattern.
24076 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
24078         PR tree-optimization/83605
24079         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
24080         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
24081         can throw.
24083 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24085         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
24086         * config/epiphany/rtems.h: New file.
24088 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24089             Uros Bizjak  <ubizjak@gmail.com>
24091         PR target/83554
24092         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
24093         QIreg_operand instead of register_operand predicate.
24094         * config/i386/i386.c (ix86_rop_should_change_byte_p,
24095         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
24096         comments instead of -fmitigate[-_]rop.
24098 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24100         PR bootstrap/81926
24101         * cgraphunit.c (symbol_table::compile): Switch to text_section
24102         before calling assembly_start debug hook.
24103         * run-rtl-passes.c (run_rtl_passes): Likewise.
24104         Include output.h.
24106 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24108         * tree-vrp.c (extract_range_from_binary_expr_1): Check
24109         range_int_cst_p rather than !symbolic_range_p before calling
24110         extract_range_from_multiplicative_op_1.
24112 2018-01-04  Jeff Law  <law@redhat.com>
24114         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
24115         redundant test in assertion.
24117 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24119         * doc/rtl.texi: Document machine_mode wrapper classes.
24121 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24123         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
24124         using tree_to_uhwi.
24126 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24128         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
24129         the VEC_PERM_EXPR fold to fail.
24131 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24133         PR debug/83585
24134         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
24135         to switched_sections.
24137 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24139         PR target/83680
24140         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
24141         test for d.testing.
24143 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
24145         PR target/83387
24146         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
24147         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
24149 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24151         PR debug/83666
24152         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
24153         is BLKmode and bitpos not zero or mode change is needed.
24155 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
24157         PR target/83675
24158         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24159         TARGET_VIS2.
24161 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
24163         PR target/83628
24164         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
24165         instead of MULT rtx.  Update all corresponding splitters.
24166         (*saddl_se): Ditto.
24167         (*ssub<modesuffix>): Ditto.
24168         (*ssubl_se): Ditto.
24169         (*cmp_sadd_di): Update split patterns.
24170         (*cmp_sadd_si): Ditto.
24171         (*cmp_sadd_sidi): Ditto.
24172         (*cmp_ssub_di): Ditto.
24173         (*cmp_ssub_si): Ditto.
24174         (*cmp_ssub_sidi): Ditto.
24175         * config/alpha/predicates.md (const23_operand): New predicate.
24176         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
24177         Look for ASHIFT, not MULT inner operand.
24178         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
24180 2018-01-04  Martin Liska  <mliska@suse.cz>
24182         PR gcov-profile/83669
24183         * gcov.c (output_intermediate_file): Add version to intermediate
24184         gcov file.
24185         * doc/gcov.texi: Document new field 'version' in intermediate
24186         file format. Fix location of '-k' option of gcov command.
24188 2018-01-04  Martin Liska  <mliska@suse.cz>
24190         PR ipa/82352
24191         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
24193 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
24195         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
24197 2018-01-03  Martin Sebor  <msebor@redhat.com>
24199         PR tree-optimization/83655
24200         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
24201         checking calls with invalid arguments.
24203 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24205         * tree-vect-stmts.c (vect_get_store_rhs): New function.
24206         (vectorizable_mask_load_store): Delete.
24207         (vectorizable_call): Return false for masked loads and stores.
24208         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
24209         instead of gimple_assign_rhs1.
24210         (vectorizable_load): Handle IFN_MASK_LOAD.
24211         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
24213 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24215         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
24216         split out from..,
24217         (vectorizable_mask_load_store): ...here.
24218         (vectorizable_load): ...and here.
24220 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24222         * tree-vect-stmts.c (vect_build_all_ones_mask)
24223         (vect_build_zero_merge_argument): New functions, split out from...
24224         (vectorizable_load): ...here.
24226 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24228         * tree-vect-stmts.c (vect_check_store_rhs): New function,
24229         split out from...
24230         (vectorizable_mask_load_store): ...here.
24231         (vectorizable_store): ...and here.
24233 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24235         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
24236         split out from...
24237         (vectorizable_mask_load_store): ...here.
24239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24241         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
24242         (vect_model_store_cost): Take a vec_load_store_type instead of a
24243         vect_def_type.
24244         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
24245         (vect_model_store_cost): Take a vec_load_store_type instead of a
24246         vect_def_type.
24247         (vectorizable_mask_load_store): Update accordingly.
24248         (vectorizable_store): Likewise.
24249         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
24251 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24253         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
24254         IFN_MASK_LOAD calls here rather than...
24255         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
24257 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24258             Alan Hayward  <alan.hayward@arm.com>
24259             David Sherwood  <david.sherwood@arm.com>
24261         * expmed.c (extract_bit_field_1): For vector extracts,
24262         fall back to extract_bit_field_as_subreg if vec_extract
24263         isn't available.
24265 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24266             Alan Hayward  <alan.hayward@arm.com>
24267             David Sherwood  <david.sherwood@arm.com>
24269         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
24270         they are variable or constant sized.
24271         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
24272         slots for constant-sized data.
24274 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24275             Alan Hayward  <alan.hayward@arm.com>
24276             David Sherwood  <david.sherwood@arm.com>
24278         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
24279         handling COND_EXPRs with boolean comparisons, try to find a better
24280         basis for the mask type than the boolean itself.
24282 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24284         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
24285         is calculated and how it can be overridden.
24286         * genmodes.c (max_bitsize_mode_any_mode): New variable.
24287         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
24288         if defined.
24289         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
24290         if nonzero.
24292 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24293             Alan Hayward  <alan.hayward@arm.com>
24294             David Sherwood  <david.sherwood@arm.com>
24296         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
24297         Remove the mode argument.
24298         (aarch64_simd_valid_immediate): Remove the mode and inverse
24299         arguments.
24300         * config/aarch64/iterators.md (bitsize): New iterator.
24301         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
24302         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
24303         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
24304         aarch64_simd_valid_immediate.
24305         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
24306         (aarch64_reg_or_bic_imm): Likewise.
24307         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
24308         with an insn_type enum and msl with a modifier_type enum.
24309         Replace element_width with a scalar_mode.  Change the shift
24310         to unsigned int.  Add constructors for scalar_float_mode and
24311         scalar_int_mode elements.
24312         (aarch64_vect_float_const_representable_p): Delete.
24313         (aarch64_can_const_movi_rtx_p)
24314         (aarch64_simd_scalar_immediate_valid_for_move)
24315         (aarch64_simd_make_constant): Update call to
24316         aarch64_simd_valid_immediate.
24317         (aarch64_advsimd_valid_immediate_hs): New function.
24318         (aarch64_advsimd_valid_immediate): Likewise.
24319         (aarch64_simd_valid_immediate): Remove mode and inverse
24320         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
24321         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
24322         and aarch64_float_const_representable_p on the result.
24323         (aarch64_output_simd_mov_immediate): Remove mode argument.
24324         Update call to aarch64_simd_valid_immediate and use of
24325         simd_immediate_info.
24326         (aarch64_output_scalar_simd_mov_immediate): Update call
24327         accordingly.
24329 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24330             Alan Hayward  <alan.hayward@arm.com>
24331             David Sherwood  <david.sherwood@arm.com>
24333         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
24334         (mode_nunits): Likewise CONST_MODE_NUNITS.
24335         * machmode.def (ADJUST_NUNITS): Document.
24336         * genmodes.c (mode_data::need_nunits_adj): New field.
24337         (blank_mode): Update accordingly.
24338         (adj_nunits): New variable.
24339         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
24340         parameter.
24341         (emit_mode_size_inline): Set need_bytesize_adj for all modes
24342         listed in adj_nunits.
24343         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
24344         listed in adj_nunits.  Don't emit case statements for such modes.
24345         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
24346         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
24347         nothing if adj_nunits is nonnull.
24348         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
24349         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
24350         (emit_mode_fbit): Update use of print_maybe_const_decl.
24351         (emit_move_size): Likewise.  Treat the array as non-const
24352         if adj_nunits.
24353         (emit_mode_adjustments): Handle adj_nunits.
24355 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24357         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
24358         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
24359         (VECTOR_MODES): Use it.
24360         (make_vector_modes): Take the prefix as an argument.
24362 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24363             Alan Hayward  <alan.hayward@arm.com>
24364             David Sherwood  <david.sherwood@arm.com>
24366         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
24367         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
24368         for MODE_VECTOR_BOOL.
24369         * machmode.def (VECTOR_BOOL_MODE): Document.
24370         * genmodes.c (VECTOR_BOOL_MODE): New macro.
24371         (make_vector_bool_mode): New function.
24372         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
24373         MODE_VECTOR_BOOL.
24374         * lto-streamer-in.c (lto_input_mode_table): Likewise.
24375         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
24376         Likewise.
24377         * stor-layout.c (int_mode_for_mode): Likewise.
24378         * tree.c (build_vector_type_for_mode): Likewise.
24379         * varasm.c (output_constant_pool_2): Likewise.
24380         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
24381         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
24382         for MODE_VECTOR_BOOL.
24383         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
24384         of mode class checks.
24385         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
24386         instead of a list of mode class checks.
24387         (expand_vector_scalar_condition): Likewise.
24388         (type_for_widest_vector_mode): Handle BImode as an inner mode.
24390 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24391             Alan Hayward  <alan.hayward@arm.com>
24392             David Sherwood  <david.sherwood@arm.com>
24394         * machmode.h (mode_size): Change from unsigned short to
24395         poly_uint16_pod.
24396         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
24397         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24398         or if measurement_type is not polynomial.
24399         (fixed_size_mode::includes_p): Check for constant-sized modes.
24400         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
24401         return a poly_uint16 rather than an unsigned short.
24402         (emit_mode_size): Change the type of mode_size from unsigned short
24403         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
24404         (emit_mode_adjustments): Cope with polynomial vector sizes.
24405         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24406         for GET_MODE_SIZE.
24407         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24408         for GET_MODE_SIZE.
24409         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
24410         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
24411         * caller-save.c (setup_save_areas): Likewise.
24412         (replace_reg_with_saved_mem): Likewise.
24413         * calls.c (emit_library_call_value_1): Likewise.
24414         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
24415         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
24416         (gen_lowpart_for_combine): Likewise.
24417         * convert.c (convert_to_integer_1): Likewise.
24418         * cse.c (equiv_constant, cse_insn): Likewise.
24419         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
24420         (cselib_subst_to_values): Likewise.
24421         * dce.c (word_dce_process_block): Likewise.
24422         * df-problems.c (df_word_lr_mark_ref): Likewise.
24423         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
24424         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
24425         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
24426         (rtl_for_decl_location): Likewise.
24427         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
24428         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
24429         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
24430         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
24431         (expand_expr_real_1): Likewise.
24432         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
24433         (pad_below): Likewise.
24434         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24435         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
24436         * ira.c (get_subreg_tracking_sizes): Likewise.
24437         * ira-build.c (ira_create_allocno_objects): Likewise.
24438         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
24439         (ira_sort_regnos_for_alter_reg): Likewise.
24440         * ira-costs.c (record_operand_costs): Likewise.
24441         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
24442         (resolve_simple_move): Likewise.
24443         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
24444         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
24445         (lra_constraints): Likewise.
24446         (CONST_POOL_OK_P): Reject variable-sized modes.
24447         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
24448         (add_pseudo_to_slot, lra_spill): Likewise.
24449         * omp-low.c (omp_clause_aligned_alignment): Likewise.
24450         * optabs-query.c (get_best_extraction_insn): Likewise.
24451         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24452         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
24453         (expand_mult_highpart, valid_multiword_target_p): Likewise.
24454         * recog.c (offsettable_address_addr_space_p): Likewise.
24455         * regcprop.c (maybe_mode_change): Likewise.
24456         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
24457         * regrename.c (build_def_use): Likewise.
24458         * regstat.c (dump_reg_info): Likewise.
24459         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
24460         (find_reloads, find_reloads_subreg_address): Likewise.
24461         * reload1.c (eliminate_regs_1): Likewise.
24462         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
24463         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
24464         (simplify_binary_operation_1, simplify_subreg): Likewise.
24465         * targhooks.c (default_function_arg_padding): Likewise.
24466         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
24467         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
24468         (verify_gimple_assign_ternary): Likewise.
24469         * tree-inline.c (estimate_move_cost): Likewise.
24470         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24471         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
24472         (get_address_cost_ainc): Likewise.
24473         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
24474         (vect_supportable_dr_alignment): Likewise.
24475         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24476         (vectorizable_reduction): Likewise.
24477         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
24478         (vectorizable_operation, vectorizable_load): Likewise.
24479         * tree.c (build_same_sized_truth_vector_type): Likewise.
24480         * valtrack.c (cleanup_auto_inc_dec): Likewise.
24481         * var-tracking.c (emit_note_insn_var_location): Likewise.
24482         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
24483         (ADDR_VEC_ALIGN): Likewise.
24485 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24486             Alan Hayward  <alan.hayward@arm.com>
24487             David Sherwood  <david.sherwood@arm.com>
24489         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
24490         unsigned short.
24491         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
24492         or if measurement_type is polynomial.
24493         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
24494         * combine.c (make_extraction): Likewise.
24495         * dse.c (find_shift_sequence): Likewise.
24496         * dwarf2out.c (mem_loc_descriptor): Likewise.
24497         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
24498         (extract_bit_field, extract_low_bits): Likewise.
24499         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
24500         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
24501         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
24502         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
24503         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
24504         * reload.c (find_reloads): Likewise.
24505         * reload1.c (alter_reg): Likewise.
24506         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
24507         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
24508         * tree-if-conv.c (predicate_mem_writes): Likewise.
24509         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
24510         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
24511         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
24512         * valtrack.c (dead_debug_insert_temp): Likewise.
24513         * varasm.c (mergeable_constant_section): Likewise.
24514         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
24516 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24517             Alan Hayward  <alan.hayward@arm.com>
24518             David Sherwood  <david.sherwood@arm.com>
24520         * expr.c (expand_assignment): Cope with polynomial mode sizes
24521         when assigning to a CONCAT.
24523 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24524             Alan Hayward  <alan.hayward@arm.com>
24525             David Sherwood  <david.sherwood@arm.com>
24527         * machmode.h (mode_precision): Change from unsigned short to
24528         poly_uint16_pod.
24529         (mode_to_precision): Return a poly_uint16 rather than an unsigned
24530         short.
24531         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
24532         or if measurement_type is not polynomial.
24533         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
24534         in which the mode is already known to be a scalar_int_mode.
24535         * genmodes.c (emit_mode_precision): Change the type of mode_precision
24536         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
24537         initializer.
24538         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24539         for GET_MODE_PRECISION.
24540         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24541         for GET_MODE_PRECISION.
24542         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
24543         as polynomial.
24544         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
24545         (expand_field_assignment, make_extraction): Likewise.
24546         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
24547         (get_last_value): Likewise.
24548         * convert.c (convert_to_integer_1): Likewise.
24549         * cse.c (cse_insn): Likewise.
24550         * expr.c (expand_expr_real_1): Likewise.
24551         * lra-constraints.c (simplify_operand_subreg): Likewise.
24552         * optabs-query.c (can_atomic_load_p): Likewise.
24553         * optabs.c (expand_atomic_load): Likewise.
24554         (expand_atomic_store): Likewise.
24555         * ree.c (combine_reaching_defs): Likewise.
24556         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
24557         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
24558         * tree.h (type_has_mode_precision_p): Likewise.
24559         * ubsan.c (instrument_si_overflow): Likewise.
24561 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24562             Alan Hayward  <alan.hayward@arm.com>
24563             David Sherwood  <david.sherwood@arm.com>
24565         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
24566         polynomial numbers of units.
24567         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
24568         (valid_vector_subparts_p): New function.
24569         (build_vector_type): Remove temporary shim and take the number
24570         of units as a poly_uint64 rather than an int.
24571         (build_opaque_vector_type): Take the number of units as a
24572         poly_uint64 rather than an int.
24573         * tree.c (build_vector_from_ctor): Handle polynomial
24574         TYPE_VECTOR_SUBPARTS.
24575         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
24576         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
24577         (build_vector_from_val): If the number of units is variable,
24578         use build_vec_duplicate_cst for constant operands and
24579         VEC_DUPLICATE_EXPR otherwise.
24580         (make_vector_type): Remove temporary is_constant ().
24581         (build_vector_type, build_opaque_vector_type): Take the number of
24582         units as a poly_uint64 rather than an int.
24583         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
24584         VECTOR_CST_NELTS.
24585         * cfgexpand.c (expand_debug_expr): Likewise.
24586         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
24587         (store_constructor, expand_expr_real_1): Likewise.
24588         (const_scalar_mask_from_tree): Likewise.
24589         * fold-const-call.c (fold_const_reduction): Likewise.
24590         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
24591         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
24592         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
24593         (fold_relational_const): Likewise.
24594         (native_interpret_vector): Likewise.  Change the size from an
24595         int to an unsigned int.
24596         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
24597         TYPE_VECTOR_SUBPARTS.
24598         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
24599         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
24600         duplicating a non-constant operand into a variable-length vector.
24601         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
24602         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
24603         * ipa-icf.c (sem_variable::equals): Likewise.
24604         * match.pd: Likewise.
24605         * omp-simd-clone.c (simd_clone_subparts): Likewise.
24606         * print-tree.c (print_node): Likewise.
24607         * stor-layout.c (layout_type): Likewise.
24608         * targhooks.c (default_builtin_vectorization_cost): Likewise.
24609         * tree-cfg.c (verify_gimple_comparison): Likewise.
24610         (verify_gimple_assign_binary): Likewise.
24611         (verify_gimple_assign_ternary): Likewise.
24612         (verify_gimple_assign_single): Likewise.
24613         * tree-pretty-print.c (dump_generic_node): Likewise.
24614         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
24615         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
24616         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
24617         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
24618         (vect_shift_permute_load_chain): Likewise.
24619         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
24620         (expand_vector_condition, optimize_vector_constructor): Likewise.
24621         (lower_vec_perm, get_compute_type): Likewise.
24622         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
24623         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
24624         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
24625         (vect_recog_mask_conversion_pattern): Likewise.
24626         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
24627         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
24628         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
24629         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
24630         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
24631         (vectorizable_shift, vectorizable_operation, vectorizable_store)
24632         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
24633         (supportable_widening_operation): Likewise.
24634         (supportable_narrowing_operation): Likewise.
24635         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
24636         Likewise.
24637         * varasm.c (output_constant): Likewise.
24639 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24640             Alan Hayward  <alan.hayward@arm.com>
24641             David Sherwood  <david.sherwood@arm.com>
24643         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
24644         so that both the length == 3 and length != 3 cases set up their
24645         own permute vectors.  Add comments explaining why we know the
24646         number of elements is constant.
24647         (vect_permute_load_chain): Likewise.
24649 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24650             Alan Hayward  <alan.hayward@arm.com>
24651             David Sherwood  <david.sherwood@arm.com>
24653         * machmode.h (mode_nunits): Change from unsigned char to
24654         poly_uint16_pod.
24655         (ONLY_FIXED_SIZE_MODES): New macro.
24656         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
24657         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
24658         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
24659         New typedefs.
24660         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
24661         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
24662         or if measurement_type is not polynomial.
24663         * genmodes.c (ZERO_COEFFS): New macro.
24664         (emit_mode_nunits_inline): Make mode_nunits_inline return a
24665         poly_uint16.
24666         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
24667         Use ZERO_COEFFS when emitting initializers.
24668         * data-streamer.h (bp_pack_poly_value): New function.
24669         (bp_unpack_poly_value): Likewise.
24670         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
24671         for GET_MODE_NUNITS.
24672         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
24673         for GET_MODE_NUNITS.
24674         * tree.c (make_vector_type): Remove temporary shim and make
24675         the real function take the number of units as a poly_uint64
24676         rather than an int.
24677         (build_vector_type_for_mode): Handle polynomial nunits.
24678         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
24679         * emit-rtl.c (const_vec_series_p_1): Likewise.
24680         (gen_rtx_CONST_VECTOR): Likewise.
24681         * fold-const.c (test_vec_duplicate_folding): Likewise.
24682         * genrecog.c (validate_pattern): Likewise.
24683         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
24684         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
24685         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
24686         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
24687         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
24688         * rtlanal.c (subreg_get_info): Likewise.
24689         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
24690         (vect_grouped_load_supported): Likewise.
24691         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
24692         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
24693         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
24694         (simplify_const_unary_operation, simplify_binary_operation_1)
24695         (simplify_const_binary_operation, simplify_ternary_operation)
24696         (test_vector_ops_duplicate, test_vector_ops): Likewise.
24697         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
24698         instead of CONST_VECTOR_NUNITS.
24699         * varasm.c (output_constant_pool_2): Likewise.
24700         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
24701         explicit-encoded elements in the XVEC for variable-length vectors.
24703 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24705         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
24707 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24708             Alan Hayward  <alan.hayward@arm.com>
24709             David Sherwood  <david.sherwood@arm.com>
24711         * coretypes.h (fixed_size_mode): Declare.
24712         (fixed_size_mode_pod): New typedef.
24713         * builtins.h (target_builtins::x_apply_args_mode)
24714         (target_builtins::x_apply_result_mode): Change type to
24715         fixed_size_mode_pod.
24716         * builtins.c (apply_args_size, apply_result_size, result_vector)
24717         (expand_builtin_apply_args_1, expand_builtin_apply)
24718         (expand_builtin_return): Update accordingly.
24720 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24722         * cse.c (hash_rtx_cb): Hash only the encoded elements.
24723         * cselib.c (cselib_hash_rtx): Likewise.
24724         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
24725         CONST_VECTOR encoding.
24727 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
24728             Jeff Law  <law@redhat.com>
24730         PR target/83641
24731         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
24732         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
24733         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
24734         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
24736         PR target/83641
24737         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
24738         explicitly probe *sp in a noreturn function if there were any callee
24739         register saves or frame pointer is needed.
24741 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
24743         PR debug/83621
24744         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24745         BLKmode for ternary, binary or unary expressions.
24747         PR debug/83645
24748         * var-tracking.c (delete_vta_debug_insn): New inline function.
24749         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
24750         insns from get_insns () to NULL instead of each bb separately.
24751         Use delete_vta_debug_insn.  No longer static.
24752         (vt_debug_insns_local, variable_tracking_main_1): Adjust
24753         delete_vta_debug_insns callers.
24754         * rtl.h (delete_vta_debug_insns): Declare.
24755         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
24756         instead of variable_tracking_main.
24758 2018-01-03  Martin Sebor  <msebor@redhat.com>
24760         PR tree-optimization/83603
24761         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
24762         arguments past the endof the argument list in functions declared
24763         without a prototype.
24764         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
24765         Avoid checking when arguments are null.
24767 2018-01-03  Martin Sebor  <msebor@redhat.com>
24769         PR c/83559
24770         * doc/extend.texi (attribute const): Fix a typo.
24771         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
24772         issuing -Wsuggest-attribute for void functions.
24774 2018-01-03  Martin Sebor  <msebor@redhat.com>
24776         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
24777         offset_int::from instead of wide_int::to_shwi.
24778         (maybe_diag_overlap): Remove assertion.
24779         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
24780         * gimple-ssa-sprintf.c (format_directive): Same.
24781         (parse_directive): Same.
24782         (sprintf_dom_walker::compute_format_length): Same.
24783         (try_substitute_return_value): Same.
24785 2018-01-03  Jeff Law  <law@redhat.com>
24787         PR middle-end/83654
24788         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
24789         non-constant residual for zero at runtime and avoid probing in
24790         that case.  Reorganize code for trailing problem to mirror handling
24791         of the residual.
24793 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
24795         PR tree-optimization/83501
24796         * tree-ssa-strlen.c (get_string_cst): New.
24797         (handle_char_store): Call get_string_cst.
24799 2018-01-03  Martin Liska  <mliska@suse.cz>
24801         PR tree-optimization/83593
24802         * tree-ssa-strlen.c: Include tree-cfg.h.
24803         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
24804         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
24805         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
24806         to false.
24807         (strlen_dom_walker::before_dom_children): Call
24808         gimple_purge_dead_eh_edges. Dump tranformation with details
24809         dump flags.
24810         (strlen_dom_walker::before_dom_children): Update call by adding
24811         new argument cleanup_eh.
24812         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
24814 2018-01-03  Martin Liska  <mliska@suse.cz>
24816         PR ipa/83549
24817         * cif-code.def (VARIADIC_THUNK): New enum value.
24818         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
24819         thunks.
24821 2018-01-03  Jan Beulich  <jbeulich@suse.com>
24823         * sse.md (mov<mode>_internal): Tighten condition for when to use
24824         vmovdqu<ssescalarsize> for TI and OI modes.
24826 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
24828         Update copyright years.
24830 2018-01-03  Martin Liska  <mliska@suse.cz>
24832         PR ipa/83594
24833         * ipa-visibility.c (function_and_variable_visibility): Skip
24834         functions with noipa attribure.
24836 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
24838         * gcc.c (process_command): Update copyright notice dates.
24839         * gcov-dump.c (print_version): Ditto.
24840         * gcov.c (print_version): Ditto.
24841         * gcov-tool.c (print_version): Ditto.
24842         * gengtype.c (create_file): Ditto.
24843         * doc/cpp.texi: Bump @copying's copyright year.
24844         * doc/cppinternals.texi: Ditto.
24845         * doc/gcc.texi: Ditto.
24846         * doc/gccint.texi: Ditto.
24847         * doc/gcov.texi: Ditto.
24848         * doc/install.texi: Ditto.
24849         * doc/invoke.texi: Ditto.
24851 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24853         * vector-builder.h (vector_builder::m_full_nelts): Change from
24854         unsigned int to poly_uint64.
24855         (vector_builder::full_nelts): Update prototype accordingly.
24856         (vector_builder::new_vector): Likewise.
24857         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
24858         (vector_builder::operator ==): Likewise.
24859         (vector_builder::finalize): Likewise.
24860         * int-vector-builder.h (int_vector_builder::int_vector_builder):
24861         Take the number of elements as a poly_uint64 rather than an
24862         unsigned int.
24863         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
24864         from unsigned int to poly_uint64.
24865         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
24866         (vec_perm_indices::new_vector): Likewise.
24867         (vec_perm_indices::length): Likewise.
24868         (vec_perm_indices::nelts_per_input): Likewise.
24869         (vec_perm_indices::input_nelts): Likewise.
24870         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
24871         number of elements per input as a poly_uint64 rather than an
24872         unsigned int.  Use the original encoding for variable-length
24873         vectors, rather than clamping each individual element.
24874         For the second and subsequent elements in each pattern,
24875         clamp the step and base before clamping their sum.
24876         (vec_perm_indices::series_p): Handle polynomial element counts.
24877         (vec_perm_indices::all_in_range_p): Likewise.
24878         (vec_perm_indices_to_tree): Likewise.
24879         (vec_perm_indices_to_rtx): Likewise.
24880         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
24881         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
24882         (tree_vector_builder::new_binary_operation): Handle polynomial
24883         element counts.  Return false if we need to know the number
24884         of elements at compile time.
24885         * fold-const.c (fold_vec_perm): Punt if the number of elements
24886         isn't known at compile time.
24888 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24890         * vec-perm-indices.h (vec_perm_builder): Change element type
24891         from HOST_WIDE_INT to poly_int64.
24892         (vec_perm_indices::element_type): Update accordingly.
24893         (vec_perm_indices::clamp): Handle polynomial element_types.
24894         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
24895         (vec_perm_indices::all_in_range_p): Likewise.
24896         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
24897         than shwi trees.
24898         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
24899         polynomial vec_perm_indices element types.
24900         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
24901         * fold-const.c (fold_vec_perm): Likewise.
24902         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
24903         * tree-vect-generic.c (lower_vec_perm): Likewise.
24904         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
24905         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
24906         element type to HOST_WIDE_INT.
24908 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24909             Alan Hayward  <alan.hayward@arm.com>
24910             David Sherwood  <david.sherwood@arm.com>
24912         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
24913         rather than an int.  Use plus_constant.
24914         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
24915         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
24917 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24918             Alan Hayward  <alan.hayward@arm.com>
24919             David Sherwood  <david.sherwood@arm.com>
24921         * calls.c (emit_call_1, expand_call): Change struct_value_size from
24922         a HOST_WIDE_INT to a poly_int64.
24924 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24925             Alan Hayward  <alan.hayward@arm.com>
24926             David Sherwood  <david.sherwood@arm.com>
24928         * calls.c (load_register_parameters): Cope with polynomial
24929         mode sizes.  Require a constant size for BLKmode parameters
24930         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
24931         forces a parameter to be padded at the lsb end in order to
24932         fill a complete number of words, require the parameter size
24933         to be ordered wrt UNITS_PER_WORD.
24935 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24936             Alan Hayward  <alan.hayward@arm.com>
24937             David Sherwood  <david.sherwood@arm.com>
24939         * reload1.c (spill_stack_slot_width): Change element type
24940         from unsigned int to poly_uint64_pod.
24941         (alter_reg): Treat mode sizes as polynomial.
24943 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24944             Alan Hayward  <alan.hayward@arm.com>
24945             David Sherwood  <david.sherwood@arm.com>
24947         * reload.c (complex_word_subreg_p): New function.
24948         (reload_inner_reg_of_subreg, push_reload): Use it.
24950 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24951             Alan Hayward  <alan.hayward@arm.com>
24952             David Sherwood  <david.sherwood@arm.com>
24954         * lra-constraints.c (process_alt_operands): Reject matched
24955         operands whose sizes aren't ordered.
24956         (match_reload): Refer to this check here.
24958 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24959             Alan Hayward  <alan.hayward@arm.com>
24960             David Sherwood  <david.sherwood@arm.com>
24962         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
24963         that the mode size is in the set {1, 2, 4, 8, 16}.
24965 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24966             Alan Hayward  <alan.hayward@arm.com>
24967             David Sherwood  <david.sherwood@arm.com>
24969         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
24970         Use plus_constant instead of gen_rtx_PLUS.
24972 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
24973             Alan Hayward  <alan.hayward@arm.com>
24974             David Sherwood  <david.sherwood@arm.com>
24976         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
24977         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
24978         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
24979         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
24980         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
24981         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
24982         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
24983         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
24984         * config/i386/i386.c (ix86_push_rounding): ...this new function.
24985         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
24986         a poly_int64.
24987         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
24988         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
24989         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
24990         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
24991         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
24992         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
24993         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
24994         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
24995         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
24996         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
24997         function.
24998         * expr.c (emit_move_resolve_push): Treat the input and result
24999         of PUSH_ROUNDING as a poly_int64.
25000         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
25001         (emit_push_insn): Likewise.
25002         * lra-eliminations.c (mark_not_eliminable): Likewise.
25003         * recog.c (push_operand): Likewise.
25004         * reload1.c (elimination_effects): Likewise.
25005         * rtlanal.c (nonzero_bits1): Likewise.
25006         * calls.c (store_one_arg): Likewise.  Require the padding to be
25007         known at compile time.
25009 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25010             Alan Hayward  <alan.hayward@arm.com>
25011             David Sherwood  <david.sherwood@arm.com>
25013         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
25014         Use plus_constant instead of gen_rtx_PLUS.
25016 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25017             Alan Hayward  <alan.hayward@arm.com>
25018             David Sherwood  <david.sherwood@arm.com>
25020         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
25021         rather than an int.
25023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25024             Alan Hayward  <alan.hayward@arm.com>
25025             David Sherwood  <david.sherwood@arm.com>
25027         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
25028         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
25029         via stack temporaries.  Treat the mode size as polynomial too.
25031 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25032             Alan Hayward  <alan.hayward@arm.com>
25033             David Sherwood  <david.sherwood@arm.com>
25035         * expr.c (expand_expr_real_2): When handling conversions involving
25036         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
25037         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
25038         as a poly_uint64 too.
25040 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25041             Alan Hayward  <alan.hayward@arm.com>
25042             David Sherwood  <david.sherwood@arm.com>
25044         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
25046 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25047             Alan Hayward  <alan.hayward@arm.com>
25048             David Sherwood  <david.sherwood@arm.com>
25050         * combine.c (can_change_dest_mode): Handle polynomial
25051         REGMODE_NATURAL_SIZE.
25052         * expmed.c (store_bit_field_1): Likewise.
25053         * expr.c (store_constructor): Likewise.
25054         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
25055         and polynomial REGMODE_NATURAL_SIZE.
25056         (gen_lowpart_common): Likewise.
25057         * reginfo.c (record_subregs_of_mode): Likewise.
25058         * rtlanal.c (read_modify_subreg_p): Likewise.
25060 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25061             Alan Hayward  <alan.hayward@arm.com>
25062             David Sherwood  <david.sherwood@arm.com>
25064         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
25065         numbers of elements.
25067 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25068             Alan Hayward  <alan.hayward@arm.com>
25069             David Sherwood  <david.sherwood@arm.com>
25071         * match.pd: Cope with polynomial numbers of vector elements.
25073 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25074             Alan Hayward  <alan.hayward@arm.com>
25075             David Sherwood  <david.sherwood@arm.com>
25077         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
25078         in a POINTER_PLUS_EXPR.
25080 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25081             Alan Hayward  <alan.hayward@arm.com>
25082             David Sherwood  <david.sherwood@arm.com>
25084         * omp-simd-clone.c (simd_clone_subparts): New function.
25085         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
25086         (ipa_simd_modify_function_body): Likewise.
25088 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25089             Alan Hayward  <alan.hayward@arm.com>
25090             David Sherwood  <david.sherwood@arm.com>
25092         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
25093         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
25094         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
25095         (expand_vector_condition, vector_element): Likewise.
25096         (subparts_gt): New function.
25097         (get_compute_type): Use subparts_gt.
25098         (count_type_subparts): Delete.
25099         (expand_vector_operations_1): Use subparts_gt instead of
25100         count_type_subparts.
25102 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25103             Alan Hayward  <alan.hayward@arm.com>
25104             David Sherwood  <david.sherwood@arm.com>
25106         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
25107         (vect_compile_time_alias): ...this new function.  Do the calculation
25108         on poly_ints rather than trees.
25109         (vect_prune_runtime_alias_test_list): Update call accordingly.
25111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25112             Alan Hayward  <alan.hayward@arm.com>
25113             David Sherwood  <david.sherwood@arm.com>
25115         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
25116         numbers of units.
25117         (vect_schedule_slp_instance): Likewise.
25119 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25120             Alan Hayward  <alan.hayward@arm.com>
25121             David Sherwood  <david.sherwood@arm.com>
25123         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
25124         constant and extern definitions for variable-length vectors.
25125         (vect_get_constant_vectors): Note that the number of units
25126         is known to be constant.
25128 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25129             Alan Hayward  <alan.hayward@arm.com>
25130             David Sherwood  <david.sherwood@arm.com>
25132         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
25133         of units as polynomial.  Choose between WIDE and NARROW based
25134         on multiple_p.
25136 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25137             Alan Hayward  <alan.hayward@arm.com>
25138             David Sherwood  <david.sherwood@arm.com>
25140         * tree-vect-stmts.c (simd_clone_subparts): New function.
25141         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
25143 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25144             Alan Hayward  <alan.hayward@arm.com>
25145             David Sherwood  <david.sherwood@arm.com>
25147         * tree-vect-stmts.c (vectorizable_call): Treat the number of
25148         vectors as polynomial.  Use build_index_vector for
25149         IFN_GOMP_SIMD_LANE.
25151 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25152             Alan Hayward  <alan.hayward@arm.com>
25153             David Sherwood  <david.sherwood@arm.com>
25155         * tree-vect-stmts.c (get_load_store_type): Treat the number of
25156         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
25157         for variable-length vectors.
25158         (vectorizable_mask_load_store): Treat the number of units as
25159         polynomial, asserting that it is constant if the condition has
25160         already been enforced.
25161         (vectorizable_store, vectorizable_load): Likewise.
25163 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25164             Alan Hayward  <alan.hayward@arm.com>
25165             David Sherwood  <david.sherwood@arm.com>
25167         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
25168         of units as polynomial.  Punt if we can't tell at compile time
25169         which vector contains the final result.
25171 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25172             Alan Hayward  <alan.hayward@arm.com>
25173             David Sherwood  <david.sherwood@arm.com>
25175         * tree-vect-loop.c (vectorizable_induction): Treat the number
25176         of units as polynomial.  Punt on SLP inductions.  Use an integer
25177         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
25178         cast of such a series for variable-length floating-point
25179         reductions.
25181 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25182             Alan Hayward  <alan.hayward@arm.com>
25183             David Sherwood  <david.sherwood@arm.com>
25185         * tree.h (build_index_vector): Declare.
25186         * tree.c (build_index_vector): New function.
25187         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
25188         of units as polynomial, forcibly converting it to a constant if
25189         vectorizable_reduction has already enforced the condition.
25190         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
25191         to create a {1,2,3,...} vector.
25192         (vectorizable_reduction): Treat the number of units as polynomial.
25193         Choose vectype_in based on the largest scalar element size rather
25194         than the smallest number of units.  Enforce the restrictions
25195         relied on above.
25197 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25198             Alan Hayward  <alan.hayward@arm.com>
25199             David Sherwood  <david.sherwood@arm.com>
25201         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
25202         number of units as polynomial.
25204 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25205             Alan Hayward  <alan.hayward@arm.com>
25206             David Sherwood  <david.sherwood@arm.com>
25208         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
25209         * target.def (autovectorize_vector_sizes): Return the vector sizes
25210         by pointer, using vector_sizes rather than a bitmask.
25211         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
25212         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
25213         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
25214         Likewise.
25215         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
25216         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
25217         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
25218         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
25219         * omp-general.c (omp_max_vf): Likewise.
25220         * omp-low.c (omp_clause_aligned_alignment): Likewise.
25221         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
25222         * tree-vect-loop.c (vect_analyze_loop): Likewise.
25223         * tree-vect-slp.c (vect_slp_bb): Likewise.
25224         * doc/tm.texi: Regenerate.
25225         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
25226         to a poly_uint64.
25227         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
25228         the vector size as a poly_uint64 rather than an unsigned int.
25229         (current_vector_size): Change from an unsigned int to a poly_uint64.
25230         (get_vectype_for_scalar_type): Update accordingly.
25231         * tree.h (build_truth_vector_type): Take the size and number of
25232         units as a poly_uint64 rather than an unsigned int.
25233         (build_vector_type): Add a temporary overload that takes
25234         the number of units as a poly_uint64 rather than an unsigned int.
25235         * tree.c (make_vector_type): Likewise.
25236         (build_truth_vector_type): Take the number of units as a poly_uint64
25237         rather than an unsigned int.
25239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25240             Alan Hayward  <alan.hayward@arm.com>
25241             David Sherwood  <david.sherwood@arm.com>
25243         * target.def (get_mask_mode): Take the number of units and length
25244         as poly_uint64s rather than unsigned ints.
25245         * targhooks.h (default_get_mask_mode): Update accordingly.
25246         * targhooks.c (default_get_mask_mode): Likewise.
25247         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
25248         * doc/tm.texi: Regenerate.
25250 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25251             Alan Hayward  <alan.hayward@arm.com>
25252             David Sherwood  <david.sherwood@arm.com>
25254         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
25255         * omp-general.c (omp_max_vf): Likewise.
25256         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
25257         (expand_omp_simd): Handle polynomial safelen.
25258         * omp-low.c (omplow_simd_context): Add a default constructor.
25259         (omplow_simd_context::max_vf): Change from int to poly_uint64.
25260         (lower_rec_simd_input_clauses): Update accordingly.
25261         (lower_rec_input_clauses): Likewise.
25263 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25264             Alan Hayward  <alan.hayward@arm.com>
25265             David Sherwood  <david.sherwood@arm.com>
25267         * tree-vectorizer.h (vect_nunits_for_cost): New function.
25268         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
25269         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
25270         (vect_analyze_slp_cost): Likewise.
25271         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
25272         (vect_model_load_cost): Likewise.
25274 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25275             Alan Hayward  <alan.hayward@arm.com>
25276             David Sherwood  <david.sherwood@arm.com>
25278         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
25279         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
25280         from an unsigned int * to a poly_uint64_pod *.
25281         (calculate_unrolling_factor): New function.
25282         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
25284 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25285             Alan Hayward  <alan.hayward@arm.com>
25286             David Sherwood  <david.sherwood@arm.com>
25288         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
25289         from an unsigned int to a poly_uint64.
25290         (_loop_vec_info::slp_unrolling_factor): Likewise.
25291         (_loop_vec_info::vectorization_factor): Change from an int
25292         to a poly_uint64.
25293         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
25294         (vect_get_num_vectors): New function.
25295         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
25296         (vect_get_num_copies): Use vect_get_num_vectors.
25297         (vect_analyze_data_ref_dependences): Change max_vf from an int *
25298         to an unsigned int *.
25299         (vect_analyze_data_refs): Change min_vf from an int * to a
25300         poly_uint64 *.
25301         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25302         than an unsigned HOST_WIDE_INT.
25303         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
25304         (vect_analyze_data_ref_dependence): Change max_vf from an int *
25305         to an unsigned int *.
25306         (vect_analyze_data_ref_dependences): Likewise.
25307         (vect_compute_data_ref_alignment): Handle polynomial vf.
25308         (vect_enhance_data_refs_alignment): Likewise.
25309         (vect_prune_runtime_alias_test_list): Likewise.
25310         (vect_shift_permute_load_chain): Likewise.
25311         (vect_supportable_dr_alignment): Likewise.
25312         (dependence_distance_ge_vf): Take the vectorization factor as a
25313         poly_uint64 rather than an unsigned HOST_WIDE_INT.
25314         (vect_analyze_data_refs): Change min_vf from an int * to a
25315         poly_uint64 *.
25316         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
25317         vfm1 as a poly_uint64 rather than an int.  Make the same change
25318         for the returned bound_scalar.
25319         (vect_gen_vector_loop_niters): Handle polynomial vf.
25320         (vect_do_peeling): Likewise.  Update call to
25321         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
25322         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
25323         be constant.
25324         * tree-vect-loop.c (vect_determine_vectorization_factor)
25325         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
25326         (vect_get_known_peeling_cost): Likewise.
25327         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
25328         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
25329         (vect_transform_loop): Likewise.  Use the lowest possible VF when
25330         updating the upper bounds of the loop.
25331         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
25332         rather than an int.
25333         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
25334         polynomial unroll factors.
25335         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
25336         (vect_make_slp_decision): Likewise.
25337         (vect_supported_load_permutation_p): Likewise, and polynomial
25338         vf too.
25339         (vect_analyze_slp_cost): Handle polynomial vf.
25340         (vect_slp_analyze_node_operations): Likewise.
25341         (vect_slp_analyze_bb_1): Likewise.
25342         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
25343         than an unsigned HOST_WIDE_INT.
25344         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
25345         (vectorizable_load): Handle polynomial vf.
25346         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
25347         a poly_uint64.
25348         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
25350 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25351             Alan Hayward  <alan.hayward@arm.com>
25352             David Sherwood  <david.sherwood@arm.com>
25354         * match.pd: Handle bit operations involving three constants
25355         and try to fold one pair.
25357 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25359         * tree-vect-loop-manip.c: Include gimple-fold.h.
25360         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
25361         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
25362         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
25363         Add a path that uses a step of VF instead of 1, but disable it
25364         for now.
25365         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
25366         and niters_no_overflow parameters.  Update calls to
25367         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
25368         Create a new SSA name if the latter choses to use a ste other
25369         than zero, and return it via niters_vector_mult_vf_var.
25370         * tree-vect-loop.c (vect_transform_loop): Update calls to
25371         vect_do_peeling, vect_gen_vector_loop_niters and
25372         slpeel_make_loop_iterate_ntimes.
25373         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
25374         (vect_gen_vector_loop_niters): Update declarations after above changes.
25376 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
25378         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
25379         128-bit round to integer instructions.
25380         (ceil<mode>2): Likewise.
25381         (btrunc<mode>2): Likewise.
25382         (round<mode>2): Likewise.
25384 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25386         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
25387         unaligned VSX load/store on P8/P9.
25388         (expand_block_clear): Allow the use of unaligned VSX
25389         load/store on P8/P9.
25391 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25393         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
25394         New function.
25395         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
25396         swap associated with both a load and a store.
25398 2018-01-02  Andrew Waterman  <andrew@sifive.com>
25400         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
25401         * config/riscv/riscv.md (clear_cache): Use it.
25403 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
25405         * web.c: Remove out-of-date comment.
25407 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25409         * expr.c (fixup_args_size_notes): Check that any existing
25410         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
25411         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
25412         (emit_single_push_insn): ...here.
25414 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25416         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
25417         (const_vector_encoded_nelts): New function.
25418         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
25419         (const_vector_int_elt, const_vector_elt): Declare.
25420         * emit-rtl.c (const_vector_int_elt_1): New function.
25421         (const_vector_elt): Likewise.
25422         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
25423         of CONST_VECTOR_ELT.
25425 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25427         * expr.c: Include rtx-vector-builder.h.
25428         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
25429         directly on the tree encoding.
25430         (const_vector_from_tree): Likewise.
25431         * optabs.c: Include rtx-vector-builder.h.
25432         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
25433         sequence of "u" values.
25434         * vec-perm-indices.c: Include rtx-vector-builder.h.
25435         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
25436         directly on the vec_perm_indices encoding.
25438 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25440         * doc/rtl.texi (const_vector): Describe new encoding scheme.
25441         * Makefile.in (OBJS): Add rtx-vector-builder.o.
25442         * rtx-vector-builder.h: New file.
25443         * rtx-vector-builder.c: Likewise.
25444         * rtl.h (rtx_def::u2): Add a const_vector field.
25445         (CONST_VECTOR_NPATTERNS): New macro.
25446         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
25447         (CONST_VECTOR_DUPLICATE_P): Likewise.
25448         (CONST_VECTOR_STEPPED_P): Likewise.
25449         (CONST_VECTOR_ENCODED_ELT): Likewise.
25450         (const_vec_duplicate_p): Check for a duplicated vector encoding.
25451         (unwrap_const_vec_duplicate): Likewise.
25452         (const_vec_series_p): Check for a non-duplicated vector encoding.
25453         Say that the function only returns true for integer vectors.
25454         * emit-rtl.c: Include rtx-vector-builder.h.
25455         (gen_const_vec_duplicate_1): Delete.
25456         (gen_const_vector): Call gen_const_vec_duplicate instead of
25457         gen_const_vec_duplicate_1.
25458         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
25459         (gen_const_vec_duplicate): Use rtx_vector_builder.
25460         (gen_const_vec_series): Likewise.
25461         (gen_rtx_CONST_VECTOR): Likewise.
25462         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
25463         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25464         Build a new vector rather than modifying a CONST_VECTOR in-place.
25465         (handle_special_swappables): Update call accordingly.
25466         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
25467         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
25468         Build a new vector rather than modifying a CONST_VECTOR in-place.
25469         (handle_special_swappables): Update call accordingly.
25471 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25473         * simplify-rtx.c (simplify_const_binary_operation): Use
25474         CONST_VECTOR_ELT instead of XVECEXP.
25476 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25478         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
25479         the selector elements to be different from the data elements
25480         if the selector is a VECTOR_CST.
25481         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
25482         ssizetype for the selector.
25484 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25486         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
25487         before testing each element individually.
25488         * tree-vect-generic.c (lower_vec_perm): Likewise.
25490 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25492         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
25493         * selftest-run-tests.c (selftest::run_tests): Call it.
25494         * vector-builder.h (vector_builder::operator ==): New function.
25495         (vector_builder::operator !=): Likewise.
25496         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
25497         (vec_perm_indices::all_from_input_p): New function.
25498         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
25499         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
25500         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
25501         instead of reading the VECTOR_CST directly.  Detect whether both
25502         vector inputs are the same before constructing the vec_perm_indices,
25503         and update the number of inputs argument accordingly.  Use the
25504         utility functions added above.  Only construct sel2 if we need to.
25506 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25508         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
25509         the broadcast of the low byte.
25510         (expand_mult_highpart): Use an explicit encoding for the permutes.
25511         * optabs-query.c (can_mult_highpart_p): Likewise.
25512         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25513         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25514         (vectorizable_bswap): Likewise.
25515         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
25516         explicit encoding for the power-of-2 permutes.
25517         (vect_permute_store_chain): Likewise.
25518         (vect_grouped_load_supported): Likewise.
25519         (vect_permute_load_chain): Likewise.
25521 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25523         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
25524         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
25525         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
25526         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
25527         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
25528         (vect_gen_perm_mask_any): Likewise.
25530 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25532         * int-vector-builder.h: New file.
25533         * vec-perm-indices.h: Include int-vector-builder.h.
25534         (vec_perm_indices): Redefine as an int_vector_builder.
25535         (auto_vec_perm_indices): Delete.
25536         (vec_perm_builder): Redefine as a stand-alone class.
25537         (vec_perm_indices::vec_perm_indices): New function.
25538         (vec_perm_indices::clamp): Likewise.
25539         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
25540         (vec_perm_indices::new_vector): New function.
25541         (vec_perm_indices::new_expanded_vector): Update for new
25542         vec_perm_indices class.
25543         (vec_perm_indices::rotate_inputs): New function.
25544         (vec_perm_indices::all_in_range_p): Operate directly on the
25545         encoded form, without computing elided elements.
25546         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
25547         encoding.  Update for new vec_perm_indices class.
25548         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
25549         the given vec_perm_builder.
25550         (expand_vec_perm_var): Update vec_perm_builder constructor.
25551         (expand_mult_highpart): Use vec_perm_builder instead of
25552         auto_vec_perm_indices.
25553         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
25554         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
25555         or double series encoding as appropriate.
25556         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
25557         vec_perm_indices instead of auto_vec_perm_indices.
25558         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25559         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25560         (vect_permute_store_chain): Likewise.
25561         (vect_grouped_load_supported): Likewise.
25562         (vect_permute_load_chain): Likewise.
25563         (vect_shift_permute_load_chain): Likewise.
25564         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25565         (vect_transform_slp_perm_load): Likewise.
25566         (vect_schedule_slp_instance): Likewise.
25567         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25568         (vectorizable_mask_load_store): Likewise.
25569         (vectorizable_bswap): Likewise.
25570         (vectorizable_store): Likewise.
25571         (vectorizable_load): Likewise.
25572         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
25573         vec_perm_indices instead of auto_vec_perm_indices.  Use
25574         tree_to_vec_perm_builder to read the vector from a tree.
25575         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
25576         vec_perm_builder instead of a vec_perm_indices.
25577         (have_whole_vector_shift): Use vec_perm_builder and
25578         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
25579         truncation to calc_vec_perm_mask_for_shift.
25580         (vect_create_epilog_for_reduction): Likewise.
25581         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
25582         from auto_vec_perm_indices to vec_perm_indices.
25583         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25584         instead of changing individual elements.
25585         (aarch64_vectorize_vec_perm_const): Use new_vector to install
25586         the vector in d.perm.
25587         * config/arm/arm.c (expand_vec_perm_d::perm): Change
25588         from auto_vec_perm_indices to vec_perm_indices.
25589         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
25590         instead of changing individual elements.
25591         (arm_vectorize_vec_perm_const): Use new_vector to install
25592         the vector in d.perm.
25593         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
25594         Update vec_perm_builder constructor.
25595         (rs6000_expand_interleave): Likewise.
25596         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
25597         (rs6000_expand_interleave): Likewise.
25599 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25601         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
25602         to qimode could truncate the indices.
25603         * optabs.c (expand_vec_perm_var): Likewise.
25605 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25607         * Makefile.in (OBJS): Add vec-perm-indices.o.
25608         * vec-perm-indices.h: New file.
25609         * vec-perm-indices.c: Likewise.
25610         * target.h (vec_perm_indices): Replace with a forward class
25611         declaration.
25612         (auto_vec_perm_indices): Move to vec-perm-indices.h.
25613         * optabs.h: Include vec-perm-indices.h.
25614         (expand_vec_perm): Delete.
25615         (selector_fits_mode_p, expand_vec_perm_var): Declare.
25616         (expand_vec_perm_const): Declare.
25617         * target.def (vec_perm_const_ok): Replace with...
25618         (vec_perm_const): ...this new hook.
25619         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
25620         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
25621         * doc/tm.texi: Regenerate.
25622         * optabs.def (vec_perm_const): Delete.
25623         * doc/md.texi (vec_perm_const): Likewise.
25624         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
25625         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
25626         expand_vec_perm for constant permutation vectors.  Assert that
25627         the mode of variable permutation vectors is the integer equivalent
25628         of the mode that is being permuted.
25629         * optabs-query.h (selector_fits_mode_p): Declare.
25630         * optabs-query.c: Include vec-perm-indices.h.
25631         (selector_fits_mode_p): New function.
25632         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
25633         is defined, instead of checking whether the vec_perm_const_optab
25634         exists.  Use targetm.vectorize.vec_perm_const instead of
25635         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
25636         fit in the vector mode before using a variable permute.
25637         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
25638         vec_perm_indices instead of an rtx.
25639         (expand_vec_perm): Replace with...
25640         (expand_vec_perm_const): ...this new function.  Take the selector
25641         as a vec_perm_indices rather than an rtx.  Also take the mode of
25642         the selector.  Update call to shift_amt_for_vec_perm_mask.
25643         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
25644         Use vec_perm_indices::new_expanded_vector to expand the original
25645         selector into bytes.  Check whether the indices fit in the vector
25646         mode before using a variable permute.
25647         (expand_vec_perm_var): Make global.
25648         (expand_mult_highpart): Use expand_vec_perm_const.
25649         * fold-const.c: Includes vec-perm-indices.h.
25650         * tree-ssa-forwprop.c: Likewise.
25651         * tree-vect-data-refs.c: Likewise.
25652         * tree-vect-generic.c: Likewise.
25653         * tree-vect-loop.c: Likewise.
25654         * tree-vect-slp.c: Likewise.
25655         * tree-vect-stmts.c: Likewise.
25656         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
25657         Delete.
25658         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
25659         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
25660         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
25661         (aarch64_vectorize_vec_perm_const): ...this new function.
25662         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25663         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25664         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
25665         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
25666         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25667         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25668         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
25669         into...
25670         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
25671         check for NEON modes.
25672         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
25673         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
25674         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
25675         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
25676         into...
25677         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
25678         the old VEC_PERM_CONST conditions.
25679         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
25680         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
25681         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
25682         (ia64_vectorize_vec_perm_const_ok): Merge into...
25683         (ia64_vectorize_vec_perm_const): ...this new function.
25684         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
25685         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
25686         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
25687         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
25688         * config/mips/mips.c (mips_expand_vec_perm_const)
25689         (mips_vectorize_vec_perm_const_ok): Merge into...
25690         (mips_vectorize_vec_perm_const): ...this new function.
25691         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
25692         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
25693         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
25694         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
25695         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
25696         (rs6000_expand_vec_perm_const): Delete.
25697         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
25698         Delete.
25699         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25700         (altivec_expand_vec_perm_const_le): Take each operand individually.
25701         Operate on constant selectors rather than rtxes.
25702         (altivec_expand_vec_perm_const): Likewise.  Update call to
25703         altivec_expand_vec_perm_const_le.
25704         (rs6000_expand_vec_perm_const): Delete.
25705         (rs6000_vectorize_vec_perm_const_ok): Delete.
25706         (rs6000_vectorize_vec_perm_const): New function.
25707         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25708         an element count and rtx array.
25709         (rs6000_expand_extract_even): Update call accordingly.
25710         (rs6000_expand_interleave): Likewise.
25711         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
25712         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
25713         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
25714         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
25715         (rs6000_expand_vec_perm_const): Delete.
25716         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
25717         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25718         (altivec_expand_vec_perm_const_le): Take each operand individually.
25719         Operate on constant selectors rather than rtxes.
25720         (altivec_expand_vec_perm_const): Likewise.  Update call to
25721         altivec_expand_vec_perm_const_le.
25722         (rs6000_expand_vec_perm_const): Delete.
25723         (rs6000_vectorize_vec_perm_const_ok): Delete.
25724         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
25725         reference to the SPE evmerge intructions.
25726         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
25727         an element count and rtx array.
25728         (rs6000_expand_extract_even): Update call accordingly.
25729         (rs6000_expand_interleave): Likewise.
25730         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
25731         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
25732         new function.
25733         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
25735 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25737         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
25738         vector mode and that that mode matches the mode of the data
25739         being permuted.
25740         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
25741         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
25742         directly using expand_vec_perm_1 when forcing selectors into
25743         registers.
25744         (expand_vec_perm_var): New function, split out from expand_vec_perm.
25746 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25748         * optabs-query.h (can_vec_perm_p): Delete.
25749         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
25750         * optabs-query.c (can_vec_perm_p): Split into...
25751         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
25752         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
25753         particular selector is valid.
25754         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25755         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25756         (vect_grouped_load_supported): Likewise.
25757         (vect_shift_permute_load_chain): Likewise.
25758         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
25759         (vect_transform_slp_perm_load): Likewise.
25760         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25761         (vectorizable_bswap): Likewise.
25762         (vect_gen_perm_mask_checked): Likewise.
25763         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
25764         implementations of variable permutation vectors into account
25765         when deciding which selector to use.
25766         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
25767         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
25768         with a false third argument.
25769         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
25770         to test whether the constant selector is valid and can_vec_perm_var_p
25771         to test whether a variable selector is valid.
25773 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25775         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
25776         * optabs-query.c (can_vec_perm_p): Likewise.
25777         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
25778         instead of vec_perm_indices.
25779         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
25780         (vect_gen_perm_mask_checked): Likewise,
25781         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
25782         (vect_gen_perm_mask_checked): Likewise,
25784 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
25786         * optabs-query.h (qimode_for_vec_perm): Declare.
25787         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
25788         (qimode_for_vec_perm): ...this new function.
25789         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
25791 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25793         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
25794         does not have a conditional at the top.
25796 2018-01-02  Richard Biener  <rguenther@suse.de>
25798         * ipa-inline.c (big_speedup_p): Fix expression.
25800 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
25802         PR target/81616
25803         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25804         for generic 4->6.
25806 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
25808         PR target/81616
25809         Generic tuning.
25810         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
25811         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
25812         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
25813         cond_taken_branch_cost 3->4.
25815 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
25817         PR tree-optimization/83581
25818         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
25819         TODO_cleanup_cfg if any changes have been made.
25821         PR middle-end/83608
25822         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
25823         convert_modes if target mode has the right side, but different mode
25824         class.
25826         PR middle-end/83609
25827         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
25828         last argument when extracting from CONCAT.  If either from_real or
25829         from_imag is NULL, use expansion through memory.  If result is not
25830         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
25831         the parts directly to inner mode, if even that fails, use expansion
25832         through memory.
25834         PR middle-end/83623
25835         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
25836         check for bswap in mode rather than HImode and use that in expand_unop
25837         too.
25839 Copyright (C) 2018 Free Software Foundation, Inc.
25841 Copying and distribution of this file, with or without modification,
25842 are permitted in any medium without royalty provided the copyright
25843 notice and this notice are preserved.