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>
29 * doc/invoke.texi (Link Options): Fix formatting and grammar.
31 2018-09-18 Richard Biener <rguenther@suse.de>
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
46 2018-09-17 David Malcolm <dmalcolm@redhat.com>
48 * diagnostic-show-locus.c (class layout_range): Add field
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
59 (gcc_rich_location::add_location_if_nearby): Use 0 for
61 * gcc-rich-location.h (text_range_label::get_text): Update for new
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
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
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
123 2018-09-17 Martin Jambor <mjambor@suse.cz>
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
140 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
142 * config/arc/arc.c: Object attributes for core4 not reflected
144 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
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>
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>
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>
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
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>
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
247 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
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
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>
290 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
293 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
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
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
322 2018-09-14 Richard Biener <rguenther@suse.de>
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
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>
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>
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>
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>
405 * config/aarch64/aarch64.md (*aarch64_bfxil):
407 * config/aarch64/constraints.md (Ulc): Define.
408 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
410 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
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>
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>
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>
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
483 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
485 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
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>
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
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
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
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>
631 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
632 to speculation_safe_value_not_needed.
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
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
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>
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.
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
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
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>
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>
751 * config/s390/s390-passes.def: New file.
752 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
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>
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".
785 (TARGET_SM35): New macro.
786 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
788 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
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>
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>
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>
849 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
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):
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):
870 2018-09-05 Martin Liska <mliska@suse.cz>
872 * doc/gcov.texi: Update documentation of humar
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>
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>
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>
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
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
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
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
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
1023 2018-09-03 Martin Liska <mliska@suse.cz>
1026 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
1027 * common/common-targhooks.c (default_get_valid_option_values):
1029 * common/common-targhooks.h (default_get_valid_option_values):
1031 * common/config/i386/i386-common.c: Move processor_target_table
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
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>
1053 * predict.c (set_even_probabilities): Add likely_edges
1054 argument and handle cases where we have precisely one
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
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
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"
1109 (vrp_can_optimize_bit_op): Likewise.
1110 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
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
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>
1163 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
1165 2018-08-31 Jakub Jelinek <jakub@redhat.com>
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
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>
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
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
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>
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>
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
1289 (c_getstr): Likewise.
1290 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
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
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>
1332 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
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
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
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>
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
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>
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
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
1453 2018-08-27 Jeff Law <law@redhat.com>
1454 PR tree-optimization/87110
1455 * tree-ssa-dse.c (compute_trims): Handle non-constant
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>
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):
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>
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>
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
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
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
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
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):
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):
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
1734 (switch_conversion::check_final_bb): Likewise.
1735 (switch_conversion::gather_default_values): Pass new argument
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
1742 (switch_decision_tree::try_switch_expansion): Use
1743 gimple_switch_default_edge.
1744 * tree-vrp.c (find_switch_asserts): Pass new argument
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):
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
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
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>
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
1821 2018-08-26 Jeff Law <law@redhat.com>
1823 * config/mips/frame-header-opt.c: Include "backend.h" rather than
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>
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
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
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
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>
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
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>
1972 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
1975 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
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
1986 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
1988 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
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'
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
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>
2047 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
2049 * config/arm/arm.c (speculation_barrier_libfunc): New static
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
2056 (speculation_barrier_insn): Only match on Armv7 or later.
2058 2018-08-23 Richard Biener <rguenther@suse.de>
2061 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
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
2071 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
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
2082 2018-08-22 Martin Sebor <msebor@redhat.com>
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
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>
2128 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
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>
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
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
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>
2198 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
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
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>
2262 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
2263 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
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>
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
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
2284 (emit_diagnostic): Add auto_diagnostic_group instance.
2286 (inform_n): Likewise.
2287 (warning): Likewise.
2288 (warning_at): Likewise.
2289 (warning_n): Likewise.
2290 (pedwarn): Likewise.
2291 (permerror): Likewise.
2293 (error_n): Likewise.
2294 (error_at): 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",
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>
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>
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
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>
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
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
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
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
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
2508 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
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>
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):
2562 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
2564 * config/rs6000/altivec.md: Don't set length attribute to the default
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>
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):
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
2649 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
2651 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
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
2659 (class layout): Add field "m_show_labels_p".
2660 (layout_range::layout_range): Add param "label" and use it to
2662 (make_range): Pass in NULL for new "label" param of layout_range's
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
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"
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
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>
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>
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>
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
2756 (expand_vec_perm_1): Try the new method.
2758 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
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()
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):
2814 (s390_decompose_address): Factor out constant pool ref
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
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
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):
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
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>
2884 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
2887 2018-08-10 Martin Liska <mliska@suse.cz>
2890 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
2892 * builtins.c (expand_builtin_expect_with_probability):
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):
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
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
2927 (strip_predictor_early): New function.
2928 (pass_strip_predict_hints::execute): Call the function to
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
2943 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
2945 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
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>
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>
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>
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>
3052 * config/s390/s390.c (s390_option_override_internal): Reduce the
3053 stack-clash-protection-probe-interval param if it would be too big
3056 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
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
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>
3082 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
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
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):
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):
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):
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
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>
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>
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
3195 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
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>
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
3238 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
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>
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
3382 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
3385 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3388 2018-08-03 Jeff Law <law@redhat.com>
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
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
3436 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
3438 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
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
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>
3480 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3481 Define to speculation_safe_value_not_needed.
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>
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>
3511 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3512 Define to speculation_safe_value_not_needed.
3515 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3516 Define to speculation_safe_value_not_needed.
3519 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3520 Define to speculation_safe_value_not_needed.
3523 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3524 Define to speculation_safe_value_not_needed.
3527 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3528 Define to speculation_safe_value_not_needed.
3531 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3532 Define to speculation_safe_value_not_needed.
3535 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3536 Define to speculation_safe_value_not_needed.
3539 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
3540 Define to speculation_safe_value_not_needed.
3543 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
3544 speculation_safe_value_not_needed.
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.
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
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>
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.
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):
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
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
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
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
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
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
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>
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
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>
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>
3840 * graphite.h: Include isl/id.h and isl/space.h to allow build
3843 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
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
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
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
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
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
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
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
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>
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
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
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
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
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
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
4097 (dump_context::dump_generic_expr_loc): Likewise, using
4099 (dump_context::dump_printf_loc_va): Likewise, using
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
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
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
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
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
4225 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
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
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
4271 * tree-vect-loop-manip.c (get_misalign_in_elems)
4272 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
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
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
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
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
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
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
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
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
4543 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
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
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
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
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
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
4681 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
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
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
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
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
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
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
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
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
4891 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
4893 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
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
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
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>
4996 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
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>
5014 * doc/extend.texi (Global Register Variables): Discourage use of type
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>
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
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
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
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
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
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>
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
5124 2018-07-26 Martin Liska <mliska@suse.cz>
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
5160 (arc_is_uncached_mem_p): uncached applies to both the variable and
5163 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
5165 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
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
5178 2018-07-25 Richard Biener <rguenther@suse.de>
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>
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>
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
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
5234 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
5235 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
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>
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>
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>
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>
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
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.
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
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
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
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
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
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.
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>
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
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
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
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
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",
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
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
5617 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
5618 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
5621 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
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
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.
5646 2018-07-19 Richard Biener <rguenther@suse.de>
5648 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
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>
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
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>
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
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
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>
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
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
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>
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
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
5829 * config/m68k/m68k.c: Do not use removed align_labels_value and
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
5847 * config/rx/rx.c (rx_align_for_label): Change return type
5849 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
5851 into align_flags class.
5852 (LABEL_ALIGN): Likewise.
5853 (LOOP_ALIGN): Likewise.
5854 * config/s390/s390.c (s390_label_align): Use align_flags
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>
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>
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>
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
6005 * config/s390/s390.opt: Add the new option.
6007 2018-07-16 Richard Biener <rguenther@suse.de>
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>
6016 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
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
6027 * config/arc/arc-cpus.def (hs4x): New cpu.
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,
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>
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
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>
6083 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
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
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>
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):
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>
6137 * match.pd (addr1 - addr2): Allow either of the operand to
6140 2018-07-13 Tom de Vries <tdevries@suse.de>
6142 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
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>
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>
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
6174 (riscv_get_interrupt_type): New function.
6175 (riscv_merge_decl_attributes): New function, checking interrupt type is
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
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>
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
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
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
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
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
6279 (gimple_match_op::set_op): Likewise.
6280 (gimple_resimplify5): Declare.
6281 * genmatch.c (decision_tree::gen): Generate simplifications for
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
6296 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
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
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
6338 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
6340 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
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
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>
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>
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>
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
6389 (vn_nary_build_or_lookup_1): Adjust.
6390 (try_to_simplify): Likewise.
6391 * gimple-match-head.c (gimple_resimplify1): Instead apply one
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
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
6418 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
6420 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
6422 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
6424 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
6426 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
6428 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
6430 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
6432 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
6434 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
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
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
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
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
6452 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
6453 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
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
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
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
6483 (write_ts_decl_non_common_tree_pointers): Do not stream
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>
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
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>
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>
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>
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.
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
6580 (zero_extendhisi2): Remove.
6581 (shift patterns): Add CC handling variants.
6585 (define_peephole2): New peephole to recognize mov that sets CC for
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,
6594 2018-07-09 Jakub Jelinek <jakub@redhat.com>
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>
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
6627 (combine_bound): Use wide-int overflow calculation instead of
6629 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
6631 * fold-const.c (int_const_binop_2): Same.
6632 (extract_muldiv_1): Same.
6633 (fold_div_compare): Same.
6634 (fold_abs_const): Same.
6636 * poly-int.h (add): 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>
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>
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>
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>
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>
6780 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6783 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6785 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6787 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
6790 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
6792 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
6794 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
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>
6815 * common.opt: Rename align options with 'str_' prefix.
6816 * common/config/i386/i386-common.c (set_malign_value): New
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
6823 * config/alpha/alpha.c (alpha_override_options_after_change):
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
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
6857 (mips_option_override): Set default values for options.
6858 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
6860 * config/rs6000/rs6000.c (rs6000_option_override_internal):
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
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
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
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
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
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
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
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>
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
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>
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
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>
7031 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
7033 2018-07-03 Richard Biener <rguenther@suse.de>
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
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
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
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.
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
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
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>
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>
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
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.
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.
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,
7362 (vect_recog_rotate_pattern): ...here.
7363 (vect_convert_input): Try to insert casts of invariants in the
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
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
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
7415 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
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
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
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>
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
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>
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
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"
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
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
7622 (bit_test_cluster::is_beneficial): Bail out for such case.
7623 (switch_decision_tree::analyze_switch_statement):
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
7632 * opt-suggestions.c (option_proposer::get_completions):
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.
7704 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
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
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.
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
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
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
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
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
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.
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>
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
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>
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
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
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
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
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
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
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
7991 (compute_branch_probabilities): Likewise.
7992 * selftest-run-tests.c (selftest::run_tests): Call
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):
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
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
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
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>
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
8050 2018-06-26 Richard Biener <rguenther@suse.de>
8052 PR tree-optimization/86287
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>
8060 * fold-const.c (fold_convertible_p): Pointer extension
8063 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
8066 * dwarf2out.c (loc_list_has_views): Adjust comments.
8067 (dw_loc_list): Split single cross-partition range with
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
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
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):
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>
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
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
8188 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
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>
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>
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>
8260 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
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):
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>
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
8315 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
8317 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
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
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
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
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
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
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
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,
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,
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,
8444 (omp_clause_code_name): Likewise.
8446 2018-06-20 Jakub Jelinek <jakub@redhat.com>
8449 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
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
8481 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
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):
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
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.
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
8692 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
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
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
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
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
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
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>
8813 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
8814 ieee128 argument takes up only one (vector) register, not two (floating
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>
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"
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
8854 (vax_output_int_subtract): Likewise.
8855 * config/vax/vax.c (vax_output_int_add): Likewise for param
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
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>
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.
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
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
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>
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
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>
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>
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>
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):
9072 2018-06-15 Jakub Jelinek <jakub@redhat.com>
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.
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.
9115 * doc/invoke.text (-mcrc): Document.
9118 2018-06-15 Nick Clifton <nickc@redhat.com>
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>
9147 * config/arc/arc.c (arc_return_address_register): Fix
9150 2018-06-15 Richard Biener <rguenther@suse.de>
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>
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>
9173 * match.pd ((A +- CST1) +- CST2): Punt if last resort
9174 unsigned_type_for returns NULL.
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>
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>
9190 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
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
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
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
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
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"
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"
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
9274 (emit_insn_after_1): Strengthen "after" param from "rtx" to
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
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
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
9314 (next_real_insn, remove_insn): Strengthen param from "rtx" to
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
9343 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
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
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
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>
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
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
9424 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
9426 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
9427 (simplify_binary_operation_1): Extend CONST_INT handling to
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
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>
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
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>
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>
9495 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
9497 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
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
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
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
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.
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>
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>
9661 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
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
9673 2018-06-08 Martin Liska <mliska@suse.cz>
9675 * tree-cfg.h (debug_function): Fix argument type to match
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
9693 (optinfo_verbosity_options): Likewise.
9694 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
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
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
9709 (opt_info_switch_p): Convert optgroup_flags from int to
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
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
9738 (opt_info_enable_passes): Likewise.
9739 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
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.
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
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
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
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
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
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
9904 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
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
9941 (hsa_function_summary::hsa_function_summary): Use the new enum
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
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.
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
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.
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.
10115 * config/i386/i386-builtin.def (BDESC_END): Likewise.
10116 (BDESC_FIRST): 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
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
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
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
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.
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
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>
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>
10518 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
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
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
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>
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
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
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
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
10621 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
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
10638 2018-06-04 Jim Wilson <jimw@sifive.com>
10640 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
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
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>
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>
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, -].
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
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
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
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
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.
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
10848 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
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
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
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>
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
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
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>
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>
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>
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.
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
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
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
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>
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
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
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>
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}
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>
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
11254 2018-05-29 Jakub Jelinek <jakub@redhat.com>
11256 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
11258 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
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
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
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
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
11302 (vec_pack<floatprefix>_float_<mode>): New expander.
11303 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
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}):
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
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
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
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
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
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
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>
11437 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
11439 * config/i386/sse.md
11440 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
11442 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
11444 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
11446 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
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
11456 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
11458 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11460 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
11462 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
11464 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
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>
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.
11507 (riscv_frflags, riscv_fsflags): Use tab after opcode.
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>
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,
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
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
11550 (vect_analyze_data_refs): Do not give up on not analyzable
11551 DRs for BB vectorization. Remove code truncating the dataref
11554 2018-05-25 Jakub Jelinek <jakub@redhat.com>
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
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>
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):
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
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):
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):
11693 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
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.
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
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
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
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
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>
11781 * varasm.c (assemble_alias): Check ifunc_resolver only on
11784 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
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
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>
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
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>
11880 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
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
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
11934 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
11936 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
11938 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
11939 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
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
11979 (slsr_process_copy): Likewise.
11980 (dump_candidate): Dump first_interp field.
11981 (replace_mult_candidate): Process all interpretations, not just
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):
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>
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
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
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>
12036 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
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
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
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):
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
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):
12094 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
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):
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):
12113 2018-05-22 Martin Sebor <msebor@redhat.com>
12116 * calls.c (maybe_warn_nonstring_arg): Use string length to set
12117 or ajust the presumed bound on an operation to avoid unnecessary
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>
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>
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
12163 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
12165 (IFKF_reg): New attributes to give the register constraints for
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>
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.
12191 (sur): Add mappings for the above.
12192 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
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
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,
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
12222 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
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
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
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
12349 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
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
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
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>
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>
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.
12444 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
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):
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
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):
12499 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
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
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
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
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
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
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
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.
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>
12731 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
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>
12782 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
12783 even if the mask is not all ones.
12786 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
12788 (ix86_gimple_fold_builtin): Likewise.
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
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
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
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
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,
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>
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):
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>
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
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.
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
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
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
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):
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
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
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):
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>
13209 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
13212 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
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
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>
13250 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
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>
13269 * doc/extend.texi: Document LLVM style format for no_sanitize
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>
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
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
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):
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>
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
13396 (vect_bb_vectorization_profitable_p): Adjust. Compute
13397 actual scalar cost using the cost vector and the add_stmt_cost
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,
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>
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>
13458 * doc/extend.texi (PowerPC Built-in Functions): Rename this
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):
13464 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13466 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13468 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13470 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
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
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):
13487 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
13489 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
13491 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
13493 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
13496 2018-05-08 Jakub Jelinek <jakub@redhat.com>
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>
13527 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
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.
13534 * config/i386/i386.c (ix86_fold_builtin): Handle
13535 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
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>
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
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
13584 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
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>
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
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
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
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>
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
13700 2018-05-05 Roland McGrath <mcgrathr@google.com>
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>
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
13742 * config.gcc (powerpc-xilinx-eabi*): Remove.
13743 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
13745 (fusion_addis_mem_combo_load): Ditto.
13746 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
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
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>
13799 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
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
13819 (fold_negate_expr_1): Likewise.
13821 2018-05-04 Jakub Jelinek <jakub@redhat.com>
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>
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>
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
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>
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>
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
13901 2018-05-02 Tom de Vries <tom@codesourcery.com>
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>
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
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
13928 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
13929 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
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>
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
13965 2018-05-02 Tom de Vries <tom@codesourcery.com>
13968 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
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>
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>
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.
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>
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
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
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
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
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"
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
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):
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
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
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
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
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
14229 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
14231 2018-04-30 Richard Biener <rguenther@suse.de>
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>
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>
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
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
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>
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.
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>
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>
14375 * config/nvptx/nvptx.c (needs_neutering_p): New function.
14376 (nvptx_single): Use needs_neutering_p to skip over insns that do not
14379 2018-04-26 Richard Biener <rguenther@suse.de>
14380 Tom de Vries <tom@codesourcery.com>
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
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
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>
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>
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>
14490 * common/config/i386/i386-common.c (ix86_handle_option): Don't
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>
14499 * doc/install.texi: Remove -mcet from bootstrap-cet.
14501 2018-04-24 Jakub Jelinek <jakub@redhat.com>
14504 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
14505 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
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>
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>
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>
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.
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):
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
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>
14643 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
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>
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>
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>
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
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>
14680 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
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
14686 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
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.
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>
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>
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
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>
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>
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>
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
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
14763 (incssp<mode>): Likewise.
14764 (nop_endbr): Likewise.
14765 * config/i386/i386.opt (mcet): Change help message to built-in
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
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
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
14832 2018-04-18 Martin Liska <mliska@suse.cz>
14834 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
14837 2018-04-18 Martin Liska <mliska@suse.cz>
14840 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
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>
14849 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
14852 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
14855 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
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>
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>
14872 * config.gcc: Obsolete powerpc*-*-*spe*.
14874 2018-04-17 Jakub Jelinek <jakub@redhat.com>
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
14882 2018-04-17 Jim Wilson <jimw@sifive.com>
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.
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):
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
14912 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
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"
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"
14932 (vec_extract_hi_v32hi): Merge all alternatives into one, use
14933 "sselog1" type instead of "sselog". Remove explicit "memory"
14935 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
14936 use "sselog1" type instead of "sselog". Remove explicit "memory"
14938 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
14939 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
14941 (vec_extract_hi_v64qi): Merge all alternatives into one, use
14942 "sselog1" type instead of "sselog". Remove explicit "memory"
14944 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
14945 use "sselog1" type instead of "sselog". Remove explicit "memory"
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
14955 2018-04-17 Martin Jambor <mjambor@suse.cz>
14958 * ipa-cp.c (create_specialized_node): Call
14959 expand_all_artificial_thunks if necessary.
14961 2018-04-17 Martin Liska <mliska@suse.cz>
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>
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
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
14991 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
14993 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
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
14998 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
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>
15018 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
15020 2018-04-17 Martin Liska <mliska@suse.cz>
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
15028 (expand_target_clones): Handle it.
15029 (ipa_target_clone): Make redirection just for target_clones
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>
15042 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
15043 vec_extract expression as having side effects to make sure it gets
15046 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
15049 * config/i386/i386.c (get_builtin_code_for_version): Check
15052 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
15055 * config.gcc: Support "skylake".
15056 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
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>
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>
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>
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>
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
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
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>
15133 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
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>
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>
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>
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
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>
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
15204 (dwarf2out_early_finish): Likewise.
15206 2018-04-12 Jakub Jelinek <jakub@redhat.com>
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>
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>
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>
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>
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>
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>
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>
15323 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
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
15330 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
15333 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
15336 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
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>
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
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>
15358 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
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>
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>
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>
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;
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
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>
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
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>
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
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
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
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
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
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
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
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
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>
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
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>
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>
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
15840 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
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
15851 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15853 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
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
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
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
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>
15910 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
15912 2018-04-05 Tom de Vries <tom@codesourcery.com>
15915 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
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
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>
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>
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
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
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>
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>
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
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
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>
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>
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>
16230 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
16232 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
16235 2018-03-30 Julia Koval <julia.koval@intel.com>
16238 * x86-tune.def (movx, partial_reg_dependency): Enable for
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>
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>.
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
16264 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
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
16276 (DIVDEU): Likewise.
16278 2018-03-28 Carl Love <cel@us.ibm.com>
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
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>
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
16309 (__builtin_divweuo): Likewise.
16310 (__builtin_divdeo): Likewise.
16311 (__builtin_divdeuo): Likewise.
16313 2018-03-28 Jakub Jelinek <jakub@redhat.com>
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>
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>
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>
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>
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>
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>
16369 * config/arm/arm.c (arm_valid_symbolic_address): Handle
16370 arm_word_relocations.
16372 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
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):
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>
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
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>
16437 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
16438 Clean up attributes.
16440 2018-03-23 Richard Biener <rguenther@suse.de>
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.
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):
16467 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
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>
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>
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>
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
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>
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>
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>
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>
16601 * Minor grammar fixes for x86 options.
16603 2018-03-20 Jakub Jelinek <jakub@redhat.com>
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>
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>
16632 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
16635 2018-03-20 Richard Biener <rguenther@suse.de>
16638 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
16639 sign-conversions as zero, fall back to standard scalar_stmt
16642 2018-03-20 Martin Liska <mliska@suse.cz>
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>
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.
16657 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
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]
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>
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>
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>
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>
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.
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.
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>
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>
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>
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
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>
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
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>
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>
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.
16829 * hsa-gen.c (mem_type_for_type): Fix comment typo.
16830 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
16832 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
16835 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
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>
16846 * multiple_target.c (create_dispatcher_calls): Redirect
16847 reference in the symbol table.
16849 2018-03-16 Martin Liska <mliska@suse.cz>
16852 * multiple_target.c (create_dispatcher_calls): Redirect also
16855 2018-03-16 Jakub Jelinek <jakub@redhat.com>
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
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
16883 2018-03-15 Jakub Jelinek <jakub@redhat.com>
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>
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
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>
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>
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>
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>
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>
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
16965 * config/doc/extend.texi: Add prototypes for vec_permxor.
16967 2018-03-14 David Malcolm <dmalcolm@redhat.com>
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
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.
17020 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
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.
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
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>
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.
17053 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
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>
17070 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
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
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
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>
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
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
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
17144 (*<su>mul<mode>3_highpart): New define_insn.
17146 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17149 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17152 2018-03-13 Martin Liska <mliska@suse.cz>
17155 * (sem_item_optimizer::sem_item_optimizer): Initialize new
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
17176 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
17177 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
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.
17185 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
17186 on the last operand.
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>
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
17279 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
17280 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
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
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>
17309 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
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
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
17329 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
17332 * config/i386/i386.opt: Replace Enforcment with Enforcement.
17334 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
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
17354 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
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>
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.
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):
17379 (builtin_access::generic_overlap): Be prepared to handle non-array
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.
17392 * recog.c (preprocess_constraints): Test address_operand for
17393 CT_ADDRESS constraints.
17395 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
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
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
17415 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
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>
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.
17443 PR tree-optimization/84178
17444 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
17446 (version_loop_for_if_conversion): Delay update_ssa call.
17447 (tree_if_conversion): Delay update_ssa until after predicate
17450 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
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>
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>
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
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
17484 (version_loop_for_if_conversion): Delay update_ssa call.
17485 (tree_if_conversion): Delay update_ssa until after predicate
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>
17517 * dwarf2out.c (struct dw_line_info_table): Update comments for
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>
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>
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>
17593 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
17594 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
17597 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
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
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>
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):
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
17653 PR tree-optimization/84687
17654 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
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>
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
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>
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.
17685 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
17687 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
17689 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
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
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
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
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
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
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
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
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,
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>
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.
17922 2018-03-02 Jakub Jelinek <jakub@redhat.com>
17925 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
17926 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
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>
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>
17957 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
17959 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
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
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
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
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>
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>
18006 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
18007 volatile arguments.
18009 2018-03-01 Richard Biener <rguenther@suse.de>
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
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
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)
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>
18080 * timevar.c (timer::print): Fix format to properly print 100%
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
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>
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>
18160 * config/i386/constraints.md (Bs): Replace
18161 ix86_indirect_branch_register with
18162 TARGET_INDIRECT_BRANCH_REGISTER.
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>
18203 * final.c (rest_of_clean_state): Also look for calls inside sequences.
18205 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
18208 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
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>
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>
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
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
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>
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
18290 * lra-constraints.c (curr_insn_transform): Set up
18291 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
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>
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.
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>
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>
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>
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>
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
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>
18415 * ipa-cp.c (determine_versionability): Do not version functions caling
18418 2018-02-21 Martin Liska <mliska@suse.cz>
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>
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
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
18447 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
18448 (gimple_fold_builtin_strlen): Pass true as last argument to
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
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
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>
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>
18522 * final.c (shorten_branches): Build align_tab array with one
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>
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
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
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):
18569 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
18570 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
18573 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
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
18581 2018-02-19 Martin Liska <mliska@suse.cz>
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
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
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>
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
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>
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
18660 2018-02-16 Carl Love <cel@us.ibm.com>
18662 * config/rs6000/altivec.h: Add builtin names vec_extract4b
18664 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
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>
18689 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
18692 2018-02-16 Richard Biener <rguenther@suse.de>
18694 PR tree-optimization/84037
18695 PR tree-optimization/84016
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
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>
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>
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.
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
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
18743 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
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
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>
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>
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>
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>
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>
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
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
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>
18872 * machmode.def: Define a complex mode for PARTIAL_INT.
18873 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
18875 * doc/rtl.texi: Document CSPImode.
18876 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
18878 (msp430_hard_regno_nregs_with_padding): Likewise.
18880 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
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>
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
18913 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
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
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>
18963 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
18965 2018-02-13 Martin Jambor <mjambor@suse.cz>
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>
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.
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
19007 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
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.
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>
19025 * config/sh/sh.c (find_barrier): Consider a sibling call
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
19040 * tree-vect-stmts.c (vect_model_simple_cost): Make early
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
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>
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
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>
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
19150 2018-02-10 Alan Modra <amodra@gmail.com>
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>
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>
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>
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
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>
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>
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>
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>
19248 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
19249 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
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
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>
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
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
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
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>
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
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>
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):
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
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
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
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
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
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>
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
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
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
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>
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>
19673 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
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
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>
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>
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>
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
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>
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>
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
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
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>
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>
19882 * config/rs6000/driver-rs6000.c: #include "diagnostic.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>
19893 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
19895 2018-02-01 Jeff Law <law@redhat.com>
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):
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
19928 2018-02-01 Renlin Li <renlin.li@arm.com>
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
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
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
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,
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>
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>
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>
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>
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>
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
20076 (arc_in_small_data_p): Do not consider for small data the uncached
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>
20087 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
20088 falign-loops=): Add Optimization flag.
20090 2018-01-30 Jeff Law <law@redhat.com>
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>
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>
20109 * lra-constraints.c (curr_insn_transform): Process AND in the
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
20125 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20128 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
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
20139 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
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
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>
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
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
20186 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
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>
20200 * tree-vect-stmts.c (vectorizable_call): Don't call
20201 targetm.vectorize_builtin_md_vectorized_function if callee is
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
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
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>
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.
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
20293 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
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
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
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
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.
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
20396 (subsf3_fpx): Likewise.
20397 (mulsf3_fpx): Likewise.
20398 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
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>
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
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
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>
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>
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>
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
20488 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
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
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
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>
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>
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) .. }
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
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>
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
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.
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
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
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>
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>
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
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
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>
20752 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
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>
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>
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>
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.
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>
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):
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>
20912 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
20914 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
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
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
20972 2018-01-18 Jakub Jelinek <jakub@redhat.com>
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>
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>
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>
21004 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
21007 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
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
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
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>
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>
21083 * ipa-inline.c (flatten_function): Do not overwrite final inlining
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>
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):
21116 * config/aarch64/constraints.md (aarch64_movti_operand):
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
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>
21150 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
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
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
21176 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
21178 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
21181 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
21184 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
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
21230 2018-01-16 Artyom Skrobov tyomitch@gmail.com
21232 * caller-save.c (insert_save): Drop unnecessary parameter. All
21235 2018-01-16 Jakub Jelinek <jakub@redhat.com>
21236 Richard Biener <rguenth@suse.de>
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>
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
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>
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
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.
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.
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>
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>
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>
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>
21453 * builtins.c (check_access): Avoid warning when the no-warning bit
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>
21464 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
21465 Use .pushsection/.popsection.
21467 2018-01-14 Martin Sebor <msebor@redhat.com>
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.
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
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
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
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.
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>
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
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):
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
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
21657 (runtime_alias_check_p): Allow runtime alias checks with
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
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
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
21693 (vect_loop_versioning): Call it.
21694 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
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
21705 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
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
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
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
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
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
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
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
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
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,
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
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):
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
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
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,
22047 (vect_analyze_loop_2): ...here. Don't check the vectorization
22048 factor against the number of loop iterations if the loop is
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
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
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
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
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
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
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
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>):
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
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
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
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
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
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
22477 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
22478 does not overlap dest if the function is frame-related. Handle
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
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
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
22544 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
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
22565 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
22566 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
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
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
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
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
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
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
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>
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
22725 (load_multiple): Delete.
22732 (store_multiple): 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.
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>
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>
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
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
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>
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
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>
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>
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
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>
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>
22922 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
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>
22930 * ipa-devirt.c (final_warning_record::grow_type_warnings):
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):
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
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
22973 (speculation_barrier): New define_insn.
22974 * doc/extend.texi: Document __builtin_speculation_barrier.
22976 2018-01-11 Jakub Jelinek <jakub@redhat.com>
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
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>
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>
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>):
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
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
23074 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
23075 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
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
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>
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
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
23135 (aarch64_builtin_vectorized_function): Check for a constant number
23137 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
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
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
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>
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>
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
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
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.
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):
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
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
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
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
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
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>
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>
23608 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
23610 * tree.c (tree_nop_conversion): Return true for location wrapper
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>
23625 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
23626 stack_realign_offset for the largest alignment of stack slot
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
23640 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
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>
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
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>
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>
23669 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
23670 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
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.
23685 * common.md (E, F): Use CONSTANT_P instead of checking for
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,
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
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>
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
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
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,
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>
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().
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,
23778 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
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>
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>
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
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
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>
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
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
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
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
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>
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
23931 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
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
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
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
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>
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>
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
24027 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
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
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.
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
24070 2018-01-05 Sudakshina Das <sudi.das@arm.com>
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
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>
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>
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.
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>
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>
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>
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>
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>
24158 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
24161 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
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
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>
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,
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,
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,
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
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
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
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,
24289 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
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
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
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
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
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
24374 * lto-streamer-in.c (lto_input_mode_table): Likewise.
24375 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
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
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
24407 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
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
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
24529 (mode_to_precision): Return a poly_uint16 rather than an unsigned
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
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
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
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):
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
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):
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
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>
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.
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>
24744 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
24745 BLKmode for ternary, binary or unary expressions.
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>
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
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
24807 (strlen_dom_walker::before_dom_children): Call
24808 gimple_purge_dead_eh_edges. Dump tranformation with details
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>
24817 * cif-code.def (VARIADIC_THUNK): New enum value.
24818 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
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>
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
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
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
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
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
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
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
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
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):
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
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
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
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
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
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
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
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
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):
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
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):
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
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
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):
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
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
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
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>
25803 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
25806 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
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
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
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
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.