1 2018-10-17 David Malcolm <dmalcolm@redhat.com>
3 * Makefile.in (SELFTEST_TARGETS): New.
4 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
5 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
6 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
8 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
9 (selftest-c++-gdb, selftest-c++-valgrind): Move to
11 * configure: Regenerate.
12 * configure.ac (selftest_languages): New.
14 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
16 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
17 overflow wraps argument.
18 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
19 wide_int_range_multiplicative_op.
20 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
21 overflow wraps argument.
22 (wide_int_range_multiplicative_op): Same.
23 (wide_int_range_lshift): Same.
24 (wide_int_range_div): Same.
25 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
26 (wide_int_range_lshift): Same.
27 (wide_int_range_div): Same.
29 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
31 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
33 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
34 wide_int_range_shift_undefined_p.
36 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
38 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
40 (@despeculate_copy<ALLI_TI:mode>): ... This.
41 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
44 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
46 * config.gcc: Obsolete *-*-solaris2.10*.
47 * doc/install.texi (Specific, *-*-solaris2*): Document it.
49 2018-10-12 Jeff Law <law@redhat.com>
51 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
52 reg + sym +- const_int addressing modes.
54 2018-10-15 David Malcolm <dmalcolm@redhat.com>
56 * common.opt (fdiagnostics-minimum-margin-width=): New option.
57 * diagnostic-show-locus.c (layout::layout): Apply the minimum
59 (layout::start_annotation_line): Only print up to 3 of the
60 margin character, to avoid touching the left-hand side.
61 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
62 minimum margin width, as set by test_diagnostic_context's ctor.
63 (selftest::test_fixit_insert_containing_newline): Likewise.
64 (selftest::test_fixit_insert_containing_newline_2): Likewise.
65 (selftest::test_line_numbers_multiline_range): Clear
67 * diagnostic.c (diagnostic_initialize): Initialize
69 * diagnostic.h (struct diagnostic_context): Add field
71 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
72 * opts.c (common_handle_option): Handle
73 OPT_fdiagnostics_minimum_margin_width_.
74 * selftest-diagnostic.c
75 (selftest::test_diagnostic_context::test_diagnostic_context):
76 Initialize min_margin_width to 6.
77 * toplev.c (general_init): Initialize global_dc->min_margin_width.
79 2018-10-15 David Malcolm <dmalcolm@redhat.com>
81 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
82 Fix usage of "error_at_rich_loc" in the comment.
84 2018-10-15 Renlin Li <renlin.li@arm.com>
87 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
88 if-conversioned loop when it contains ifn with types not
90 * internal-fn.c (expand_direct_optab_fn): Add an assert.
91 (direct_internal_fn_supported_p): New helper function.
92 * internal-fn.h (direct_internal_fn_supported_p): Declare.
94 2018-10-15 Jakub Jelinek <jakub@redhat.com>
97 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
98 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
99 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
101 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
103 PR tree-optimization/87022
104 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
105 bits in dist vector rather than the first one.
107 2018-10-15 Richard Biener <rguenther@suse.de>
110 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
111 (visit_loadstore): When a used restrict tag escaped verify that
112 the points-to solution of "other" pointers do not include
114 (compute_dependence_clique): If a used restrict tag escaped
115 communicated that down to visit_loadstore.
117 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
119 * config/s390/s390.c (s390_expand_vec_init): Force vector element
120 into reg if it isn't a general operand.
122 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
125 * config/i386/sse.md (*vec_dupv2di): Add register source to
128 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
131 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
132 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
133 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
134 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
136 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
138 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
139 (notice_args_size): Set it in the current trace if no insn that can
140 throw internally has been seen yet.
141 (connect_traces): When connecting args_size between traces, allow the
142 incoming values not to match if there is an insn setting it before the
143 first insn that can throw internally; in that case, force the creation
144 of a CFI note on this latter insn.
146 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
148 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
149 the base class, because using the injected-class-name was not clearly
150 specified until DR 176.
152 2018-10-12 Paul Koning <ni1d@arrl.net>
154 * config/pdp11/pdp11.md (doloop_end): New expander.
155 (doloop_end_insn): renamed from "doloop_end".
156 (addqi3): New pattern.
157 (subqi3): New pattern.
158 * config/pdp11/predicates.md (incdec_operand): New predicate.
160 2018-10-12 Yury Gribov <tetra2005@gmail.com>
163 * real.c (format_helper::can_represent_integral_type_p): New function
164 * real.h (format_helper::can_represent_integral_type_p): Ditto.
165 * match.pd: New pattern.
167 2018-10-12 Alexandre Oliva <oliva@adacore.com>
169 * configure.ac: Introduce --enable-large-address-aware
170 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
171 * doc/install.texi: Document it.
172 * configure, config.in: Rebuilt.
173 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
174 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
175 (LINK_SPEC): Insert it.
176 * config/i386/mingw-264.h: Likewise.
178 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
180 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
182 PR rtl-optimization/87600
183 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
185 2018-10-12 Paul Koning <ni1d@arrl.net>
187 * doc/md.texi (doloop_end): Document that the pattern code may
188 need to check operand mode.
190 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
192 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
193 to zero-extend between int and floating-point registers.
194 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
195 ldp into floating-point registers. Add type and arch attributes.
196 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
197 Use f_loads for type attribute.
199 2018-10-11 Martin Sebor <msebor@redhat.com>
201 * doc/extend.texi (attribute packed): Correct typos.
203 2018-10-11 Martin Sebor <msebor@redhat.com>
205 * doc/extend.texi (attribute flatten): Mention interaction with
208 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
211 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
212 Set new_decl virtual flag to zero.
214 2018-10-11 Martin Sebor <msebor@redhat.com>
217 * doc/extend.texi (attribute format_arg): Discuss using multiple
218 attributes on a single function.
220 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
222 PR tree-optimization/86829
223 * match.pd (sin (atan (x))): New simplification rules.
224 (cos (atan (x))): Likewise.
225 * real.c (build_sinatan_real): New function.
226 * real.h (build_sinatan_real): Prototype.
228 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
230 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
232 (fold_mergeeo_helper): New helper function.
233 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
234 intrinsics. Correct some whitespace indentation issues.
236 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
239 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
240 Use HOST_WIDE_INT_1U for shift.
242 2018-10-11 Doug Rupp <rupp@adacore.com>
243 Olivier Hainque <hainque@adacore.com>
245 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
246 Pass --relax to the linker for RTPs.
247 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
249 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
250 Jan Hubicka <jh@suse.cz>
251 Martin Jambor <mjambor@suse.cz>
253 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
254 the same elements are repeated rather than printing all of them.
255 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
257 * read-rtl-function.c (test_loading_repeat): New function.
258 (read_rtl_function_c_tests): Call test_loading_repeat.
259 * rtl-tests.c (test_dumping_repeat): New function.
260 (rtl_tests_c_tests): Call test_dumping_repeat.
262 2018-10-11 Richard Biener <rguenther@suse.de>
264 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
265 bdver?_cost): Unify to ...
266 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
267 * config/i386/i386.c (processor_cost_table): Adjust.
269 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
272 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
273 the thunk when expanding to GIMPLE.
275 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
277 * varasm.c (mergeable_string_section): Don't try to move zero-length
278 strings to the merge section.
280 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
283 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
285 2018-10-10 Jakub Jelinek <jakub@redhat.com>
288 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
291 2018-10-10 Richard Biener <rguenther@suse.de>
293 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
294 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
295 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
296 and recursing and pattern terminating the recursion on SSE
297 vector width using ix86_expand_reduc.
298 (reduc_sminmax_scal_<mode>): Split into part reducing to half
299 width and recursing and SSE2 vector variant doing the final
300 reduction with ix86_expand_reduc.
301 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
302 with terminating the recursion at AVX level, splitting that
305 2018-10-09 David Malcolm <dmalcolm@redhat.com>
307 * genmatch.c (error_cb): Rename to...
308 (diagnostic_cb): ...this, converting int params to enums.
309 (fatal_at): Update for renaming.
310 (warning_at): Likewise.
312 * input.c (selftest::ebcdic_execution_charset::apply):
313 Update for renaming of...
314 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
316 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
317 converting level and reason to enums.
318 (class selftest::lexer_error_sink): Rename to...
319 (class selftest::lexer_test_options): ...this, renaming field
320 "m_errors" to "m_diagnostics".
321 (selftest::lexer_test_options::apply): Update for renaming of...
322 (selftest::lexer_test_options::on_error): ...this, renaming to...
323 (selftest::lexer_test_options::on_diagnostic): ...this
324 converting level and reason to enums.
325 (selftest::test_lexer_string_locations_raw_string_unterminated):
326 Update for renamings.
327 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
330 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
332 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
333 * config/rs6000/pmmintrin.h: New file.
335 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
337 PR tree-optimization/86659
338 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
340 2018-10-09 Richard Biener <rguenther@suse.de>
342 PR tree-optimization/63155
343 * tree-ssa-structalias.c: Include tree-ssa.h.
344 (get_constraint_for_ssa_var): For undefs return nothing_id.
345 (find_func_aliases): Cleanup PHI handling.
347 2018-10-09 Richard Biener <rguenther@suse.de>
349 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
352 2018-10-09 Martin Liska <mliska@suse.cz>
354 * asan.c (asan_emit_stack_protection): If a stack variable
355 is located in a same file as current function, then emit
356 line info into variable definition string.
358 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
360 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
363 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
365 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
368 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
370 PR tree-optimization/86659
371 * gimple-match.h (struct gimple_match_op): Add reverse field.
372 (gimple_match_op::set_op): New overloaded method.
373 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
374 the REF_REVERSE_STORAGE_ORDER flag on the value.
375 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
376 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
378 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
381 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
382 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
384 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
387 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
388 Defined with __builtin_ia32_vfmaddsubpd512_mask.
390 2018-10-08 Richard Biener <rguenther@suse.de>
392 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
393 cost the same as AVX128 ones.
395 2018-10-08 Paul Koning <ni1d@arrl.net>
397 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
398 (expand_block_move): New function.
399 * config/pdp11/pdp11.c (output_block_move): Remove.
400 (expand_block_move): New function.
401 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
402 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
403 (*movmemhi1): Remove.
405 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
407 * config/s390/2827.md: Increase latencies for some FP instructions.
409 2018-10-08 Richard Biener <rguenther@suse.de>
411 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
413 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
414 * tree-vectorizer.h (dump_stmt_cost): Adjust.
415 (add_stmt_cost): Dump return value of the hook.
417 2018-10-08 Richard Biener <rguenther@suse.de>
419 PR tree-optimization/63155
420 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
421 (ssa_propagation_engine::ssa_propagate): Remove redundant
424 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
426 PR rtl-optimization/86939
427 PR rtl-optimization/87479
428 * ira.h (non_conflicting_reg_copy_p): New prototype.
429 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
430 (make_hard_regno_dead): Don't add conflicts for register
431 ignore_reg_for_conflicts.
432 (make_object_dead): Likewise.
433 (non_conflicting_reg_copy_p): New function.
434 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
435 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
436 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
437 (make_hard_regno_dead): Don't add conflicts for register
438 ignore_reg_for_conflicts. Remove special conflict handling of
439 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
440 check_pic_pseudo_p and update callers.
441 (mark_pseudo_dead): Don't add conflicts for register
442 ignore_reg_for_conflicts.
443 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
445 2018-10-05 Andrew Waterman <andrew@sifive.com>
446 Jim Wilson <jimw@sifive.com>
448 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
449 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
450 new pattern using HONOR_SNANS that emits one extra instruction.
452 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
454 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
455 patterns): Merge SI and DI patterns to a GPR pattern.
456 (unnamed define_insn and define_split for record form of that): Merge
457 to a single define_insn_and_split pattern.
459 2018-10-05 David Malcolm <dmalcolm@redhat.com>
462 * input.c (expand_location_to_spelling_point): Add param "aspect"
463 and use rather than hardcoding LOCATION_ASPECT_CARET.
464 (get_substring_ranges_for_loc): Handle the case of a single token
465 within a macro expansion.
466 * input.h (expand_location_to_spelling_point): Add "aspect" param,
467 defaulting to LOCATION_ASPECT_CARET.
469 2018-10-05 Paul Koning <ni1d@arrl.net>
471 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
472 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
473 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
474 (pdp11_guard_type): New function.
476 2018-10-05 Paul Koning <ni1d@arrl.net>
478 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
479 * config/pdp11/pdp11.opt (mfloat32): Remove.
481 * doc/invoke.texi (pdp11 -mfloat32): Remove:
482 (pdp11 -mfloat64): Remove.
484 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
486 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
487 (*cmp<mode>_cc_i387): Ditto.
488 (*cmpu<mode>_cc_i387): Ditto.
489 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
490 * config/i386/i386.c (ix86_expand_fp_compare): Remove
492 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
493 Emit x86_sahf_1 pattern.
494 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
495 (ix86_expand_carry_flag_compare): Ditto.
497 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
499 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
500 to reg_or_0_operand. Add "C" constraint.
501 (*cmpxf_cc_i387): Ditto.
502 (*cmp<mode>_i387): Change operand 2 predicate
503 to nonimm_or_0_operand. Add "C" constraint.
504 (*cmp<mode>_cc_i387): Ditto.
505 (*cmp<mode>_0_i387): Remove insn pattern.
506 (*cmp<mode>_0_cc_i387): Ditto.
508 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
510 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
511 * config/i386/predicates.md (nonimm_or_0_operand): Rename
512 from vector_move_operand. Update all uses.
514 2018-10-05 Martin Sebor <msebor@redhat.com>
516 PR tree-optimization/87490
517 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
520 2018-10-05 Richard Biener <rguenther@suse.de>
522 PR tree-optimization/63155
523 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
524 vertical space in dumpfiles.
525 * tree-ssa-propagate.h
526 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
527 * tree-ssa-propagate.c (cfg_blocks_back): New global.
528 (ssa_edge_worklist_back): Likewise.
529 (curr_order): Likewise.
530 (cfg_blocks_get): Remove abstraction.
531 (cfg_blocks_add): Likewise.
532 (cfg_blocks_empty_p): Likewise.
533 (add_ssa_edge): Add to current or next worklist based on
535 (add_control_edge): Likewise.
536 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
538 (ssa_propagation_engine::ssa_propagate): ... here. Unify
539 iteration from CFG and SSA edge worklist so we process
540 everything in RPO order, prioritizing forward progress
542 (ssa_prop_init): Allocate new worklists, do not dump
544 (ssa_prop_fini): Free new worklists.
546 2018-10-05 Richard Biener <rguenther@suse.de>
548 * tree-core.h (tree_block::abstract_flag): Remove.
549 (tree_block::block_num): Make full 32bits.
550 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
551 * tree.h (BLOCK_ABSTRACT): Remove.
552 * dwarf2out.c (gen_lexical_block_die): Remove dead code
553 resulting from BLOCK_ABSTRACT being always false.
554 (gen_inlined_subroutine_die): Likewise.
555 (gen_block_die): Likewise.
556 * tree.c (block_ultimate_origin): Likewise.
557 * tree-pretty-print.c (dump_block_node): Remove code dealing
559 * tree-ssa-live.c (dump_scope_block): Likewise.
560 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
561 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
563 2018-10-05 Richard Biener <rguenther@suse.de>
565 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
566 is asked for initialize mode to the component mode of the
569 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
572 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
574 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
576 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
579 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
581 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
582 to be DImode. When using mffscrn, force the operand to a register.
584 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
586 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
587 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
588 X87MODEF mode iterator.
589 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
590 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
591 X87MODEF mode iterator.
593 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
595 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
598 2018-10-04 David Malcolm <dmalcolm@redhat.com>
600 * Makefile.in (OBJS): Add opt-problem.o.
601 * dump-context.h: Include "selftest.h.
602 (selftest::temp_dump_context): New forward decl.
603 (class dump_context): Make friend of class
604 selftest::temp_dump_context.
605 (dump_context::dump_loc_immediate): New decl.
606 (class dump_pretty_printer): Move here from dumpfile.c.
607 (class temp_dump_context): Move to namespace selftest.
608 (temp_dump_context::temp_dump_context): Add param
609 "forcibly_enable_dumping".
610 (selftest::verify_dumped_text):
611 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
612 (selftest::verify_item):
613 (ASSERT_IS_TEXT): Move here from dumpfile.c.
614 (ASSERT_IS_TREE): Likewise.
615 (ASSERT_IS_GIMPLE): Likewise.
616 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
618 (dump_context::dump_loc_immediate): ...this new function.
619 (class dump_pretty_printer): Move to dump-context.h.
620 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
621 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
622 (temp_dump_context::temp_dump_context): Move to "selftest"
623 namespace. Add param "forcibly_enable_dumping", and use it to
624 conditionalize the use of m_pp;
625 (selftest::verify_dumped_text): Make non-static.
626 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
627 (selftest::verify_item): Make non-static.
628 (ASSERT_IS_TEXT): Move to dump-context.h.
629 (ASSERT_IS_TREE): Likewise.
630 (ASSERT_IS_GIMPLE): Likewise.
631 (selftest::test_capture_of_dump_calls): Pass "true" for new
632 param of temp_dump_context.
633 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
634 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
636 * opt-problem.cc: New file.
637 * opt-problem.h: New file.
638 * optinfo-emit-json.cc
639 (selftest::test_building_json_from_dump_calls): Pass "true" for
640 new param of temp_dump_context.
641 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
642 (optinfo::emit_for_opt_problem): New function.
643 (optinfo::emit): Clarity which emit_item is used.
644 * optinfo.h (optinfo::get_dump_location): New accessor.
645 (optinfo::emit_for_opt_problem): New decl.
646 (optinfo::emit): Make const.
647 * selftest-run-tests.c (selftest::run_tests): Call
648 selftest::opt_problem_cc_tests.
649 * selftest.h (selftest::opt_problem_cc_tests): New decl.
650 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
651 bool to opt_result, converting fprintf messages to
652 opt_result::failure_at calls. Add "stmt" param for use by the
654 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
655 (runtime_alias_check_p): Convert return type from bool to
656 opt_result, converting dump_printf calls to
657 opt_result::failure_at, using the statement DDR_A for their
659 (find_data_references_in_stmt): Convert return type from bool to
660 opt_result, converting "return false" to opt_result::failure_at
662 * tree-data-ref.h: Include "opt-problem.h".
663 (dr_analyze_innermost): Convert return type from bool to opt_result,
664 and add a const gimple * param.
665 (find_data_references_in_stmt): Convert return type from bool to
667 (runtime_alias_check_p): Likewise.
668 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
669 dr_analyze_innermost.
670 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
671 Convert return type from bool to opt_result, adding a message for
672 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
673 (vect_analyze_data_ref_dependence): Convert return type from bool
674 to opt_result. Change sense of return type from "false"
675 effectively meaning "no problems" to "false" meaning a problem,
676 so that "return false" becomes "return opt_result::success".
677 Convert "return true" calls to opt_result::failure_at, using
678 the location of statement A rather than vect_location.
679 (vect_analyze_data_ref_dependences): Convert return type from bool
681 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
682 calls to opt_result::failure_at, using the stmt location rather
684 (vect_verify_datarefs_alignment): Convert return type from bool
686 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
687 into multiple more-tightly-scoped copies.
688 (vect_analyze_data_refs_alignment): Convert return type from bool
690 (vect_analyze_data_ref_accesses): Likewise, converting a
691 "return false" to a "return opt_result::failure_at", adding a
693 (vect_prune_runtime_alias_test_list): Convert return type from
694 bool to opt_result, converting dump_printf_loc to
695 opt_result::failure_at. Add a %G to show the pertinent statement,
696 and use the stmt's location rather than vect_location.
697 (vect_find_stmt_data_reference): Convert return type from
698 bool to opt_result, converting dump_printf_loc to
699 opt_result::failure_at, using stmt's location.
700 (vect_analyze_data_refs): Convert return type from bool to
701 opt_result. Convert "return false" to "return
702 opt_result::failure_at", adding messages as needed.
703 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
704 type from bool to opt_result.
705 (vect_determine_vf_for_stmt): Likewise.
706 (vect_determine_vectorization_factor): Likewise, converting
707 dump_printf_loc to opt_result::failure_at, using location of phi
708 rather than vect_location.
709 (vect_analyze_loop_form_1): Convert return type from bool to
710 opt_result, converting dump_printf_loc calls, retaining the use of
712 (vect_analyze_loop_form): Convert return type from loop_vec_info
713 to opt_loop_vec_info.
714 (vect_analyze_loop_operations): Convert return type from bool to
715 opt_result, converting dump_printf_loc calls, using the location
716 of phi/stmt rather than vect_location where available. Convert
717 various "return false" to "return opt_result::failure_at" with
718 "unsupported phi" messages.
719 (vect_get_datarefs_in_loop): Convert return type from bool to
720 opt_result. Add a message for the
721 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
722 (vect_analyze_loop_2): Convert return type from bool to
723 opt_result. Ensure "ok" is set to a opt_result::failure_at before
724 each "goto again;", adding new messages where needed.
725 Add "unsupported grouped {store|load}" messages.
726 (vect_analyze_loop): Convert return type from loop_vec_info to
728 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
730 * tree-vect-stmts.c (process_use): Likewise, converting
731 dump_printf_loc call and using stmt location, rather than
733 (vect_mark_stmts_to_be_vectorized): Likeise.
734 (vect_analyze_stmt): Likewise, adding a %G.
735 (vect_get_vector_types_for_stmt): Convert return type from bool to
736 opt_result, converting dump_printf_loc calls and using stmt
737 location, rather than vect_location.
738 (vect_get_mask_type_for_stmt): Convert return type from tree to
739 opt_tree, converting dump_printf_loc calls and using stmt location.
740 * tree-vectorizer.c: Include "opt-problem.h.
741 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
742 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
743 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
744 enabled, use it to report at the top level "couldn't vectorize
745 loop" followed by the problem.
746 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
747 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
749 (vect_analyze_stmt): Likewise.
750 (vect_get_vector_types_for_stmt): Likewise.
751 (tree vect_get_mask_type_for_stmt): Likewise.
752 (vect_analyze_data_ref_dependences): Likewise.
753 (vect_enhance_data_refs_alignment): Likewise.
754 (vect_analyze_data_refs_alignment): Likewise.
755 (vect_verify_datarefs_alignment): Likewise.
756 (vect_analyze_data_ref_accesses): Likewise.
757 (vect_prune_runtime_alias_test_list): Likewise.
758 (vect_find_stmt_data_reference): Likewise.
759 (vect_analyze_data_refs): Likewise.
760 (vect_analyze_loop): Convert return type from loop_vec_info to
762 (vect_analyze_loop_form): Likewise.
763 (vect_analyze_slp): Convert return type from bool to opt_result.
765 2018-10-04 David Malcolm <dmalcolm@redhat.com>
767 * doc/invoke.texi (-fopt-info): Document new "internals"
769 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
770 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
772 (optinfo_verbosity_options): Add "internals".
773 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
774 (dump_context::apply_dump_filter_p): New member function.
775 (dump_context::dump_loc): Use apply_dump_filter_p rather than
776 explicitly masking the dump_kind.
777 (dump_context::begin_scope): Increment the scope depth first. Use
778 apply_dump_filter_p rather than explicitly masking the dump_kind.
779 (dump_context::emit_item): Use apply_dump_filter_p rather than
780 explicitly masking the dump_kind.
781 (dump_dec): Likewise.
782 (dump_hex): Likewise.
783 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
784 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
785 (opt_info_switch_p): Update handling of default
786 MSG_OPTIMIZED_LOCATIONS to cope with default of
787 MSG_PRIORITY_USER_FACING.
788 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
789 masking the dump_kind.
790 (selftest::test_capture_of_dump_calls): Update test_dump_context
791 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
792 than MSG_ALL. Generalize scope test to be run at all four
793 combinations of with/without MSG_PRIORITY_USER_FACING and
794 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
795 for each of the two values.
796 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
797 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
798 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
799 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
800 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
802 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
803 dump messages as MSG_PRIORITY_USER_FACING.
804 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
805 about the interaction with MSG_PRIORITY_*.
807 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
809 * varasm.c (output_constant): Add new parameter merge_strings.
810 Make strings properly zero terminated in merge string sections.
811 (mergeable_string_section): Don't fail if the last char is non-zero.
812 (assemble_variable_contents): Handle merge string sections.
813 (assemble_variable): Likewise.
814 (assemble_constant_contents): Likewise.
815 (output_constant_def_contents): Likewise.
816 (output_constructor_array_range,
817 output_constructor_regular_field): Adjust call to output_constant.
818 (output_object_block): Adjust call to assemble_constant_contents
819 and assemble_variable_contents.
821 2018-10-04 Martin Liska <mliska@suse.cz>
824 * cgraphunit.c (process_function_and_variable_attributes):
825 Warn about a function with alias attribute and a body.
827 2018-10-04 Martin Liska <mliska@suse.cz>
830 * multiple_target.c (redirect_to_specific_clone): New function.
831 (ipa_target_clone): Use it.
832 * tree-inline.c: Fix comment.
834 2018-10-04 David Malcolm <dmalcolm@redhat.com>
836 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
838 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
839 (gcc::dump_manager::register_pass): New member function, adapted
840 from loop body in gcc::pass_manager::register_pass, adding a
841 call to update_dfi_for_opt_info.
842 (gcc::dump_manager::opt_info_enable_passes): Store the
843 -fopt-info options into the new fields. Move the loop
845 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
847 * dumpfile.h (struct opt_pass): New forward decl.
848 (gcc::dump_manager::register_pass): New decl.
849 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
850 (class gcc::dump_manager): Add fields "m_optgroup_flags",
851 "m_optinfo_flags", and "m_optinfo_filename".
852 * passes.c (gcc::pass_manager::register_pass): Move all of the
853 dump-handling code to gcc::dump_manager::register_pass.
855 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
857 PR rtl-optimization/87466
858 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
859 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
860 * doc/tm.texi: Regenerate.
861 * ira-lives.c (process_bb_node_lives): Use the new target hook.
862 * lra-lives.c (process_bb_lives): Likewise.
863 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
866 2018-10-04 Tamar Christina <tamar.christina@arm.com>
868 * params.c (add_params): Fix initialization.
870 2018-10-04 Martin Liska <mliska@suse.cz>
872 PR gcov-profile/84107
873 * tree-profile.c (init_ic_make_global_vars):
874 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
875 Come up with new ic_tuple* variables. Emit
876 __gcov_indirect_call{,_topn} variables.
877 (gimple_gen_ic_profiler): Access the variable
879 (gimple_gen_ic_func_profiler): Access
880 __gcov_indirect_call.callee field.
881 (gimple_init_gcov_profiler): Use ptr_type_node.
882 * value-prof.c (gimple_ic): Use ptr_type_node.
884 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
886 PR tree-optimization/85787
887 * ipa-pure-const.c (malloc_candidate_p_1): Move most of malloc_candidate_p
888 into this function and add support for detecting multiple phis.
889 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
891 2018-10-04 Martin Liska <mliska@suse.cz>
894 * ipa-inline.c (inline_to_all_callers_1):
895 Call ultimate_alias_target for node being inlined.
897 2018-10-03 Jeff Law <law@redhat.com>
899 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
901 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
902 * tree.h (get_typenode_from_name): Prototype.
904 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
906 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
907 Change operand 2 predicate to nonimmediate_operand.
908 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
910 2018-10-03 Martin Sebor <msebor@redhat.com>
911 Jeff Law <law@redhat.com>
913 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
915 (get_string_length): Detect unterminated arrays.
916 (format_string): Same.
917 (format_directive): Warn about unterminated arrays.
918 (handle_gimple_call): Mark statements with no_warning as needed.
920 2018-10-03 Jim Wilson <jimw@sifive.com>
922 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
923 also define __riscv_abi_rve. Delete trailing white space.
925 2018-10-03 Paul Koning <ni1d@arrl.net>
927 Enable LRA register allocator for PDP11.
928 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
931 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
932 * config/pdp11/pdp11.opt (-mlra): New option.
933 * doc/invoke.texi (PDP-11 Options): Document -mlra.
935 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
937 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
938 (*<absneg:code>extend<mode>xf2): Ditto.
940 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
942 PR tree-optimization/87415
943 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
946 2018-10-02 Jeff Law <law@redhat.com>
948 * gimple-fold.c (get_range_strlen): Only set *nonstr when
949 an unterminated string is discovered. Bubble up range
950 even for unterminated strings.
951 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
952 indicates the string was not terminated via NONSTR.
954 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
956 * tree-vrp.c (extract_range_from_unary_expr): Special case all
958 Do not do anything special for anti-ranges.
960 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
962 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
963 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
965 2018-10-03 Martin Liska <mliska@suse.cz>
967 PR gcov-profile/86109
968 * coverage.c (coverage_begin_function): Do not
969 mark lambdas as artificial.
970 * tree-core.h (struct GTY): Remove tm_clone_flag
971 and introduce new lambda_function.
972 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
974 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
977 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
978 P8_VECTOR and VSX are enabled.
980 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
982 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
983 0x3907 as CPU model number.
985 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
987 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
988 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
989 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
990 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
991 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
992 * config/s390/s390.md: Likewise. Rename also the cpu attribute
993 value from arch12 to z14.
995 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
997 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
999 (isinf<mode>2): Ditto.
1001 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
1003 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
1004 before emitting fxam. Perform calculations in XFmode.
1006 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
1008 * match.pd (((X /[ex] A) +- B) * A): New transformation.
1010 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
1013 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
1014 * tree.c (signed_or_unsigned_type_for): Handle complex.
1016 2018-10-02 Jeff Law <law@redhat.com>
1018 * gimple-fold.c (get_range_strlen): Remove dead code.
1020 2018-10-02 Martin Sebor <msebor@redhat.com>
1021 Jeff Law <law@redhat.com>
1023 * builtins.c (unterminated_array): Add new arguments.
1024 If argument is not terminated, bubble up size and exact
1026 (expand_builtin_strnlen): Detect, avoid expanding
1027 and diagnose unterminated arrays.
1028 (c_strlen): Fill in offset of start of unterminated strings.
1029 * builtins.h (unterminated_array): Update prototype.
1031 2018-10-02 Richard Biener <rguenther@suse.de>
1033 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
1034 of haddv4df, first reduce to SSE width and exploit the fact
1035 that we only need element zero with the reduction result.
1036 (reduc_plus_scal_v2df): Likewise.
1038 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
1040 * dojump.h (do_jump): Delete.
1041 (do_jump_1): Likewise.
1042 (split_comparison): Move around.
1043 * dojump.c (do_jump): Make static.
1044 (do_jump_1): Likewise.
1045 (jumpifnot): Move around.
1046 (jumpifnot_1): Likewise.
1048 (jumpif_1): Likewise.
1049 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
1051 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
1053 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
1054 insns in the delay slot and add_insn_after for the jump insn.
1056 2018-10-02 Richard Biener <rguenther@suse.de>
1058 * tree-inline.c (expand_call_inline): Use the location of
1059 the callee declaration for the inline-entry marker.
1060 * final.c (notice_source_line): Remove special-casing of
1061 NOTE_INSN_INLINE_ENTRY.
1063 2018-10-01 Carl Love <cel@us.ibm.com>
1066 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
1067 (__builtin_mtfsb0): New.
1068 (__builtin_mtfsb1): New.
1069 ( __builtin_set_fpscr_rn): New.
1070 (__builtin_set_fpscr_drn): New.
1071 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
1072 (rs6000_expand_set_fpscr_rn_builtin): Add.
1073 (rs6000_expand_set_fpscr_drn_builtin): Add.
1074 (rs6000_expand_builtin): Add case statement entries for
1075 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
1076 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
1077 RS6000_BUILTIN_MFFSL.
1078 (rs6000_init_builtins): Add ftype initialization and def_builtin
1079 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
1080 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
1081 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
1082 rs6000_mffscdrn): Add define_insn.
1083 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
1084 * doc/extend.texi: Add documentation for the builtins.
1086 2018-10-01 Richard Biener <rguenther@suse.de>
1088 PR tree-optimization/87465
1089 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
1090 causing branch miscounts.
1092 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1094 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
1095 aarch64_option_default_param): New.
1096 (params.h): Include.
1097 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
1098 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
1099 stack-clash protection validation code.
1101 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1103 * params.c (validate_param): New.
1104 (add_params): Use it.
1105 (set_param_value): Refactor param validation into validate_param.
1106 (diagnostic.h): Include.
1107 * diagnostic.h (diagnostic_ready_p): New.
1109 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1111 * params.c (set_param_value):
1112 Add index of parameter being validated.
1113 * common/common-target.def (option_validate_param): New.
1114 * common/common-targhooks.h (default_option_validate_param): New.
1115 * common/common-targhooks.c (default_option_validate_param): New.
1116 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
1117 * doc/tm.texi: Regenerate.
1119 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1122 * config/aarch64/aarch64.c (aarch64_override_options_internal):
1123 Add validation for stack-clash parameters and set defaults.
1125 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1128 * configure.ac: Add stack-clash-protection-guard-size.
1129 * doc/install.texi: Document it.
1130 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
1131 * params.def: Update comment for guard-size.
1132 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
1133 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
1134 * configure: Regenerate.
1136 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1139 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
1140 STACK_DYNAMIC_OFFSET): New.
1141 * config/aarch64/aarch64.c (aarch64_layout_frame):
1142 Update outgoing args size.
1143 (aarch64_stack_clash_protection_alloca_probe_range,
1144 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1146 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1149 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
1151 * target.def (stack_clash_protection_alloca_probe_range): New.
1152 (stack_clash_protection_final_dynamic_probe): Remove.
1153 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
1154 (default_stack_clash_protection_final_dynamic_probe): Remove.
1155 * targhooks.c: Likewise.
1156 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
1157 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
1158 * doc/tm.texi: Regenerate.
1160 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1163 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
1164 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
1165 aarch64_clamp_to_uimm12_shift): New.
1166 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
1167 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
1169 2018-10-01 Tamar Christina <tamar.christina@arm.com>
1172 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
1174 2018-10-01 Jeff Law <law@redhat.com>
1175 Richard Sandiford <richard.sandiford@linaro.org>
1176 Tamar Christina <tamar.christina@arm.com>
1179 * config/aarch64/aarch64.md
1180 (probe_stack_range): Add k (SP) constraint.
1181 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
1182 STACK_CLASH_MAX_UNROLL_PAGES): New.
1183 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
1184 stack probes for stack clash.
1185 (aarch64_allocate_and_probe_stack_space): New.
1186 (aarch64_expand_prologue): Use it.
1187 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
1188 (aarch64_sub_sp): Add emit_move_imm optional param.
1190 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
1192 PR tree-optimization/87261
1193 * match.pd: Remove trailing whitespace.
1194 Add (x & y) | ~(x | y) -> ~(x ^ y),
1195 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
1197 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1199 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
1201 (ashlsi3_insn): Update instruction constraints.
1202 (ashrsi3_insn): Likewise.
1203 (rotrsi3): Likewise.
1204 (add_shift): Likewise.
1205 * config/arc/constraints.md (Csz): New 32 bit constraint. It
1206 avoids placing in the limm field small constants which, otherwise,
1207 could end into a small instruction.
1209 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
1211 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
1212 destination register is not odd-even.
1213 (umaddsidi4_split): Likewise.
1215 2018-10-01 Richard Biener <rguenther@suse.de>
1217 * tree-inline.c (expand_call_inline): Store origin of fn
1218 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
1219 * tree.c (block_ultimate_origin): Simplify and do some
1222 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
1224 * config/i386/mmx.md (EMMS): New int iterator.
1225 (emms): New int attribute.
1226 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
1227 EMMS int iterator. Explicitly declare clobbers.
1228 (mmx_emms): Remove expander.
1230 * config/i386/predicates.md (emms_operation): Remove predicate.
1231 (vzeroall_pattern): New predicate.
1232 (vzeroupper_pattern): Rename from vzeroupper_operation.
1233 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
1234 vzeroupper_pattern and vzeroall_pattern predicates.
1236 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
1238 PR rtl-optimization/86939
1239 * ira-lives.c (make_hard_regno_born): Rename from this...
1240 (make_hard_regno_live): ... to this. Remove update to conflict
1241 information. Update function comment.
1242 (make_hard_regno_dead): Add conflict information update. Update
1244 (make_object_born): Rename from this...
1245 (make_object_live): ... to this. Remove update to conflict information.
1246 Update function comment.
1247 (make_object_dead): Add conflict information update. Update function
1249 (mark_pseudo_regno_live): Call make_object_live.
1250 (mark_pseudo_regno_subword_live): Likewise.
1251 (mark_hard_reg_dead): Update function comment.
1252 (mark_hard_reg_live): Call make_hard_regno_live.
1253 (process_bb_node_lives): Likewise.
1254 * lra-lives.c (make_hard_regno_born): Rename from this...
1255 (make_hard_regno_live): ... to this. Remove update to conflict
1256 information. Remove now uneeded check_pic_pseudo_p argument.
1257 Update function comment.
1258 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
1259 to conflict information. Update function comment.
1260 (mark_pseudo_live): Remove update to conflict information. Update
1262 (mark_pseudo_dead): Add conflict information update.
1263 (mark_regno_live): Call make_hard_regno_live.
1264 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
1265 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
1267 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
1270 * config/i386/i386.c (construct_container): Use TImode for
1271 BLKmode values in 2 integer registers.
1273 2018-09-29 Jeff Law <law@redhat.com>
1275 * builtins.c (unterminated_array): Pass in c_strlen_data * to
1276 c_strlen rather than just a tree *.
1277 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
1278 Update recursive calls appropriately. If caller did not provide a
1279 suitable data pointer, create a local one. When a non-terminated
1280 string is discovered, bubble up information about the string via the
1281 c_strlen_data object.
1282 * builtins.h (c_strlen): Update prototype.
1283 (c_strlen_data): New structure.
1284 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
1285 For a type 2 call, if c_strlen indicates a non-terminated string
1286 use the length of the non-terminated string.
1287 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
1289 2018-09-29 Jakub Jelinek <jakub@redhat.com>
1292 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
1293 __m512d type for __A argument rather than __m512.
1295 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
1297 * match.pd (simple_comparison): Don't optimize if either operand is
1298 a function pointer when target needs function pointer canonicalization.
1300 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1302 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
1303 power5 .. power9 to remove indirection.
1304 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
1305 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
1306 ASM_CPU_476_SPEC): Delete.
1307 (ASM_CPU_SPEC): Adjust.
1308 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
1309 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
1311 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1313 * config.in: Delete HAVE_AS_DCI.
1314 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
1315 * config/rs6000/rs6000.h: Ditto.
1316 * configure.ac: Delete HAVE_AS_DCI.
1317 * configure: Regenerate.
1319 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
1321 * config.in (HAVE_AS_LWSYNC): Delete.
1322 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1323 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
1325 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
1326 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
1328 * configure.ac: Delete HAVE_AS_LWSYNC.
1329 * configure: Regenerate.
1331 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1332 Pierre-Marie de Rodat <derodat@adacore.com>
1334 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
1335 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
1336 (cgraph_node::create_thunk): Add indirect_offset parameter.
1337 (thunk_adjust): Likewise.
1338 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
1339 and initialize the corresponding field with it.
1340 (cgraph_node::dump): Dump indirect_offset field.
1341 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
1342 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
1343 (thunk_adjust): Add indirect_offset parameter and deal with it.
1344 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
1345 pass it to thunk_adjust. Do not call the target hook if it's non-zero
1346 or if the thunk is external or local. Fix formatting. Do not chain
1347 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
1348 if any, in the GIMPLE representation.
1349 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
1350 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
1351 (input_node): Read indirect_offset field.
1352 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
1353 call to thunk_adjust.
1354 * tree-nested.c (struct nesting_info): Add thunk_p field.
1355 (create_nesting_tree): Set it.
1356 (convert_all_function_calls): Copy static chain from targets to thunks.
1357 (finalize_nesting_tree_1): Return early for thunks.
1358 (unnest_nesting_tree_1): Do not finalize thunks.
1359 (gimplify_all_functions): Do not gimplify thunks.
1361 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1363 * opt-suggestions.c (option_proposer::build_option_suggestions):
1364 Release "option_values".
1366 2018-09-28 David Malcolm <dmalcolm@redhat.com>
1368 * coverage.c (get_coverage_counts): Convert problem-reporting dump
1369 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
1370 * dumpfile.c (kind_as_string): New function.
1371 (dump_loc): Rather than a hardcoded prefix of "note: ", use
1372 kind_as_string to vary the prefix based on dump_kind.
1373 (selftest::test_capture_of_dump_calls): Update for above.
1375 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1377 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
1378 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
1380 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
1382 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
1383 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
1384 INVALID_REGNUM instead of FPSR_REG.
1385 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
1386 * config/i386/i386.md: Update comment of FP compares.
1387 (fldenv): Do not clobber FPSR_REG.
1389 2018-09-28 Richard Biener <rguenther@suse.de>
1391 * tree.h (BLOCK_ORIGIN): New.
1392 * omp-expand.c (grid_expand_target_grid_body): Assign
1393 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
1394 * tree-inline.c (remap_block): Likewise.
1395 * auto-profile.c (get_function_decl_from_block): Simplify
1396 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
1397 * langhooks.c (lhd_print_error_function): Likewise.
1398 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
1400 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1401 * tree.c (block_nonartificial_location): Likewise.
1402 (block_ultimate_origin): Likewise.
1403 * tree-pretty-print.c (percent_K_format): Likewise. Remove
1404 no longer needed LTO case.
1406 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
1407 Jan Hubicka <jh@suse.cz>
1408 Martin Jambor <mjambor@suse.cz>
1410 * simplify-rtx.c (simplify_merge_mask): New function.
1411 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
1412 same masks are used in op1 or op2.
1413 (test_vec_merge): New function.
1414 (test_vector_ops): Call test_vec_merge.
1416 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
1418 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
1419 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
1420 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
1421 (cypress_costs): Set it.
1422 (supersparc_costs): Likewise.
1423 (hypersparc_costs): Likewise.
1424 (leon_cost): Likewise.
1425 (leon3_costs): Likewise.
1426 (sparclet_costs): Likewise.
1427 (ultrasparc_costs): Likewise.
1428 (ultrasparc_costs): Likewise.
1429 (niagara_costs): Likewise.
1430 (niagara2_costs): Likewise.
1431 (niagara3_costs): Likewise.
1432 (niagara4_costs): Likewise.
1433 (niagara7_costs): Likewise.
1434 (m8_costs): Likewise.
1435 (TARGET_CAN_FOLLOW_JUMP): Define.
1436 (pass_work_around_errata::gate): Minor tweak.
1437 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
1438 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
1439 Automaitcally clear MASK_FSMULD mask for V7 processors.
1440 (sparc_can_follow_jump): New static function.
1441 (output_ubranch): Deal with CROSSING_JUMP_P.
1442 (sparc_use_sched_lookahead): Rewrite using switch statement.
1443 (sparc_issue_rate): Reorder.
1444 (sparc_branch_cost): New function.
1446 2018-09-27 Martin Sebor <msebor@redhat.com>
1448 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
1449 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
1450 (int_fits_type_p): Same.
1452 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
1454 * config/i386/i386.md (FPCR_REG): Remove.
1455 (UNSPEC_FLDCW): Remove.
1456 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
1457 (x86_fldcw_1): Remove insn pattern.
1458 (fnstenv): Do not clobber FPCR_REG.
1460 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
1461 (CALL_USED_REGISTERS): Ditto.
1462 (REG_ALLOC_ORDER): Ditto.
1463 (REG_CLASS_CONTENTS): Ditto.
1464 (HI_REGISTER_NAMES): Ditto.
1465 (ADDITIONAL_REGISTER_NAMES): Use defines instead
1466 of numerical constants.
1467 * config/i386/i386.c (regclass_map): Remove fpsr register.
1468 (dbx_register_map): Ditto.
1469 (dbx64_register_map): Ditto.
1470 (svr4_dbx_register_map): Ditto.
1471 (print_reg): Do not handle FPCR_REG.
1473 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
1476 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
1477 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
1478 Delete, always treat as true.
1479 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1480 Ditto. Simplify remaining code.
1481 * config/powerpcspe/powerpcspe.h: Ditto.
1482 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
1483 Simplify remaining code.
1484 (rs6000_expand_builtin): Ditto.
1485 * config/rs6000/rs6000.h: Ditto.
1486 * configure.ac: Ditto.
1487 * configure: Regenerate.
1489 2018-09-27 Martin Liska <mliska@suse.cz>
1491 * coverage.c (get_coverage_counts): Revert the formatting
1492 of missing profile opt info.
1494 2018-09-27 Richard Biener <rguenther@suse.de>
1498 * dwarf2out.c (set_block_origin_self): Do not mark outermost
1499 block as we do not output that.
1500 (gen_inlined_subroutine_die): Elide the originally outermost
1501 block, matching what we do for concrete instances.
1502 (decls_for_scope): Add parameter specifying whether to recurse
1505 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
1506 Tom de Vries <tom@codesourcery.com>
1510 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
1511 STORE_FLAG_VALUE == 1.
1513 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
1515 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
1516 constant definitions.
1517 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
1518 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
1519 ("speculation_barrier"): New expander definition.
1521 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
1523 PR gcov-profile/86957
1524 * common.opt: New warning option -Wmissing-profile.
1525 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
1526 * doc/invoke.texi: Document -Wmissing-profile.
1528 2018-09-26 Jim Wilson <jimw@sifive.com>
1530 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
1531 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
1533 2018-09-26 Martin Sebor <msebor@redhat.com>
1535 * tree.c (zerop): Change return type to bool.
1536 (integer_zerop, integer_onep, integer_each_onep): Same.
1537 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1538 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1539 (real_onep, real_minus_onep, chain_index): Same.
1540 (print_type_hash_statistics, type_list_equal): Same.
1541 * tree.h (zerop): Same.
1542 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
1543 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
1544 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
1545 (real_onep, real_minus_onep, chain_index): Same.
1546 (print_type_hash_statistics, type_list_equal): Same.
1548 2018-09-26 Jim Wilson <jimw@sifive.com>
1550 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
1552 2018-09-26 Jakub Jelinek <jakub@redhat.com>
1555 * config/i386/i386.c: Include debug.h and dwarf2out.h.
1556 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
1559 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
1561 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
1563 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1565 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
1566 and FP_SECOND_SSE_REGS.
1567 (REG_CLASS_NAMES): Ditto.
1568 (REG_CLASS_CONTENTS): Ditto.
1569 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
1570 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
1571 (ix86_preferred_output_reload_class): Ditto.
1572 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
1573 clobber constraint to "=&f".
1574 (fix_truncdi_i387): Ditto.
1575 (lrintxfdi2): Ditto.
1576 (fistdi2_<rounding>): Ditto.
1577 (fpremxf4_i387): Change "=u" constraint to "=f".
1578 (fprem1xf4_i387): Ditto.
1580 (fptanxf4_i387): Ditto.
1581 (fxtractxf3_i387): Ditto.
1582 (fscalexf4_i387): Ditto.
1583 (atan2xf3): Change "u" constraint to "f".
1584 (fyl2xxf3_i387): Ditto.
1585 (fyl2xp1xf3_i387): Ditto.
1587 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
1590 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
1591 for removed I387_MASK_PM entity.
1594 2018-09-26 Jeff Law <law@redhat.com>
1596 2018-09-26 Alexey Neyman <stilor@att.net>
1598 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1599 headers are no longer pulled in by <isl/val.h>.
1601 2018-09-26 Richard Biener <rguenther@suse.de>
1604 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
1605 or concrete instance DIE to the tree. Create abstract origin
1606 attributes also for concrete instances.
1608 2018-09-26 Alexey Neyman <stilor@att.net>
1610 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
1611 headers are no longer pulled in by <isl/val.h>.
1613 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
1615 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
1616 Use new helper functions.
1617 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
1618 Use new helper functions.
1619 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
1620 aarch_mm_needs_release): New declarations.
1621 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
1622 aarch_mm_needs_release): New.
1624 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
1626 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
1627 (arm32_output_mi_thunk): Deal with long calls.
1629 2018-09-26 Richard Biener <rguenther@suse.de>
1633 * tree-inline.c (expand_call_inline): When the location
1634 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
1635 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
1636 the inserted BLOCK to make inlined_function_outer_scope_p
1638 * dwarf2out.c (add_call_src_coords_attributes): Do not add
1639 coords for reserved locations.
1641 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1643 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
1644 (*call_indirect_nonlocal_sysv<mode>): Ditto.
1645 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
1646 (*sibcall_nonlocal_sysv<mode>): Ditto.
1647 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
1648 (<bd>_<mode>): Ditto.
1649 (<bd>tf_<mode>): Ditto.
1651 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1653 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
1654 control string as a list of templates instead of as C code.
1655 (*altivec_movti): Ditto.
1656 * config/rs6000/darwin.md (movdf_low_di): Ditto.
1658 2018-09-25 Jim Wilson <jimw@sifive.com>
1660 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
1661 when target symbol is weak.
1663 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
1666 * builtins.c (unterminated_array): Simplify.
1667 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
1668 where pointer arithmetic is safe.
1670 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
1673 * config/rs6000/altivec.md (altivec_vspltb): Use
1674 const_0_to_15_operand instead of u5bit_cint_operand.
1675 (*altivec_vspltb_internal): Ditto.
1676 (altivec_vspltb_direct): Ditto.
1677 (altivec_vsplth): Use const_0_to_7_operand instead of
1679 (*altivec_vsplth_internal): Ditto.
1680 (altivec_vsplth_direct): Ditto.
1681 (altivec_vspltw): Use const_0_to_3_operand instead of
1683 (*altivec_vspltw_internal): Ditto.
1684 (altivec_vspltw_direct): Ditto.
1685 (altivec_vspltsf): Ditto.
1686 (*altivec_vspltsf_internal): Ditto.
1687 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
1688 various splats with the proper size immediate. Reorder the various
1689 cases by ascending size of immediate, and put all such together.
1691 2018-09-25 Richard Biener <rguenther@suse.de>
1694 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
1696 (maybe_create_die_with_external_ref): Do not create
1697 DW_TAG_imported_unit here.
1698 (add_abstract_origin_attribute): Handle external BLOCK refs.
1699 (dwarf2out_abstract_function): Simplify LTO case.
1700 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
1701 rather than using maybe_create_die_with_external_ref.
1703 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
1706 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
1707 (frndintxf2_mask_pm_i387): Ditto.
1708 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
1709 Enable for !flag_trapping_math.
1710 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
1711 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
1712 Change operand 1 predicate to nonimmediate_operand.
1713 (attr "i387_cw"): Remove mask_pm.
1714 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
1715 (enum ix86_entity): Remove I387_MASK_PM.
1716 * config/i386/i386.c (ix86_i387_mode_needed): Do not
1717 handle I387_MASK_PM.
1718 (ix86_mode_needed): Ditto.
1719 (ix86_mode_after): Ditto.
1720 (ix86_mode_entry): Ditto.
1721 (ix86_mode_exit): Ditto.
1722 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
1724 2018-09-25 Jakub Jelinek <jakub@redhat.com>
1726 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
1727 to_update_switch_stmts to vNULL instead of calling create on them
1730 2018-09-25 Richard Biener <rguenther@suse.de>
1732 PR tree-optimization/87402
1733 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
1734 (visit_phi): Re-instantiate handling of supposed to be VARYING
1735 but non-VARYING backedge value.
1737 2018-09-25 Richard Biener <rguenther@suse.de>
1740 * dwarf2out.c (struct sym_off_pair): New.
1741 (external_die_map): New global.
1742 (lookup_decl_die): When in LTO create DIEs lazily from the
1744 (lookup_block_die): New function, create DIEs lazily in LTO.
1745 (equate_block_to_die): New function.
1746 (dwarf2out_die_ref_for_decl): During WPA get the association
1747 from the external DIE map.
1748 (dwarf2out_register_external_die): Record mapping into the
1750 (maybe_create_die_with_external_ref): New function split out from
1751 DIE generation part of old dwarf2out_register_external_die.
1752 (add_abstract_origin_attribute): Do not return the DIE. When
1753 in LTO reference externals directly.
1754 (dwarf2out_abstract_function): When in LTO ignore calls for
1755 decls with external DIEs (already present abstract instances).
1756 (gen_call_site_die): Adjust.
1757 (add_high_low_attributes): Likewise.
1758 (gen_lexical_block_die): Likewise.
1759 (gen_inlined_subroutine_die): Likewie.
1760 (gen_block_die): Likewise.
1761 (dwarf2out_inline_entry): Likewise.
1762 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
1765 2018-09-25 Martin Liska <mliska@suse.cz>
1767 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
1768 integers and not by a float value.
1770 2018-09-25 Martin Liska <mliska@suse.cz>
1773 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
1775 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
1776 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
1777 * godump.c (go_format_type): Remove extra parenthesis.
1779 2018-09-25 Martin Liska <mliska@suse.cz>
1781 * alias.c (set_dest_equal_p): Remove unused function.
1782 * config/i386/i386.c (def_builtin_pure2): Likewise.
1783 * diagnostic-show-locus.c (class layout): Remove
1785 (layout::layout): Likewise here.
1786 * dump-context.h (class temp_dump_context): Likewise.
1787 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
1788 (add_AT_loclistsptr): Likewise.
1789 (add_AT_offset): Likewise.
1790 (get_AT_hi_pc): Likewise.
1791 (is_comdat_die): Likewise.
1792 (type_is_enum): Likewise.
1793 (ceiling): Likewise.
1794 (add_AT_vms_delta): Likewise.
1795 (is_class_die): Likewise.
1796 * edit-context.c (class line_event): Remove unused field.
1797 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
1799 * ipa-cp.c (ipa_get_vr_lat): Likewise.
1800 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
1801 (ok_for_base_p_nonstrict): Likewise.
1802 * tree-chrec.c (is_not_constant_evolution): Likewise.
1803 (chrec_fold_poly_cst): Likewise.
1804 * tree-if-conv.c (has_pred_critical_p): Likewise.
1805 * tree-ssa-coalesce.c (print_exprs): Likewise.
1806 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
1807 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
1808 * tree-vrp.c (value_ranges_intersect_p): Likewise.
1809 (value_range_nonnegative_p): Likewise.
1811 2018-09-25 Martin Liska <mliska@suse.cz>
1813 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
1814 Do not handle "GNU Pascal".
1815 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
1817 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
1818 from documentation. Likewise.
1819 * dbxout.c (dbxout_range_type): Likewise.
1820 * doc/cpp.texi: Likewise.
1821 * doc/extend.texi: Likewise.
1822 * doc/frontends.texi: Likewise.
1823 * doc/invoke.texi: Remove Pascal entry.
1824 * tree.def (CLEANUP_POINT_EXPR): Likewise.
1825 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
1827 2018-09-25 Martin Liska <mliska@suse.cz>
1830 * doc/invoke.texi: Document all parameters and remove default
1833 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
1836 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
1837 contains HOST_WIDE_INTs when computing its size.
1839 2018-09-24 Jim Wilson <jimw@sifive.com>
1842 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
1844 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
1846 2018-09-24 Andrew Pinski <apinski@marvell.com>
1848 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
1849 access prev before checking it for NULLness in the
1850 AARCH64_FUSE_CMP_BRANCH case.
1852 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
1855 * config/i386/i386.c (rest_of_insert_endbranch): Set
1856 endbr_queued_at_entrance to true and don't insert ENDBR if
1857 x86_function_profiler will be called.
1858 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
1860 * config/i386/i386.h (machine_function): Add
1861 endbr_queued_at_entrance.
1863 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1865 * genattrtab.c (mk_attr_alt): Use alternative_mask.
1866 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
1868 (check_attr_test): Use alternative_mask.
1869 (get_attr_value): Likewise.
1870 (compute_alternative_mask): Use alternative_mask and XWINT.
1871 (make_alternative_compare): Use alternative_mask.
1872 (attr_alt_subset_p): Use XWINT.
1873 (attr_alt_subset_of_compl_p): Likewise.
1874 (attr_alt_intersection): Use alternative_mask and XWINT.
1875 (attr_alt_union): Likewise.
1876 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
1877 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
1878 (simplify_test_exp): Use alternative_mask and XWINT.
1879 (write_test_expr): Use alternative_mask and XWINT, adjust bit
1880 number calculation to support 64 bits. Generate code that
1881 checks 64-bit masks.
1882 (main): Use alternative_mask.
1883 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
1885 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
1888 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
1889 RETURN+USE when returning via %r14.
1891 2018-09-24 Martin Liska <mliska@suse.cz>
1893 * gcov.c (output_lines): Print colorization legend
1894 for both flag_use_colors and flag_use_hotness_colors.
1897 2018-09-24 Martin Liska <mliska@suse.cz>
1899 * coverage.c (get_coverage_counts): Use warning_at
1900 with current_function_decl location. Use %qD in warning
1903 2018-09-24 Martin Liska <mliska@suse.cz>
1905 * memory-block.h (memory_block_pool::release): Annotate with
1906 valgrind that the memory is not accessible.
1908 2018-09-24 Martin Liska <mliska@suse.cz>
1911 * asan.c: Make asan_handled_variables extern.
1913 * cfgexpand.c (expand_stack_vars): Make sure
1914 a representative is unpoison if another
1915 variable in the partition is handled by
1916 use-after-scope sanitization.
1918 2018-09-24 Richard Biener <rguenther@suse.de>
1920 PR tree-optimization/63155
1921 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
1922 the worklist when the edge of the respective argument isn't
1925 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1927 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
1928 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
1929 (MASK_CLASS_P): Update for rename.
1930 (MAYBE_MASK_CLASS_P): Ditto.
1931 (REG_CLASS_NAMES): Update.
1932 (REG_CLASS_CONTENT): Update.
1933 * config/i386/i386.c (regclass_map): Update for MASK_REG
1934 and ALL_MASK_REGS rename.
1935 * config/i386/constraints.md (Yk): Update for rename.
1938 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1940 * config/i386/i386.h (enum reg_class): Remove
1941 EVEX_SSE_REGS and MOD4_SSE_REGS.
1942 (REG_CLASS_NAMES): Update.
1943 (REG_CLASS_CONTENT): Update.
1944 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
1945 registers as ALL_SSE_REGS.
1946 (ix86_additional_allocno_class_p): Remove.
1947 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
1948 (ix86_register_priority): Lower priority of EVEX SSE registers.
1949 Use IN_RANGE macro where appropriate.
1950 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
1951 AVX-5124VNNIW checks.
1952 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
1953 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
1954 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
1955 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
1956 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
1957 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
1958 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
1959 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
1960 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
1961 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
1962 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
1963 * config/i386/constraints.md (Yh): Remove.
1965 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1967 * config/i386/i386.c (regclass_map): Declare integer REX registers
1970 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
1972 * doc/service.texi (Service): Switch the fsf.org link to https.
1974 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
1977 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1978 Define to speculation_safe_value_not_needed.
1980 2018-09-21 Florian Weimer <fweimer@redhat.com>
1983 * doc/extend.texi (Common Function Attributes): Mention that
1984 noreturn suppresses tail call optimization.
1986 2018-09-21 Jeff Law <law@redhat.com>
1988 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1989 vr_values::cleanup_edges_and_switches.
1990 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1992 (identify_jump_threads): Remove EDGE_IGNORE handling.
1993 (execute_vrp): Move handling of to_remove_edges and
1994 to_update_switch_stmts into vr_values class member functions.
1995 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1996 (to_update_switch_stmts): Likewise.
1997 * vr-values.c: Include cfghooks.h.
1998 (vr_values::vr_values): Initialize to_remove_edges and
1999 to_update_switch_stmts.
2000 (vr_values::~vr_values): Verify to_remove_edges and
2001 to_update_switch_stmts are empty.
2002 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
2003 (vr_values::cleanup_edges_and_switches): New member function.
2004 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
2005 function. Add new data members.
2007 2018-09-21 David Malcolm <dmalcolm@redhat.com>
2009 PR tree-optimization/87309
2010 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
2011 calls with pflags and alt_flags.
2012 (selftest::test_capture_of_dump_calls): Add test of interaction of
2013 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
2015 2018-09-21 Olivier Hainque <hainque@adacore.com>
2017 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
2019 2018-09-21 Olivier Hainque <hainque@adacore.com>
2021 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
2023 2018-09-21 Olivier Hainque <hainque@adacore.com>
2025 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
2026 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
2027 (VXWORKS_LIBS_RTP): Minor reordering.
2029 2018-09-21 Olivier Hainque <hainque@adacore.com>
2031 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
2032 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
2034 2018-09-21 Olivier Hainque <hainque@adacore.com>
2036 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
2037 (PTRDIFF_TYPE): Likewise.
2039 2018-09-21 Olivier Hainque <hainque@adacore.com>
2041 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
2042 triplet, similar to support for VxWorks7.
2043 * config/vxworks-dummy.h: Provide a default definition
2044 of TARGET_VXWORKS64 to 0.
2046 2018-09-21 Olivier Hainque <hainque@adacore.com>
2048 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
2049 * config/vxworks-dummy.h: here.
2051 2018-09-21 Olivier Hainque <hainque@adacore.com>
2053 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
2055 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
2056 Bo Zhou <zbo.zhou@hisilicon.com>
2058 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
2059 * config/aarch64/aarch64-tune.md: Regenerated.
2060 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
2061 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
2062 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
2064 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
2065 Julian Brown <julian@codesourcery.com>
2067 * builtins.c (get_builtin_sync_mem): Handle address spaces.
2069 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2071 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
2072 if the call takes a static chain.
2074 2018-09-21 Martin Liska <mliska@suse.cz>
2076 * auto-profile.c (autofdo_source_profile::read): Do not
2078 (read_profile): Do not add working sets.
2079 (read_autofdo_file): Remove sum_all.
2080 (afdo_callsite_hot_enough_for_early_inline): Remove const
2082 * coverage.c (struct counts_entry): Remove gcov_summary.
2083 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
2084 do not support GCOV_TAG_PROGRAM_SUMMARY.
2085 (get_coverage_counts): Remove summary and expected
2087 * coverage.h (get_coverage_counts): Likewise.
2088 * doc/gcov-dump.texi: Remove -w option.
2089 * gcov-dump.c (dump_working_sets): Remove.
2090 (main): Do not support '-w' option.
2091 (print_usage): Likewise.
2092 (tag_summary): Likewise.
2093 * gcov-io.c (gcov_write_summary): Do not dump
2095 (gcov_read_summary): Likewise.
2096 (gcov_histo_index): Remove.
2097 (gcov_histogram_merge): Likewise.
2098 (compute_working_sets): Likewise.
2099 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
2101 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
2102 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
2103 (GCOV_HISTOGRAM_SIZE): Remove.
2104 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
2105 (struct gcov_summary): Simplify rapidly just
2106 to runs and sum_max fields.
2107 (gcov_histo_index): Remove.
2108 (NUM_GCOV_WORKING_SETS): Likewise.
2109 (compute_working_sets): Likewise.
2110 * gcov-tool.c (print_overlap_usage_message): Remove
2111 trailing empty line.
2112 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
2113 (output_lines): Remove program related line.
2114 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
2115 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
2117 (input_profile_summary): Do not read it.
2118 (merge_profile_summaries): And do not merge it.
2119 (input_symtab): Do not call removed function.
2120 * modulo-sched.c (sms_schedule): Do not print sum_max.
2121 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
2122 removed when histogram method was invented.
2123 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
2125 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
2127 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
2128 and dump selected value.
2129 * profile.c (add_working_set): Remove.
2130 (get_working_sets): Likewise.
2131 (find_working_set): Likewise.
2132 (get_exec_counts): Do not work with working sets.
2133 (read_profile_edge_counts): Do not inform as sum_max is removed.
2134 (compute_branch_probabilities): Likewise.
2135 (compute_value_histograms): Remove argument for call of
2136 get_coverage_counts.
2137 * profile.h: Do not make gcov_summary const.
2139 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
2141 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
2143 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
2145 PR tree-optimization/86990
2146 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
2147 Check that the entire merged store group is made of constants only for
2150 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
2152 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
2153 (VTABLE_VERIFICATION_SPEC): Likewise.
2154 (SANITIZER_EARLY_SPEC): Likewise.
2155 (SANITIZER_SPEC): Likewise.
2156 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
2157 * doc/invoke.texi (Link Options): Document -r.
2159 2018-09-20 Richard Biener <rguenther@suse.de>
2162 * gimplify.c (gimplify_expr): Retain alignment of
2163 addressable lvalue in dereference.
2165 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
2168 * configure.ac: Check for .loc is_stmt support.
2169 * configure, config.in: Rebuilt.
2170 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
2173 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
2175 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
2176 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
2179 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
2181 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
2183 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
2186 2018-09-20 Marek Polacek <polacek@redhat.com>
2188 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
2190 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2192 PR tree-optimization/87288
2193 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
2194 into account when determining PEELING_FOR_NITERS.
2196 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
2198 PR tree-optimization/86877
2199 * tree-vect-loop.c (vect_analyze_loop_2): Call
2200 vect_verify_datarefs_alignment.
2202 2018-09-19 Marek Polacek <polacek@redhat.com>
2204 * doc/invoke.texi: Document -Wclass-conversion.
2206 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
2208 * config/pa/pa.c (pa_adjust_priority): Delete.
2209 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
2211 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
2212 (atomic_storehi): Likewise.
2213 (atomic_storesi): Likewise.
2214 (atomic_loaddi): Restore compare and swap exchange loop code.
2216 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
2218 PR rtl-optimization/86902
2219 * combine.c (try_combine): When changing the CC mode used, don't change
2220 an unrelated mode in other_insn to that new CC mode.
2222 2018-09-19 David Malcolm <dmalcolm@redhat.com>
2224 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
2225 with %T in place of calls to dump_generic_expr.
2226 (prune_runtime_alias_test_list): Likewise.
2227 (create_runtime_alias_checks): Likewise.
2228 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
2229 (vect_analyze_data_ref_dependence): Likewise.
2230 (vect_slp_analyze_data_ref_dependence): Likewise.
2231 (vect_record_base_alignment): Likewise. Use %G in place of call
2232 to dump_gimple_stmt.
2233 (vect_compute_data_ref_alignment): Likewise.
2234 (verify_data_ref_alignment): Likewise.
2235 (vect_find_same_alignment_drs): Likewise.
2236 (vect_analyze_group_access_1): Likewise.
2237 (vect_analyze_data_ref_accesses): Likewise.
2238 (dependence_distance_ge_vf): Likewise.
2239 (dump_lower_bound): Likewise.
2240 (vect_prune_runtime_alias_test_list): Likewise.
2241 (vect_find_stmt_data_reference): Likewise.
2242 (vect_analyze_data_refs): Likewise.
2243 (vect_create_addr_base_for_vector_ref): Likewise.
2244 (vect_create_data_ref_ptr): Likewise.
2245 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
2246 (vect_can_advance_ivs_p): Likewise.
2247 (vect_update_ivs_after_vectorizer): Likewise.
2248 (vect_gen_prolog_loop_niters): Likewise.
2249 (vect_prepare_for_masked_peels): Likewise.
2250 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
2251 (vect_determine_vectorization_factor): Likewise.
2252 (vect_is_simple_iv_evolution): Likewise.
2253 (vect_analyze_scalar_cycles_1): Likewise.
2254 (vect_analyze_loop_operations): Likewise.
2255 (report_vect_op): Likewise.
2256 (vect_is_slp_reduction): Likewise.
2257 (check_reduction_path): Likewise.
2258 (vect_is_simple_reduction): Likewise.
2259 (vect_create_epilog_for_reduction): Likewise.
2260 (vect_finalize_reduction:): Likewise.
2261 (vectorizable_induction): Likewise.
2262 (vect_transform_loop_stmt): Likewise.
2263 (vect_transform_loop): Likewise.
2264 (optimize_mask_stores): Likewise.
2265 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
2266 (vect_split_statement): Likewise.
2267 (vect_recog_over_widening_pattern): Likewise.
2268 (vect_recog_average_pattern): Likewise.
2269 (vect_determine_min_output_precision_1): Likewise.
2270 (vect_determine_precisions_from_range): Likewise.
2271 (vect_determine_precisions_from_users): Likewise.
2272 (vect_mark_pattern_stmts): Likewise.
2273 (vect_pattern_recog_1): Likewise.
2274 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
2275 (vect_record_max_nunits): Likewise.
2276 (vect_build_slp_tree_1): Likewise.
2277 (vect_build_slp_tree_2): Likewise.
2278 (vect_print_slp_tree): Likewise.
2279 (vect_analyze_slp_instance): Likewise.
2280 (vect_detect_hybrid_slp_stmts): Likewise.
2281 (vect_detect_hybrid_slp_1): Likewise.
2282 (vect_slp_analyze_operations): Likewise.
2283 (vect_slp_analyze_bb_1): Likewise.
2284 (vect_transform_slp_perm_load): Likewise.
2285 (vect_schedule_slp_instance): Likewise.
2286 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
2287 (vect_mark_stmts_to_be_vectorized): Likewise.
2288 (vect_init_vector_1): Likewise.
2289 (vect_get_vec_def_for_operand): Likewise.
2290 (vect_finish_stmt_generation_1): Likewise.
2291 (vect_check_load_store_mask): Likewise.
2292 (vectorizable_call): Likewise.
2293 (vectorizable_conversion): Likewise.
2294 (vectorizable_operation): Likewise.
2295 (vectorizable_load): Likewise.
2296 (vect_analyze_stmt): Likewise.
2297 (vect_is_simple_use): Likewise.
2298 (vect_get_vector_types_for_stmt): Likewise.
2299 (vect_get_mask_type_for_stmt): Likewise.
2300 * tree-vectorizer.c (increase_alignment): Likewise.
2302 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
2304 * doc/rtl.texi: Adjust vec_select description.
2305 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
2306 non-constant selectors.
2308 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
2310 * config/aarch64/aarch64-protos.h
2311 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
2312 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
2313 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
2314 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
2315 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
2316 (AARCH64_FL_PROFILE): Move index so flags are ordered.
2317 (AARCH64_ISA_RCPC8_4): New flag.
2318 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
2319 to aarch64_offset_9bit_signed_unscaled_p.
2320 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
2322 * config/aarch64/constraints.md (Ust): New constraint.
2323 * config/aarch64/predicates.md.
2324 (aarch64_9bit_offset_memory_operand): New predicate.
2325 (aarch64_rcpc_memory_operand): New predicate.
2327 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
2329 PR rtl-optimization/87361
2330 * rtlanal.c (nonzero_bits1): Revert accidental change.
2332 2018-09-19 Richard Biener <rguenther@suse.de>
2334 PR tree-optimization/87349
2335 PR tree-optimization/87342
2336 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
2338 2018-09-18 Marek Polacek <polacek@redhat.com>
2340 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
2341 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
2343 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2345 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
2347 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
2349 PR rtl-optimization/86882
2350 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
2352 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
2354 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
2355 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
2357 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
2360 * doc/invoke.texi (Link Options): Fix formatting and grammar.
2362 2018-09-18 Richard Biener <rguenther@suse.de>
2365 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
2366 (compute_samebase_partition_bases): Likewise.
2367 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
2368 (gimple_can_coalesce_p): Simplify.
2370 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
2372 Handle a library implementation of ffs calling __builtin_ffs.
2373 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
2374 (mmix_init_libfuncs): New function: make __builtin_ffs expand
2377 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2379 * diagnostic-show-locus.c (class layout_range): Add field
2381 (layout_range::layout_range): Add "original_idx" param and use it
2382 to initialize new field.
2383 (make_range): Use 0 for original_idx.
2384 (layout::layout): Pass in index to calls to
2385 maybe_add_location_range.
2386 (layout::maybe_add_location_range): Add param "original_idx" and
2387 pass it on to layout_range.
2388 (layout::print_any_labels): Pass on range->m_original_idx to
2390 (gcc_rich_location::add_location_if_nearby): Use 0 for
2392 * gcc-rich-location.h (text_range_label::get_text): Update for new
2394 (range_label_for_type_mismatch::get_text): Likewise.
2396 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2398 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
2400 2018-09-17 David Malcolm <dmalcolm@redhat.com>
2402 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
2403 format_string_diagnostic_t.
2404 (fmtwarn_n): Likewise.
2405 * substring-locations.c
2406 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
2407 (format_warning_n_va): Convert to...
2408 (format_string_diagnostic_t::emit_warning_n_va): ...this.
2409 (format_warning_va): Convert to...
2410 (format_string_diagnostic_t::emit_warning_va): ...this.
2411 (format_warning_at_substring): Convert to...
2412 (format_string_diagnostic_t::emit_warning): ...this.
2413 (format_warning_at_substring_n): Convert to...
2414 (format_string_diagnostic_t::emit_warning_n): ...this.
2415 * substring-locations.h (class format_string_diagnostic_t): New
2417 (format_warning_va): Convert to
2418 format_string_diagnostic_t::emit_warning_va.
2419 (format_warning_n_va): Convert to
2420 format_string_diagnostic_t::emit_warning_n_va.
2421 (format_warning_at_substring): Convert to
2422 format_string_diagnostic_t::emit_warning.
2423 (format_warning_at_substring_n): Convert to
2424 format_string_diagnostic_t::emit_warning_n.
2426 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
2427 Bernd Schmidt <bernds_cb1@t-online.de>
2429 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
2432 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
2434 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
2435 operand 0 predicate to nonimmediate operand.
2436 (rint<mode>2_frndint): Remove insn pattern.
2437 (rint<mode>2): Change operand 1 predicate to general_operand.
2438 Extend operand 1 to XFmode and generate rintxf2 insn.
2439 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
2440 Do not use X87MODEF mode macro.
2441 (frndintxf2_<rounding>_i387): Rename from
2442 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
2443 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
2444 to XFmode and generate significandxf3 insn.
2446 2018-09-17 Richard Biener <rguenther@suse.de>
2448 PR tree-optimization/87328
2449 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
2450 visiting unexecutable backedges when not iterating.
2451 (do_rpo_vn): Mark all edges not executable even when not
2454 2018-09-17 Martin Jambor <mjambor@suse.cz>
2457 * doc/invoke.texi (Warning Options): Likewise.
2459 2018-09-17 Richard Biener <rguenther@suse.de>
2461 PR tree-optimization/87301
2462 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
2463 clean EH info from leftover copy assignments.
2465 2018-09-17 Martin Liska <mliska@suse.cz>
2467 PR gcov-profile/85871
2468 * gcov.c (output_intermediate_file): Fix out of bounds
2471 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
2473 * config/arc/arc.c: Object attributes for core4 not reflected
2475 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
2478 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
2480 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
2482 2018-09-17 Martin Liska <mliska@suse.cz>
2484 * doc/gcov.texi: Document new option --use-hotness-colors.
2485 * gcov.c (struct source_info): Declare new field.
2486 (source_info::source_info): Set default for maximum_count.
2487 (print_usage): Add new -q option.
2488 (process_args): Process it.
2489 (accumulate_line_info): Save src->maximum_count.
2490 (output_line_beginning): Make color line number if
2491 flag_use_hotness_colors is set.
2492 (output_line_details): Pass default argument value.
2493 (output_lines): Pass src->maximum_count.
2495 2018-09-17 Martin Liska <mliska@suse.cz>
2497 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
2498 Use processor_names table.
2499 * config/i386/i386.c (ix86_default_align): Use
2500 processor_cost_table for alignment values.
2501 (ix86_option_override_internal): Use processor_names.
2502 (ix86_function_specific_print): Likewise.
2503 * config/i386/i386.h (struct processor_costs):
2504 Add alignment values.
2505 (struct ptt): Remove and replace with const char *.
2506 * config/i386/x86-tune-costs.h (struct processor_costs):
2507 Declare default alignments for all costs.
2509 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
2511 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
2512 symbolics or VR_VARYING ranges for ABS_EXPR.
2513 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
2514 when range will wrap.
2516 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
2519 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
2520 before and after a JUMP_TABLE_DATA.
2522 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
2525 * dojump.c (do_compare_and_jump): Canonicalize function pointers
2526 when one operand is a function pointer. Use POINTER_TYPE_P and
2527 FUNC_OR_METHOD_TYPE_P.
2528 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
2529 * fold-const.c (build_range_check): Likewise.
2530 * match.pd (simple_comparison): Likewise.
2532 2018-09-14 David Malcolm <dmalcolm@redhat.com>
2535 * spellcheck.c (get_edit_distance_cutoff): New function.
2536 (selftest::test_edit_distance_unit_test_oneway): Rename to...
2537 (selftest::test_get_edit_distance_one_way): ...this.
2538 (selftest::test_get_edit_distance_unit): Rename to...
2539 (selftest::test_get_edit_distance_both_ways): ...this.
2540 (selftest::test_edit_distances): Move tests to this new function,
2541 and test some more pairs of strings. Update for above renaming.
2542 (selftest::get_old_cutoff): New function.
2543 (selftest::test_get_edit_distance_cutoff): New function.
2544 (selftest::assert_suggested_for): New function.
2545 (ASSERT_SUGGESTED_FOR): New macro.
2546 (selftest::assert_not_suggested_for): New function.
2547 (ASSERT_NOT_SUGGESTED_FOR): New macro.
2548 (selftest::test_suggestions): New function.
2549 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
2550 tests to selftest::test_edit_distances and call it. Add calls to
2551 selftest::test_get_edit_distance_cutoff and
2552 selftest::test_suggestions.
2553 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
2554 (best_match::consider): Replace hard-coded cutoff calculation with
2556 (best_match::get_cutoff): New declaration.
2557 (best_match::get_best_meaningful_candidate): Likewise.
2559 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2561 * builtins.c (fold_builtin_strlen): Remove TODO comment.
2563 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
2566 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
2568 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
2569 terminated string literal.
2571 2018-09-14 Martin Sebor <msebor@redhat.com>
2573 * builtins.c (unterminated_array): Handle ARRAY_REF.
2574 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
2575 * builtins.h (unterminated_array): Declare extern.
2576 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
2578 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
2581 2018-09-14 Martin Sebor <msebor@redhat.com>
2582 Jeff Law <law@redhat.com>
2584 * builtins.c (unterminated_array): New.
2585 (expand_builtin_strcpy): Adjust.
2586 (expand_builtin_strcpy_args): Detect unterminated arrays.
2587 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
2588 unterminated arrays.
2589 * gimple-fold.h (get_maxval_strlen): Add argument.
2590 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
2592 * gimple-fold.c (get_range_strlen): Add argument.
2593 (get_maxval_strlen): Adjust.
2594 * gimple-fold.h (get_range_strlen): Add argument.
2596 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
2598 * config/i386/movdirintrin.h: Fix copyright year.
2600 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
2602 * reg-stack.c: Include regs.h.
2603 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
2604 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
2605 FIRST_STACK_REG, not DFmode.
2606 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
2607 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
2608 (change stack): Default register mode to the reg_raw_mode of
2609 FIRST_STACK_REG, not DFmode.
2610 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
2611 (*swapxf): Rename from swapxf.
2613 2018-09-14 Carl Love <cel@us.ibm.com>
2615 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
2616 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
2618 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
2621 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
2624 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
2627 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
2629 2018-09-14 Jason Merrill <jason@redhat.com>
2631 Fix --enable-gather-detailed-mem-stats.
2632 * hash-table.c (hash_table_usage): Change from variable to function.
2633 * hash-table.h: Adjust.
2634 * Makefile.in: Add missing dependencies on hash-table.h.
2636 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2638 PR tree-optimization/87259
2640 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
2641 execute_cse_reciprocals_1 has tried transforming.
2643 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
2645 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
2646 VR_VARYING for PLUS/MINUS_EXPR.
2648 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
2650 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
2653 2018-09-14 Richard Biener <rguenther@suse.de>
2656 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
2657 bits for the merged partition.
2659 2018-09-13 Martin Sebor <msebor@redhat.com>
2660 Bernd Edlinger <bernd.edlinger@hotmail.de>
2662 * builtins.h (c_srlen): Add argument.
2663 * builtins.c (warn_string_no_nul): New function.
2664 (c_strlen): Add argument and use it. Update recursive calls.
2665 Pass DECL argument to string_constant to get info on non
2666 terminated strings. Update *NONSTR as needed.
2667 (fold_builtin_strlen): Add argument to calls to c_strlen.
2668 Warn for unterminated arrays.
2669 (warn_string_no_null): Add prototype.
2670 * expr.c (string_constant): Update arguments. Update recursive
2671 calls appropriately. Detect missing NUL terminator and outermost
2672 declaration its missing in.
2673 Improve checks for arrays with nonzero lower bound or elements
2674 that are not a single byte. Simplify offset computation.
2675 Simplify checks for non-NUL terminated strings.
2676 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
2677 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
2679 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
2681 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
2683 * fold-const.c (c_getstr): Fix function comment. Remove unused third
2684 argument. Fix range checks.
2685 * fold-const.h (c_getstr): Adjust protoype.
2686 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
2687 string is constant but contains no NUL byte.
2689 * expr.c (string_constant): Adjust function comment.
2690 Remove bogus check for zero termination.
2692 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
2694 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
2695 (get_constant_size): Don't make STRING_CSTs larger than they are.
2696 (check_string_literal): New check function for STRING_CSTs.
2697 (output_constant): Use it.
2699 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
2702 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
2704 2018-09-13 Richard Biener <rguenther@suse.de>
2706 PR tree-optimization/87263
2707 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
2708 (struct unwind_state): Add max_rpo field.
2709 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
2710 Compute max_rpo, the max RPO number a block can be backwards reached
2711 from. Re-write non-iterating mode to a RPO ordered worklist approach,
2712 separating it from the iterating mode.
2714 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2716 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
2717 (rfs_decision): New scheduling decision.
2719 2018-09-13 Richard Biener <rguenther@suse.de>
2722 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
2723 (vn_nary_op_insert_pieces_predicated): Do not write useless
2724 valid_dominated_by_p entry outside of the allocated storage.
2726 2018-09-13 Omar Sandoval <osandov@osandov.com>
2727 Tom de Vries <tdevries@suse.de>
2730 * dwarf2out.c (is_c): New function.
2731 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
2733 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
2736 * config/aarch64/aarch64.md (*aarch64_bfxil):
2738 * config/aarch64/constraints.md (Ulc): Define.
2739 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
2741 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
2744 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
2746 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
2747 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
2748 aarch64_layout_frame call.
2749 (aarch64_expand_epilogue): Likewise.
2750 (aarch64_initial_elimination_offset): Likewise.
2751 (aarch64_get_separate_components): Likewise.
2752 (aarch64_use_return_insn_p): Likewise.
2753 (aarch64_layout_frame): Remove unneeded check.
2755 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2757 * configure.ac: Only append
2758 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
2759 gcc_config_arguments if it was never reconfigured or last reconfigure
2760 was with different arguments.
2761 * configure: Regenerated.
2763 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2764 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2767 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
2768 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
2770 2018-09-13 Jakub Jelinek <jakub@redhat.com>
2772 PR tree-optimization/87287
2773 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
2774 X % C == 0 to X % (unsigned) C == 0 optimization to ...
2775 * match.pd (X % C == 0): ... here. New optimization.
2777 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2780 * expr.h (maybe_optimize_mod_cmp): Declare.
2781 * expr.c (mod_inv): New function.
2782 (maybe_optimize_mod_cmp): New function.
2783 (do_store_flag): Use it.
2784 * cfgexpand.c (expand_gimple_cond): Likewise.
2786 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
2787 Julian Brown <julian@codesourcery.com>
2790 * gimplify.c (gimplify_scan_omp_clauses): Set
2791 target_firstprivatize_array_bases in OpenACC parallel and kernels
2792 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
2793 OpenACC data regions.
2795 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
2797 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
2798 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
2800 2018-09-12 Richard Biener <rguenther@suse.de>
2802 PR tree-optimization/87280
2803 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
2804 edge but unreachable target.
2805 (do_rpo_vn): For conservatively handling a PHI only mark
2806 the backedge executable but not the block reachable.
2808 2018-09-12 Richard Biener <rguenther@suse.de>
2810 PR tree-optimization/87266
2811 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
2814 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
2816 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
2818 ("trunc<BFP:mode><DFP_ALL:mode>2")
2819 ("trunc<DFP_ALL:mode><BFP:mode>2")
2820 ("extend<BFP:mode><DFP_ALL:mode>2")
2821 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
2822 according to the target operand type.
2824 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2825 Andreas Krebbel <krebbel@linux.ibm.com>
2827 PR tree-optimization/86844
2828 * gimple-ssa-store-merging.c
2829 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
2830 there are any overlapping stores in between them, make sure they are
2831 also coalesced or we give up completely.
2833 2018-09-12 Jakub Jelinek <jakub@redhat.com>
2836 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
2837 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
2839 2018-09-12 Tom de Vries <tdevries@suse.de>
2841 * common.opt (gdescribe-dies): Add option.
2842 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
2843 attribute for artifical and nameless decls.
2844 (dwarf2out_register_external_die): Add description attribute to
2845 external reference die.
2846 (add_desc_attribute): New functions.
2847 (gen_subprogram_die): Add description attribute to
2848 DW_TAG_call_site_parameter.
2849 * tree-pretty-print.c (print_generic_expr_to_str): New function.
2850 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
2851 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
2853 (@item -gdescribe-dies): Add.
2855 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2857 * tree-vrp.c (vrp_shift_undefined_p): Remove.
2858 (extract_range_from_binary_expr_1: Call
2859 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
2860 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
2863 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
2865 * gimple-ssa-warn-alloca.c
2866 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
2867 field for ALLOCA_BOUND_*_LARGE.
2869 2018-09-11 Nathan Sidwell <nathan@acm.org>
2871 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
2873 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2875 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
2876 for clobbers. Remove obsolete comment.
2878 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2880 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
2881 mpxchk, mpxld and mpxst types.
2882 (define_attr length_immediate): Remove all processing of mpx types.
2883 (define_attr prefix_0f): Ditto.
2884 (define_attr memory): Ditto.
2886 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
2888 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
2889 (log<mode>2): Change operand 1 predicate to general_operand.
2890 Extend operand 1 to XFmode and generate logxf3 insn.
2891 (log10<mode>2): Change operand 1 predicate to general_operand.
2892 Extend operand 1 to XFmode and generate log10xf3 insn.
2893 (log2<mode>2): Change operand 1 predicate to general_operand.
2894 Extend operand 1 to XFmode and generate log2xf3 insn.
2895 (fyl2xp1_extend<mode>xf3_i387): Remove.
2896 (log1p<mode>2): Change operand 1 predicate to general_operand.
2897 Extend operand 1 to XFmode and generate log1pxf3 insn.
2898 (fxtract_extend<mode>xf3_i387): Remove.
2899 (logb<mode>2): Change operand 1 predicate to general_operand.
2900 Extend operand 1 to XFmode and generate logbxf3 insn.
2901 (ilogb<mode>2): Change operand 1 predicate to general_operand.
2902 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
2903 (significand<mode>2): Change operand 1 predicate to general_operand.
2904 Extend operand 1 to XFmode and generate significandxf3 insn.
2906 2018-09-11 Nathan Sidwell <nathan@acm.org>
2908 * gcc.c (perror_with_name, pfatal_with_name): Delete.
2909 (load_specs): Use fatal_error.
2910 (DELETE_IF_ORDINARY, process_command): Use error.
2911 (execute, run_attempt): Use fatal_error.
2913 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
2915 * diagnostic-core.h (sorry_at): New prototype.
2916 * diagnostic.c (sorry_at): New function.
2918 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
2920 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
2921 by zero as VR_UNDEFINED.
2923 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2925 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
2926 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
2927 (<sincos>mode2): New expander.
2928 (sincos_extend<mode>xf3_i387): Remove insn pattern.
2929 (sincos -> sin, cos splitters): Remove splitter patterns.
2930 (sincos<mode>3): Change operand 2 predicate to general_operand.
2931 Extend operand 2 to XFmode and generate sincosxf3 insn.
2932 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
2933 Change operand 3 predicate to const1_operand.
2934 (fptan_extend<mode>xf4_i387): Remove insn pattern.
2935 (tanxf2): Update operands in the call to fptanxf4_i387.
2936 (tan<mode>2): Change operand 1 predicate to general_operand.
2937 Extend operand 1 to XFmode and generate tanxf3 insn.
2938 (atan2xf3): Rename from *fpatanxf3_i387.
2939 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
2940 (atan2xf3): Remove expander.
2941 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
2942 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
2943 (atan<mode>2): Change operand 1 predicate to general_operand.
2944 Extend operand 1 to XFmode and generate atanxf3 insn.
2946 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
2948 * config/i386/i386.md (x87/SSE constant load splitter): Use
2949 memory_operand instead of nonimmediate_operand for input operand
2952 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
2954 * config/i386/i386.md (float partial SSE register stall splitter): Move
2955 splitter near its instruction pattern.
2956 (float_extend partial SSE register stall splitter): Ditto.
2957 (float_truncate partial SSE register stall splitter): Ditto.
2959 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
2962 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2963 to speculation_safe_value_not_needed.
2966 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2967 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2968 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2969 leaf_function_p, instead use has_hard_reg_initial_val.
2971 2018-09-09 Nathan Sidwell <nathan@acm.org>
2973 * gcc.h (pfatal_with_name): Don't declare here.
2974 * gcc.c (pfatal_with_name): Make static.
2976 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
2978 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2981 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
2983 PR rtl-optimization/85458
2984 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2985 priority hook to reduce the priority of EXPR.
2987 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
2989 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2990 DImode for x87 on 32bit targets. Conditionally disable x87 modes
2991 with X87_ENABLE_FLOAT. Remove preparation code.
2992 (*float<SWI48:mode><MODEF:mode>2): Rename from
2993 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
2994 math using "enabled" attribute.
2995 (*floatdi<MODEF:mode>2_i387): Rename from
2996 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
2997 enable for 32bit targets only.
2998 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
3000 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
3001 as operand 1 predicate. Rewrite as define_insn_and_split.
3002 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
3004 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
3006 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
3007 to fallthru to FLOAT case.
3009 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3012 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
3013 around folding of vec_sl to handle out of range shift values.
3015 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
3017 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
3018 Update callers to gen_fix_trunc<mode>_i387_fisttp
3019 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
3020 nonimmediate_operand.
3021 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
3022 and corresponding splitters.
3023 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
3024 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
3025 (fix_truncdi_i387_with_temp): Remove insn pattern
3026 and corresponding splitters.
3027 (fix_trunc<mode>_i387): Change operand 0 predicate to
3028 nonimmediate_operand.
3029 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
3030 and corresponding splitters.
3031 (*fistdi2_1): Remove.
3033 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
3034 (lrintxfdi2): Remove expander. Reimplement as define_insn.
3035 (*fist<mode>2_1): Remove.
3036 (fist<mode>2): Ditto.
3037 (fist<mode>2_with_temp): Remove insn pattern and corresponding
3039 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
3040 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
3041 (fistdi2_<rounding>): Change operand 0 predicate to
3042 nonimmediate_operand.
3043 (fistdi2_<rounding>_with_temp): Remove insn pattern
3044 and corresponding splitters.
3045 (fist<mode>2_<rounding>): Change operand 0 predicate to
3046 nonimmediate_operand.
3047 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
3048 and corresponding splitters.
3050 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
3052 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
3054 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
3057 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3059 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
3060 early gimple folding of vec_splat().
3061 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
3062 * gimple-fold.h: Add an extern define for tree_vec_extract().
3064 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
3066 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
3067 wrappers around TREE_TYPE comparisons.
3069 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3072 * config/s390/predicates.md: Add nonsym_memory_operand.
3073 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
3074 contains a SYMBOL_REF, load it into an intermediate pseudo.
3075 (s390_emit_compare_and_swap): Legitimize operand.
3076 * config/s390/s390.md: Use the new nonsym_memory_operand
3077 with UNSPECV_CAS patterns.
3079 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3082 * config/s390/s390-passes.def: New file.
3083 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
3085 (make_pass_s390_early_mach): Add declaration.
3086 * config/s390/s390.c (make_pass_s390_early_mach):
3087 (s390_option_override): Remove dynamic registration.
3088 * config/s390/t-s390: Add s390-passes.def.
3090 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
3092 * config/s390/s390.c (s390_decompose_constant_pool_ref):
3093 Remove UNSPEC_LTREL_BASE check.
3094 (annotate_constant_pool_refs): Likewise.
3095 (find_constant_pool_ref): Likewise.
3096 (find_ltrel_base): Removed.
3097 (replace_ltrel_base): Removed.
3098 (s390_mainpool_finish): Remove replace_ltrel_base call.
3099 (s390_chunkify_start): Remove pending LTREL_BASE logic.
3100 (s390_chunkify_finish): Remove replace_ltrel_base call.
3101 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
3103 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
3106 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
3107 to speculation_safe_value_not_needed.
3109 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
3110 Bernd Schmidt <bernds_cb1@t-online.de>
3112 * config/nvptx/nvptx-opts.h: New file.
3113 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
3114 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
3116 (TARGET_SM35): New macro.
3117 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
3119 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
3121 (misa=): New option.
3122 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
3124 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
3126 * config/i386/i386.md (truncdfsf2): Remove expander.
3127 (truncdfsf2_with_temp): Ditto.
3128 (truncxf<mode>2): Ditto.
3129 (*truncdfsf_fast_mixed): Remove insn pattern.
3130 (*truncdfsf_fast_i387): Ditto.
3131 (*truncdfsf_mixed): Ditto.
3132 (*truncdfsf_i387): Ditto.
3133 (*truncdfsf2_i387_1): Ditto.
3134 (*truncxfsf2_mixed): Ditto.
3135 (*truncxfdf2_mixed): Ditto.
3136 (*truncxf<mode>2_i387_noop): Ditto. Update callers
3137 to call gen_truncxf<mode>2 instead.
3138 (*truncxf<mode>2_i387): Remove.
3139 (reg->reg splitters): Remove splitter pattern.
3140 (reg->mem splitters): Ditto.
3142 (truncdfsf2): New insn pattern.
3143 (truncxf<mode>2): Ditto.
3145 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3147 * tree-ssa-math-opts.c (is_mult_by): New function.
3148 (is_square_of): Use the above.
3149 (optimize_recip_sqrt): New function.
3150 (pass_cse_reciprocals::execute): Use the above.
3152 2018-09-05 Richard Biener <rguenther@suse.de>
3155 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
3156 to zero-init the emplaced vec.
3158 2018-09-05 Martin Liska <mliska@suse.cz>
3160 PR tree-optimization/87205
3161 * tree-switch-conversion.c (pass_lower_switch::execute):
3162 Group cases for switch statements.
3164 2018-09-05 Richard Biener <rguenther@suse.de>
3166 PR tree-optimization/87217
3167 * tree-ssa-sccvn.c (vuse_valueize): New.
3168 (vn_reference_lookup_pieces): Use it.
3169 (vn_reference_lookup): Likewise.
3171 2018-09-05 Nathan Sidwell <nathan@acm.org>
3174 * stor-layout.c (place_field): Scan forwards to check last
3175 bitfield when ms_bitfield_placement is in effect.
3177 2018-09-05 Richard Biener <rguenther@suse.de>
3180 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
3183 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
3184 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
3186 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
3187 * config.gcc (extra_objs): Build it.
3188 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
3190 * config/aarch64/aarch64-passes.def
3191 (pass_tag_collision_avoidance): New pass.
3192 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
3193 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
3194 (aarch64_classify_address): Remove static qualifier.
3195 (aarch64_address_info, aarch64_address_type): Move to...
3196 * config/aarch64/aarch64-protos.h: ... here.
3197 (make_pass_tag_collision_avoidance): New function.
3198 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
3201 2018-09-05 Martin Liska <mliska@suse.cz>
3203 * doc/gcov.texi: Update documentation of humar
3205 * gcov.c (format_count): Print one decimal place, it provides
3206 more fine number of situations like '1G' vs. '1.4G'.
3208 2018-09-05 Martin Liska <mliska@suse.cz>
3211 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
3212 * optc-gen.awk: Allow 'Var' for Deprecated options in order
3213 to generate a MASK value.
3215 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
3218 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
3219 if frame pointer isn't used.
3220 (compute_frame_pointer_to_fb_displacement): Likewise.
3222 2018-09-04 Jakub Jelinek <jakub@redhat.com>
3225 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
3226 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
3227 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
3228 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
3229 and OPTION_MASK_ISA_XSAVEC_UNSET.
3231 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
3233 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
3234 XOR operations in NAND case.
3236 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3238 * wide-int-range.cc (wide_int_range_convert): New.
3239 * wide-int-range.h (wide_int_range_convert): New.
3240 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
3241 code into wide_int_range_convert.
3242 (extract_range_into_wide_ints): Do not munge anti range constants
3243 into the entire domain. Just return the range back.
3245 2018-09-04 Martin Liska <mliska@suse.cz>
3247 * genmatch.c (output_line_directive): Add new argument
3249 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
3251 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3253 * doc/invoke.texi (Option Summary): Add whitespace.
3255 * doc/invoke.texi (Option Summary): Add -Waligned-new.
3257 2018-09-04 Richard Biener <rguenther@suse.de>
3259 PR tree-optimization/87211
3260 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
3261 backedge value we're supposed to treat as VARYING also number
3262 the PHI to VARYING in case it got a different value-number already.
3264 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
3266 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
3267 (extract_range_from_binary_expr_1): Do not call
3268 vrp_can_optimize_bit_op.
3269 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
3271 (wide_int_range_get_mask_and_bounds): New.
3272 (wide_int_range_optimize_bit_op): New.
3273 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
3274 (wide_int_range_bit_and): Same.
3275 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
3276 (wide_int_range_optimize_bit_op): New.
3277 (wide_int_range_get_mask_and_bounds): New.
3279 2018-09-04 Richard Biener <rguenther@suse.de>
3281 PR tree-optimization/87176
3282 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
3283 variable. When value-numbering a virtual PHI node make sure
3284 to not value-number to the backedge value.
3286 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
3288 * doc/extend.texi (Long Long, Hex Floats): Document support for
3289 long long and hex floats in more recent versions of ISO C++.
3291 2018-09-03 Richard Biener <rguenther@suse.de>
3293 PR tree-optimization/87177
3294 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
3297 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3299 * bb-reorder.c (edge_order): Convert to C-qsort-style
3300 tri-state comparator.
3301 (reorder_basic_blocks_simple): Change std::stable_sort to
3304 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3306 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
3307 tri-state comparator.
3308 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
3310 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3312 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
3313 (mergesort): ... here as maximum count for using netsort.
3314 (gcc_qsort): Set nlim to 3 if stable sort is requested.
3315 (gcc_stablesort): New.
3316 * system.h (gcc_stablesort): Declare.
3318 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
3320 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
3321 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
3322 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
3324 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
3326 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
3327 lxsdx and stxsdx alternatives.
3328 (*mov<mode>_hardfloat64): Ditto.
3329 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
3331 2018-09-03 Richard Biener <rguenther@suse.de>
3333 PR tree-optimization/87200
3334 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
3337 2018-09-03 Martin Liska <mliska@suse.cz>
3339 PR tree-optimization/87201
3340 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3341 Fix parenthesis in an expression.
3343 2018-09-03 Richard Biener <rguenther@suse.de>
3345 PR tree-optimization/87197
3346 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
3347 visited. CSE the VN_INFO hashtable lookup.
3349 PR tree-optimization/87169
3350 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
3351 iterating make sure there's no extra backedges from irreducible
3352 regions feeding the header. Mark the destination block
3355 2018-09-03 Martin Liska <mliska@suse.cz>
3358 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
3359 * common/common-targhooks.c (default_get_valid_option_values):
3361 * common/common-targhooks.h (default_get_valid_option_values):
3363 * common/config/i386/i386-common.c: Move processor_target_table
3365 (ix86_get_valid_option_values): New function.
3366 (TARGET_GET_VALID_OPTION_VALUES): New macro.
3367 * config/i386/i386.c (struct ptt): Move to i386-common.c.
3368 (PTA_*): Move all defined masks into i386-common.c.
3369 (ix86_function_specific_restore): Use new processor_cost_table.
3370 * config/i386/i386.h (struct ptt): Moved from i386.c.
3371 (struct pta): Likewise.
3372 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
3373 * doc/tm.texi.in: Likewise.
3374 * opt-suggestions.c (option_proposer::suggest_option):
3375 Pass prefix to build_option_suggestions.
3376 (option_proposer::get_completions): Likewise.
3377 (option_proposer::build_option_suggestions): Use the new target
3379 * opts.c (struct option_help_tuple): New struct.
3380 (print_filtered_help): Use the new target hook.
3382 2018-09-03 Martin Liska <mliska@suse.cz>
3385 * predict.c (set_even_probabilities): Add likely_edges
3386 argument and handle cases where we have precisely one
3388 (combine_predictions_for_bb): Catch also likely_edges.
3389 (tree_predict_by_opcode): Handle gswitch statements.
3390 * tree-cfg.h (find_case_label_for_value): New declaration.
3391 (find_taken_edge_switch_expr): Likewise.
3392 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
3393 Find pivot in decision tree based on probabily, not by number of
3396 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
3398 * doc/standards.texi (Standards): Update Objective-C reference.
3400 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3402 * doc/install.texi (Prerequisites): Update link for MPC.
3404 2018-09-01 Michael Matz <matz@suse.de>
3406 PR tree-optimization/87074
3407 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
3408 PHIs for outer-loop uses.
3410 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3412 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
3413 * doc/invoke.texi (C Dialect Options): Ditto.
3415 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
3417 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
3419 2018-08-31 Richard Biener <rguenther@suse.de>
3421 PR tree-optimization/87168
3422 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
3423 (rpo_elim::eliminate_avail): When OP was not visited it must
3426 2018-08-31 David Malcolm <dmalcolm@redhat.com>
3428 * tree-vrp.c (copy_value_range): Convert param "from" from
3429 "value_range *" to "const value_range *".
3430 (range_is_null): Likewise for param "vr".
3431 (range_int_cst_p): Likewise.
3432 (range_int_cst_singleton_p): Likewise.
3433 (symbolic_range_p): Likewise.
3434 (value_ranges_intersect_p): Likewise for both params.
3435 (value_range_nonnegative_p): Likewise for param "vr".
3436 (value_range_constant_singleton): Likewise.
3437 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
3438 (extract_range_into_wide_ints): Likewise for param "vr".
3439 (extract_range_from_multiplicative_op): Likewise for params "vr0"
3441 (vrp_can_optimize_bit_op): Likewise.
3442 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
3444 (extract_range_from_unary_expr): Likewise.
3445 (debug_value_range): Likewise for param "vr".
3446 (value_range::dump): Add "const" qualifier.
3447 (vrp_prop::check_array_ref): Convert local "vr" from
3448 "value_range *" to "const value_range *".
3449 (vrp_prop::check_mem_ref): Likewise.
3450 (vrp_prop::visit_stmt): Likewise for local "old_vr".
3451 (vrp_intersect_ranges_1): Likewise for param "vr_1".
3452 (vrp_intersect_ranges): Likewise.
3453 (simplify_stmt_for_jump_threading): Likewise for local "vr".
3454 (vrp_prop::vrp_finalize): Likewise.
3455 * tree-vrp.h (value_range::dump): Add "const" qualifier.
3456 (vrp_intersect_ranges): Add "const" qualifier to params as above.
3457 (extract_range_from_unary_expr): Likewise.
3458 (value_range_constant_singleton): Likewise.
3459 (symbolic_range_p): Likewise.
3460 (copy_value_range): Likewise.
3461 (extract_range_from_binary_expr_1): Likewise.
3462 (range_int_cst_p): Likewise.
3463 (vrp_set_zero_nonzero_bits): Likewise.
3464 (range_int_cst_singleton_p): Likewise.
3466 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
3468 * config/aarch64/arm_neon.h (vabsd_s64): New.
3469 (vnegd_s64): Likewise.
3471 2018-08-31 Martin Jambor <mjambor@suse.cz>
3473 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
3475 2018-08-31 Martin Liska <mliska@suse.cz>
3477 * ipa-icf.c (sem_item::add_type): Use
3478 sem_item::m_type_hash_cache.
3479 * ipa-icf.h: Move the cache from sem_item_optimizer
3482 2018-08-31 Nathan Sidwell <nathan@acm.org>
3484 * doc/extend.texi (Backwards Compatibility): Remove implicit
3485 extern C leeway of () being (...).
3487 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3489 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
3491 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
3495 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
3497 2018-08-31 Jakub Jelinek <jakub@redhat.com>
3500 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
3501 gen_int_mode. Formatting fixes.
3503 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
3505 * target.def (custom_function_descriptors): Improve documentation.
3506 * doc/tm.texi.in (Trampolines): Expand discussion of function
3507 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
3508 beginning of the section.
3509 * doc/tm.texi: Regenerated.
3511 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
3513 * cfg.h (class auto_edge_flag): Spell out the template-id of the
3514 base class in the initializer list. This is a workaround for
3515 building with older GCC.
3516 (class auto_bb_flag): Likewise.
3518 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
3520 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
3521 (altivec_vcmpequ<VI_char>_p): Remove star.
3522 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
3524 (expand_strncmp_vec_sequence): New function.
3525 (emit_final_str_compare_vec): New function.
3526 (expand_strn_compare): Add support for vector strncmp.
3527 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
3528 length specification to bytes.
3529 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
3530 (vcmpnezb_p): New pattern.
3531 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
3532 for option -mstring-compare-inline-limit.
3534 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
3536 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
3537 (PTA_SKYLAKE): Add PTA_AES.
3538 (PTA_GOLDMONT): Likewise.
3540 2018-08-29 Jan Hubicka <jh@suse.cz>
3543 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
3544 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
3546 2018-08-29 Jan Hubicka <jh@suse.cz>
3548 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
3550 (hash_tree): Do not visit TYPE_STUB_DECL.
3551 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
3552 stream TYPE_STUB_DECL.
3553 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
3554 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
3555 after free_lang_data.
3556 (type_in_anonymous_namespace_p): Likewise.
3558 2018-08-29 Jan Hubicka <jh@suse.cz>
3560 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3561 comment that it has to be even number.
3562 (class sreal): Change m_sig type to int32_t.
3563 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3564 int64_t for temporary calculations.
3565 (sreal_verify_basics): Drop one bit from minimum and maximum.
3567 2018-08-30 Richard Biener <rguenther@suse.de>
3569 PR tree-optimization/87147
3570 * tree-ssa-sccvn.c (SSA_VISITED): New function.
3571 (visit_phi): When the degenerate result is from the backedge and
3572 we didn't visit its definition yet drop to VARYING.
3573 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
3575 2018-08-29 Jan Hubicka <jh@suse.cz>
3577 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
3579 (hash_tree): Likewise.
3581 2018-08-29 Jan Hubicka <jh@suse.cz>
3583 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
3584 and TYPE_NEXT_REF_TO.
3586 2018-08-29 Jan Hubicka <jh@suse.cz>
3588 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
3589 comment that it has to be even number.
3590 (class sreal): Change m_sig type to int32_t.
3591 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
3592 int64_t for temporary calculations.
3593 (sreal_verify_basics): Drop one bit from minimum and maximum.
3595 2018-08-30 Tamar Christina <tamar.christina@arm.com>
3597 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
3599 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
3602 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
3603 if to_add is negative.
3605 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
3608 * builtins.c (c_strlen): Improve range checks.
3610 2018-08-29 Martin Sebor <msebor@redhat.com>
3611 Jeff Law <law@redhat.com>
3613 PR tree-optimization/86714
3614 PR tree-optimization/86711
3615 * builtins.c (c_strlen): Add arguments to call to string_constant.
3616 * expr.c (string_constant): Add argument. Detect missing nul
3617 terminator and outermost declaration it's missing in.
3618 * expr.h (string_constant): Add argument.
3619 * fold-const.c (read_from_constant_string): Add arguments to call to
3621 (c_getstr): Likewise.
3622 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3624 * tree-ssa-strlen.c (get_stridx): Likewise.
3626 2018-08-29 Jan Hubicka <jh@suse.cz>
3628 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
3629 Do not stream DECL_VINDEX.
3630 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
3631 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
3632 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
3634 2018-08-29 Richard Biener <rguenther@suse.de>
3636 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
3637 virtual operands that are not default defs to honor region
3639 (rpo_vn_valueize): Remove ineffective code here.
3641 2018-08-29 Richard Biener <rguenther@suse.de>
3643 PR tree-optimization/87132
3644 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
3645 when skipping defs reachable over backedges.
3647 2018-08-29 Richard Biener <rguenther@suse.de>
3649 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
3650 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
3651 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
3652 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
3653 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
3654 (vn_reference_lookup_3): Remove use of const_parms.
3655 (free_rpo_vn): Do not free const_parms.
3656 (do_rpo_vn): Do not call init_const_parms.
3657 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
3658 SSA_NAME_POINTS_TO_READONLY_MEMORY.
3659 (call_may_clobber_ref_p_1): Likewise.
3661 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
3664 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
3666 (-ftree-scev-cprop): Document.
3668 2018-08-29 Jan Hubicka <jh@suse.cz>
3670 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
3672 (sreal constructor): Update.
3673 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
3674 sreal:operator/): Update.
3676 2018-08-29 Martin Liska <mliska@suse.cz>
3678 * tree-switch-conversion.c (switch_conversion::expand):
3679 Strenghten assumption about gswitch statements.
3681 2018-08-29 Richard Biener <rguenther@suse.de>
3683 PR tree-optimization/87117
3684 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
3685 re-value-number released SSA VDEFs.
3687 2018-08-29 Richard Biener <rguenther@suse.de>
3689 PR tree-optimization/87126
3690 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
3692 2018-08-28 Jim Wilson <jimw@sifive.com>
3694 * config/riscv/pic.md: Rewrite.
3695 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
3697 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
3698 (SOFTF, default_load, softload, softstore): New.
3700 2018-08-28 Jeff Law <law@redhat.com>
3702 * fold-const.c (fold_binary_loc): Remove recently added assert.
3704 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
3706 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
3707 to OP parmeter of generated function.
3709 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
3711 PR tree-optimization/87009
3712 * match.pd: Add boolean optimizations.
3714 2018-08-28 Martin Sebor <msebor@redhat.com>
3717 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
3718 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
3719 (pass_walloca::gate): Use it.
3720 (alloca_call_type): Same.
3721 (pass_walloca::execute): Same.
3722 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
3724 2018-08-28 David Malcolm <dmalcolm@redhat.com>
3726 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
3727 GCC_VERSION for usage of "__gcc_dump_printf__" format from
3730 2018-08-28 Richard Biener <rguenther@suse.de>
3732 PR tree-optimization/87124
3733 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
3734 constants before looking up avail.
3736 2018-08-28 Jakub Jelinek <jakub@redhat.com>
3739 * calls.c (maybe_warn_nonstring_arg): Punt early if
3740 warn_stringop_overflow is zero. Don't call get_range_strlen
3741 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
3742 Swap comparison operands to have constants on rhs. Only use
3743 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
3744 increment lenrng[0].
3746 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
3748 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
3749 use of tree_to_shwi. Remove duplicated test for the size being
3750 a whole number of bytes.
3752 2018-08-28 Richard Biener <rguenther@suse.de>
3754 PR tree-optimization/87117
3755 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
3756 Handle removed stmt without LHS (GIMPLE_NOP).
3758 2018-08-28 Richard Biener <rguenther@suse.de>
3760 PR tree-optimization/87117
3761 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
3762 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
3764 2018-08-28 Richard Biener <rguenther@suse.de>
3766 PR tree-optimization/87117
3767 * tree-ssa-pre.c (compute_avail): Do not make expressions
3768 with predicated values available.
3769 (get_expr_value_id): Assert we do not run into predicated value
3772 2018-08-28 Richard Biener <rguenther@suse.de>
3774 PR tree-optimization/87117
3775 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
3776 get virtual operands.
3777 (get_expr_operands): Handle STRING_CST like other decls.
3779 2018-08-28 Martin Liska <mliska@suse.cz>
3781 * tree.h: Update documentation of fndecl_built_in_p
3785 2018-08-27 Jeff Law <law@redhat.com>
3786 PR tree-optimization/87110
3787 * tree-ssa-dse.c (compute_trims): Handle non-constant
3790 2018-08-27 Martin Sebor <msebor@redhat.com>
3792 PR tree-optimization/86914
3793 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
3795 2018-08-27 Martin Sebor <msebor@redhat.com>
3797 PR tree-optimization/87112
3798 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
3799 the type of the bound argument.
3801 2018-08-27 Jeff Law <law@redhat.com>
3803 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
3804 type does not have a TYPE_SIZE_UNIT.
3806 2018-08-27 Steve Ellcey <sellcey@cavium.com>
3808 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
3809 with include of backend.h.
3811 2018-08-27 Richard Biener <rguenther@suse.de>
3813 PR tree-optimization/86927
3814 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
3815 use const cond reduction code.
3817 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
3819 PR tree-optimization/85758
3820 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
3822 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3825 * diagnostic-show-locus.c (class layout_range): Update for
3826 conversion of show_caret_p to a tri-state.
3827 (layout_range::layout_range): Likewise.
3828 (make_range): Likewise.
3829 (layout::maybe_add_location_range): Likewise.
3830 (layout::should_print_annotation_line_p): Don't show annotation
3831 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
3832 (layout::get_state_at_point): Update for conversion of
3833 show_caret_p to a tri-state. Bail out early for
3834 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
3835 underlining or source colorization.
3836 (gcc_rich_location::add_location_if_nearby): Update for conversion
3837 of show_caret_p to a tri-state.
3838 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
3839 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
3841 (selftest::test_one_liner_labels): Likewise.
3842 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
3843 conversion of show_caret_p to a tri-state.
3844 * pretty-print.c (text_info::set_location): Likewise.
3845 * pretty-print.h (text_info::set_location): Likewise.
3846 * substring-locations.c (format_warning_n_va): Likewise.
3847 * tree-diagnostic.c (default_tree_printer): Likewise.
3848 * tree-pretty-print.c (newline_and_indent): Likewise.
3850 2018-08-27 David Malcolm <dmalcolm@redhat.com>
3853 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
3854 line above for line-insertion fix-it hints.
3855 (selftest::test_fixit_insert_containing_newline): Update the
3856 expected results, and add a test with line-numbering enabled.
3858 2018-08-27 Martin Liska <mliska@suse.cz>
3861 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
3862 params with DECL_HAS_VALUE_EXPR_P.
3864 2018-08-27 Martin Liska <mliska@suse.cz>
3866 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
3867 selected expansion strategy.
3869 2018-08-27 Martin Liska <mliska@suse.cz>
3871 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
3872 * builtins.c (is_builtin_fn): Likewise.
3873 * attribs.c (diag_attr_exclusions): Use new function
3874 fndecl_built_in_p and remove check for FUNCTION_DECL if
3876 (builtin_mathfn_code): Likewise.
3877 (fold_builtin_expect): Likewise.
3878 (fold_call_expr): Likewise.
3879 (fold_builtin_call_array): Likewise.
3880 (fold_call_stmt): Likewise.
3881 (set_builtin_user_assembler_name): Likewise.
3882 (is_simple_builtin): Likewise.
3883 * calls.c (gimple_alloca_call_p): Likewise.
3884 (maybe_warn_nonstring_arg): Likewise.
3885 * cfgexpand.c (expand_call_stmt): Likewise.
3886 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
3887 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
3888 (cgraph_node::verify_node): Likewise.
3889 * cgraphclones.c (build_function_decl_skip_args): Likewise.
3890 (cgraph_node::create_clone): Likewise.
3891 * config/arm/arm.c (arm_insert_attributes): Likewise.
3892 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
3893 * dse.c (scan_insn): Likewise.
3894 * expr.c (expand_expr_real_1): Likewise.
3895 * fold-const.c (operand_equal_p): Likewise.
3896 (fold_binary_loc): Likewise.
3897 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
3898 * gimple-low.c (lower_stmt): Likewise.
3899 * gimple-pretty-print.c (dump_gimple_call): Likewise.
3900 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
3902 * gimple.c (gimple_build_call_from_tree): Likewise.
3903 (gimple_call_builtin_p): Likewise.
3904 (gimple_call_combined_fn): Likewise.
3905 * gimplify.c (gimplify_call_expr): Likewise.
3906 (gimple_boolify): Likewise.
3907 (gimplify_modify_expr): Likewise.
3908 (gimplify_addr_expr): Likewise.
3909 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
3910 * ipa-cp.c (determine_versionability): Likewise.
3911 * ipa-fnsummary.c (compute_fn_summary): Likewise.
3912 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
3913 * ipa-split.c (visit_bb): Likewise.
3914 (split_function): Likewise.
3915 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
3916 * lto-cgraph.c (input_node): Likewise.
3917 * lto-streamer-out.c (write_symbol): Likewise.
3918 * omp-low.c (setjmp_or_longjmp_p): Likewise.
3919 (lower_omp_1): Likewise.
3920 * predict.c (strip_predict_hints): Likewise.
3921 * print-tree.c (print_node): Likewise.
3922 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
3923 * trans-mem.c (is_tm_irrevocable): Likewise.
3924 (is_tm_load): Likewise.
3925 (is_tm_simple_load): Likewise.
3926 (is_tm_store): Likewise.
3927 (is_tm_simple_store): Likewise.
3928 (is_tm_abort): Likewise.
3929 (tm_region_init_1): Likewise.
3930 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
3931 * tree-cfg.c (verify_gimple_call): Likewise.
3932 (move_stmt_r): Likewise.
3933 (stmt_can_terminate_bb_p): Likewise.
3934 * tree-eh.c (lower_eh_constructs_2): Likewise.
3935 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
3936 * tree-inline.c (remap_gimple_stmt): Likewise.
3937 (copy_bb): Likewise.
3938 (estimate_num_insns): Likewise.
3939 (fold_marked_statements): Likewise.
3940 * tree-sra.c (scan_function): Likewise.
3941 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
3942 (optimize_stack_restore): Likewise.
3943 (pass_fold_builtins::execute): Likewise.
3944 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
3945 (mark_all_reaching_defs_necessary_1): Likewise.
3946 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
3947 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
3948 (pass_forwprop::execute): Likewise.
3949 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3950 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3951 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
3952 * tree-ssa-strlen.c (get_string_length): Likewise.
3953 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
3954 (find_func_aliases_for_call): Likewise.
3955 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
3956 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
3957 * tree-tailcall.c (find_tail_calls): Likewise.
3958 * tree.c (need_assembler_name_p): Likewise.
3959 (free_lang_data_in_decl): Likewise.
3960 (get_call_combined_fn): Likewise.
3961 * ubsan.c (is_ubsan_builtin_p): Likewise.
3962 * varasm.c (incorporeal_function_p): Likewise.
3963 * tree.h (DECL_BUILT_IN): Remove and replace with
3965 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3966 (fndecl_built_in_p): New.
3968 2018-08-27 Martin Liska <mliska@suse.cz>
3970 PR tree-optimization/86847
3971 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3972 Dump also subtree probability.
3973 (switch_decision_tree::do_jump_if_equal): New function.
3974 (switch_decision_tree::emit_case_nodes): Handle special
3975 situations in balanced tree that can be emitted much simpler.
3976 Fix calculation of probabilities that happen in tree expansion.
3977 * tree-switch-conversion.h (struct cluster): Add
3979 (struct simple_cluster): Likewise.
3980 (struct case_tree_node): Add new function has_child.
3981 (do_jump_if_equal): New.
3983 2018-08-27 Martin Liska <mliska@suse.cz>
3985 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3986 Add new argument to bit_test_cluster constructor.
3987 (bit_test_cluster::emit): Set bits really number of values
3989 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3990 probability argument.
3991 * tree-switch-conversion.h (struct bit_test_cluster):
3992 Add m_handles_entire_switch.
3994 2018-08-27 Martin Liska <mliska@suse.cz>
3996 PR tree-optimization/86702
3997 * tree-switch-conversion.c (jump_table_cluster::emit):
3998 Make probabilities even for values in jump table
3999 according to number of cases handled.
4000 (switch_decision_tree::compute_cases_per_edge): Pass
4001 argument to reset_out_edges_aux function.
4002 (switch_decision_tree::analyze_switch_statement): Likewise.
4003 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
4006 2018-08-27 Martin Liska <mliska@suse.cz>
4008 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
4009 cfun argument explicitly.
4010 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
4011 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
4012 function gimple_switch_default_bb.
4013 (convert_switch_statements):
4015 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
4016 * stmt.c (label_to_block_fn): Use label_to_block and pass
4017 cfun argument explicitly and use gimple_switch_label_bb.
4018 (expand_case): Likewise.
4019 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
4020 cfun argument explicitly. Likewise.
4021 (make_edges_bb): Likewise.
4022 (make_cond_expr_edges): Likewise.
4023 (get_cases_for_edge): Likewise.
4024 (make_gimple_switch_edges): Likewise.
4025 (label_to_block_fn): Likewise.
4026 (label_to_block): Likewise.
4027 (make_goto_expr_edges): Likewise.
4028 (make_gimple_asm_edges): Likewise.
4029 (main_block_label): Likewise.
4030 (group_case_labels_stmt): Likewise.
4031 (find_taken_edge_computed_goto): Likewise.
4032 (find_taken_edge_switch_expr): Likewise.
4033 (gimple_verify_flow_info): Likewise.
4034 (gimple_redirect_edge_and_branch): Likewise.
4035 (gimple_switch_label_bb): New function.
4036 (gimple_switch_default_bb): Likewise.
4037 (gimple_switch_edge): Likewise.
4038 (gimple_switch_default_edge): Likewise.
4039 * tree-cfg.h (label_to_block_fn): Remove and replace ...
4040 (label_to_block): ... with this.
4041 (gimple_switch_label_bb): New.
4042 (gimple_switch_default_bb): Likewise.
4043 (gimple_switch_edge): Likewise.
4044 (gimple_switch_default_edge): Likewise.
4045 * tree-cfgcleanup.c (convert_single_case_switch): Use
4046 new gimple functions and pass new argument to label_to_block.
4047 (cleanup_control_flow_bb):
4048 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
4049 cfun argument explicitly.
4050 (make_eh_edges): Likewise.
4051 (redirect_eh_dispatch_edge): Likewise.
4052 (lower_resx): Likewise.
4053 (lower_eh_dispatch): Likewise.
4054 (maybe_remove_unreachable_handlers): Likewise.
4055 (unsplit_eh): Likewise.
4056 (cleanup_empty_eh): Likewise.
4057 (verify_eh_edges): Likewise.
4058 (verify_eh_dispatch_edge): Likewise.
4059 * tree-ssa-dom.c (record_edge_info): Likewise.
4060 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
4061 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
4062 (thread_through_normal_block): Likewise.
4063 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
4064 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
4065 * tree-switch-conversion.c (switch_conversion::collect): Use new
4067 (switch_conversion::check_final_bb): Likewise.
4068 (switch_conversion::gather_default_values): Pass new argument
4070 (switch_conversion::build_constructors): Likewise.
4071 (switch_decision_tree::compute_cases_per_edge): Use new
4072 gimple_switch_edge function.
4073 (switch_decision_tree::analyze_switch_statement): Pass new argument
4075 (switch_decision_tree::try_switch_expansion): Use
4076 gimple_switch_default_edge.
4077 * tree-vrp.c (find_switch_asserts): Pass new argument
4079 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
4080 (vr_values::simplify_switch_using_ranges): Likewise.
4082 2018-08-27 Richard Biener <rguenther@suse.de>
4084 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
4085 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
4087 * tree-ssa-sccvn.h (struct vn_pval): New structure.
4088 (struct vn_nary_op_s): Add unwind_to member. Add
4089 predicated_values flag and put result into a union together
4090 with a linked list of vn_pval.
4091 (struct vn_ssa_aux): Add name member to make maintaining
4092 a map of SSA name to vn_ssa_aux possible. Remove no longer
4093 needed info, dfsnum, low, visited, on_sccstack, use_processed
4094 and range_info_anti_range_p members.
4095 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
4096 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
4098 (vn_valueize): New global.
4099 (vn_context_bb): Likewise.
4100 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
4101 VN_INFO_PTR_INFO): Remove.
4102 * tree-ssa-sccvn.c: ... (rewrite)
4103 (pass_fre::execute): For -O2+ initialize loops and run
4104 RPO VN in optimistic mode (iterating). For -O1 and -Og
4105 run RPO VN in non-optimistic mode.
4106 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
4107 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
4108 * doc/invoke.texi (sccvn-max-scc-size): Remove.
4109 (rpo-vn-max-loop-depth): Document.
4110 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
4111 when valuezing the VUSE signals we walked out of the region.
4112 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
4113 (phi_translate): Set VN context block to use for availability
4115 (compute_avail): Likewise.
4116 (pre_valueize): New function.
4117 (pass_pre::execute): Adjust to the RPO VN API.
4119 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
4120 (propagate_constants_for_unrolling): Remove.
4121 (tree_unroll_loops_completely): Perform value-numbering
4122 on the unrolled bodies loop parent.
4124 2018-08-27 Richard Biener <rguenther@suse.de>
4126 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
4127 for partial antic compute.
4129 2018-08-27 Jakub Jelinek <jakub@redhat.com>
4131 PR rtl-optimization/87065
4132 * combine.c (simplify_if_then_else): Formatting fix.
4133 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
4135 (known_cond): Don't return const_true_rtx for vector modes. Use
4136 CONST0_RTX instead of const0_rtx. Formatting fixes.
4138 2018-08-27 Martin Liska <mliska@suse.cz>
4140 PR gcov-profile/87069
4141 * gcov.c (process_file): Record files already processed
4142 and warn about a file being processed multiple times.
4144 2018-08-27 Martin Liska <mliska@suse.cz>
4147 * config/aarch64/aarch64.c (aarch64_override_options_internal):
4148 Set default values for x_aarch64_*_string strings.
4149 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
4150 prefix. For -mabi do not print '=ABI' in help and use
4151 <option_value> format for -msve-vector-bits and -moverride
4154 2018-08-26 Jeff Law <law@redhat.com>
4156 * config/mips/frame-header-opt.c: Include "backend.h" rather than
4159 2018-08-26 Marek Polacek <polacek@redhat.com>
4161 PR c++/87029, Implement -Wredundant-move.
4162 * doc/invoke.texi: Document -Wredundant-move.
4164 2018-08-25 Martin Sebor <msebor@redhat.com>
4166 PR tree-optimization/87059
4167 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
4168 to the same type as the other.
4169 * fold-const.c (fold_binary_loc): Assert expectation.
4171 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
4173 * config/darwin.c (machopic_legitimize_pic_address): Clean up
4174 extraneous parentheses, dead code section and formatting.
4176 2018-08-24 David Malcolm <dmalcolm@redhat.com>
4179 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
4180 wide enough for jumps in the line-numbering to be visible.
4181 (layout::print_gap_in_line_numbering): New member function.
4182 (layout::calculate_line_spans): When using line numbering, merge
4183 line spans that are only 1 line apart.
4184 (diagnostic_show_locus): When printing line numbers, show gaps in
4185 line numbering directly, rather than printing headers.
4186 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
4187 line-numbering with multiple line spans.
4188 (selftest::test_fixit_insert_containing_newline_2): Add test of
4189 line-numbering, in which the spans are close enough to be merged.
4191 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4193 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
4194 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
4195 * tree-vrp.c (range_is_nonnull): Remove.
4196 (range_includes_zero_p): Accept value_range instead of min/max.
4197 (extract_range_from_binary_expr_1): Do not early bail on
4199 Use range_includes_zero_p instead of range_is_nonnull.
4200 (extract_range_from_unary_expr): Use range_includes_zero_p instead
4201 of range_is_nonnull.
4202 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
4203 special case VR_ANTI_RANGE.
4204 (vrp_finalize): Same.
4205 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
4207 (range_is_nonnull): Remove.
4208 * vr-values.c (vrp_stmt_computes_nonzero): Use
4209 range_includes_zero_p instead of range_is_nonnull.
4210 (extract_range_basic): Pass value_range to range_includes_zero_p
4211 instead of range_is_nonnull.
4213 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
4215 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
4216 * emit-rtl.h (rtl_data): Remove return_bnd.
4217 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
4218 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
4219 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
4220 (POINTER_BOUNDS_MODE): Remove definition.
4221 (make_pointer_bounds_mode): Remove.
4222 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
4223 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
4224 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
4225 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
4226 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
4227 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
4228 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
4230 * config/i386/i386-modes.def (BND32, BND64): Remove.
4231 * config/i386/i386.c (dbx_register_map): Remove bound registers.
4232 (dbx64_register_map): Ditto.
4233 (svr4_dbx_register_map): Ditto.
4234 (indirect_thunk_bnd_needed): Remove.
4235 (indirect_thunks_bnd_used): Ditto.
4236 (indirect_return_bnd_needed): Ditto.
4237 (indirect_return_via_cx_bnd): Ditto.
4238 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
4239 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
4240 (output_indirect_thunk): Ditto. Remove need_prefix argument.
4241 (output_indirect_thunk_function): Remove handling of
4242 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
4243 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
4244 (ix86_save_reg): Remove handling of crtl->return_bnd.
4245 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
4246 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
4247 and UNSPEC_BNDLX_ADDR.
4248 (ix86_output_indirect_branch_via_reg): Remove handling of
4249 indirect_thunk_prefix_bnd.
4250 (ix86_output_indirect_branch_via_push): Ditto.
4251 (ix86_output_function_return): Ditto.
4252 (ix86_output_indirect_function_return): Ditto.
4253 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
4254 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
4255 (CALL_USED_REGISTERS): Ditto.
4256 (REG_ALLOC_ORDER): Update for removal of bound registers.
4257 (HI_REGISTER_NAMES): Ditto.
4258 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
4259 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
4260 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
4261 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
4262 (FIRST_PSEUDO_REG): Update.
4263 (BND): Remove mode iterator.
4264 * config/i386/predicates.md (bnd_mem_operator): Remove.
4266 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4268 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
4271 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
4273 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
4274 the case in which the permute needs only a single element and
4275 repeats for every vector of the result. Extend that case to
4276 handle variable-length vectors.
4277 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
4279 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
4282 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
4283 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
4285 2018-08-24 Richard Biener <rguenther@suse.de>
4287 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
4288 bb_flags_allocated members.
4289 (auto_flag): New RAII class for allocating flags.
4290 (auto_edge_flag): New RAII class for allocating edge flags.
4291 (auto_bb_flag): New RAII class for allocating bb flags.
4292 * cfgloop.c (verify_loop_structure): Allocate temporary edge
4294 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
4295 in favor of temporarily allocated BB flag.
4296 * hsa-brig.c: Re-order includes.
4297 * hsa-dump.c: Likewise.
4298 * hsa-regalloc.c: Likewise.
4299 * print-rtl.c: Likewise.
4300 * profile-count.c: Likewise.
4302 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
4305 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
4308 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
4311 * wide-int-range.cc (wide_int_range_div): Do not ignore result
4312 from wide_int_range_multiplicative_op.
4314 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4316 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
4319 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
4321 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
4324 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
4326 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
4327 full GENERIC expression used for replacement.
4329 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
4331 * tree-vrp.c (abs_extent_range): Remove.
4332 (extract_range_into_wide_ints): Pass wide ints by reference.
4333 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
4334 Pass wide ints by reference in all calls to
4335 extract_range_into_wide_ints.
4336 * wide-int-range.cc (wide_int_range_div): New.
4337 * wide-int-range.h (wide_int_range_div): New.
4338 (wide_int_range_includes_zero_p): New.
4339 (wide_int_range_zero_p): New.
4341 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
4343 * config/aarch64/aarch64.md (arches): New enum.
4344 (arch): New enum attr.
4345 (arch_enabled): New attr.
4346 (enabled): Now uses arch_enabled only.
4347 (simd, sve, fp16): Removed attribute.
4348 (fp): Attr now defined in terms of 'arch'.
4349 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
4350 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
4351 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
4352 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
4353 attributes into 'arch'.
4354 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
4355 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
4356 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
4357 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
4358 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
4359 'simd' attribute into 'arch'.
4360 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
4361 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
4362 Convert use of 'fp' attribute to 'arch'.
4363 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
4364 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
4366 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
4367 (different modes) Merge 'fp' and 'simd' into 'arch'.
4368 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
4371 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
4373 PR rtl-optimization/87026
4374 * expmed.c (canonicalize_comparison): If we can no longer create
4375 pseudoregisters, don't.
4377 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
4380 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
4382 * config/arm/arm.c (speculation_barrier_libfunc): New static
4384 (arm_init_libfuncs): Initialize it.
4385 (arm_emit_speculation_barrier): New function.
4386 * config/arm/arm.md (speculation_barrier): Call
4387 arm_emit_speculation_barrier for architectures that do not have
4389 (speculation_barrier_insn): Only match on Armv7 or later.
4391 2018-08-23 Richard Biener <rguenther@suse.de>
4394 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
4397 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4399 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
4400 of single-vector TBLs.
4401 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
4404 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
4407 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
4408 aarch64_evpc_tbl guard.
4410 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
4412 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
4415 2018-08-22 Martin Sebor <msebor@redhat.com>
4418 * tree-pretty-print.c (pretty_print_string): Add argument.
4419 (dump_generic_node): Call to pretty_print_string with string size.
4421 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
4423 PR rtl-optimization/86771
4424 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
4425 of two SETs into those two SETs, one to be placed at i2, if that SETs
4426 destination is modified between i2 and i3.
4428 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4430 PR tree-optimization/86725
4431 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
4433 (vect_analyze_scalar_cycles_1): Check it.
4435 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4437 PR tree-optimization/86725
4438 * tree-vect-loop.c (vect_is_simple_reduction): When treating
4439 an outer loop phi as a double reduction, make sure that the
4440 single user of the phi result is an inner loop phi.
4442 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4444 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
4445 grouped stores with gaps to a strided group.
4447 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
4449 * tree-vect-stmts.c (get_group_load_store_type)
4450 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
4451 first statement in a group.
4453 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4455 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
4456 the sequence used in gcc/gcc.c.
4458 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4461 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
4464 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4466 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
4467 Darwin10-specific unwinder-shim.
4468 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
4469 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
4470 New to cater for Darwin10 Rosetta.
4472 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4474 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
4475 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
4477 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
4482 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
4483 (dwarf2out_switch_text_section): Generate a local label for the
4484 second function sub-section and apply it as the second FDE start
4486 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
4487 second sub-section start.
4489 2018-08-22 Richard Biener <rguenther@suse.de>
4491 PR tree-optimization/86988
4492 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
4494 2018-08-22 Richard Biener <rguenther@suse.de>
4496 PR tree-optimization/86945
4497 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
4499 2018-08-22 Alexandre Oliva <oliva@adacore.com>
4501 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
4502 a comment about how uses of r2 for .sdata2 come about.
4504 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
4506 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
4508 2018-08-21 Marek Polacek <polacek@redhat.com>
4510 PR c++/86981, Implement -Wpessimizing-move.
4511 * doc/invoke.texi: Document -Wpessimizing-move.
4513 2018-08-21 Jan Hubicka <jh@suse.cz>
4515 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
4516 * tree.h (is_redundant_typedef): Remove.
4517 * dwarf2out.c (is_redundant_typedef): Turn into static function.
4519 2018-08-21 Jan Hubicka <jh@suse.cz>
4521 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
4524 2018-08-21 Tamar Christina <tamar.christina@arm.com>
4526 * expmed.c (extract_low_bits): Reject invalid subregs early.
4528 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4531 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
4534 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4536 * config/vxworks.h: Guard vxworks_asm_out_constructor and
4537 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
4538 * config/vxworks.c: Likewise.
4540 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4542 * config/vxworks.c: Set targetm.have_ctors_dtors
4543 if HAVE_INITFINI_ARRAY_SUPPORT.
4544 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
4545 if HAVE_INITFINI_ARRAY_SUPPORT.
4547 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
4549 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
4550 default search path for VxWorks < 7.
4552 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
4554 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
4555 (get_format_string): Refer to c_getstr.
4557 2018-08-21 Tom de Vries <tdevries@suse.de>
4559 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4560 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4561 (debuginfo_early_stop): Declare.
4562 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
4563 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
4564 (debuginfo_early_stop): New function.
4565 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
4566 and debuginfo_early_stop.
4567 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
4568 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
4569 (general_init): Call debuginfo_early_init.
4570 (finalize): Call debuginfo_fini.
4571 (do_compile): Call debuginfo_init.
4572 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
4574 (@item -fdump-debug, @item -fdump-earlydebug): Add.
4576 2018-08-21 Tom de Vries <tdevries@suse.de>
4578 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
4579 flag_dump_noaddr and flag_dump_unnumbered.
4581 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
4583 * wide-int-range.cc (wide_int_range_abs): New.
4584 (wide_int_range_order_set): Rename from wide_int_range_min_max.
4585 * wide-int-range.h (wide_int_range_abs): New.
4586 (wide_int_range_min_max): New.
4587 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
4588 case to call wide_int_range_abs.
4589 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
4590 (extract_range_from_abs_expr): Delete.
4592 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
4595 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
4596 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
4599 2018-08-20 Nathan Sidwell <nathan@acm.org>
4600 Jeff Law <law@redhat.com>
4602 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
4603 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
4605 2018-08-20 David Malcolm <dmalcolm@redhat.com>
4608 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
4609 (decl_attributes): Likewise.
4610 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
4612 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
4613 * diagnostic-core.h (class auto_diagnostic_group): New class.
4614 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
4615 (diagnostic_report_diagnostic): Handle the first diagnostics within
4617 (emit_diagnostic): Add auto_diagnostic_group instance.
4619 (inform_n): Likewise.
4620 (warning): Likewise.
4621 (warning_at): Likewise.
4622 (warning_n): Likewise.
4623 (pedwarn): Likewise.
4624 (permerror): Likewise.
4626 (error_n): Likewise.
4627 (error_at): Likewise.
4629 (fatal_error): Likewise.
4630 (internal_error): Likewise.
4631 (internal_error_no_backtrace): Likewise.
4632 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
4633 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
4634 * diagnostic.h (struct diagnostic_context): Add fields
4635 "diagnostic_group_nesting_depth",
4636 "diagnostic_group_emission_count", "begin_group_cb",
4638 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
4639 Add auto_diagnostic_group instance(s).
4640 (find_explicit_erroneous_behavior): Likewise.
4641 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
4642 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
4643 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
4644 (gimplify_va_arg_expr): Likewise.
4645 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
4646 (HSA_SORRY_AT): Likewise.
4647 * ipa-devirt.c (compare_virtual_tables): Likewise.
4648 (warn_odr): Likewise.
4649 * multiple_target.c (expand_target_clones): Likewise.
4650 * opts-common.c (cmdline_handle_error): Likewise.
4651 * reginfo.c (globalize_reg): Likewise.
4652 * substring-locations.c (format_warning_n_va): Likewise.
4653 * tree-inline.c (expand_call_inline): Likewise.
4654 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
4655 * tree-ssa-loop-niter.c
4656 (do_warn_aggressive_loop_optimizations): Likewise.
4657 * tree-ssa-uninit.c (warn_uninit): Likewise.
4658 * tree.c (warn_deprecated_use): Likewise.
4660 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
4663 * config/i386/i386.md (eh_return): Always update EH return
4664 address in word_mode.
4666 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
4668 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
4669 TARGET_SPLIT_COMPLEX_ARG is defined.
4671 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4673 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
4675 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
4678 * expr.c (expand_assignment): Assert that bitpos is positive.
4679 (store_field): Likewise
4680 (expand_expr_real_1): Make sure that bitpos is positive.
4681 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
4684 2018-08-20 Nathan Sidwell <nathan@acm.org>
4686 * Makefile.in (CPP_ID_DATA_H): Delete.
4687 (CPP_INTERNAL_H): Don't add it.
4688 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
4689 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
4691 2018-08-20 Richard Biener <rguenther@suse.de>
4693 PR tree-optimization/78655
4694 * tree-vrp.c (extract_range_from_binary_expr_1): Make
4695 pointer + offset nonnull if either operand is nonnull work.
4697 2018-08-20 Tom de Vries <tdevries@suse.de>
4699 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
4700 unless the referenced die describes the added property using
4701 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
4702 Otherwise, add a DW_AT_location to the referenced die.
4704 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
4707 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
4708 register_operand when calling ix86_set_reg_reg_cost.
4709 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
4710 Set *total to 0 for operands that satisfy x86_64_immediate_operand
4711 predicate and to 1 otherwise.
4713 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
4715 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
4716 emit a diagnostic that it is not supported and reset the option.
4717 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
4718 supported and consume the option. (ASM_FINAL_SPEC): New.
4720 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4722 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
4725 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
4727 C-SKY port: Documentation
4729 * doc/extend.texi (C-SKY Function Attributes): New section.
4730 * doc/invoke.texi (Option Summary): Add C-SKY options.
4731 (C-SKY Options): New section.
4732 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
4734 2018-08-17 Jojo <jijie_rong@c-sky.com>
4735 Huibin Wang <huibin_wang@c-sky.com>
4736 Sandra Loosemore <sandra@codesourcery.com>
4737 Chung-Lin Tang <cltang@codesourcery.com>
4739 C-SKY port: Backend implementation
4741 * config/csky/*: New.
4742 * common/config/csky/*: New.
4744 2018-08-17 Jojo <jijie_rong@c-sky.com>
4745 Huibin Wang <huibin_wang@c-sky.com>
4746 Sandra Loosemore <sandra@codesourcery.com>
4747 Chung-Lin Tang <cltang@codesourcery.com>
4748 Andrew Jenner <andrew@codesourcery.com>
4750 C-SKY port: Configury
4752 * config.gcc (csky-*-*): New.
4753 * configure.ac: Add csky to targets for dwarf2 debug_line support.
4754 * configure: Regenerated.
4756 2018-08-17 David Malcolm <dmalcolm@redhat.com>
4758 * dump-context.h: Include "dumpfile.h".
4759 (dump_context::dump_printf_va): Convert final param from va_list
4760 to va_list *. Convert from ATTRIBUTE_PRINTF to
4761 ATTRIBUTE_GCC_DUMP_PRINTF.
4762 (dump_context::dump_printf_loc_va): Likewise.
4763 * dumpfile.c: Include "stringpool.h".
4764 (make_item_for_dump_printf_va): Delete.
4765 (make_item_for_dump_printf): Delete.
4766 (class dump_pretty_printer): New class.
4767 (dump_pretty_printer::dump_pretty_printer): New ctor.
4768 (dump_pretty_printer::emit_items): New member function.
4769 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
4771 (dump_pretty_printer::emit_item): New member function.
4772 (dump_pretty_printer::stash_item): New member function.
4773 (dump_pretty_printer::format_decoder_cb): New member function.
4774 (dump_pretty_printer::decode_format): New member function.
4775 (dump_context::dump_printf_va): Reimplement in terms of
4776 dump_pretty_printer.
4777 (dump_context::dump_printf_loc_va): Convert final param from va_list
4779 (dump_context::begin_scope): Reimplement call to
4780 make_item_for_dump_printf.
4781 (dump_printf): Update for change to dump_printf_va.
4782 (dump_printf_loc): Likewise.
4783 (selftest::test_capture_of_dump_calls): Convert "stmt" from
4784 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
4785 with %T, %E, and %G.
4786 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
4787 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
4788 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
4789 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
4790 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
4791 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
4792 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
4793 within a dump_printf_loc call to "%wu".
4794 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
4795 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
4796 missing space after "=".
4797 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
4798 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
4799 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
4800 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
4801 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
4802 duplicate "vectorized" from message.
4804 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
4806 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
4807 polyNxK_t element's TYPE_STRING_FLAG.
4809 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
4811 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
4812 (they were unnamed before). Fix comments.
4814 2018-08-17 Nathan Sidwell <nathan@acm.org>
4816 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
4818 2018-08-17 Richard Biener <rguenther@suse.de>
4820 PR tree-optimization/86841
4821 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
4823 2018-08-17 Martin Liska <mliska@suse.cz>
4825 * common.opt: Remove Warn, Init and Report for options with
4826 Ignore/Deprecated flag. Warning is done automatically for
4828 * config/i386/i386.opt: Likewise.
4829 * config/ia64/ia64.opt: Likewise.
4830 * config/rs6000/rs6000.opt: Likewise.
4831 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
4832 Remove usage of flag_check_pointer_bounds.
4833 * lto-wrapper.c (merge_and_complain): Do not handle
4834 OPT_fcheck_pointer_bounds.
4835 (append_compiler_options): Likewise.
4836 * opt-functions.awk: Do not handle Deprecated.
4837 * optc-gen.awk: Check that Var, Report and Init are not
4838 used for an option with Ignore/Deprecated flag.
4839 * opts-common.c (decode_cmdline_option): Do not report
4841 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
4843 * opts.h (struct cl_option): Remove cl_deprecated flag.
4844 (CL_ERR_DEPRECATED): Remove error enum value.
4846 2018-08-17 Richard Biener <rguenther@suse.de>
4849 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
4850 across a va-arg-pack using call adjust its return value accordingly.
4852 2018-08-16 Martin Sebor <msebor@redhat.com>
4854 PR tree-optimization/86853
4855 * gimple-ssa-sprintf.c (struct format_result): Rename member.
4856 (struct fmtresult): Add member and initialize it in ctors.
4857 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
4858 (format_string): Handle %S the same as %ls. Set MAYFAIL.
4859 (format_directive): Set POSUNDER4K when MAYFAIL is set.
4860 (parse_directive): Handle %C same as %c.
4861 (sprintf_dom_walker::compute_format_length): Adjust.
4862 (is_call_safe): Adjust.
4864 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
4866 * builtins.c (c_strlen): Add new parameter eltsize. Use it
4867 for determining how to count the elements.
4868 * builtins.h (c_strlen): Adjust prototype.
4869 * expr.c (string_constant): Add new parameter mem_size.
4870 Set *mem_size appropriately.
4871 * expr.h (string_constant): Adjust protoype.
4872 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
4873 * gimple-fold.h (get_range_strlen): Adjust prototype.
4874 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
4875 (format_string): Call get_string_length with eltsize.
4877 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4879 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
4880 to emit the span, rather than setting it as the prefix.
4882 2018-08-16 David Malcolm <dmalcolm@redhat.com>
4884 * diagnostic-show-locus.c (layout::start_annotation_line): Add
4885 "margin_char" parameter, defaulting to space. Use it in place
4886 of pp_space for the initial part of the margin.
4887 (layout::print_leading_fixits): Use '+' when filling the margin
4888 of line-insertion fix-it hints.
4890 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4892 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
4895 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4897 * config/rs6000/altivec.md: Don't set length attribute to the default
4899 * config/rs6000/darwin.md: Ditto.
4900 * config/rs6000/dfp.md: Ditto.
4901 * config/rs6000/htm.md: Ditto.
4902 * config/rs6000/rs6000.md: Ditto.
4903 * config/rs6000/sync.md: Ditto.
4904 * config/rs6000/vsx.md: Ditto.
4906 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4908 * config/rs6000/altivec.md: Don't set length attribute to the default
4909 value, for branch instructions.
4910 * config/rs6000/darwin.md: Ditto.
4911 * config/rs6000/rs6000.md: Ditto.
4913 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
4915 * config/rs6000/rs6000.md (length): Always define as const_int 4.
4916 (unnamed conditional branch define_insn): Set length to 4 or 8
4917 depending on offset.
4918 (<bd>_<mode>): Similar, for alternative 0.
4919 (<bd>tf_<mode>): Ditto.
4921 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4923 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
4925 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
4927 * doc/rtl.texi: Replace old RTX class names with new names.
4930 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
4932 * expmed.h (canonicalize_comparison): New declaration.
4933 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
4934 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
4935 * optabs.c (prepare_cmp_insn): Likewise.
4936 * rtl.h (unsigned_condition_p): New function which checks if a
4937 comparison operator is unsigned.
4939 2018-08-16 Nathan Sidwell <nathan@acm.org>
4941 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
4942 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
4944 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4947 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
4948 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
4949 (mov<mov>): ..this and enable unconditionally.
4951 2018-08-16 Tamar Christina <tamar.christina@arm.com>
4953 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
4955 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
4957 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
4958 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
4959 "Common" with "Target".
4961 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4963 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4964 * doc/invoke.texi (mmitigate-rop): Remove.
4965 * config/i386/i386.c: Do not include "regrename.h".
4966 (ix86_rop_should_change_byte_p, reg_encoded_number)
4967 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4969 (ix86_reorg): Remove call to ix86_mitigate_rop.
4970 * config/i386/i386.md (attr "modrm_class"): Remove.
4971 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4972 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4973 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4975 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
4977 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4978 allow folding of mergeh() and mergel() for the float and double types.
4979 (fold_mergehl_helper): Rework to handle building a permute tree
4982 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4984 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4987 2018-08-15 David Malcolm <dmalcolm@redhat.com>
4989 * common.opt (fdiagnostics-show-labels): New option.
4990 * diagnostic-show-locus.c (class layout_range): Add field
4992 (class layout): Add field "m_show_labels_p".
4993 (layout_range::layout_range): Add param "label" and use it to
4995 (make_range): Pass in NULL for new "label" param of layout_range's
4997 (layout::layout): Initialize m_show_labels_p.
4998 (layout::maybe_add_location_range): Pass in loc_range->m_label
4999 when constructing layout_range instances.
5000 (struct line_label): New struct.
5001 (layout::print_any_labels): New member function.
5002 (layout::print_line): Call it if label-printing is enabled.
5003 (selftest::test_one_liner_labels): New test.
5004 (selftest::test_diagnostic_show_locus_one_liner): Call it.
5005 * diagnostic.c (diagnostic_initialize): Initialize
5006 context->show_labels_p.
5007 * diagnostic.h (struct diagnostic_context): Add field
5009 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5010 -fno-diagnostics-show-labels.
5011 * dwarf2out.c (gen_producer_string): Add
5012 OPT_fdiagnostics_show_labels to the ignored options.
5013 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
5015 (gcc_rich_location::maybe_add_expr): Likewise.
5016 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
5017 label" param, defaulting to NULL.
5018 (gcc_rich_location::add_expr): Add "label" param.
5019 (gcc_rich_location::maybe_add_expr): Likewise.
5020 (class text_range_label): New class.
5021 (class range_label_for_type_mismatch): New class.
5022 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
5023 of format_warning_va.
5024 (fmtwarn_n): Likewise for new params of format_warning_n_va.
5025 * lto-wrapper.c (merge_and_complain): Add
5026 OPT_fdiagnostics_show_labels to the "pick one setting" options.
5027 (append_compiler_options): Likewise to the dropped options.
5028 (append_diag_options): Likewise to the passed-on options.
5029 * opts.c (common_handle_option): Handle the new option.
5030 * selftest-diagnostic.c
5031 (test_diagnostic_context::test_diagnostic_context): Enable
5033 * substring-locations.c: Include "gcc-rich-location.h".
5034 (format_warning_n_va): Add "fmt_label" and "param_label" params
5035 and use them as appropriate.
5036 (format_warning_va): Add "fmt_label" and "param_label" params,
5037 passing them on to format_warning_n_va.
5038 (format_warning_at_substring): Likewise.
5039 (format_warning_at_substring_n): Likewise.
5040 * substring-locations.h (format_warning_va): Add "fmt_label" and
5041 "param_label" params.
5042 (format_warning_n_va): Likewise.
5043 (format_warning_at_substring): Likewise.
5044 (format_warning_at_substring_n): Likewise.
5045 * toplev.c (general_init): Initialize global_dc->show_labels_p.
5047 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
5050 * builtins.c (expand_builtin_memcmp): Do not expand the call
5051 when overflow is detected.
5053 2018-08-15 Martin Sebor <msebor@redhat.com>
5055 PR tree-optimization/71625
5056 * config/aarch64/aarch64-builtins.c
5057 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
5059 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
5061 * config/s390/s390.c (s390_reorg): Remove loop.
5063 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
5066 (darwin_function_switched_text_sections): Delete.
5067 * gcc/config/darwin.h
5068 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
5070 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
5073 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
5074 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
5075 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
5077 2018-08-15 Martin Liska <mliska@suse.cz>
5079 PR tree-optimization/86925
5080 * predict.c (expr_expected_value_1): When taking
5081 later predictor, assign also probability.
5082 Use fold_build2_initializer_loc in order to fold
5083 the expression in -frounding-math.
5085 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
5087 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
5089 (expand_vec_perm_1): Try the new method.
5091 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
5094 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
5095 Check whether lra_live_max_point is 0 before dividing.
5097 2018-08-14 Martin Sebor <msebor@redhat.com>
5099 PR tree-optimization/86650
5100 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
5101 (vrp_prop::check_mem_ref): Same.
5103 2018-08-13 Liu Hao <lh_mouse@126.com>
5105 * pretty-print.c (eat_esc_sequence): Swap the foreground and
5106 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
5107 and clear it thereafter, as it only works for DBCS.
5109 2018-08-13 Liu Hao <lh_mouse@126.com>
5111 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
5112 handle returned by _get_osf_handle().
5114 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5116 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
5117 for folding vec_perm.
5119 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5121 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
5122 Add support for gimple-folding of vec_pack() and vec_unpack()
5125 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
5127 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
5128 vec_xst variants to the list.
5129 (rs6000_gimple_fold_builtin): Add support for folding unaligned
5130 vector loads and stores.
5132 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
5134 * config.gcc (rs6000-ibm-aix4.x): Delete.
5135 (rs6000-ibm-aix5.1): Delete.
5136 (rs6000-ibm-aix5.2): Delete.
5137 (rs6000-ibm-aix5.3): Delete.
5138 * config/rs6000/aix43.h: Delete.
5139 * config/rs6000/aix51.h: Delete.
5140 * config/rs6000/aix52.h: Delete.
5141 * config/rs6000/t-aix43: Delete.
5143 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
5145 * config/s390/s390.c (s390_decompose_constant_pool_ref):
5147 (s390_decompose_address): Factor out constant pool ref
5150 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5152 * config/nds32/nds32-predicates.c
5153 (nds32_can_use_bclr_p): Change return type as bool.
5154 (nds32_can_use_bset_p): Ditto.
5155 (nds32_can_use_btgl_p): Ditto.
5156 (nds32_can_use_bitci_p): Ditto.
5157 * config/nds32/nds32-protos.h
5158 (nds32_can_use_bclr_p): Change declaration.
5159 (nds32_can_use_bset_p): Ditto.
5160 (nds32_can_use_btgl_p): Ditto.
5161 (nds32_can_use_bitci_p): Ditto.
5163 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5165 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
5166 Support -msched-prolog-epilog option.
5167 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
5169 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5171 * common/config/nds32/nds32-common.c
5172 (nds32_option_optimization_table): Enalbe -malways-align.
5174 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
5176 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
5178 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
5179 OPT_misr_secure_ case.
5180 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
5181 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
5182 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
5184 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
5185 (nds32_isr_info): New field security_level.
5186 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
5187 * config/nds32/nds32.md (return_internal): Consider critical attribute.
5188 * config/nds32/nds32.opt (misr-secure): New option.
5189 * config/nds32/nds32_init.inc: New file.
5190 * config/nds32/nds32_isr.h: New file.
5192 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
5194 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
5195 Update comment for atomic instructions.
5196 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
5197 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
5199 (atomic_loaddi): Revise fence expansion to only emit fence prior to
5200 load for __ATOMIC_SEQ_CST model.
5201 (atomic_loaddi_1): Remove float register target.
5202 (atomic_storedi): Handle CONST_INT values.
5203 (atomic_storedi_1): Remove float register source. Add special case
5205 (memory_barrier): New expander and insn.
5207 2018-08-11 Jakub Jelinek <jakub@redhat.com>
5209 PR tree-optimization/86835
5210 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
5211 new_stmt after def_gsi, make sure to insert new_square_stmt after
5212 that stmt, not 2 stmts before it.
5214 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
5217 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
5220 2018-08-10 Martin Liska <mliska@suse.cz>
5223 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
5225 * builtins.c (expand_builtin_expect_with_probability):
5227 (expand_builtin_expect_with_probability): New function.
5228 (build_builtin_expect_predicate): Add new argumnet probability
5229 for BUILT_IN_EXPECT_WITH_PROBABILITY.
5230 (fold_builtin_expect):
5233 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
5234 * builtins.h (fold_builtin_expect): Set new argument.
5235 * doc/extend.texi: Document __builtin_expect_with_probability.
5236 * doc/invoke.texi: Likewise.
5237 * gimple-fold.c (gimple_fold_call): Pass new argument.
5238 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
5239 also BUILT_IN_EXPECT_WITH_PROBABILITY.
5240 * predict.c (get_predictor_value): New function.
5241 (expr_expected_value): Add new argument probability. Assume
5242 that predictor and probability are always non-null.
5243 (expr_expected_value_1): Likewise. For __builtin_expect and
5244 __builtin_expect_with_probability set probability. Handle
5245 combination in binary expressions.
5246 (tree_predict_by_opcode): Simplify code by simply calling
5247 get_predictor_value.
5248 (pass_strip_predict_hints::execute): Add handling of
5249 BUILT_IN_EXPECT_WITH_PROBABILITY.
5250 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
5252 * tree.h (DECL_BUILT_IN_P): New function.
5254 2018-08-10 Martin Liska <mliska@suse.cz>
5256 PR tree-optimization/85799
5257 * passes.def: Add argument for pass_strip_predict_hints.
5258 * predict.c (class pass_strip_predict_hints): Add new argument
5260 (strip_predictor_early): New function.
5261 (pass_strip_predict_hints::execute): Call the function to
5263 (strip_predict_hints): New function.
5264 * predict.def: Fix comment.
5266 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
5268 * Makefile.in: Clarify which tm.texi to copy over to assert the
5269 right to grant a GFDL license for all.
5271 2018-08-09 Jeff Law <law@redhat.com>
5273 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
5276 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
5278 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
5281 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5283 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
5284 reductions for variable-length vectors.
5286 2018-08-09 David Malcolm <dmalcolm@redhat.com>
5289 * common.opt (fdiagnostics-show-line-numbers): New option.
5290 * diagnostic-show-locus.c (class layout): Add fields
5291 "m_show_line_numbers_p" and "m_linenum_width";
5292 (num_digits): New function.
5293 (test_num_digits): New function.
5294 (layout::layout): Initialize new fields. Update m_x_offset
5295 logic to handle any left margin.
5296 (layout::print_source_line): Print line number when requested.
5297 (layout::start_annotation_line): New member function.
5298 (layout::print_annotation_line): Call it.
5299 (layout::print_leading_fixits): Likewise.
5300 (layout::print_trailing_fixits): Likewise. Update calls to
5301 move_to_column for new parameter.
5302 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
5303 it to potentially call start_annotation_line.
5304 (layout::show_ruler): Call start_annotation_line.
5305 (selftest::test_line_numbers_multiline_range): New selftest.
5306 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
5307 and selftest::test_line_numbers_multiline_range.
5308 * diagnostic.c (diagnostic_initialize): Initialize
5309 show_line_numbers_p.
5310 * diagnostic.h (struct diagnostic_context): Add field
5311 "show_line_numbers_p".
5312 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
5313 -fno-diagnostics-show-line-numbers.
5314 * dwarf2out.c (gen_producer_string): Add
5315 OPT_fdiagnostics_show_line_numbers to the ignored options.
5316 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
5317 one setting" options.
5318 (append_compiler_options): Likewise to the dropped options.
5319 (append_diag_options): Likewise to the passed-on options.
5320 * opts.c (common_handle_option): Handle the new option.
5321 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
5323 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
5325 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
5326 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
5327 third argument of type "const signed char" to descriptions of
5328 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
5329 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
5330 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
5331 __builtin_bcdsub_ov functions.
5333 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5335 PR tree-optimization/86858
5336 * tree-vect-loop.c (vect_is_simple_reduction): Restore
5337 flow_bb_inside_loop_p calls.
5339 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
5341 PR tree-optimization/86871
5342 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
5343 instead of gimple_assign_lhs.
5345 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
5348 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
5349 register constraint to operand 0.
5350 (add<mode>3_carryinC): Likewise.
5351 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
5353 2018-08-09 Martin Liska <mliska@suse.cz>
5356 * common.opt: Remove extra line.
5358 2018-08-09 Martin Liska <mliska@suse.cz>
5360 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
5361 at the end of a line, make first letter capital and end up
5362 a sentence with a dot.
5363 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
5364 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
5365 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
5366 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
5367 (PARAM_MAX_ISL_OPERATIONS): Likewise.
5368 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
5369 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
5370 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
5371 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
5372 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
5373 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
5374 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
5375 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
5376 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
5377 (PARAM_TREE_REASSOC_WIDTH): Likewise.
5378 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
5379 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
5380 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
5382 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
5385 * config/s390/s390.c (s390_option_override_internal): Reduce the
5386 stack-clash-protection-probe-interval param if it would be too big
5389 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
5392 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
5393 * config/m68k/m68k.c (handle_move_double): Don't call
5394 m68k_final_prescan_insn.
5395 (m68k_adjust_decorated_operand): Renamed from
5396 m68k_final_prescan_insn, remove first and third operand and
5398 (print_operand): Call it.
5399 (print_operand_address): Call it.
5401 2018-08-08 Nathan Sidwell <nathan@acm.org>
5403 * diagnostic.c (diagnostic_report_current_module): Use
5404 linemap_included_from & linemap_included_from_linemap.
5406 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
5408 * config/aarch64/aarch64-cores.def: Add phecda core.
5409 * config/aarch64/aarch64-tune.md: Regenerate.
5410 * doc/invoke.texi: Add phecda core.
5412 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5415 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
5417 * config/s390/s390.md ("movti"): Add more alternatives for
5418 constant to GPR copies.
5420 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
5422 * config/s390/s390.c: Fix whitespace damage throughout the file.
5423 * config/s390/s390.h: Likewise.
5424 * config/s390/tpf.h: Likewise.
5426 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5428 * config/s390/s390.c (s390_loadrelative_operand_p):
5429 Remove TARGET_CPU_ZARCH usages.
5430 (s390_rtx_costs): Likewise.
5431 (s390_legitimate_constant_p): Likewise.
5432 (s390_cannot_force_const_mem): Likewise.
5433 (legitimate_reload_constant_p): Likewise.
5434 (s390_preferred_reload_class): Likewise.
5435 (legitimize_pic_address): Likewise.
5436 (legitimize_tls_address): Likewise.
5437 (s390_split_branches): Removed.
5438 (s390_add_execute): Removed.
5439 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
5440 (s390_mainpool_start): Likewise.
5441 (s390_mainpool_finish): Likewise.
5442 (s390_mainpool_cancel): Removed.
5443 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
5444 (s390_chunkify_cancel): Likewise.
5445 (s390_return_addr_rtx): Likewise.
5446 (s390_register_info): Remove split_branches_pending_p uages.
5447 (s390_optimize_register_info): Likewise.
5448 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
5449 split_branches_pending_p usages.
5450 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
5451 (s390_load_got): Likewise.
5452 (s390_expand_split_stack_prologue): Likewise.
5453 (output_asm_nops): Likewise.
5454 (s390_function_profiler): Likewise.
5455 (s390_emit_call): Likewise.
5456 (s390_conditional_register_usage): Likewise.
5457 (s390_optimize_prologue): Likewise.
5458 (s390_reorg): Remove TARGET_CPU_ZARCH and
5459 split_branches_pending_p usages.
5460 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
5462 (s390_output_indirect_thunk_function): Likewise.
5463 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
5464 (TARGET_CPU_ZARCH_P): Removed.
5465 (struct machine_function): Remove split_branches_pending_p.
5466 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
5468 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
5470 * common/config/s390/s390-common.c (processor_flags_table):
5472 * config.gcc: Remove with_arch/with_tune support.
5473 * config/s390/2064.md: Remove cpu attribute comparisons.
5474 * config/s390/driver-native.c (s390_host_detect_local_cpu):
5476 * config/s390/linux.h (ASM_SPEC):
5477 Remove -march support.
5478 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
5479 Use a table to get an arch level.
5480 * config/s390/s390-opts.h (enum processor_type):
5482 * config/s390/s390.c
5483 (processor_table): Remove entries, add arch_level values.
5484 (s390_issue_rate): Remove cases.
5485 (s390_option_override): Adjust
5486 s390_option_override_internal() call.
5487 (s390_option_override_internal): Remove deprecation warning.
5488 (s390_valid_target_attribute_tree): Adjust
5489 s390_option_override_internal() call.
5490 * config/s390/s390.h (struct s390_processor):
5491 Share with s390-c.c, add arch_level field.
5492 * config/s390/s390.md:
5493 Remove occurrences in cpu attribute.
5494 * config/s390/s390.opt: Remove -march/-mtune support.
5495 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
5496 * doc/invoke.texi: Remove deprecation warning.
5498 2018-08-08 Luis Machado <luis.machado@linaro.org>
5500 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
5502 (qdf24xx_tunings): Set vector cost structure to
5503 qdf24xx_vector_cost.
5505 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
5506 <register_sextend>: Set to 3.
5508 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
5511 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
5512 * config/aarch64/aarch64-simd.md
5513 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
5514 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
5515 (aarch64_frecpx<mode>): ...this new pattern.
5516 * config/aarch64/aarch64-simd-builtins.def: Remove comment
5517 about aarch64_frecp<FRECP:frecp_suffix><mode>.
5519 2018-08-07 Martin Liska <mliska@suse.cz>
5522 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
5523 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
5524 * predict.def (PRED_MALLOC_NONNULL): New predictor.
5525 * doc/extend.texi: Document that malloc attribute adds
5528 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
5531 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5532 Define to speculation_safe_value_not_needed.
5534 2018-08-06 Jeff Law <law@redhat.com>
5536 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
5537 the vr_values instance to cprop_into_stmt.
5538 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
5539 (cprop_operand): Also query EVRP to determine if OP is a constant.
5541 2018-08-06 Nathan Sidwell <nathan@acm.org>
5543 * diagnostic.c (diagnostic_report_current_module): Reroll
5544 included-at loop. Translate text.
5546 2018-08-06 David Malcolm <dmalcolm@redhat.com>
5548 * function-tests.c (selftest::test_expansion_to_rtl): Call
5549 free_after_compilation.
5551 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5553 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
5555 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
5557 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
5558 loops with memory block operations from getting unrolled.
5560 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
5563 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5564 Define to speculation_safe_value_not_needed.
5566 2018-08-06 Jeff Law <law@redhat.com>
5568 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
5571 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5574 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
5575 with all enabled __intN types.
5577 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
5579 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5581 * alias.c (record_set): Check for clobber high.
5582 * cfgexpand.c (expand_gimple_stmt): Likewise.
5583 * combine-stack-adj.c (single_set_for_csa): Likewise.
5584 * combine.c (find_single_use_1): Likewise.
5585 (set_nonzero_bits_and_sign_copies): Likewise.
5586 (get_combine_src_dest): Likewise.
5587 (is_parallel_of_n_reg_sets): Likewise.
5588 (try_combine): Likewise.
5589 (record_dead_and_set_regs_1): Likewise.
5590 (reg_dead_at_p_1): Likewise.
5591 (reg_dead_at_p): Likewise.
5592 * dce.c (deletable_insn_p): Likewise.
5593 (mark_nonreg_stores_1): Likewise.
5594 (mark_nonreg_stores_2): Likewise.
5595 * df-scan.c (df_find_hard_reg_defs): Likewise.
5596 (df_uses_record): Likewise.
5597 (df_get_call_refs): Likewise.
5598 * dwarf2out.c (mem_loc_descriptor): Likewise.
5599 * haifa-sched.c (haifa_classify_rtx): Likewise.
5600 * ira-build.c (create_insn_allocnos): Likewise.
5601 * ira-costs.c (scan_one_insn): Likewise.
5602 * ira.c (equiv_init_movable_p): Likewise.
5603 (rtx_moveable_p): Likewise.
5604 (interesting_dest_for_shprep): Likewise.
5605 * jump.c (mark_jump_label_1): Likewise.
5606 * postreload-gcse.c (record_opr_changes): Likewise.
5607 * postreload.c (reload_cse_simplify): Likewise.
5608 (struct reg_use): Add source expr.
5609 (reload_combine): Check for clobber high.
5610 (reload_combine_note_use): Likewise.
5611 (reload_cse_move2add): Likewise.
5612 (move2add_note_store): Likewise.
5613 * print-rtl.c (print_pattern): Likewise.
5614 * recog.c (decode_asm_operands): Likewise.
5615 (store_data_bypass_p): Likewise.
5616 (if_test_bypass_p): Likewise.
5617 * regcprop.c (kill_clobbered_value): Likewise.
5618 (kill_set_value): Likewise.
5619 * reginfo.c (reg_scan_mark_refs): Likewise.
5620 * reload1.c (maybe_fix_stack_asms): Likewise.
5621 (eliminate_regs_1): Likewise.
5622 (elimination_effects): Likewise.
5623 (mark_not_eliminable): Likewise.
5624 (scan_paradoxical_subregs): Likewise.
5625 (forget_old_reloads_1): Likewise.
5626 * reorg.c (find_end_label): Likewise.
5627 (try_merge_delay_insns): Likewise.
5628 (redundant_insn): Likewise.
5629 (own_thread_p): Likewise.
5630 (fill_simple_delay_slots): Likewise.
5631 (fill_slots_from_thread): Likewise.
5632 (dbr_schedule): Likewise.
5633 * resource.c (update_live_status): Likewise.
5634 (mark_referenced_resources): Likewise.
5635 (mark_set_resources): Likewise.
5636 * rtl.c (copy_rtx): Likewise.
5637 * rtlanal.c (reg_referenced_p): Likewise.
5638 (single_set_2): Likewise.
5639 (noop_move_p): Likewise.
5640 (note_stores): Likewise.
5641 * sched-deps.c (sched_analyze_reg): Likewise.
5642 (sched_analyze_insn): Likewise.
5644 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5646 * cse.c (invalidate_reg): New function extracted from...
5647 (invalidate): ...here.
5648 (canonicalize_insn): Check for clobber high.
5649 (invalidate_from_clobbers): invalidate clobber highs.
5650 (invalidate_from_sets_and_clobbers): Likewise.
5651 (count_reg_usage): Check for clobber high.
5652 (insn_live_p): Likewise.
5653 * cselib.c (cselib_expand_value_rtx_1):Likewise.
5654 (cselib_invalidate_regno): Check for clobber in setter.
5655 (cselib_invalidate_rtx): Pass through setter.
5656 (cselib_invalidate_rtx_note_stores):
5657 (cselib_process_insn): Check for clobber high.
5658 * cselib.h (cselib_invalidate_rtx): Add operand.
5660 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5662 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
5663 (mark_not_eliminable): Likewise.
5664 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
5665 * lra-lives.c (process_bb_lives): Check for clobber high.
5666 * lra.c (new_insn_reg): Remember clobber highs.
5667 (collect_non_operand_hard_regs): Check for clobber high.
5668 (lra_set_insn_recog_data): Likewise.
5669 (add_regs_to_insn_regno_info): Likewise.
5670 (lra_update_insn_regno_info): Likewise.
5672 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5674 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
5675 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
5677 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5679 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
5680 (copy_insn_1): Likewise.
5681 (gen_hard_reg_clobber_high): New gen function.
5682 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
5683 * genemit.c (gen_exp): Likewise.
5684 (gen_emit_seq): Pass through info.
5685 (gen_insn): Check for CLOBBER_HIGH.
5686 (gen_expand): Pass through info.
5687 (gen_split): Likewise.
5688 (output_add_clobbers): Likewise.
5689 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
5690 (remove_clobbers): Likewise.
5691 * rtl.h (gen_hard_reg_clobber_high): New declaration.
5693 2018-08-06 Alan Hayward <alan.hayward@arm.com>
5695 * doc/rtl.texi (clobber_high): Add.
5696 (parallel): Add in clobber high
5697 * rtl.c (rtl_check_failed_code3): Add function.
5698 * rtl.def (CLOBBER_HIGH): Add expression.
5699 * rtl.h (RTL_CHECKC3): Add macro.
5700 (rtl_check_failed_code3): Add declaration.
5701 (XC3EXP): Add macro.
5703 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
5706 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
5707 cfun->machine->max_used_stack_alignment if needed.
5709 2018-08-04 Martin Sebor <msebor@redhat.com>
5711 PR tree-optimization/86571
5712 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
5715 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
5718 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5721 2018-08-03 Jeff Law <law@redhat.com>
5724 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5725 Define to speculation_safe_value_not_needed.
5727 2018-08-03 David Malcolm <dmalcolm@redhat.com>
5729 * doc/gcov.texi (-x): Remove duplicate "to".
5730 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
5731 (-Wif-not-aligned): Remove duplicate "is".
5732 (-flto): Remove duplicate "the".
5733 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
5734 duplicate "v5.00.b".
5735 (MSP430 Options): Remove duplicate "and" from the description
5736 of "-mgprel-sec=regexp".
5737 (x86 Options): Remove duplicate copies of "vmldLog102" and
5738 vmlsLog104 from description of "-mveclibabi=type".
5740 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
5742 * internal-fn.h (first_commutative_argument): Declare.
5743 * internal-fn.c (first_commutative_argument): New function.
5744 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
5745 restrictions for pattern statements. Use first_commutative_argument
5746 to look for commutative operands in calls to internal functions.
5748 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
5750 * Makefile.in (wide-int-range.o): New.
5751 * tree-vrp.c: Move all the wide_int_* functions to...
5752 * wide-int-range.cc: ...here.
5753 * tree-vrp.h: Move all the wide_int_* prototypes to...
5754 * wide-int-range.h: ...here.
5756 2018-08-03 Tom de Vries <tdevries@suse.de>
5758 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
5760 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
5761 * except.c (output_function_exception_table): Do early exit if
5762 targetm_common.except_unwind_info (&global_options) == UI_NONE.
5764 2018-08-03 Martin Liska <mliska@suse.cz>
5766 * predict.c (dump_prediction): Change to 2 digits
5769 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
5771 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
5773 (falkor_am_1_gtov_gtov): ... a new insn reservation.
5775 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
5777 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
5778 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
5779 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
5780 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
5781 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
5783 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5785 * diagnostic-show-locus.c (diagnostic_show_locus): Use
5786 pp_take_prefix when saving the existing prefix.
5787 * diagnostic.c (diagnostic_append_note): Likewise.
5788 * langhooks.c (lhd_print_error_function): Likewise.
5789 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
5790 param's type. Free the existing prefix.
5791 (pp_take_prefix): New function.
5792 (pretty_printer::pretty_printer): Drop the prefix parameter.
5793 Rename the length parameter to match the comment.
5794 (pretty_printer::~pretty_printer): Free the prefix.
5795 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
5797 (struct pretty_printer): Drop the "const" from "prefix" field's
5798 type and clarify memory management.
5799 (pp_set_prefix): Drop the "const" from the 2nd param.
5800 (pp_take_prefix): New decl.
5802 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
5804 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
5805 for word_mode_ok here instead of passing as argument.
5806 (expand_block_compare): Change select_block_compare_mode() call.
5807 (expand_strncmp_gpr_sequence): New function.
5808 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
5810 2018-08-02 Jeff Law <law@redhat.com>
5813 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5814 Define to speculation_safe_value_not_needed.
5817 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5818 Define to speculation_safe_value_not_needed.
5820 2018-08-02 Tom de Vries <tdevries@suse.de>
5823 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
5824 function. Return UI_TARGET unconditionally.
5825 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
5826 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
5828 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5830 * genemit.c (print_overload_test): Fix typo.
5832 2018-08-02 Richard Biener <rguenther@suse.de>
5834 PR tree-optimization/86816
5835 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
5836 which checks for value availability before querying it.
5837 (gvn_uses_equal): Use it.
5838 (same_succ_hash): Likewise.
5839 (gimple_equal_p): Likewise.
5841 2018-08-02 Nick Clifton <nickc@redhat.com>
5844 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5845 Define to speculation_safe_value_not_needed.
5848 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5849 Define to speculation_safe_value_not_needed.
5852 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5853 Define to speculation_safe_value_not_needed.
5856 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5857 Define to speculation_safe_value_not_needed.
5860 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5861 Define to speculation_safe_value_not_needed.
5864 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5865 Define to speculation_safe_value_not_needed.
5868 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5869 Define to speculation_safe_value_not_needed.
5872 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
5873 Define to speculation_safe_value_not_needed.
5876 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5877 speculation_safe_value_not_needed.
5880 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
5881 to speculation_safe_value_not_needed.
5883 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5885 * doc/md.texi: Expand the documentation of instruction names
5886 to mention port-local uses. Document '@' in pattern names.
5887 * read-md.h (overloaded_instance, overloaded_name): New structs.
5889 (md_reader::handle_overloaded_name): New member function.
5890 (md_reader::get_overloads): Likewise.
5891 (md_reader::m_first_overload): New member variable.
5892 (md_reader::m_next_overload_ptr): Likewise.
5893 (md_reader::m_overloads_htab): Likewise.
5894 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
5895 m_next_overload_ptr and m_overloads_htab.
5896 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
5897 (get_mode_token, get_code_token, get_int_token): New functions.
5898 (map_attr_string): Add an optional argument that passes back
5899 the associated iterator.
5900 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
5901 (md_reader::handle_overloaded_name, add_overload_instance): New
5903 (apply_iterators): Handle '@' names. Report an error if '@'
5904 is used without iterators.
5905 (initialize_iterators): Initialize the new iterator_group fields.
5906 * genopinit.c (handle_overloaded_code_for)
5907 (handle_overloaded_gen): New functions.
5908 (main): Use them to print declarations of maybe_code_for_* and
5909 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
5910 * genemit.c (print_overload_arguments, print_overload_test)
5911 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
5912 (main): Use it to print definitions of maybe_code_for_* and
5913 maybe_gen_* functions.
5914 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
5915 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
5916 instead of explicit mode checks.
5917 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
5918 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
5919 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
5920 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
5921 (aarch64_expand_compare_and_swap): Likewise
5922 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
5923 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
5924 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
5925 (aarch64_constant_pool_reload_icode): Delete.
5926 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
5927 instead of aarch64_constant_pool_reload_icode. Use
5928 code_for_aarch64_reload_mov instead of explicit mode checks.
5929 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
5930 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
5931 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
5932 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
5933 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
5934 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
5935 (aarch64_atomic_load_op_code): Delete.
5936 (aarch64_emit_atomic_load_op): Likewise.
5937 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
5938 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
5939 instead of aarch64_emit_atomic_load_op.
5940 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
5941 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
5942 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
5943 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
5944 character before the pattern name.
5945 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
5946 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
5947 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
5948 (aarch64_frecps<mode>): Likewise.
5949 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
5950 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
5951 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
5952 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
5953 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
5955 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
5957 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
5958 Allow HFmode constants if TARGET_FP_F16INST.
5960 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
5963 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5964 No longer check last store for clobber of address register.
5966 2018-08-02 Martin Liska <mliska@suse.cz>
5968 PR gcov-profile/86817
5969 * gcov.c (process_all_functions): New function.
5971 (process_file): Move functions processing to
5972 process_all_functions.
5974 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5976 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5977 "const" to the "gimple *" and "rtx_insn *" parameters.
5978 * dumpfile.h (dump_user_location_t::dump_user_location_t):
5980 (dump_location_t::dump_location_t): Likewise.
5982 2018-08-01 Martin Sebor <msebor@redhat.com>
5984 PR tree-optimization/86650
5985 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5986 rather than a "gcall *". Directly pass the data of interest
5987 to percent_K_format, rather than building a temporary CALL_EXPR
5989 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5990 (gimple_fold_builtin_strncat): Adjust.
5991 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5992 gcall* argument with gimple*.
5993 * gimple-ssa-warn-restrict.c (check_call): Same.
5994 (wrestrict_dom_walker::before_dom_children): Same.
5995 (builtin_access::builtin_access): Same.
5996 (check_bounds_or_overlap): Same
5997 (maybe_diag_overlap): Same.
5998 (maybe_diag_offset_bounds): Same.
5999 * tree-diagnostic.c (default_tree_printer): Move usage of
6000 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
6002 * tree-pretty-print.c (percent_K_format): Add argument.
6003 * tree-pretty-print.h: Add argument.
6004 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
6005 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
6006 (maybe_diag_stxncpy_trunc): Same.
6007 (handle_builtin_stxncpy): Same.
6008 (handle_builtin_strcat): Same.
6010 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6012 * match.pd: Optimise pointer range checks.
6014 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6016 PR tree-optimization/86758
6017 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
6018 to remove pattern statements.
6020 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6022 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
6023 result of dfs_enumerate_from when constructing stmt_vec_infos,
6024 instead of additionally calling get_loop_body.
6026 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6028 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
6030 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
6031 When creating an iv, assert that the step is not known to be zero.
6032 (vect_setup_realignment): Update call accordingly.
6033 * tree-vect-stmts.c (vectorizable_store): Likewise.
6034 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
6036 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6038 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
6039 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
6040 (vectorizable_reduction): Likewise.
6041 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6042 (vect_detect_hybrid_slp_stmts): Likewise.
6043 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6045 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
6047 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
6048 (wide_int_set_zero_nonzero_bits): ...this.
6049 (zero_nonzero_bits_from_vr): Rename to...
6050 (vrp_set_zero_nonzero_bits): ...this.
6051 (extract_range_from_multiplicative_op_1): Abstract wide int
6053 (wide_int_range_multiplicative_op): ...here.
6054 (extract_range_from_binary_expr_1): Extract wide int binary
6055 operations into their own functions.
6056 (wide_int_range_lshift): New.
6057 (wide_int_range_can_optimize_bit_op): New.
6058 (wide_int_range_shift_undefined_p): New.
6059 (wide_int_range_bit_xor): New.
6060 (wide_int_range_bit_ior): New.
6061 (wide_int_range_bit_and): New.
6062 (wide_int_range_trunc_mod): New.
6063 (extract_range_into_wide_ints): New.
6064 (vrp_shift_undefined_p): New.
6065 (extract_range_from_multiplicative_op): New.
6066 (vrp_can_optimize_bit_op): New.
6067 * tree-vrp.h (value_range::dump): New.
6068 (wide_int_range_multiplicative_op): New.
6069 (wide_int_range_lshift):New.
6070 (wide_int_range_shift_undefined_p): New.
6071 (wide_int_range_bit_xor): New.
6072 (wide_int_range_bit_ior): New.
6073 (wide_int_range_bit_and): New.
6074 (wide_int_range_trunc_mod): New.
6075 (zero_nonzero_bits_from_bounds): Rename to...
6076 (wide_int_set_zero_nonzero_bits): ...this.
6077 (zero_nonzero_bits_from_vr): Rename to...
6078 (vrp_set_zero_nonzero_bits): ...this.
6079 (range_easy_mask_min_max): Rename to...
6080 (wide_int_range_can_optimize_bit_op): this.
6081 * vr-values.c (simplify_bit_ops_using_ranges): Rename
6082 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
6084 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6086 * tree-vectorizer.h (vect_orig_stmt): New function.
6087 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
6088 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
6089 (vect_create_epilog_for_reduction): Likewise.
6090 (vectorizable_live_operation): Likewise.
6091 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6092 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
6093 * tree-vect-stmts.c (vectorizable_call): Likewise.
6094 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
6096 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6098 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
6100 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
6101 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
6102 (vect_transform_loop): Likewise.
6103 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6105 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6107 * tree-vectorizer.h (vect_schedule_slp): Return void.
6108 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
6109 (vect_schedule_slp): Likewise.
6111 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6113 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
6115 (vect_transform_loop): Update calls accordingly. Schedule SLP
6116 instances before the main loop, if any exist.
6118 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
6120 PR tree-optimization/86749
6121 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
6122 If the lhs is used in a COND_EXPR, check that it is being used
6123 as the "then" or "else" value.
6125 2018-08-01 Tom de Vries <tdevries@suse.de>
6128 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
6129 speculation_safe_value_not_needed.
6131 2018-08-01 Richard Biener <rguenther@suse.de>
6133 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
6136 2018-08-01 Martin Liska <mliska@suse.cz>
6138 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
6139 format how successful transformation is dumped.
6140 (gimple_mod_pow2_value_transform): Likewise.
6141 (gimple_mod_subtract_transform): Likewise.
6142 (gimple_stringops_transform): Likewise.
6144 2018-08-01 Martin Liska <mliska@suse.cz>
6147 * value-prof.c (interesting_stringop_to_profile_p):
6148 Simplify the code and add BUILT_IN_MEMMOVE.
6149 (gimple_stringops_transform): Likewise.
6151 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
6153 * config/aarch64/aarch64-simd.md
6154 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
6155 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
6156 use GPI iterator instead of SI mode.
6158 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6160 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
6161 rs6000_speculation_barrier.
6162 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
6163 new barrier pattern name.
6165 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
6167 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
6168 (speculation_barrier): New insn.
6170 2018-08-01 Richard Biener <rguenther@suse.de>
6173 * graphite.h: Include isl/id.h and isl/space.h to allow build
6176 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
6179 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6180 mode for COFF targets.
6181 * defaults.h (TARGET_COFF): Define.
6182 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
6183 TARGET_COFF): Define.
6184 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
6185 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
6186 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
6188 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6190 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
6191 * gimple-streamer-out.c (output_bb): Save it.
6192 * lto-streamer-in.c (input_struct_function_base): Restore
6193 instance discriminator if available. Create map on demand.
6194 * lto-streamer-out.c (output_struct_function_base): Save it if
6196 * final.c (decl_to_instance_map): Document LTO strategy.
6198 2018-07-31 Alexandre Oliva <oliva@adacore.com>
6199 Olivier Hainque <hainque@adacore.com>
6201 * debug.h (decl_to_instance_map_t): New type.
6202 (decl_to_instance_map): Declare.
6203 (maybe_create_decl_to_instance_map): New inline function.
6204 * final.c (bb_discriminator, last_bb_discriminator): New statics,
6205 to track basic block discriminators.
6206 (final_start_function_1): Initialize them.
6207 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
6209 (decl_to_instance_map): New variable.
6210 (map_decl_to_instance, maybe_set_discriminator): New functions.
6211 (notice_source_line): Set discriminator.
6213 2018-07-31 Ian Lance Taylor <iant@golang.org>
6215 * targhooks.c (default_have_speculation_safe_value): Add
6218 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6220 * dump-context.h: Include "pretty-print.h".
6221 (dump_context::refresh_dumps_are_enabled): New decl.
6222 (dump_context::emit_item): New decl.
6223 (class dump_context): Add fields "m_test_pp" and
6225 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6226 (temp_dump_context::get_dumped_text): New decl.
6227 (class temp_dump_context): Add field "m_pp".
6228 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
6229 (dump_context::refresh_dumps_are_enabled): ...and add a test for
6231 (set_dump_file): Update for above change.
6232 (set_alt_dump_file): Likewise.
6233 (dump_loc): New overload, taking a pretty_printer *.
6234 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
6235 to any test pretty-printer.
6236 (make_item_for_dump_gimple_stmt): New function, adapted from
6237 optinfo::add_gimple_stmt.
6238 (dump_context::dump_gimple_stmt): Call it, and use the result,
6239 eliminating the direct usage of dump_file and alt_dump_file in
6240 favor of indirectly using them via emit_item.
6241 (make_item_for_dump_gimple_expr): New function, adapted from
6242 optinfo::add_gimple_expr.
6243 (dump_context::dump_gimple_expr): Call it, and use the result,
6244 eliminating the direct usage of dump_file and alt_dump_file in
6245 favor of indirectly using them via emit_item.
6246 (make_item_for_dump_generic_expr): New function, adapted from
6248 (dump_context::dump_generic_expr): Call it, and use the result,
6249 eliminating the direct usage of dump_file and alt_dump_file in
6250 favor of indirectly using them via emit_item.
6251 (make_item_for_dump_printf_va): New function, adapted from
6252 optinfo::add_printf_va.
6253 (make_item_for_dump_printf): New function.
6254 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
6255 and use the result, eliminating the direct usage of dump_file and
6256 alt_dump_file in favor of indirectly using them via emit_item.
6257 (make_item_for_dump_dec): New function.
6258 (dump_context::dump_dec): Call it, and use the result,
6259 eliminating the direct usage of dump_file and alt_dump_file in
6260 favor of indirectly using them via emit_item.
6261 (make_item_for_dump_symtab_node): New function, adapted from
6262 optinfo::add_symtab_node.
6263 (dump_context::dump_symtab_node): Call it, and use the result,
6264 eliminating the direct usage of dump_file and alt_dump_file in
6265 favor of indirectly using them via emit_item.
6266 (dump_context::begin_scope): Reimplement, avoiding direct usage
6267 of dump_file and alt_dump_file in favor of indirectly using them
6269 (dump_context::emit_item): New member function.
6270 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
6271 Set up test pretty-printer on the underlying context. Call
6272 refresh_dumps_are_enabled.
6273 (temp_dump_context::~temp_dump_context): Call
6274 refresh_dumps_are_enabled.
6275 (temp_dump_context::get_dumped_text): New member function.
6276 (selftest::verify_dumped_text): New function.
6277 (ASSERT_DUMPED_TEXT_EQ): New macro.
6278 (selftest::test_capture_of_dump_calls): Run all tests twice, with
6279 and then without optinfo enabled. Add uses of
6280 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
6281 * dumpfile.h: Update comment for the dump_* API.
6282 * optinfo-emit-json.cc
6283 (selftest::test_building_json_from_dump_calls): Update for new
6284 param for temp_dump_context ctor.
6285 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
6286 and "m_owned" field.
6287 (optinfo_item::~optinfo_item): Likewise.
6288 (optinfo::add_item): New member function.
6289 (optinfo::emit): Update comment.
6290 (optinfo::add_string): Delete.
6291 (optinfo::add_printf): Delete.
6292 (optinfo::add_printf_va): Delete.
6293 (optinfo::add_gimple_stmt): Delete.
6294 (optinfo::add_gimple_expr): Delete.
6295 (optinfo::add_tree): Delete.
6296 (optinfo::add_symtab_node): Delete.
6297 (optinfo::add_dec): Delete.
6298 * optinfo.h (class dump_context): New forward decl.
6299 (optinfo::add_item): New decl.
6300 (optinfo::add_string): Delete.
6301 (optinfo::add_printf): Delete.
6302 (optinfo::add_printf_va): Delete.
6303 (optinfo::add_gimple_stmt): Delete.
6304 (optinfo::add_gimple_expr): Delete.
6305 (optinfo::add_tree): Delete.
6306 (optinfo::add_symtab_node): Delete.
6307 (optinfo::add_dec): Delete.
6308 (optinfo::add_poly_int): Delete.
6309 (optinfo_item::optinfo_item): Remove "owned" param.
6310 (class optinfo_item): Remove field "m_owned".
6312 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6315 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
6316 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
6317 requested variable alignment.
6318 (expand_one_ssa_partition): Likewise.
6319 (expand_one_var): Likewise.
6321 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6323 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
6324 to speculation_safe_value_not_needed.
6326 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6328 * targhooks.h (speculation_safe_value_not_needed): New prototype.
6329 * targhooks.c (speculation_safe_value_not_needed): New function.
6330 * target.def (have_speculation_safe_value): Update documentation.
6331 * doc/tm.texi: Regenerated.
6333 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6335 * config/aarch64/iterators.md (ALLI_TI): New iterator.
6336 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
6338 (despeculate_copy<ALLI:mode>_insn): New insn.
6339 (despeculate_copyti_insn): New insn.
6340 (despeculate_simple<ALLI:mode>): New insn
6341 (despeculate_simpleti): New insn.
6342 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
6344 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
6345 aarch64_speculation_safe_value.
6346 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
6348 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6350 * config/aarch64/aarch64-speculation.cc: New file.
6351 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
6352 before pass_reorder_blocks.
6353 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
6355 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
6356 X14 and X15 when tracking speculation.
6357 * config/aarch64/aarch64.md (register name constants): Add
6358 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
6359 (unspec): Add UNSPEC_SPECULATION_TRACKER.
6360 (speculation_barrier): New insn attribute.
6361 (cmp<mode>): Allow SP in comparisons.
6362 (speculation_tracker): New insn.
6363 (speculation_barrier): Add speculation_barrier attribute.
6364 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
6365 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
6366 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
6368 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6370 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
6371 aarch64_track_speculation is true.
6372 (tb<optab><mode>1): Likewise.
6373 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
6374 generate CB[N]Z when tracking speculation.
6375 (aarch64_split_compare_and_swap): Likewise.
6376 (aarch64_split_atomic_op): Likewise.
6378 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6380 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
6382 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6384 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
6385 (speculation_barrier): New insn.
6387 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6389 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
6390 * config/arm/arm.md (speculation_barrier): New expand.
6391 (speculation_barrier_insn): New pattern.
6393 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
6395 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
6396 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
6397 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
6398 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
6400 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
6401 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
6402 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
6403 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
6404 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
6405 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
6406 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
6407 * builtins.c (expand_speculation_safe_value): New function.
6408 (expand_builtin): Call it.
6409 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
6410 * doc/extend.texi: Document __builtin_speculation_safe_value.
6411 * doc/md.texi: Document "speculation_barrier" pattern.
6412 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
6413 TARGET_HAVE_SPECULATION_SAFE_VALUE.
6414 * doc/tm.texi: Regenerated.
6415 * target.def (have_speculation_safe_value, speculation_safe_value): New
6417 * targhooks.c (default_have_speculation_safe_value): New function.
6418 (default_speculation_safe_value): New function.
6419 * targhooks.h (default_have_speculation_safe_value): Add prototype.
6420 (default_speculation_safe_value): Add prototype.
6422 2018-07-31 David Malcolm <dmalcolm@redhat.com>
6424 * dump-context.h (dump_context::dump_loc): New decl.
6425 * dumpfile.c (dump_context::dump_loc): New member function.
6426 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
6427 and dump_gimple_stmt.
6428 (dump_context::dump_gimple_expr_loc): Likewise, using
6430 (dump_context::dump_generic_expr_loc): Likewise, using
6432 (dump_context::dump_printf_loc_va): Likewise, using
6434 (dump_context::begin_scope): Explicitly using the global function
6435 "dump_loc", rather than the member function.
6437 2018-07-31 Martin Sebor <msebor@redhat.com>
6439 PR tree-optimization/86741
6440 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
6442 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
6444 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
6445 depend on whether prefetch instructions will be emitted or not.
6446 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
6447 will be emitted or not.
6448 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
6449 (TARGET_SETMEM_PFD): New macros.
6451 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6453 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
6454 (NULL_STMT_VEC_INFO): Delete.
6455 (stmt_vec_info::operator*): Likewise.
6456 (stmt_vec_info::operator gimple *): Likewise.
6457 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
6458 of NULL_STMT_VEC_INFO.
6459 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6460 (vect_reassociating_reduction_p): Likewise.
6461 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6462 (vectorizable_store): Likewise.
6463 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
6464 (vec_info::free_stmt_vec_infos): Likewise.
6466 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6468 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
6469 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
6471 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6473 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
6474 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6475 (vec_info::free_stmt_vec_info): New private member functions.
6476 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
6477 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
6478 * tree-parloops.c (gather_scalar_reductions): Remove calls to
6479 set_stmt_vec_info_vec and free_stmt_vec_infos.
6480 * tree-vect-loop.c (_loop_vec_info): Remove call to
6481 set_stmt_vec_info_vec.
6482 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
6483 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
6484 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
6485 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
6486 (vec_info::free_stmt_vec_info): ...these new functions. Remove
6487 assignments in {vec_info::,}new_stmt_vec_info that are redundant
6488 with the clearing in the xcalloc.
6489 (stmt_vec_info_vec): Delete.
6490 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
6491 (vectorize_loops): Likewise.
6492 (vec_info::~vec_info): Remove argument from call to
6493 free_stmt_vec_infos.
6494 (vec_info::add_stmt): Remove vinfo argument from call to
6497 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6499 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
6500 rather than a gimple stmt.
6501 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
6502 information for pattern statements when passed the original
6503 statement; instead wait to be passed the pattern statement itself.
6504 Don't call set_vinfo_for_stmt here.
6505 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
6506 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
6507 stmt_vec_infos here.
6508 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
6509 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
6510 stmt_vec_infos entry.
6512 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6514 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
6515 * tree-vectorizer.c (vec_info::replace_stmt): New function.
6516 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
6517 * tree-vect-stmts.c (vectorizable_call): Likewise.
6518 (vectorizable_simd_clone_call): Likewise.
6520 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6522 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
6523 * tree-vectorizer.c (vec_info::remove_stmt): New function.
6524 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
6525 * tree-vect-loop.c (vect_transform_loop): Likewise.
6526 * tree-vect-slp.c (vect_schedule_slp): Likewise.
6527 * tree-vect-stmts.c (vect_remove_stores): Likewise.
6529 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6531 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
6532 (vect_dr_stmt): Delete.
6533 * tree-vectorizer.c (vec_info::lookup_dr): New function.
6534 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
6536 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
6537 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
6538 (vect_verify_datarefs_alignment, vect_peeling_supportable)
6539 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6540 (vect_analyze_data_refs): Likewise.
6541 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
6543 (vect_find_same_alignment_drs): Likewise.
6544 (vect_slp_analyze_node_dependences): Update calls accordingly.
6545 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6546 instead of DR_VECT_AUX.
6547 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
6548 of a vector data references. Use vec_info::lookup_dr instead of
6550 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
6551 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
6552 instead of DR_VECT_AUX.
6554 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6556 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
6558 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6560 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6561 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6562 (vect_gen_prolog_loop_niters): Likewise.
6564 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6566 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
6567 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
6568 (vect_known_alignment_in_bytes, vect_dr_behavior)
6569 (vect_get_scalar_dr_size): Take references as dr_vec_infos
6570 instead of data_references. Update calls to other routines for
6571 which the same change has been made.
6572 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
6573 dr_vec_infos instead of stmt_vec_infos.
6574 (vect_analyze_data_ref_dependence): Update call accordingly.
6575 (vect_slp_analyze_data_ref_dependence)
6576 (vect_record_base_alignments): Use DR_VECT_AUX.
6577 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
6578 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
6579 (vector_alignment_reachable_p, vect_get_data_access_cost)
6580 (vect_peeling_supportable, vect_analyze_group_access_1)
6581 (vect_analyze_group_access, vect_analyze_data_ref_access)
6582 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
6583 (vect_compile_time_alias, vect_small_gap_p)
6584 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
6585 (vect_supportable_dr_alignment): Take references as dr_vec_infos
6586 instead of data_references. Update calls to other routines for
6587 which the same change has been made.
6588 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
6589 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6590 (vect_slp_analyze_and_verify_node_alignment)
6591 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
6592 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6593 (vect_setup_realignment): Use dr_vec_infos. Update calls after
6595 (_vect_peel_info::dr): Replace with...
6596 (_vect_peel_info::dr_info): ...this new field.
6597 (vect_peeling_hash_get_most_frequent)
6598 (vect_peeling_hash_choose_best_peeling): Update accordingly.
6599 (vect_peeling_hash_get_lowest_cost):
6600 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
6601 routines for which the same change has been made.
6602 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
6604 * tree-vect-loop-manip.c (get_misalign_in_elems)
6605 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
6607 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6608 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
6609 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
6610 (get_group_load_store_type, get_negative_load_store_type)
6611 (vect_get_data_ptr_increment, vectorizable_store)
6612 (vectorizable_load): Likewise.
6613 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
6614 Update calls to other routines for which the same change has been made.
6616 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6618 * tree-vectorizer.h (vec_info::move_dr): New member function.
6619 (dataref_aux): Rename to...
6620 (dr_vec_info): ...this and add "dr" and "stmt" fields.
6621 (_stmt_vec_info::dr_aux): Update accordingly.
6622 (_stmt_vec_info::data_ref_info): Delete.
6623 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
6624 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
6625 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
6626 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
6628 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
6629 (STMT_VINFO_DR_INFO): New macro.
6630 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
6631 (set_dr_misalignment): Update after rename of dataref_aux.
6632 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
6633 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
6634 initialization of STMT_VINFO_DATA_REF.
6635 * tree-vectorizer.c (vec_info::move_dr): New function.
6636 * tree-vect-patterns.c (vect_recog_bool_pattern)
6637 (vect_recog_mask_conversion_pattern)
6638 (vect_recog_gather_scatter_pattern): Use it.
6639 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
6640 the "dr" and "stmt" fields of dr_vec_info instead of
6641 STMT_VINFO_DATA_REF.
6643 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6645 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
6646 (is_pattern_stmt_p): Use it.
6647 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
6648 on pattern statements.
6650 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6652 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
6653 original stmt as a stmt_vec_info rather than a gimple stmt.
6654 (vect_pattern_recog_1): Take the statement directly as a
6655 stmt_vec_info, rather than via a gimple_stmt_iterator.
6656 Update call to vect_mark_pattern_stmts.
6657 (vect_pattern_recog): Update calls accordingly.
6659 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6661 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
6662 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
6663 a vect_def_type for the first argument.
6664 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
6665 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
6666 operand if it isn't defined by a vectorized statement.
6667 (vect_build_gather_load_calls): Remove the mask_dt argument and
6668 update calls to vect_get_vec_def_for_stmt_copy.
6669 (vectorizable_bswap): Likewise the dt argument.
6670 (vectorizable_call): Update calls to vectorizable_bswap and
6671 vect_get_vec_def_for_stmt_copy.
6672 (vectorizable_simd_clone_call, vectorizable_assignment)
6673 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
6674 (vectorizable_comparison): Update calls to
6675 vect_get_vec_def_for_stmt_copy.
6676 (vectorizable_store): Likewise. Remove now-unnecessary calls to
6678 (vect_get_loop_based_defs): Remove dt argument and update call
6679 to vect_get_vec_def_for_stmt_copy.
6680 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
6681 and vect_get_vec_def_for_stmt_copy.
6682 (vectorizable_load): Update calls to vect_build_gather_load_calls
6683 and vect_get_vec_def_for_stmt_copy.
6684 * tree-vect-loop.c (vect_create_epilog_for_reduction)
6685 (vectorizable_reduction, vectorizable_live_operation): Update calls
6686 to vect_get_vec_def_for_stmt_copy.
6688 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6690 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
6691 and gimple stmt arguments with a stmt_vec_info.
6692 (vect_record_base_alignments): Update calls accordingly.
6693 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
6694 and gimple stmt arguments with a stmt_vec_info.
6695 (vect_build_slp_tree_1): Remove vinfo argument and update call
6696 to vect_record_max_nunits.
6697 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
6698 and vect_record_max_nunits.
6700 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6702 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
6703 file and take a stmt_vec_info instead of a gimple stmt.
6704 (supportable_widening_operation, vect_finish_replace_stmt)
6705 (vect_finish_stmt_generation, vect_get_store_rhs)
6706 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
6707 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
6708 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
6709 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
6710 (vect_create_data_ref_ptr, bump_vector_ptr)
6711 (vect_permute_store_chain, vect_setup_realignment)
6712 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
6713 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
6714 (vectorizable_reduction, vectorizable_induction)
6715 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
6716 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
6717 than gimple stmts as arguments.
6718 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
6719 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
6720 (can_group_stmts_p, vect_check_gather_scatter)
6721 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
6722 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
6723 (vect_permute_load_chain, vect_shift_permute_load_chain)
6724 (vect_transform_grouped_load)
6725 (vect_record_grouped_load_vectors): Likewise.
6726 * tree-vect-loop.c (vect_fixup_reduc_chain)
6727 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
6728 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
6729 (vectorizable_reduction, vectorizable_induction)
6730 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
6731 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
6732 (vect_get_load_store_mask): Likewise.
6733 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
6734 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
6735 * tree-vect-stmts.c (vect_mark_relevant)
6736 (is_simple_and_all_uses_invariant)
6737 (exist_non_indexing_operands_for_use_p, process_use)
6738 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
6739 (vect_get_vec_def_for_operand, vect_get_vec_defs)
6740 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
6741 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
6742 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
6743 (get_negative_load_store_type, get_load_store_type)
6744 (vect_check_load_store_mask, vect_check_store_rhs)
6745 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
6746 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6747 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6748 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6749 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
6750 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
6751 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
6752 (supportable_widening_operation): Likewise.
6754 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6756 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
6757 a stmt_vec_info instead of a gcall.
6758 (vect_check_gather_scatter): Update call accordingly.
6759 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
6761 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
6762 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
6763 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
6764 instead of a gimple stmt.
6765 (vect_transform_loop): Update calls accordingly.
6766 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
6767 stmt_vec_infos instead of gimple stmts.
6768 (vect_analyze_slp_instance): Update use accordingly.
6769 * tree-vect-stmts.c (read_vector_array, write_vector_array)
6770 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
6771 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
6772 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
6773 (vect_gen_widened_results_half, vect_get_loop_based_defs)
6774 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
6775 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
6776 down to subroutines.
6778 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6780 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
6781 of the worklist from a vector of gimple stmts to a vector of
6783 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6784 (vect_mark_stmts_to_be_vectorized): Likewise
6786 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6788 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
6789 statement before passing it to vect_analyze_stmt.
6790 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
6791 the chain of phi vector definitions. Track the exit phi via its
6793 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
6794 STMT_VINFO_REDUC_DEF.
6795 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
6796 stmt_vec_infos to handle the statement chains.
6797 (vect_get_slp_defs): Record the first statement in the node
6798 using a stmt_vec_info.
6799 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
6800 statements here and pass their stmt_vec_info down to subroutines.
6801 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
6802 down to vect_finish_stmt_generation.
6803 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
6804 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
6805 stmt_vec_infos to subroutines.
6806 (vect_remove_stores): Use stmt_vec_infos to handle the statement
6809 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6811 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
6812 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
6813 (vect_permute_store_chain, vect_setup_realignment)
6814 (vect_permute_load_chain, vect_shift_permute_load_chain)
6815 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
6816 stmts internally, and when passing values to other vectorizer routines.
6817 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6818 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
6819 (vect_analyze_loop_operations, get_initial_def_for_reduction)
6820 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
6821 (vectorizable_reduction, vectorizable_induction)
6822 (vectorizable_live_operation, vect_transform_loop_stmt)
6823 (vect_transform_loop): Likewise.
6824 * tree-vect-patterns.c (vect_reassociating_reduction_p)
6825 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
6826 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
6827 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6828 (vect_slp_analyze_node_operations_1): Likewise.
6829 * tree-vect-stmts.c (vect_mark_relevant, process_use)
6830 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
6831 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
6832 (vect_finish_stmt_generation_1, get_group_load_store_type)
6833 (get_load_store_type, vect_build_gather_load_calls)
6834 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6835 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6836 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6837 (vectorizable_store, vectorizable_load, vectorizable_condition)
6838 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
6839 (supportable_widening_operation): Likewise.
6840 (vect_get_vector_types_for_stmt): Likewise.
6841 * tree-vectorizer.h (vect_dr_behavior): Likewise.
6843 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6845 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6846 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
6847 (vect_permute_store_chain, vect_permute_load_chain)
6848 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
6849 repeated stmt_vec_info lookups.
6850 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
6851 (vect_update_ivs_after_vectorizer): Likewise.
6852 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
6853 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6854 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
6855 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6856 (vect_bb_slp_scalar_cost): Likewise.
6857 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
6859 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6861 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
6862 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
6863 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
6864 to get gassigns and gcalls, rather than operating on generc gimple
6866 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
6867 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
6868 (vectorizable_load, vect_analyze_stmt): Likewise.
6869 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
6871 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6873 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
6874 return stmt_vec_infos rather than gimple stmts. Do not accept
6876 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
6878 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
6879 Update use of get_later_stmt.
6880 (vect_get_constant_vectors): Update call accordingly.
6881 (vect_schedule_slp_instance): Likewise
6882 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
6883 (vect_slp_analyze_instance_dependence): Likewise.
6884 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
6886 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6888 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
6889 (stmt_info_for_cost::stmt_info): ...this new field.
6890 (add_stmt_costs): Update accordingly.
6891 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6892 (vect_get_known_peeling_cost): Likewise.
6893 (vect_estimate_min_profitable_iters): Likewise.
6894 * tree-vect-stmts.c (record_stmt_cost): Likewise.
6896 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6898 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
6899 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6900 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
6902 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
6904 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6906 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
6907 a gimple stmt to a stmt_vec_info.
6908 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
6910 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6912 * tree-vectorizer.h (vec_info::grouped_stores): Change from
6913 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6914 (_loop_vec_info::reduction_chains): Likewise.
6915 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
6917 * tree-vect-slp.c (vect_analyze_slp): Likewise.
6919 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6921 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
6922 a gimple stmt to a stmt_vec_info.
6923 (_stmt_vec_info::next_element): Likewise.
6924 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
6925 (vect_slp_analyze_and_verify_node_alignment)
6926 (vect_analyze_group_access_1, vect_analyze_group_access)
6927 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
6928 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
6929 (vect_supportable_dr_alignment): Update accordingly.
6930 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
6931 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
6932 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
6933 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
6934 * tree-vect-slp.c (vect_build_slp_tree_1)
6935 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
6936 (vect_split_slp_store_group, vect_analyze_slp_instance)
6937 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
6938 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
6939 (get_group_load_store_type, get_load_store_type)
6940 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
6941 (vect_transform_stmt, vect_remove_stores): Likewise.
6943 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6945 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
6947 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
6948 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
6949 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
6950 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
6951 (vector_alignment_reachable_p, vect_get_data_access_cost)
6952 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
6953 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
6954 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
6955 (vect_analyze_group_access_1, vect_analyze_group_access)
6956 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
6957 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
6958 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
6959 result of vect_dr_stmt and use the stmt_vec_info instead of
6960 the associated gimple stmt.
6961 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
6962 (vect_gen_prolog_loop_niters): Likewise.
6963 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6965 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6967 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6968 to a vec<stmt_vec_info>.
6969 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6970 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6972 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6973 to a vec<stmt_vec_info>.
6974 (bst_traits::value_type, bst_traits::value_type): Likewise.
6975 (bst_traits::hash): Update accordingly.
6976 (vect_get_and_check_slp_defs): Change the stmts parameter from
6977 a vec<gimple *> to a vec<stmt_vec_info>.
6978 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6979 (vect_build_slp_tree): Likewise.
6980 (vect_build_slp_tree_2): Likewise. Update uses of
6981 SLP_TREE_SCALAR_STMTS.
6982 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6983 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6984 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6985 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6986 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6987 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6988 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6989 (vect_get_constant_vectors, vect_get_slp_defs)
6990 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6991 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6992 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6993 instead of gimple stmts.
6994 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6995 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6996 (vect_slp_analyze_instance_dependence): Update uses of
6997 SLP_TREE_SCALAR_STMTS.
6998 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6999 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
7000 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
7001 (get_initial_defs_for_reduction): Likewise.
7002 (vect_create_epilog_for_reduction): Likewise.
7003 (vectorize_fold_left_reduction): Likewise.
7004 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
7005 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
7006 (can_vectorize_live_stmts): Likewise.
7008 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7010 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
7011 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
7012 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
7014 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
7016 (gather_scalar_reductions): Update after above interface changes.
7017 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
7018 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
7020 (vect_force_simple_reduction): Likewise.
7021 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
7022 LOOP_VINFO_REDUCTIONS.
7023 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7025 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7027 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
7028 a gimple stmt to a stmt_vec_info.
7029 * tree-vect-loop.c (vect_active_double_reduction_p)
7030 (vect_force_simple_reduction, vectorizable_reduction): Update
7033 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7035 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
7036 vec<gimple *> to a vec<stmt_vec_info>.
7037 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
7038 the reduction_phis argument from a vec<gimple *> to a
7040 (vectorizable_reduction): Likewise the phis local variable that
7041 is passed to vect_create_epilog_for_reduction. Update for new type
7042 of SLP_TREE_VEC_STMTS.
7043 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
7044 (vectorizable_live_operation): Likewise.
7045 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
7046 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
7048 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7050 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
7051 a gimple stmt to a stmt_vec_info.
7052 (vectorizable_condition, vectorizable_live_operation)
7053 (vectorizable_reduction, vectorizable_induction): Pass back the
7054 vectorized statement as a stmt_vec_info.
7055 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
7056 use of STMT_VINFO_VEC_STMT.
7057 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
7058 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
7059 as stmt_vec_infos rather than gimple stmts.
7060 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
7062 (vectorizable_live_operation): Likewise.
7063 (vectorizable_reduction, vectorizable_induction): Likewise,
7064 updating use of STMT_VINFO_VEC_STMT.
7065 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
7066 of STMT_VINFO_VEC_STMT.
7067 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
7068 (vectorizable_simd_clone_call, vectorizable_conversion)
7069 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7070 (vectorizable_store, vectorizable_load, vectorizable_condition)
7071 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
7072 from a gimple stmt to a stmt_vec_info.
7073 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
7074 pointer to a stmt_vec_info to the vectorizable_* routines.
7076 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7078 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
7079 a gimple stmt to a stmt_vec_info.
7080 (is_pattern_stmt_p): Update accordingly.
7081 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
7082 (vect_record_grouped_load_vectors): Likewise.
7083 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
7084 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
7085 (vect_model_reduction_cost): Likewise.
7086 (vect_create_epilog_for_reduction): Likewise.
7087 (vectorizable_reduction, vectorizable_induction): Likewise.
7088 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7089 Return the stmt_vec_info for the pattern statement.
7090 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
7091 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
7092 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7093 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
7094 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
7095 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
7096 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
7097 (free_stmt_vec_info, vect_is_simple_use): Likewise.
7099 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7101 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
7102 (vect_finish_stmt_generation): Likewise.
7103 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7104 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
7105 (vect_build_gather_load_calls): Use the return value of the above
7106 functions instead of a separate call to vinfo_for_stmt. Use narrow
7107 scopes for the input gimple stmt and wider scopes for the associated
7108 stmt_vec_info. Use vec_info::lookup_def when setting these
7109 stmt_vec_infos from an SSA_NAME definition.
7110 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
7111 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
7112 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
7113 (vectorizable_store, vectorizable_load, vectorizable_condition)
7114 (vectorizable_comparison): Likewise.
7115 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
7116 (vectorizable_reduction): Likewise.
7118 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7120 * tree-vectorizer.h (vect_is_simple_use): Add an optional
7121 stmt_vec_info * parameter before the optional gimple **.
7122 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7123 (process_use, vect_get_vec_def_for_operand_1): Update callers.
7124 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
7125 * tree-vect-loop.c (vectorizable_reduction): Likewise.
7126 (vectorizable_live_operation): Likewise.
7127 * tree-vect-patterns.c (type_conversion_p): Likewise.
7128 (vect_look_through_possible_promotion): Likewise.
7129 (vect_recog_rotate_pattern): Likewise.
7130 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
7132 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7134 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
7135 a typedef to a wrapper class.
7136 (NULL_STMT_VEC_INFO): New macro.
7137 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
7138 (stmt_vec_info::operator*): New function.
7139 (stmt_vec_info::operator gimple *): Likewise.
7140 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
7141 (add_stmt_costs): Likewise.
7142 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
7143 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
7144 (vect_get_known_peeling_cost): Likewise.
7145 (vect_estimate_min_profitable_iters): Likewise.
7146 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7147 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
7148 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
7149 (vectorizable_store, free_stmt_vec_infos): Likewise.
7150 (new_stmt_vec_info): Change return type of xcalloc to
7153 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7155 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
7156 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
7157 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
7158 a single_imm_use-based sequence.
7159 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
7161 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7163 * tree-vectorizer.h (vec_info::lookup_def): Declare.
7164 * tree-vectorizer.c (vec_info::lookup_def): New function.
7165 * tree-vect-patterns.c (vect_get_internal_def): Use it.
7166 (vect_widened_op_tree): Likewise.
7167 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
7168 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
7169 (vectorizable_reduction): Likewise.
7170 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
7172 (vect_is_slp_reduction): Update calls accordingly. Use
7173 vec_info::lookup_def.
7174 (vect_is_simple_reduction): Likewise
7175 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
7177 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7179 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
7180 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
7181 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
7183 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
7184 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
7185 (vect_update_vf_for_slp, vect_analyze_loop_operations)
7186 (vect_is_slp_reduction, vectorizable_induction)
7187 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
7188 * tree-vect-patterns.c (vect_init_pattern_stmt):
7189 (vect_determine_min_output_precision_1, vect_determine_precisions)
7190 (vect_pattern_recog): Likewise.
7191 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
7192 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
7193 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
7194 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
7195 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
7196 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
7197 info field from a loop to a loop_vec_info.
7199 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7201 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
7202 (vec_info::add_stmt): Declare.
7203 * tree-vectorizer.c (vec_info::add_stmt): New function.
7204 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
7205 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
7206 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
7207 (vectorizable_induction): Likewise.
7208 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
7209 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
7210 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
7211 (vectorizable_load): Likewise.
7212 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
7213 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
7214 (vect_recog_gather_scatter_pattern): Likewise.
7215 (append_pattern_def_seq): Likewise. Remove a check that is
7216 performed by add_stmt itself.
7218 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7220 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
7221 which make_ssa_name was called with new_stmt before new_stmt
7224 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7226 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
7228 (vect_is_slp_reduction): ...here...
7229 (vect_is_simple_reduction): ...and here. Remove repetition of tests
7230 that are already known to be false.
7232 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7234 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
7235 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
7236 STMT_VINFO_NUM_SLP_USES when it's true.
7237 (vect_free_slp_instance): Add a final_p parameter and pass it to
7239 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
7240 (vect_analyze_slp_instance): Likewise.
7241 (vect_slp_analyze_operations): Likewise.
7242 (vect_slp_analyze_bb_1): Likewise.
7243 * tree-vectorizer.c (vec_info): Likewise.
7244 * tree-vect-loop.c (vect_transform_loop): Likewise.
7246 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7248 * tree-vect-loop.c (vectorizable_reduction): Assert that the
7249 function is not called for second and subsequent members of
7252 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
7254 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
7255 cases for nested loops from here to ...
7256 (vect_create_epilog_for_reduction): ...here. Only call
7257 vect_is_simple_use for inner-loop reductions.
7259 2018-07-31 Martin Liska <mliska@suse.cz>
7261 PR gcov-profile/85338
7262 PR gcov-profile/85350
7263 PR gcov-profile/85372
7264 * profile.c (struct location_triplet): New.
7265 (struct location_triplet_hash): Likewise.
7266 (output_location): Do not output a BB that
7267 is already recorded for a line.
7268 (branch_prob): Use streamed_locations.
7270 2018-07-31 Martin Liska <mliska@suse.cz>
7272 PR gcov-profile/85370
7273 * coverage.c (coverage_begin_function): Do not mark target
7274 clones as artificial functions.
7276 2018-07-31 Martin Liska <mliska@suse.cz>
7278 PR gcov-profile/83813
7279 PR gcov-profile/84758
7280 PR gcov-profile/85217
7281 PR gcov-profile/85332
7282 * profile.c (branch_prob): Do not record GOTO expressions
7283 for GIMPLE statements which locations are already streamed.
7285 2018-07-31 Olivier Hainque <hainque@adacore.com>
7287 * gcc.c (handle_spec_function): Accept a soft_matched_part
7288 argument, as do_spec_1. Pass it down to ...
7289 (eval_spec_function): Accept a soft_matched_part argument,
7290 and pass it down to ...
7291 (do_spec_2): Accept a soft_matched_part argument, and pass
7292 it down to do_spec_1.
7293 (do_spec_1): Pass soft_matched_part to handle_spec_function.
7294 (handle_braces): Update call to handle_spec_function.
7295 (driver::set_up_specs): Update calls to do_spec_2.
7296 (compare_debug_dump_opt_spec_function): Likewise.
7297 (compare_debug_self_opt_spec_function): Likewise.
7299 2018-07-31 Olivier Hainque <hainque@adacore.com>
7301 * common.opt (nolibc): New option.
7302 * doc/invoke.texi (Link Options): Document it.
7303 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
7304 * config/alpha/linux.h: Likewise.
7305 * config/arc/elf.h: Likewise.
7306 * config/arm/uclinux-elf.h: Likewise.
7307 * config/arm/unknown-elf.h: Likewise.
7308 * config/avr/avrlibc.h: Likewise.
7309 * config/bfin/bfin.h: Likewise.
7310 * config/bfin/linux.h: Likewise.
7311 * config/bfin/uclinux.h: Likewise.
7312 * config/darwin.h: Likewise.
7313 * config/darwin10.h: Likewise.
7314 * config/darwin12.h: Likewise.
7315 * config/gnu-user.h: Likewise.
7316 * config/lm32/uclinux-elf.h: Likewise.
7317 * config/pa/pa-hpux11.h: Likewise.
7318 * config/pa/pa64-hpux.h: Likewise.
7319 * config/sparc/sparc.h: Likewise.
7321 2018-07-31 Olivier Hainque <hainque@adacore.com>
7323 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
7324 undefined variables.
7326 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7329 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
7332 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7334 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7335 terminated string literal.
7337 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
7339 PR rtl-optimization/85160
7340 * combine.c (is_just_move): New function.
7341 (try_combine): Allow combining two instructions into two if neither of
7342 the original instructions was a move.
7344 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
7347 * doc/extend.texi (Global Register Variables): Discourage use of type
7349 (Local Register Variables): Likewise.
7351 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
7353 PR tree-optimization/86506
7354 * hwint.h (ceil_log2): Resync with hwint.c implementation.
7356 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
7359 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
7360 hard_regno, make sure no insn between `from` and `to` clobbers it.
7362 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
7363 Tom de Vries <tdevries@suse.de>
7365 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
7366 (PTX_DEFAULT_RUNTIME_DIM): ... this.
7367 (nvptx_goacc_validate_dims): Set default worker and gang dims to
7368 PTX_DEFAULT_RUNTIME_DIM.
7369 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
7371 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
7373 * config/pa/pa.c (pa_output_addr_vec): Align address table.
7374 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
7375 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
7377 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
7379 * config/rs6000/constraints.md (wG constraint): Delete, no longer
7381 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
7382 predicate to reflect toc fusion has been deleted.
7383 (toc_fusion_mem_raw): Delete, no longer used.
7384 (toc_fusion_mem_wrapped): Likewise.
7385 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
7387 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
7388 Delete, no longer used.
7389 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
7390 meant to be used for toc fusion.
7391 (rs6000_debug_print_mode): Delete toc fusion debugging.
7392 (rs6000_debug_reg_global): Likewise.
7393 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
7394 fusion and secondary reload support that were never used.
7395 (rs6000_option_override_internal): Delete TOC fusion, that was only
7396 partially defined, and it did not work unless you also used the
7398 (rs6000_legitimate_address_p): Delete TOC fusion support.
7399 (rs6000_opt_masks): Likewise.
7400 (fusion_wrap_memory_address): Delete function, no longer used.
7401 (fusion_split_address); Delete TOC fusion support.
7402 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
7403 longer used with toc fusion being deleted.
7404 (TARGET_TOC_FUSION_FP): Likewise.
7405 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
7407 (toc fusion spliter): Delete TOC fusion support.
7408 (toc_fusionload_<mode>): Likewise.
7409 (toc_fusionload_di): Likewise.
7410 (fusion_gpr_load_<mode>): Delete generator function, this insn no
7411 longer needs to be named. Rename predicate to delete TOC fusion.
7412 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7413 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7414 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
7415 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
7416 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
7418 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
7420 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
7421 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
7422 __int128 in built-in function prototypes.
7423 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
7424 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
7426 2018-07-27 Martin Sebor <msebor@redhat.com>
7428 PR tree-optimization/86696
7429 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
7430 types, including enums.
7431 (handle_char_store): Be prepared for the above function to fail.
7433 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
7435 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
7436 when optimization level is lower than 2 or optimize for size.
7438 2018-07-26 Martin Sebor <msebor@redhat.com>
7440 PR tree-optimization/86043
7441 PR tree-optimization/86042
7442 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
7443 (get_string_cst_length): Rename...
7444 (get_min_string_length): ...to this. Add argument.
7445 (handle_char_store): Extend to handle multi-character stores by
7447 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
7448 * tree.h (initializer_zerop): Add argument.
7450 2018-07-26 Jakub Jelinek <jakub@redhat.com>
7453 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
7454 declare target to variables if they have always,{to,from,tofrom} map
7457 2018-07-26 Martin Liska <mliska@suse.cz>
7460 * lto-wrapper.c: Add linker_output as prefix
7461 for ltrans_output_file.
7463 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
7465 PR rtl-optimization/85805
7466 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
7467 value for hard registers if that was written in the same mode.
7469 2018-07-26 Martin Liska <mliska@suse.cz>
7471 PR gcov-profile/86536
7472 * gcov.c (format_gcov): Use printf format %.*f directly
7473 and do not handle special values.
7475 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7477 * common/config/arc/arc-common.c (arc_option_optimization_table):
7478 Update default optimizations for size.
7480 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7482 * config/arc/arc.md (movsf_insn): Add short instruction selection.
7483 * config/arc/constraints.md (CfZ): New constraint.
7484 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
7485 (subsf3_fpu): Likewise.
7486 (cmpsf_fpu): Likewise.
7487 (cmpsf_fpu_uneq): Likewise.
7489 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7491 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
7493 (arc_is_uncached_mem_p): uncached applies to both the variable and
7496 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
7498 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
7501 2018-07-25 David Malcolm <dmalcolm@redhat.com>
7503 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
7504 field "m_scopes" from vec to auto_vec.
7506 2018-07-25 Martin Liska <mliska@suse.cz>
7508 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
7511 2018-07-25 Richard Biener <rguenther@suse.de>
7514 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
7515 special wrt context_die late.
7516 (gen_subprogram_die): Re-use DIEs in local scope.
7518 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
7520 PR tree-optimization/86644
7521 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
7523 2018-07-25 Martin Liska <mliska@suse.cz>
7526 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
7527 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
7529 2018-07-25 Martin Liska <mliska@suse.cz>
7532 * params.def: Explain ASan abbreviation and provide
7533 a documentation link.
7535 2018-07-24 Martin Sebor <msebor@redhat.com>
7537 PR tree-optimization/86622
7538 PR tree-optimization/86532
7539 * builtins.h (string_length): Declare.
7540 * builtins.c (c_strlen): Correct handling of non-constant offsets.
7541 (check_access): Be prepared for non-constant length ranges.
7542 (string_length): Make extern.
7543 * expr.c (string_constant): Only handle the minor non-constant
7544 array index. Use string_constant to compute the length of
7545 a generic string constant.
7547 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
7549 PR tree-optimization/86618
7550 * tree-vect-stmts.c (vectorizable_call): Don't take the address
7551 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
7553 2018-07-24 David Malcolm <dmalcolm@redhat.com>
7555 PR tree-optimization/86636
7556 * json.cc (json::object::set): Fix comment. Add assertions.
7557 (json::array::append): Move here from json.h. Add comment and an
7559 (json::string::string): Likewise.
7560 * json.h (json::array::append): Move to json.cc.
7561 (json::string::string): Likewise.
7562 * optinfo-emit-json.cc
7563 (optrecord_json_writer::impl_location_to_json): Assert that we
7564 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
7565 wrapper around it. Expand the location once, rather than three
7567 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
7568 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
7570 (optrecord_json_writer::optinfo_to_json): Likewise, in four
7571 places. Fix some overlong lines.
7573 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
7575 * config/aarch64/aarch64-simd.md
7576 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
7577 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
7578 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
7579 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
7580 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
7581 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
7582 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
7583 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
7584 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
7586 2018-07-24 Jakub Jelinek <jakub@redhat.com>
7589 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
7590 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
7591 and abs_d == d, do the power of two handling if profitable.
7593 2018-07-24 Richard Biener <rguenther@suse.de>
7595 * match.pd: Add BIT_FIELD_REF canonicalizations.
7597 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7600 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
7602 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7604 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
7605 terminated STRING_CST object.
7607 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
7609 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
7611 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7613 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
7614 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
7615 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
7616 the elements into a register.
7617 (rs6000_split_v4si_init_di_reg): Delete.
7618 (rs6000_split_v4si_init): Delete.
7619 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
7620 (vsx_init_v4si): Rewrite as a define_expand.
7622 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
7624 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
7625 zero_extend argument from memory): New.
7627 2018-07-22 Martin Sebor <msebor@redhat.com>
7630 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
7631 diagnosing calls with unknown arguments unless -Walloca-larger-than
7632 is restricted to less than PTRDIFF_MAX bytes.
7634 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
7636 * doc/gcov.texi (Invoking Gcov): Editorial changes.
7638 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7640 * pretty-print.c (text_info::set_location): Remove redundant
7641 "line_table" parameter from call to rich_location::set_range.
7643 2018-07-20 Martin Sebor <msebor@redhat.com>
7646 * builtins.c (expand_builtin_alloca): Adjust.
7647 * calls.c (alloc_max_size): Simplify.
7648 * cgraphunit.c (cgraph_node::expand): Adjust.
7649 * common.opt (larger_than_size, warn_frame_larger_than): Remove
7651 (frame_larger_than_size): Same.
7652 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
7653 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
7655 * doc/invoke.texi (GCC Command Options): Document option arguments.
7656 Explain byte-size arguments and suffixes.
7657 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
7658 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
7659 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
7660 * doc/options.texi (UInteger): Expand.
7661 (Host_Wide_Int, ByteSize): Document new properties.
7662 * final.c (final_start_function_1): Include sizes in an error message.
7663 * function.c (frame_offset_overflow): Same.
7664 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
7665 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
7666 Diagnose unbounded alloca calls only for limits of less than
7668 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
7669 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
7671 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
7672 only for limits of less than PTRDIFF_MAX.
7673 * langhooks-def.h (lhd_handle_option): Change function argument
7675 * langhooks.c (lhd_handle_option): Same.
7676 * langhooks.h (handle_option): Same.
7677 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
7679 (var_type, var_type_struct): Same.
7680 (var_set): Handle ByteSize flag.
7681 * optc-gen.awk: Add comments to output to ease debugging. Make
7682 use of HOST_WIDE_INT where appropriate.
7683 * opts-gen-save.awk: Use %lx to format unsigned long.
7684 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
7685 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
7686 arguments. Parse bytes-size suffixes.
7687 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
7688 (enum_value_to_arg): Same.
7689 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
7690 (handle_option): Adjust.
7691 (generate_option): Change function argument to HOST_WIDE_INT.
7692 (cmdline_handle_error): Adjust.
7693 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
7694 (set_option): Change function argument to HOST_WIDE_INT.
7695 (option_enabled): Handle cl_host_wide_int.
7696 (get_option_state): Handle CLVC_SIZE.
7697 (control_warning_option): Same.
7698 * opts.c (common_handle_option): Change function argument to
7699 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
7700 OPT_Wvla_larger_than_.
7701 * opts.h (enum cl_var_type): Add an enumerator.
7702 * stor-layout.c (layout_decl): Print a more meaningful warning.
7703 * toplev.c (output_stack_usage): Adjust.
7705 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
7707 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
7708 call to inline_expand_builtin_string_cmp.
7709 (expand_builtin_strcmp): Likewise.
7710 (expand_builtin_strncmp): Likewise.
7711 (inline_string_cmp): Delete the last parameter, change char_type_node
7712 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
7714 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
7715 the inlining expansion on target where the type of the call has same or
7716 narrower precision than unsigned char.
7718 2018-07-20 David Malcolm <dmalcolm@redhat.com>
7720 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
7721 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
7722 * common.opt (fsave-optimization-record): New option.
7723 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
7724 * doc/invoke.texi (-fsave-optimization-record): New option.
7725 * dumpfile.c: Include "optinfo-emit-json.h".
7726 (struct kv_pair): Move to coretypes.h.
7727 (optgroup_options): Make non-static.
7728 (dump_context::end_scope): Call
7729 optimization_records_maybe_pop_dump_scope.
7730 * dumpfile.h (optgroup_options): New decl.
7731 * json.cc: New file.
7733 * optinfo-emit-json.cc: New file.
7734 * optinfo-emit-json.h: New file.
7735 * optinfo.cc: Include "optinfo-emit-json.h".
7736 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
7737 (optinfo_enabled_p): Check optimization_records_enabled_p.
7738 (optinfo_wants_inlining_info_p): Likewise.
7739 * optinfo.h: Update comment.
7740 * profile-count.c (profile_quality_as_string): New function.
7741 * profile-count.h (profile_quality_as_string): New decl.
7742 (profile_count::quality): New accessor.
7743 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
7744 and optinfo_emit_json_cc_tests.
7745 * selftest.h (selftest::json_cc_tests): New decl.
7746 (selftest::optinfo_emit_json_cc_tests): New decl.
7747 * toplev.c: Include "optinfo-emit-json.h".
7748 (compile_file): Call optimization_records_finish.
7749 (do_compile): Call optimization_records_start.
7750 * tree-ssa-live.c: Include optinfo.h.
7751 (remove_unused_scope_block_p): Retain inlining information if
7752 optinfo_wants_inlining_info_p returns true.
7754 2018-07-20 Richard Biener <rguenther@suse.de>
7757 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
7758 to cover -flto-partition=none.
7760 2018-07-20 Martin Liska <mliska@suse.cz>
7762 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
7763 (get_decl_source_range): Remove unused function.
7765 2018-07-20 Richard Biener <rguenther@suse.de>
7767 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
7768 (struct vn_phi_s): Likewise.
7769 (struct vn_reference_s): Likewise.
7770 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
7771 for searching the slot of an entry known to be in the hash itself.
7772 (vn_phi_hasher::equal): Likewise.
7773 (vn_reference_hasher::equal): Likewise.
7774 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
7776 (optimistic_info, current_info): Remove, keeping only valid_info.
7777 (vn_reference_lookup_1): Remove fallback lookup.
7778 (vn_reference_lookup_2): Likewise.
7779 (vn_nary_op_lookup_1): Likewise.
7780 (vn_phi_lookup): Likewise.
7781 (vn_nary_build_or_lookup_1): Make sure to not chain the built
7783 (vn_reference_insert): Adjust, chain the inserted hash element
7784 at last_inserted_ref.
7785 (vn_reference_insert_pieces): Likewise.
7786 (visit_reference_op_call): Likewise.
7787 (vn_nary_op_insert_into): Chain the inserted hash element at
7789 (vn_nary_op_insert_pieces): Adjust.
7790 (vn_nary_op_insert): Likewise.
7791 (vn_nary_op_insert_stmt): Likewise.
7792 (vn_phi_insert): Adjust, chain the inserted hash element at
7794 (process_scc): Remove clearing and copying the optimistic
7795 table. Instead remove elements inserted during an optimistic
7796 iteration from the single table we maintain.
7797 (init_scc_vn): Adjust.
7798 (free_scc_vn): Likewise.
7799 (sccvn_dom_walker::record_cond): Likewise.
7800 (sccvn_dom_walker::after_dom_children): Likewise.
7802 2018-07-19 Martin Sebor <msebor@redhat.com>
7804 PR tree-optimization/84047
7805 PR tree-optimization/83776
7806 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
7807 (check_array_bounds): Call it.
7809 2018-07-19 Martin Sebor <msebor@redhat.com>
7811 * align.h (align_flags): Use member initialization.
7813 2018-07-19 David Malcolm <dmalcolm@redhat.com>
7815 * Makefile.in (OBJS): Add optinfo.o.
7816 * coretypes.h (class symtab_node): New forward decl.
7817 (struct cgraph_node): New forward decl.
7818 (class varpool_node): New forward decl.
7819 * dump-context.h: New file.
7820 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
7822 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
7823 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
7824 (set_alt_dump_file): Likewise.
7825 (dump_context::~dump_context): New dtor.
7826 (dump_gimple_stmt): Move implementation to...
7827 (dump_context::dump_gimple_stmt): ...this new member function.
7828 Add the stmt to any pending optinfo, creating one if need be.
7829 (dump_gimple_stmt_loc): Move implementation to...
7830 (dump_context::dump_gimple_stmt_loc): ...this new member function.
7831 Start a new optinfo and add the stmt to it.
7832 (dump_gimple_expr): Move implementation to...
7833 (dump_context::dump_gimple_expr): ...this new member function.
7834 Add the stmt to any pending optinfo, creating one if need be.
7835 (dump_gimple_expr_loc): Move implementation to...
7836 (dump_context::dump_gimple_expr_loc): ...this new member function.
7837 Start a new optinfo and add the stmt to it.
7838 (dump_generic_expr): Move implementation to...
7839 (dump_context::dump_generic_expr): ...this new member function.
7840 Add the tree to any pending optinfo, creating one if need be.
7841 (dump_generic_expr_loc): Move implementation to...
7842 (dump_context::dump_generic_expr_loc): ...this new member
7843 function. Add the tree to any pending optinfo, creating one if
7845 (dump_printf): Move implementation to...
7846 (dump_context::dump_printf_va): ...this new member function. Add
7847 the text to any pending optinfo, creating one if need be.
7848 (dump_printf_loc): Move implementation to...
7849 (dump_context::dump_printf_loc_va): ...this new member function.
7850 Start a new optinfo and add the stmt to it.
7851 (dump_dec): Move implementation to...
7852 (dump_context::dump_dec): ...this new member function. Add the
7853 value to any pending optinfo, creating one if need be.
7854 (dump_context::dump_symtab_node): New member function.
7855 (dump_context::get_scope_depth): New member function.
7856 (dump_context::begin_scope): New member function.
7857 (dump_context::end_scope): New member function.
7858 (dump_context::ensure_pending_optinfo): New member function.
7859 (dump_context::begin_next_optinfo): New member function.
7860 (dump_context::end_any_optinfo): New member function.
7861 (dump_context::s_current): New global.
7862 (dump_context::s_default): New global.
7863 (dump_scope_depth): Delete global.
7864 (dumpfile_ensure_any_optinfo_are_flushed): New function.
7865 (dump_symtab_node): New function.
7866 (get_dump_scope_depth): Reimplement in terms of dump_context.
7867 (dump_begin_scope): Likewise.
7868 (dump_end_scope): Likewise.
7869 (selftest::temp_dump_context::temp_dump_context): New ctor.
7870 (selftest::temp_dump_context::~temp_dump_context): New dtor.
7871 (selftest::verify_item): New function.
7872 (ASSERT_IS_TEXT): New macro.
7873 (ASSERT_IS_TREE): New macro.
7874 (ASSERT_IS_GIMPLE): New macro.
7875 (selftest::test_capture_of_dump_calls): New test.
7876 (selftest::dumpfile_c_tests): Call it.
7877 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
7878 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
7879 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
7880 descriptive comment.
7881 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
7882 (dump_node, dump_bb): Move these unrelated decls.
7883 (class dump_manager): Add leading comment.
7884 * optinfo.cc: New file.
7885 * optinfo.h: New file.
7887 2018-07-19 Michael Collison <michael.collison@arm.com>
7888 Richard Henderson <rth@redhat.com>
7890 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
7891 (subti): Handle op1 zero.
7892 (subvti4, usub4ti4): New.
7893 (*sub<GPI>3_compare1_imm): New.
7894 (sub<GPI>3_carryinCV): New.
7895 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
7896 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
7898 2018-07-19 Michael Collison <michael.collison@arm.com>
7899 Richard Henderson <rth@redhat.com>
7901 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
7902 (addti3): Create simpler code if low part is already known to be 0.
7903 (addvti4, uaddvti4): New.
7904 (*add<GPI>3_compareC_cconly_imm): New.
7905 (*add<GPI>3_compareC_cconly): New.
7906 (*add<GPI>3_compareC_imm): New.
7907 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
7908 handle constants within this pattern..
7909 (*add<GPI>3_compareV_cconly_imm): New.
7910 (*add<GPI>3_compareV_cconly): New.
7911 (*add<GPI>3_compareV_imm): New.
7912 (add<GPI>3_compareV): New.
7913 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
7914 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
7915 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
7916 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
7917 with 'comparison' operator.
7918 (*add<GPI>3_compareV_cconly_imm): Ditto.
7919 (*add<GPI>3_compareV_cconly): Ditto.
7920 (*add<GPI>3_compareV_imm): Ditto.
7921 (add<GPI>3_compareV): Ditto.
7922 (add<mode>3_carryinC): Ditto.
7923 (*add<mode>3_carryinC_zero): Ditto.
7924 (*add<mode>3_carryinC): Ditto.
7925 (add<mode>3_carryinV): Ditto.
7926 (*add<mode>3_carryinV_zero): Ditto.
7927 (*add<mode>3_carryinV): Ditto.
7929 2018-07-19 Michael Collison <michael.collison@arm.com>
7930 Richard Henderson <rth@redhat.com>
7932 * config/aarch64/aarch64-modes.def (CC_V): New.
7933 * config/aarch64/aarch64-protos.h
7934 (aarch64_addti_scratch_regs): Declare
7935 (aarch64_subvti_scratch_regs): Declare.
7936 (aarch64_expand_subvti): Declare.
7937 (aarch64_gen_unlikely_cbranch): Declare
7938 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
7939 for signed overflow using CC_Vmode.
7940 (aarch64_get_condition_code_1): Handle CC_Vmode.
7941 (aarch64_gen_unlikely_cbranch): New function.
7942 (aarch64_addti_scratch_regs): New function.
7943 (aarch64_subvti_scratch_regs): New function.
7944 (aarch64_expand_subvti): New function.
7946 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7948 * config/aarch64/aarch64-option-extensions.def: New entry for profile
7950 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
7951 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
7954 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7957 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
7958 address check not strict.
7960 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
7962 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7964 (store_pair_lanes<mode>): Likewise.
7965 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7966 enum value 'ADDR_QUERY_LDP_STP_N'.
7967 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7968 (aarch64_print_address_internal): Add declaration.
7969 (aarch64_print_ldpstp_address): Remove.
7970 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7971 (aarch64_print_operand): Change printing of 'y'.
7972 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7973 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7974 'true' rather than '1'.
7975 * config/aarch64/constraints.md (Uml): Likewise.
7976 (Uml): Rename to Umn.
7979 2018-07-19 Richard Biener <rguenther@suse.de>
7981 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7983 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7984 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7985 (vn_reference_hasher): Likewise.
7986 (struct vn_tables_s): Remove obstack and alloc-pool members.
7987 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7988 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7989 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7990 (vn_reference_insert_pieces): Likewise.
7991 (alloc_vn_nary_op_noinit): Adjust.
7992 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7993 (vn_phi_eq): Adjust.
7994 (shared_lookup_phiargs): Remove.
7995 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7996 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7997 (visit_reference_op_call): Likewise.
7998 (copy_nary, copy_phi, copy_reference): Remove.
7999 (process_scc): Rewind the obstack when iterating. Do not
8000 copy the elements to valid_info but just move them from one
8001 hashtable to the other.
8002 (allocate_vn_table): Adjust.
8003 (free_vn_table): Likewise.
8004 (init_scc_vn): Likewise.
8005 (free_scc_vn): Likewise.
8007 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
8010 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
8011 indirect_return as function type attribute.
8012 (ix86_attribute_table): Change indirect_return to function
8014 * doc/extend.texi: Update indirect_return attribute.
8016 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
8018 * wide-int.h (widest2_int): New.
8019 * gimple-fold.c (arith_overflowed_p): Use it.
8020 * tree.h (widest2_int_cst): New.
8021 * tree-vrp.c (wide_int_binop_overflow): Rename from
8022 vrp_int_const_binop.
8023 Rewrite to work on trees.
8024 (extract_range_from_multiplicative_op_1): Abstract code to...
8025 (wide_int_range_min_max): ...here.
8026 (wide_int_range_cross_product): ...and here.
8027 (extract_range_from_binary_expr_1): Abstract overflow code to...
8028 (wide_int_range_mult_wrapping): ...here.
8029 * tree-vrp.h (wide_int_range_cross_product): New.
8030 (wide_int_range_mult_wrapping): New.
8032 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
8033 Julia Koval <julia.koval@intel.com>
8035 * config/i386/x86-tune-costs.h (skylake_memcpy,
8036 skylake_memset): Replace rep_prefix with unrolling for size 512.
8038 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
8041 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
8042 comparision with EQ_EXPR in last stmt.
8044 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
8046 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
8047 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
8048 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
8049 previously known as "PowerPC AltiVec Built-in Functions". Move
8050 some material to new subsubsections "PowerPC AltiVec Built-in
8051 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
8053 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
8054 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
8055 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
8056 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
8058 2018-07-18 Richard Biener <rguenther@suse.de>
8060 PR tree-optimization/86557
8061 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
8064 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
8066 * config/s390/s390.c (s390_function_profiler): Generate CFI.
8068 2018-07-17 Jeff Law <law@redhat.com>
8070 * config/arm/arm.c (get_label_padding): Update for recent
8071 changes to label_to_alignment.
8073 PR tree-optimization/86010
8074 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
8076 * config/mips/mips.c (vr4130_align_insns): Update for recent
8077 changes to label_to_alignment.
8079 * config/frv/frv.c (frv_label_align): Update for recent changes
8080 to label_to_alignment.
8082 * config/nios2/nios2.c (nios2_label_align): Update for recent
8083 changes which dropped ALIGN_LABELS_LOG.
8085 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
8087 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
8088 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
8090 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
8092 * config/arc/arc.c (arc_label_align): Use align_labels instead of
8093 deprecated align_labels_log.
8095 2018-07-17 Richard Biener <rguenther@suse.de>
8098 * dwarf2out.c (init_sections_and_labels): Always generate
8099 a debug_line_str_section for early LTO debug.
8100 (dwarf2out_finish): Reset debug_line_str_hash output early.
8101 Bump counter for extra dwarf5 .debug_loc labels to not conflict
8103 (dwarf2out_early_finish): Output debug_line_str.
8105 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8107 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
8108 index register on z196 or later.
8110 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
8112 * config/s390/s390.c (s390_default_align): Set default function
8114 (s390_override_options_after_change): Call s390_default align.
8115 (s390_option_override_internal): Call s390_default_align.
8116 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
8118 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8121 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
8122 to _looptemp_ clauses, other than the first two.
8124 2018-07-17 Martin Liska <mliska@suse.cz>
8126 * opts.c: Do not enable OPT_falign_* for -Os.
8128 2018-07-17 Martin Liska <mliska@suse.cz>
8130 * align.h (MAX_CODE_ALIGN): New.
8131 (MAX_CODE_ALIGN_VALUE): New.
8132 * common/config/i386/i386-common.c (ix86_handle_option):
8133 (MAX_CODE_ALIGN): Moved to align.h.
8134 * final.c (MAX_CODE_ALIGN): Likewise.
8135 * opts.c (parse_and_check_align_values):
8136 (MAX_CODE_ALIGN): Likewise.
8137 (MAX_CODE_ALIGN_VALUE): Likewise.
8139 2018-07-17 Martin Liska <mliska@suse.cz>
8141 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
8142 in order to fulfil coding style.
8143 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
8144 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8145 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8146 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
8147 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
8148 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
8149 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
8150 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
8151 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8153 2018-07-17 Martin Liska <mliska@suse.cz>
8155 * align.h: New file.
8156 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
8158 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
8159 align_flags of label_to_alignment.
8160 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
8162 * config/m68k/m68k.c: Do not use removed align_labels_value and
8164 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
8165 (LOOP_ALIGN): Likewise.
8166 (LABEL_ALIGN): Likewise.
8167 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8168 Remove not used macro.
8169 (rs6000_loop_align): Change return type to align_flags.
8170 (rs6000_loop_align_max_skip): Remove.
8171 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
8172 Change return type to align_flags.
8173 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
8174 Remove not used macro.
8175 (rs6000_loop_align): Change return type to align_flags.
8176 (rs6000_loop_align_max_skip): Remove.
8177 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
8178 * config/rx/rx-protos.h (rx_align_for_label): Make it
8180 * config/rx/rx.c (rx_align_for_label): Change return type
8182 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
8184 into align_flags class.
8185 (LABEL_ALIGN): Likewise.
8186 (LOOP_ALIGN): Likewise.
8187 * config/s390/s390.c (s390_label_align): Use align_flags
8189 (s390_asm_output_function_label): Likewise.
8190 * config/sh/sh.c (sh_override_options_after_change):
8191 Use align_flags class directly without macros.
8192 (find_barrier): Likewise.
8193 (barrier_align): Likewise.
8194 (sh_loop_align): Likewise.
8195 * config/spu/spu.c (spu_option_override):
8196 Use align_flags_tuple::get_value instead of removed macros.
8197 (spu_sched_init): Likewise.
8198 * config/spu/spu.h (GTY): Likewise.
8199 * config/visium/visium.c (visium_option_override):
8200 Set "8" as default secondary alignment.
8201 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
8202 in order to guarantee secondary alignment of 8.
8203 * coretypes.h: Include align.h header file.
8204 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
8205 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
8206 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
8207 * doc/tm.texi.in: Likewise.
8208 * final.c (struct label_alignment): Remove not used structure.
8209 (LABEL_ALIGN): Change type to align_flags.
8210 (LOOP_ALIGN): Likewise.
8211 (JUMP_ALIGN): Likewise.
8212 (default_loop_align_max_skip): Remove.
8213 (default_label_align_max_skip): Likewise.
8214 (default_jump_align_max_skip): Likewise.
8215 (default_label_align_after_barrier_max_skip):
8216 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
8217 (LABEL_TO_MAX_SKIP): Remove.
8218 (label_to_alignment): Return align_flags type instead of integer.
8219 (label_to_max_skip): Remove.
8220 (align_fuzz): Use align_flags type.
8221 (compute_alignments): Use align_flags type and use align_flags::max
8222 to combine multiple alignments.
8223 (grow_label_align): Grow vec instead of C array.
8224 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
8225 (shorten_branches): Use align_flags type and use align_flags::max
8226 to combine multiple alignments.
8227 (final_scan_insn_1): Remove usage of secondary alignment that comes
8228 from label alignment, but instead use proper secondary alignment
8229 which is computed in grow_label_align.
8230 * flags.h (struct align_flags_tuple): Move to align.h.
8231 (struct align_flags): Likewise.
8232 (state_align_loops): Rename to align_loops.
8233 (state_align_jumps): Rename to align_jumps.
8234 (state_align_labels): Rename to align_labels.
8235 (state_align_functions): Rename to align_functions.
8236 (align_loops_log): Remove.
8237 (align_jumps_log): Remove.
8238 (align_labels_log): Remove.
8239 (align_functions_log): Remove.
8240 (align_loops_max_skip): Remove.
8241 (align_jumps_max_skip): Remove.
8242 (align_labels_max_skip): Remove.
8243 (align_functions_max_skip): Remove.
8244 (align_loops_value): Remove.
8245 (align_jumps_value): Remove.
8246 (align_labels_value): Remove.
8247 (align_functions_value): Remove.
8248 * output.h (label_to_alignment): Change return type to align_flags.
8249 (label_to_max_skip): Remove.
8250 * target.def: Remove loop_align_max_skip, label_align_max_skip,
8251 jump_align_max_skip macros.
8252 * targhooks.h (default_loop_align_max_skip): Remove.
8253 (default_label_align_max_skip): Likewise.
8254 (default_jump_align_max_skip): Likewise.
8255 (default_label_align_after_barrier_max_skip): Remove.
8256 * toplev.c (read_log_maxskip): Use ::normalize function.
8257 (parse_N_M): Remove not used argument and also call ::normalize.
8258 (parse_alignment_opts): Do not pass unused arguments.
8259 * varasm.c (assemble_start_function): Use directly align_functions
8260 instead of removed macros.
8261 * system.h: Do not poison removed macros.
8263 2018-07-17 Jakub Jelinek <jakub@redhat.com>
8266 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
8267 and cond temporaries don't have reference type if iterator has
8268 pointer type. For init use &for_pre_body instead of pre_p if
8269 for_pre_body is non-empty.
8271 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8273 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
8274 double-double modes to SFmode directly directly.
8275 (trunc<mode>sf2_fprs): Delete.
8277 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8279 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
8280 for conversions between IFmode and the decimal floating point modes.
8281 (init_float128_ieee): Use the correct names for conversions between
8282 KFmode and the decimal floating point modes.
8284 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
8286 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
8287 for the conversions between TDmode and IFmode.
8288 (init_float128_ieee): Use more correct names for the conversions
8289 between TDmode and KFmode.
8291 2018-07-16 Jakub Jelinek <jakub@redhat.com>
8293 PR tree-optimization/86526
8294 * builtins.c (expand_builtin_memcmp): Formatting fixes.
8295 (inline_expand_builtin_string_cmp): Likewise.
8296 (inline_string_cmp): Likewise. Use c_readstr instead of
8297 builtin_memcpy_read_str. Add unit_mode temporary.
8299 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
8302 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
8303 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
8305 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
8307 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
8308 Alphabetize prototypes of built-in functions, separating out
8309 built-in functions that are listed in this section but should be
8310 described elsewhere.
8312 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
8315 * expmed.c (emit_store_flag): Do not emit setcc followed by a
8316 conditional move when trapping comparison was split to a
8317 non-trapping one (and vice versa).
8319 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8321 * config/s390/s390.c (s390_function_profiler): Generate nops
8322 instead of profiler call sequences.
8323 * config/s390/s390.opt: Add the new option.
8325 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8327 * config/s390/s390.c (s390_function_profiler): Generate
8328 __mcount_loc section.
8329 * config/s390/s390.opt: Add the new option.
8331 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
8333 * common.opt: Add the new warning.
8334 * config/s390/s390.c (s390_function_profiler): Emit "brasl
8335 %r0,__fentry__" when -mfentry is specified.
8336 (s390_option_override_internal): Disallow -mfentry for 31-bit
8338 * config/s390/s390.opt: Add the new option.
8340 2018-07-16 Richard Biener <rguenther@suse.de>
8343 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
8344 for function-local FUNCTION_DECL and RESULT_DECL immediately.
8346 2018-07-16 Martin Liska <mliska@suse.cz>
8349 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
8352 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8354 * config/arc/arcHS.md: Update ARCHS scheduling rules.
8356 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
8358 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
8360 * config/arc/arc-cpus.def (hs4x): New cpu.
8362 * config/arc/arc-tables.opt: Regenerate.
8363 * config/arc/arc.c (arc_sched_issue_rate): New function.
8364 (TARGET_SCHED_ISSUE_RATE): Define.
8365 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
8366 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
8368 (attr tune): Add ARCHS4x tune values.
8369 (attr tune_dspmpy): Define.
8370 (*tst): Correct instruction type.
8371 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
8372 * config/arc/arcHS4x.md: New file.
8373 * config/arc/fpu.md: Update instruction type attributes.
8374 * config/arc/t-multilib: Regenerate.
8376 2018-07-16 Tom de Vries <tdevries@suse.de>
8379 * var-tracking.c (vt_initialize): Fix pre_dec handling.
8381 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
8383 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
8386 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
8388 PR tree-optimization/86514
8389 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
8390 conversion to a boolean type from a type with greater precision.
8392 2018-07-16 Tom de Vries <tdevries@suse.de>
8394 * var-tracking.c (vt_initialize): Print adjusted insn slim if
8395 dump_flags request TDF_SLIM.
8397 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
8399 * fold-const.c (int_const_binop_1): Abstract...
8400 (wide_int_binop): ...wide int code here.
8401 (poly_int_binop): ...poly int code here.
8402 Abstract the rest of int_const_binop_1 into int_const_binop.
8403 * fold-const.h (wide_int_binop): New.
8404 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
8405 Remove useless PLUS/MINUS_EXPR case.
8406 (zero_nonzero_bits_from_vr): Move wide int code...
8407 (zero_nonzero_bits_from_bounds): ...here.
8408 (extract_range_from_binary_expr_1): Move mask optimization code...
8409 (range_easy_mask_min_max): ...here.
8410 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
8411 (range_easy_mask_min_max): New.
8413 2018-07-15 Jeff Law <law@redhat.com>
8416 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
8419 2018-07-14 Jim Wilson <jimw@sifive.com>
8421 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
8423 2018-07-14 Paul Koning <ni1d@arrl.net>
8425 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
8427 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
8429 * lto-streamer-out.c (copy_function_or_variable): Dump info about
8432 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
8433 Steve Munroe <munroesj52@gmail.com>
8435 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
8436 (_mm_andnot_si128): Likewise.
8437 (_mm_or_si128): Likewise.
8438 (_mm_xor_si128): Likewise.
8440 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
8443 * builtins.c (expand_builtin_memcmp): Inline the calls first
8444 when result_eq is false.
8445 (expand_builtin_strcmp): Inline the calls first.
8446 (expand_builtin_strncmp): Likewise.
8447 (inline_string_cmp): New routine. Expand a string compare
8448 call by using a sequence of char comparison.
8449 (inline_expand_builtin_string_cmp): New routine. Inline expansion
8450 a call to str(n)cmp/memcmp.
8451 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
8453 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
8455 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
8457 * config/arm/driver-arm.c: Include arm-native.h.
8458 (host_detect_local_cpu): Use auto-generated data tables.
8459 (vendors, arm_cpu_table): Delete. Move part information to ...
8460 * config/arm/arm-cpus.in: ... here.
8461 * config/arm/parsecpu.awk (gen_native): New function.
8462 (vendor, part): New CPU fields.
8463 (END): Add support for building the native CPU detection tables.
8464 * config/arm/t-arm (arm-native.h): Add build rule.
8465 (driver-arm.o): Add dependency on arm-native.h.
8467 2018-07-13 Richard Biener <rguenther@suse.de>
8470 * match.pd (addr1 - addr2): Allow either of the operand to
8473 2018-07-13 Tom de Vries <tdevries@suse.de>
8475 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
8478 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
8480 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
8481 arrays instead of numbered variables.
8483 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
8485 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
8486 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
8488 2018-07-13 Richard Biener <rguenther@suse.de>
8491 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
8492 instead of get_context_die.
8494 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
8495 Richard Biener <rguenther@suse.de>
8498 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
8499 that the loop latch destination where phi is defined.
8501 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
8503 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
8504 (riscv_expand_epilogue): Add assertion to check interrupt mode.
8505 (riscv_set_current_function): Extract getting interrupt type to new
8507 (riscv_get_interrupt_type): New function.
8508 (riscv_merge_decl_attributes): New function, checking interrupt type is
8510 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
8512 2018-07-12 Paul Koning <ni1d@arrl.net>
8514 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
8517 2018-07-12 Paul Koning <ni1d@arrl.net>
8519 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
8520 zero reference, add doloop_end instead.
8521 * doc/md.texi (decrement_and_branch_until_zero): Remove.
8522 (Looping patterns): Remove decrement_and_branch_until_zero. Add
8523 detail for doloop_end.
8525 2018-07-12 Martin Sebor <msebor@redhat.com>
8528 * attribs.c (decl_attributes): Reject conflicting attributes before
8529 calling attribute handlers.
8531 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8533 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
8535 (gcc::dump_manager::get_dump_file_name): likewise.
8536 (dump_begin): Likewise.
8537 * dumpfile.h (dump_begin): Update prototype.
8538 (gcc::dump_manager::get_dump_file_name,
8539 gcc::dump_manager::get_dump_file_name): Update prototype.
8541 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8543 * internal-fn.h (vectorizable_internal_fn_p): New function.
8544 * tree-vect-slp.c (compatible_calls_p): Likewise.
8545 (vect_build_slp_tree_1): Remove nops argument. Handle calls
8546 to internal functions.
8547 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
8549 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8551 * fold-const.h (inverse_conditions_p): Declare.
8552 * fold-const.c (inverse_conditions_p): New function.
8553 * match.pd: Use inverse_conditions_p. Add folds of view_converts
8554 that test the inverse condition of a conditional internal function.
8555 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
8556 * internal-fn.c (internal_fn_mask_index): Handle conditional
8558 (vectorized_internal_fn_supported_p): New function.
8559 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
8560 (any_pred_load_store): Replace with...
8561 (need_to_predicate): ...this new variable.
8562 (redundant_ssa_names): New variable.
8563 (ifcvt_can_use_mask_load_store): Move initial checks to...
8564 (ifcvt_can_predicate): ...this new function. Handle tree codes
8565 for which a conditional internal function exists.
8566 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
8567 instead of ifcvt_can_use_mask_load_store. Update after variable
8569 (predicate_load_or_store): New function, split out from
8570 predicate_mem_writes.
8571 (check_redundant_cond_expr): New function.
8572 (value_available_p): Likewise.
8573 (predicate_rhs_code): Likewise.
8574 (predicate_mem_writes): Rename to...
8575 (predicate_statements): ...this. Use predicate_load_or_store
8576 and predicate_rhs_code.
8577 (combine_blocks, tree_if_conversion): Update after above name changes.
8578 (ifcvt_local_dce): Handle redundant_ssa_names.
8579 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
8580 general conditional functions.
8581 * tree-vect-stmts.c (vectorizable_call): Likewise.
8583 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8584 Alan Hayward <alan.hayward@arm.com>
8585 David Sherwood <david.sherwood@arm.com>
8587 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
8588 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
8589 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
8590 plus and minus and convert them into IFN_COND_FMA-based sequences.
8591 (convert_mult_to_fma): Handle conditional plus and minus.
8593 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8595 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
8596 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
8597 (cond_fnms_optab): New optabs.
8598 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
8600 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
8601 * internal-fn.h (get_conditional_internal_fn): Declare.
8602 (get_unconditional_internal_fn): Likewise.
8603 * internal-fn.c (cond_ternary_direct): New macro.
8604 (expand_cond_ternary_optab_fn): Likewise.
8605 (direct_cond_ternary_optab_supported_p): Likewise.
8606 (FOR_EACH_COND_FN_PAIR): Likewise.
8607 (get_conditional_internal_fn): New function.
8608 (get_unconditional_internal_fn): Likewise.
8609 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
8610 (gimple_match_op::gimple_match_op): Add a new overload for 5
8612 (gimple_match_op::set_op): Likewise.
8613 (gimple_resimplify5): Declare.
8614 * genmatch.c (decision_tree::gen): Generate simplifications for
8616 * gimple-match-head.c (gimple_simplify): Define an overload for
8617 5 operands. Handle calls with 5 arguments in the top-level overload.
8618 (convert_conditional_op): Handle conversions from unconditional
8619 internal functions to conditional ones.
8620 (gimple_resimplify5): New function.
8621 (build_call_internal): Pass a fifth operand.
8622 (maybe_push_res_to_seq): Likewise.
8623 (try_conditional_simplification): Try converting conditional
8624 internal functions to unconditional internal functions.
8625 Handle 3-operand unconditional forms.
8626 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
8627 Define ternary equivalents of the current rules for binary conditional
8629 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
8631 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
8632 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
8633 (optab): Handle them.
8634 (SVE_COND_FP_TERNARY): New int iterator.
8635 (sve_fmla_op, sve_fmad_op): New int attributes.
8636 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
8637 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
8638 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
8640 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
8642 * target.def (preferred_else_value): New target hook.
8643 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
8644 * doc/tm.texi: Regenerate.
8645 * targhooks.h (default_preferred_else_value): Declare.
8646 * targhooks.c (default_preferred_else_value): New function.
8647 * internal-fn.h (conditional_internal_fn_code): Declare.
8648 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
8649 (get_conditional_internal_fn): Use it.
8650 (conditional_internal_fn_code): New function.
8651 * gimple-match.h (gimple_match_cond): New struct.
8652 (gimple_match_op): Add a cond member function.
8653 (gimple_match_op::gimple_match_op): Update all forms to take a
8655 * genmatch.c (expr::gen_transform): Use the same condition as res_op
8656 for the suboperation, but don't specify a particular else_value.
8657 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
8658 (visit_nary_op, visit_reference_op_load): Pass
8659 gimple_match_cond::UNCOND to the gimple_match_op constructor.
8660 * gimple-match-head.c: Include tree-eh.h
8661 (convert_conditional_op): New function.
8662 (maybe_resimplify_conditional_op): Likewise.
8663 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
8664 (gimple_resimplify2): Likewise.
8665 (gimple_resimplify3): Likewise.
8666 (gimple_resimplify4): Likewise.
8667 (maybe_push_res_to_seq): Return null for conditional operations.
8668 (try_conditional_simplification): New function.
8669 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
8671 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
8673 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
8675 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
8677 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
8679 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
8681 (hash_tree): Do not hash DECL_FCONTEXT
8682 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
8683 Do not stream DECL_FCONTEXT.
8684 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
8685 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
8687 2018-07-12 Richard Biener <rguenther@suse.de>
8690 * dwarf2out.c (gen_block_die): Only output blocks when they have
8691 at least one !DECL_IGNORED_P variable.
8693 2018-07-12 Richard Biener <rguenther@suse.de>
8696 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
8697 Remove -mieee-fp handling.
8699 2018-07-12 Richard Biener <rguenther@suse.de>
8701 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
8702 left-over from last patch.
8704 2018-07-12 Jakub Jelinek <jakub@redhat.com>
8706 PR tree-optimization/86492
8707 * gimple-ssa-store-merging.c
8708 (imm_store_chain_info::coalesce_immediate_stores): Call
8709 check_no_overlap even for the merge_overlapping case. Formatting fix.
8711 2018-07-12 Richard Biener <rguenther@suse.de>
8714 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
8715 move possibly trapping operations into the conditional.
8717 2018-07-12 Richard Biener <rguenther@suse.de>
8719 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
8720 (vn_lookup_simplify_result): Remove recursion limit applied
8722 (vn_nary_build_or_lookup_1): Adjust.
8723 (try_to_simplify): Likewise.
8724 * gimple-match-head.c (gimple_resimplify1): Instead apply one
8726 (gimple_resimplify2): Likewise.
8727 (gimple_resimplify3): Likewise.
8728 (gimple_resimplify4): Likewise.
8730 2018-07-11 Jakub Jelinek <jakub@redhat.com>
8732 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
8733 Use __mmask64 type instead of __mmask8 for __M argument.
8734 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
8735 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
8737 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
8738 __mmask16 for __M argument.
8739 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
8740 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
8741 to __mmask16 instead of __mmask8.
8742 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
8743 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
8744 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
8745 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
8746 instead of __mmask16 for __U argument.
8747 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
8748 __mmask16 instead of __mmask8 for __U argument.
8749 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
8751 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
8753 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
8755 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
8757 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
8759 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
8761 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
8763 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
8765 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
8767 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
8769 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
8770 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
8771 return type as well as __M argument type and all casts from __mmask8
8773 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
8774 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
8775 return type as well as __M argument type and all casts from __mmask8
8777 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
8778 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
8779 return type as well as __M argument type and all casts from __mmask8
8781 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
8782 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
8783 return type as well as __M argument type and all casts from __mmask8
8785 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
8786 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
8789 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
8791 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
8792 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
8795 2018-07-11 Paul Koning <ni1d@arrl.net>
8797 * doc/md.texi (define_subst): Document how multiple occurrences of
8798 the same argument in the replacement pattern are handled.
8800 2018-07-11 Paul Koning <ni1d@arrl.net>
8802 * doc/extend.texi (Common Variable Attributes): Move "mode" into
8804 (Common Type Attributes): Add "mode" attribute.
8806 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
8808 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
8809 stream DECL_ORIGINAL_TYPE.
8810 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
8811 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
8812 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
8813 Do not walk original type.
8814 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
8816 (write_ts_decl_non_common_tree_pointers): Do not stream
8818 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
8819 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
8821 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8823 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
8824 thread twice from the same starting edge.
8826 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
8828 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
8829 * gimple.c (gimple_call_nonnull_result_p): ...here...
8830 (gimple_call_nonnull_arg): ...and here.
8831 * gimple.h (gimple_call_nonnull_result_p): New.
8832 (gimple_call_nonnull_arg): New.
8834 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
8836 * config/arm/arm-cpus.in: Move information from fpu field of each
8837 cpu definition to the isa field.
8838 * config/arm/parsecpu.awk (fpu): Delete match rule.
8839 (gen_comm_data): Don't add bits from the CPU's FPU entry.
8841 2018-07-11 Richard Biener <rguenther@suse.de>
8844 * dwarf2out.c (init_sections_and_labels): Use
8845 output_asm_line_debug_info consistently.
8846 (dwarf2out_early_finish): Likewise.
8847 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
8850 2018-07-11 Richard Biener <rguenther@suse.de>
8852 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
8853 Rework father_bb setting in a way to avoid propagating constants
8854 multiple times on a loop body.
8856 2018-07-10 Mark Wielaard <mark@klomp.org>
8859 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
8861 2018-07-10 Richard Biener <rguenther@suse.de>
8863 * hash-map.h (hash_map::iterator::operator*): Return
8864 references to key and value.
8866 2018-07-10 Jakub Jelinek <jakub@redhat.com>
8869 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
8870 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
8871 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
8872 (gimplify_omp_for): For composite loops, move outer
8873 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
8874 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
8875 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
8876 TREE_LIST for both the original class iterator and the "last" helper
8877 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
8878 loop, remember has_decl_expr from outer composite loops for the
8879 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
8881 2018-07-09 Martin Sebor <msebor@redhat.com>
8885 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
8886 accessing implicitly initialized array elements.
8887 * expr.c (string_constant): Handle string initializers of
8888 character arrays within aggregates.
8889 * gimple-fold.c (fold_array_ctor_reference): Add argument.
8890 Store element offset. As a special case, handle zero size.
8891 (fold_nonarray_ctor_reference): Same.
8892 (fold_ctor_reference): Add argument. Store subobject offset.
8893 * gimple-fold.h (fold_ctor_reference): Add argument.
8895 2018-07-09 Paul Koning <ni1d@arrl.net>
8897 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
8898 (pdp11_insn_cost): New function.
8899 (pdp11_md_asm_adjust): New function.
8900 (TARGET_INVALID_WITHIN_DOLOOP): Define.
8901 (pdp11_rtx_costs): Update to match machine better.
8902 (output_addr_const_pdp11): Correct format mismatch warnings.
8903 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
8904 * config/pdp11/pdp11.md: General change to add base_cost and/or
8905 length attributes for use by new pdp11_insn_cost function.
8906 (MIN_BRANCH): Correct definition.
8908 (doloop_end): Use standard pattern name for looping pattern.
8909 (doloop_end_nocc): New.
8910 (movsf): Add another constraint alternative.
8911 (zero_extendqihi2): Add constraint alternatives for not in place
8913 (zero_extendhisi2): Remove.
8914 (shift patterns): Add CC handling variants.
8918 (define_peephole2): New peephole to recognize mov that sets CC for
8921 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8923 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
8924 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
8927 2018-07-09 Jakub Jelinek <jakub@redhat.com>
8930 * real.c (real_nextafter): Return true if result is denormal.
8932 2018-07-09 Martin Liska <mliska@suse.cz>
8934 * common.opt: Add back wrongly removed attribute.
8936 2018-07-09 Richard Biener <rguenther@suse.de>
8939 * dwarf2out.c (gen_block_die): For an early generated DIE
8940 always output high/low PC attributes.
8942 2018-07-09 Tom de Vries <tdevries@suse.de>
8944 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
8945 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
8946 onto VAR_DECL with abstract origin.
8948 2018-07-07 Jim Wilson <jimw@sifive.com>
8950 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
8952 2018-07-07 Tom de Vries <tdevries@suse.de>
8954 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
8956 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
8958 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
8960 (combine_bound): Use wide-int overflow calculation instead of
8962 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8964 * fold-const.c (int_const_binop_2): Same.
8965 (extract_muldiv_1): Same.
8966 (fold_div_compare): Same.
8967 (fold_abs_const): Same.
8969 * poly-int.h (add): Same.
8973 * predict.c (predict_iv_comparison): Same.
8974 * profile-count.c (slow_safe_scale_64bit): Same.
8975 * simplify-rtx.c (simplify_const_binary_operation): Same.
8976 * tree-chrec.c (tree_fold_binomial): Same.
8977 * tree-data-ref.c (split_constant_offset_1): Same.
8978 * tree-if-conv.c (idx_within_array_bound): Same.
8979 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8980 * tree-ssa-phiopt.c (minmax_replacement): Same.
8981 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8982 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8983 * vr-values.c (vr_values::adjust_range_with_scev): Same.
8984 * wide-int.cc (wi::add_large): Same.
8985 (wi::mul_internal): Same.
8986 (wi::sub_large): Same.
8987 (wi::divmod_internal): Same.
8988 * wide-int.h: Change overflow type to overflow_type for neg, add,
8989 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8990 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8991 mul_internal, divmod_internal.
8992 (overflow_type): New enum.
8993 (accumulate_overflow): New.
8995 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
8997 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8998 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
9000 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
9002 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
9003 argument is checked for zero before entering loop, avoid checking again.
9005 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
9007 * gimplify.h (generic_expr_could_trap_p): Set as global function.
9008 * gimplify.h (generic_expr_could_trap_p): Likwise.
9009 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
9011 2018-07-06 Jakub Jelinek <jakub@redhat.com>
9013 PR tree-optimization/86401
9014 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
9015 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
9016 (fold_bit_and_mask): ... here. New helper function for match.pd.
9017 * fold-const.h (fold_bit_and_mask): Declare.
9018 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
9020 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
9023 * target.def (translate_mode_attribute): New hook.
9024 * targhooks.h (default_translate_mode_attribute): Declare.
9025 * targhooks.c (default_translate_mode_attribute): New function.
9026 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
9027 * doc/tm.texi: Regenerate.
9028 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
9029 (rs6000_translate_mode_attribute): New function.
9031 2018-07-06 Paul Koning <ni1d@arrl.net>
9033 * doc/md.texi (define_split): Document DONE and FAIL.
9034 (define_peephole2): Ditto.
9036 2018-07-05 Jeff Law <law@redhat.com>
9038 PR tree-optimization/86010
9039 * tree-ssa-dse.c (compute_trims): More aggressively trim at
9040 both the head and tail of mem* and str* calls.
9042 2018-07-05 Jim Wilson <jimw@sifive.com>
9044 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
9046 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
9048 * config/aarch64/aarch64-simd.md: correct flags text for
9049 MIN_EXPR replacement.
9051 2018-07-05 James Clarke <jrtc27@jrtc27.com>
9053 * configure: Regenerated.
9055 2018-07-05 Carl Love <cel@us.ibm.com>
9057 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
9058 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
9059 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
9060 VSX_BUILTIN_DOUBLEL_V4SF.
9062 2018-07-05 Martin Sebor <msebor@redhat.com>
9065 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
9066 than its domain to compute its the upper bound of a char array.
9068 2018-07-05 Nathan Sidwell <nathan@acm.org>
9070 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
9071 * doc/cpp.texi: Update comment.
9072 * doc/tm.texi: Rebuilt.
9073 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
9074 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
9075 * doc/extend.texi (Backwards Compatibility): Clarify it is system
9076 headers affected by extern "C".
9077 * system.h: Poison NO_IMPLICIT_EXTERN_C.
9078 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
9079 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
9080 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
9081 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
9082 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
9083 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
9084 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
9085 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
9086 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
9087 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
9088 config/sparc/sp64-elf.h, config/spu/spu.h,
9089 config/stormy16/stormy16.h, config/v850/v850.h,
9090 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
9091 define NO_IMPLICIT_EXTERN_C.
9092 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
9094 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9097 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
9098 instead of GET_MODE_SIZE when comparing Units.
9100 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9103 * rtlanal.c (set_noop_p): Constrain on mode change,
9104 include hard-reg-set.h
9106 2018-07-05 Tamar Christina <tamar.christina@arm.com>
9108 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
9110 2018-07-05 Jakub Jelinek <jakub@redhat.com>
9113 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9116 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9118 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9120 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
9123 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
9125 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
9127 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
9130 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
9132 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
9134 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
9135 neg_*_op* variables.
9137 2018-07-04 Martin Liska <mliska@suse.cz>
9139 * tree-switch-conversion.c: Define
9140 max_ratio_for_speed and max_ratio_for_size constants.
9142 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9143 Martin Liska <mliska@suse.cz>
9148 * common.opt: Rename align options with 'str_' prefix.
9149 * common/config/i386/i386-common.c (set_malign_value): New
9151 (ix86_handle_option): Use it to set -falign-* options/
9152 * config/aarch64/aarch64-protos.h (struct tune_params): Change
9153 type from int to string.
9154 * config/aarch64/aarch64.c: Update default values from int
9156 * config/alpha/alpha.c (alpha_override_options_after_change):
9158 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
9159 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9160 max skip conditionally.
9161 * config/i386/freebsd.h (SUBALIGN_LOG): New.
9162 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9163 max skip conditionally.
9164 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9165 max skip conditionally.
9166 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
9167 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9168 max skip conditionally.
9169 * config/i386/i386.c (struct ptt): Change type from int to
9171 (ix86_default_align): Set default values.
9172 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
9173 max skip conditionally.
9174 * config/i386/iamcu.h (SUBALIGN_LOG): New.
9175 (ASM_OUTPUT_MAX_SKIP_ALIGN):
9176 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
9177 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9178 max skip conditionally.
9179 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
9180 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
9181 * config/i386/x86-64.h (SUBALIGN_LOG): New.
9182 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
9183 max skip conditionally.
9184 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9185 * config/ia64/ia64.c (ia64_option_override): Set default values
9186 for alignment options.
9187 * config/m68k/m68k.c: Handle new str_align_* options.
9188 * config/mips/mips.c (mips_set_compression_mode): Change
9190 (mips_option_override): Set default values for options.
9191 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
9193 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9195 * config/rx/rx.c (rx_option_override): Likewise.
9196 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
9197 (LABEL_ALIGN): Use align_labels_log.
9198 (LOOP_ALIGN): Use align_loops_align.
9199 * config/s390/s390.c (s390_asm_output_function_label): Use new
9201 * config/sh/sh.c (sh_override_options_after_change):
9202 Change type of constants.
9203 * config/spu/spu.c (spu_sched_init): Likewise.
9204 * config/sparc/sparc.c (sparc_option_override): Set default
9206 * config/visium/visium.c (visium_option_override): Likewise.
9207 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
9208 emit p2align format with last argument if it's not needed.
9209 * doc/invoke.texi: Document extended format of -falign-*.
9210 * final.c: Use align_labels alignment.
9211 * flags.h (struct target_flag_state): Change type to use
9213 (struct align_flags_tuple): New.
9214 (struct align_flags): Likewise.
9215 (align_loops_log): Redefine macro to use new types.
9216 (align_loops_max_skip): Redefine macro to use new types.
9217 (align_jumps_log): Redefine macro to use new types.
9218 (align_jumps_max_skip): Redefine macro to use new types.
9219 (align_labels_log): Redefine macro to use new types.
9220 (align_labels_max_skip): Redefine macro to use new types.
9221 (align_functions_log): Redefine macro to use new types.
9222 (align_loops): Redefine macro to use new types.
9223 (align_jumps): Redefine macro to use new types.
9224 (align_labels): Redefine macro to use new types.
9225 (align_functions): Redefine macro to use new types.
9226 (align_functions_max_skip): Redefine macro to use new types.
9227 (align_loops_value): New macro.
9228 (align_jumps_value): New macro.
9229 (align_labels_value): New macro.
9230 (align_functions_value): New macro.
9231 * function.c (invoke_set_current_function_hook): Propagate
9232 alignment values from flags to global variables default in
9234 * ipa-icf.c (sem_function::equals_wpa): Use
9235 cl_optimization_option_eq instead of memcmp.
9236 * lto-streamer.h (cl_optimization_stream_out): Support streaming
9238 (cl_optimization_stream_in): Likewise.
9239 * optc-save-gen.awk: Support strings in cl_optimization.
9240 * opth-gen.awk: Likewise.
9241 * opts.c (finish_options): Remove error checking of invalid
9243 (MAX_CODE_ALIGN): Remove.
9244 (MAX_CODE_ALIGN_VALUE): Likewise.
9245 (parse_and_check_align_values): New function.
9246 (check_alignment_argument): Likewise.
9247 (common_handle_option): Use check_alignment_argument.
9248 * opts.h (parse_and_check_align_values): Declare.
9249 * toplev.c (init_alignments): Remove.
9250 (read_log_maxskip): New.
9251 (parse_N_M): Likewise.
9252 (parse_alignment_opts): Likewise.
9253 (backend_init_target): Remove usage of init_alignments.
9254 * toplev.h (parse_alignment_opts): Declare.
9255 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
9257 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
9258 * tree.c (cl_option_hasher::equal): New.
9259 * varasm.c: Use new global macros.
9261 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
9263 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
9264 Use a simpler align directive also if MAXSKIP = ALIGN-1.
9265 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9266 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9267 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9268 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
9269 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
9270 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
9271 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
9272 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9273 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9274 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9275 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
9277 2018-07-04 Martin Liska <mliska@suse.cz>
9278 Jonathan Wakely <jwakely@redhat.com>
9280 * coverage.c: Use correct type.
9281 * doc/invoke.texi: Language correction.
9283 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
9286 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
9287 ENDBRANCH for non-tail call which may return via indirect branch.
9288 * doc/extend.texi: Document indirect_return attribute.
9290 2018-07-03 Martin Sebor <msebor@redhat.com>
9292 PR tree-optimization/86274
9293 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
9295 (format_floating): Correct handling of infinities and NaNs.
9297 2018-07-03 Martin Sebor <msebor@redhat.com>
9299 * print-tree.c (print_real_cst): New function.
9300 (print_node_brief): Call it.
9301 (print_node): Ditto.
9303 2018-07-03 Jeff Law <law@redhat.com>
9305 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
9306 into a single pattern.
9308 * config/h8300/h8300.md (ors code_iterator): New.
9309 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
9310 a single pattern and single splitter.
9311 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
9312 (iorqi3_1, xorqi3_1): Likewise.
9313 (iorqi3, xorqi3 expanders): Similarly.
9315 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
9316 (movmd_internal) into a single pattern using the P mode iterator.
9317 (movmd splitters): Similarly.
9318 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
9319 (movsd splitters): Similarly.
9321 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
9322 ADDB, ADDW and ADDL into a single ADD attribute which selects the
9323 right table based on the size of the operand.
9324 * config/h8300/h8300.md (length_table): Corresponding changes. All
9325 references to "addb", "addw" and "addl" changed to "add".
9326 (btst patterns): Merge two variants into a single pattern.
9327 (tstqi, tsthi): Likewise.
9328 (addhi3_incdec, addsi3_incdec): Likewise.
9329 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
9330 (mulhi3, mulsi3): Likewise.
9331 (udivhi3, udivsi3): Likewise.
9332 (divhi3, divsi3): Likewise.
9333 (andorqi3, andorhi3, andorsi3): Likewise.
9335 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
9338 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
9339 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
9341 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9343 PR tree-optimization/85694
9344 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
9345 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
9347 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
9348 (<u>avg<mode>3_ceil): New patterns.
9350 2018-07-03 David Malcolm <dmalcolm@redhat.com>
9352 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
9353 scan-tree-dump directive.
9354 * gcc.dg/vect/slp-perm-2.c: Likewise.
9355 * gcc.dg/vect/slp-perm-3.c: Likewise.
9356 * gcc.dg/vect/slp-perm-5.c: Likewise.
9357 * gcc.dg/vect/slp-perm-6.c: Likewise.
9358 * gcc.dg/vect/slp-perm-7.c: Likewise.
9359 * gcc.dg/vect/slp-perm-8.c: Likewise.
9361 2018-07-03 Marek Polacek <polacek@redhat.com>
9364 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
9366 2018-07-03 Richard Biener <rguenther@suse.de>
9369 * tree-ssa-structalias.c (find_func_clobbers): Properly
9370 handle indirect calls.
9372 2018-07-03 Jeff Law <law@redhat.com>
9374 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
9375 (shifts): New code iterator.
9376 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
9377 expander. Fix HImode handling on H8/SX.
9378 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
9379 (subqi3, subhi3, subsi3 expanders): Likewise.
9380 (andqi3, andhi3, andsi3 expanders): Likewise.
9381 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
9382 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
9383 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
9384 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
9385 (zero_extendqihi2, zero_extendqisi2): Likewise.
9386 (extendqihi2, extendqisi2): Likewise.
9387 (rotlqi3, rotlhi3, rotlsi3): Likewise.
9388 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
9389 (rotlqi3_1, rotlhi3_1): Likewise.
9390 (logicalhi3_sn, logicalsi3_sn): Likewise.
9391 (logicalhi3, logicalsi3): Likewise.
9393 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9395 * tree-vect-patterns.c (vect_recog_rotate_pattern)
9396 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9397 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
9398 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
9399 type to append_pattern_def_seq instead of creating a stmt_vec_info
9401 (build_mask_conversion): Likewise. Remove vinfo argument.
9402 (vect_add_conversion_to_patterm): Likewise, renaming to...
9403 (vect_add_conversion_to_pattern): ...this.
9404 (vect_recog_mask_conversion_pattern): Update call to
9405 build_mask_conversion. Pass the vector type to
9406 append_pattern_def_seq here too.
9407 (vect_recog_gather_scatter_pattern): Update call to
9408 vect_add_conversion_to_pattern.
9410 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9412 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
9413 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9414 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
9415 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
9416 STMT_VINFO_PATTERN_DEF_SEQ to null here.
9417 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
9418 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
9419 append_pattern_def_seq instead of new_pattern_def_seq.
9420 (vect_recog_divmod_pattern): Do both of the above.
9421 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
9424 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9426 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
9427 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
9428 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
9429 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
9430 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
9431 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9432 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
9433 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9434 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9435 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
9436 parameter with a single stmt_vec_info.
9437 (vect_recog_func_ptr): Likewise.
9438 (vect_recog_gather_scatter_pattern): Likewise, folding in...
9439 (vect_try_gather_scatter_pattern): ...this.
9440 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
9441 the stmt_vec_info of the statement to be matched. Don't clear
9442 STMT_VINFO_RELATED_STMT.
9443 (vect_pattern_recog): Update call accordingly.
9445 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9447 PR tree-optimization/85694
9448 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
9449 (uavgM3_ceil): Document new optabs.
9450 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
9451 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
9453 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
9454 (savg_ceil_optab): New optabs.
9455 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
9456 (vect_vect_recog_func_ptrs): Add it.
9457 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
9458 constant directly from the associated lhs.
9460 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9462 * tree-vect-patterns.c (vect_split_statement): New function.
9463 (vect_convert_input): Use it to try to split an existing cast.
9465 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9467 * poly-int.h (print_hex): New function.
9468 * dumpfile.h (dump_dec, dump_hex): Declare.
9469 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
9470 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
9471 min_input_precision, operation_precision and operation_sign.
9472 * tree-vect-patterns.c (vect_get_range_info): New function.
9473 (vect_same_loop_or_bb_p, vect_single_imm_use)
9474 (vect_operation_fits_smaller_type): Delete.
9475 (vect_look_through_possible_promotion): Add an optional
9476 single_use_p parameter.
9477 (vect_recog_over_widening_pattern): Rewrite to use new
9478 stmt_vec_info infomration. Handle one operation at a time.
9479 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
9480 (vect_truncatable_operation_p, vect_set_operation_type)
9481 (vect_set_min_input_precision): New functions.
9482 (vect_determine_min_output_precision_1): Likewise.
9483 (vect_determine_min_output_precision): Likewise.
9484 (vect_determine_precisions_from_range): Likewise.
9485 (vect_determine_precisions_from_users): Likewise.
9486 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
9487 (vect_vect_recog_func_ptrs): Put over_widening first.
9489 (vect_pattern_recog): Call vect_determine_precisions.
9491 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
9493 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
9494 statements that have been replaced by further pattern statements.
9495 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
9497 2018-07-03 Richard Biener <rguenther@suse.de>
9499 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
9500 always set *dt. Dump vectype in vectype overload.
9501 * dumpfile.h (dump_gimple_expr): New function.
9502 (dump_gimple_expr_loc): Likewise.
9503 * dumpfile.c (dump_gimple_expr): New function.
9504 (dump_gimple_expr_loc): Likewise.
9506 2018-07-02 Jeff Law <law@redhat.com>
9508 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
9509 the H8/300, H8/300H and H8/S variants into a single pattern.
9510 (movhi_h8300, movqi_h8300hs): Similarly.
9511 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
9512 (QHI mode iterator): New.
9514 * config/h8300/h8300.md: Remove trailing whitespace.
9516 2018-07-02 Jim Wilson <jimw@sifive.com>
9518 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
9519 instead of emit_insn for interrupt returns.
9520 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
9521 (riscv_sret, riscv_uret): Likewise.
9523 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9525 * pretty-print.c (selftest::test_pp_format): Move save and restore
9526 of quotes to class auto_fix_quotes, and add an instance.
9527 * selftest.c: Include "intl.h".
9528 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
9529 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
9530 * selftest.h (selftest::auto_fix_quotes): New class.
9532 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
9534 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
9535 (aarch64_sve_prepare_conditional_op): Remove.
9536 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
9537 Allow aarch64_simd_reg_or_zero as select operand; remove
9538 the aarch64_sve_prepare_conditional_op call.
9539 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
9540 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9541 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
9542 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
9543 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
9544 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
9545 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
9546 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
9547 and a splitters to match all of the *_any patterns.
9548 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
9550 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
9551 (SVE_COND_FP_BINARY_REV): Remove.
9552 (sve_int_op_rev, sve_fp_op_rev): New.
9553 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
9554 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
9555 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
9556 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
9557 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
9558 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
9559 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
9560 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
9561 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
9563 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
9564 Remove match_dup 1 from the inner unspec.
9565 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
9567 * config/aarch64/aarch64.md (movprfx): New attr.
9568 (length): Default movprfx to 8.
9569 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
9570 (*madd<SVE_I>, *msub<SVE_I): Likewise.
9571 (*<su>mul<SVE_I>3_highpart): Likewise.
9572 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
9573 (*v<ASHIFT><SVE_I>3): Likewise.
9574 (*<su><MAXMIN><SVE_I>3): Likewise.
9575 (*<su><MAXMIN><SVE_F>3): Likewise.
9576 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
9577 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
9578 (*div<SVE_F>4): Likewise.
9580 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
9582 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
9585 2018-07-02 Richard Biener <rguenther@suse.de>
9587 PR tree-optimization/86363
9588 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
9589 memset argument refers to a non-variable address.
9591 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9593 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
9594 {PLUS,MINUS}_EXPR code to...
9595 (adjust_symbolic_bound): ...here,
9596 (combine_bound): ...here,
9597 (set_value_range_with_overflow): ...and here.
9599 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
9601 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
9603 (extract_range_from_abs_expr): ...here.
9605 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
9607 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
9608 -fno-omit-frame-pointer when not optimizing.
9610 2018-07-02 Martin Liska <mliska@suse.cz>
9613 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
9614 (propagate_nothrow): Likewise.
9616 2018-07-02 Martin Liska <mliska@suse.cz>
9619 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
9621 2018-07-02 David Malcolm <dmalcolm@redhat.com>
9623 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
9624 function in r262149, changing "loc" param from source_location to
9625 const dump_location_t &.
9626 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
9627 declaration, as above.
9629 2018-07-01 Paul Koning <ni1d@arrl.net>
9631 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9632 -munit-asm, -mgnu-asm, -mdec-asm.
9633 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
9634 (pdp11_output_labelref): New.
9635 (pdp11_output_def): New.
9636 (pdp11_output_addr_vec_elt): New.
9637 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
9638 %# and %@ format codes.
9639 (pdp11_option_override): New.
9640 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
9641 (pdp11_output_ident): New.
9642 (pdp11_asm_named_section): New.
9643 (pdp11_asm_init_sections): New.
9644 (pdp11_file_start): New.
9645 (pdp11_file_end): New.
9646 (output_ascii): Use .ascii/.asciz for -mdec-asm.
9647 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
9648 %o, like %c but octal.
9649 (pdp11_option_override): New.
9650 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
9652 (DATA_SECTION_ASM_OP): Ditto.
9653 (READONLY_DATA_SECTION_ASM_OP): New.
9654 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
9655 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
9656 (ASM_OUTPUT_LABELREF): Ditto.
9657 (ASM_OUTPUT_DEF): Ditto.
9658 (ASM_OUTPUT_EXTERNAL): New.
9659 (ASM_OUTPUT_SOURCE_FILENAME): New.
9660 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
9661 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
9662 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
9663 %# and %@ format codes.
9664 * config/pdp11/pdp11.opt (mgnu-asm): New.
9665 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
9666 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
9667 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
9669 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
9671 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
9672 dereferencing path[] beyond its length.
9674 (debug_all_paths): New.
9675 (rewire_first_differing_edge): New.
9676 (adjust_paths_after_duplication): New.
9677 (duplicate_thread_path): Call adjust_paths_after_duplication.
9679 (thread_through_all_blocks): Add new argument to
9680 duplicate_thread_path.
9682 2018-06-30 Jim Wilson <jimw@sifive.com>
9684 * config/riscv/predicates.md (p2m1_shift_operand): New.
9685 (high_mask_shift_operand): New.
9686 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
9687 pattern using p2m1_shift_operand.
9688 (lshsi3_zero_extend_3+2): New combiner pattern using
9689 high_mask_shift_operand.
9691 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9693 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
9695 (vect_recog_rotate_pattern): ...here.
9696 (vect_convert_input): Try to insert casts of invariants in the
9698 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
9699 preheader to be empty.
9701 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9703 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
9704 vector type. If given, install it in the new statement's
9706 (vect_element_precision): New function.
9707 (vect_unpromoted_value): New struct.
9708 (vect_unpromoted_value::vect_unpromoted_value): New function.
9709 (vect_unpromoted_value::set_op): Likewise.
9710 (vect_look_through_possible_promotion): Likewise.
9711 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
9712 (vect_widened_op_tree, vect_convert_input): Likewise.
9713 (vect_convert_inputs, vect_convert_output): Likewise.
9714 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
9715 to handle the optional cast of the multiplication result and
9716 vect_widened_op_tree to detect the widened multiplication itself.
9717 Do not require the input and output of promotion casts to have
9718 the same sign, but base the signedness of the operation on the
9719 input rather than the result. If the pattern includes two
9720 promotions, check that those promotions have the same sign.
9721 Do not restrict the MULT_EXPR handling to a double-width result;
9722 handle quadruple-width results and wider. Use vect_convert_inputs
9723 to convert the inputs to the common type.
9724 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
9725 to handle the optional cast of the ABS result. Also allow a sign
9726 change or a sign extension between the ABS and MINUS.
9727 Use vect_widened_op_tree to detect the widened subtraction and use
9728 vect_convert_inputs to convert the inputs to the common type.
9729 (vect_handle_widen_op_by_const): Delete.
9730 (vect_recog_widen_op_pattern): New function.
9731 (vect_recog_widen_mult_pattern): Use it.
9732 (vect_recog_widen_shift_pattern): Likewise.
9733 (vect_recog_widen_sum_pattern): Use
9734 vect_look_through_possible_promotion to handle the promoted
9737 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9739 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
9740 the containing gimple_seq *.
9741 * gimple-iterator.h (gsi_for_stmt): Declare it.
9742 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
9743 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
9744 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
9745 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
9746 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
9748 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
9750 (vect_mark_pattern_stmts): ...here. Handle cases in which the
9751 statement being replaced is part of an existing pattern
9752 definition sequence, inserting the new pattern statements before
9754 (vect_pattern_recog_1): Don't return a bool. If the statement
9755 is already part of a pattern, instead apply pattern matching
9756 to the pattern definition statements. Don't clear the
9757 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
9758 (vect_pattern_recog): Don't break after the first match;
9759 continue processing the pattern definition statements instead.
9760 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
9762 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9764 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
9765 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9766 (vect_recog_widen_sum_pattern): Use it.
9768 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9770 * tree-vect-loop.c (vectorizable_reduction): Assert that the
9771 phi is not a pattern statement and has not been replaced by
9772 a pattern statement.
9773 * tree-vect-patterns.c (type_conversion_p): Don't check
9774 STMT_VINFO_IN_PATTERN_P.
9775 (vect_recog_vector_vector_shift_pattern): Likewise.
9776 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
9777 the pattern statement rather than the original statement; check
9778 directly for a WIDEN_MULT_EXPR here.
9779 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
9780 vect_is_simple_use to return the pattern statement rather
9781 than the original statement; use is_pattern_stmt_p to check
9782 for such a pattern statement.
9783 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
9784 to return the pattern statement rather than the original statement;
9785 don't do the same transformation here.
9786 (vect_is_simple_use): If the defining statement has been replaced
9787 by a pattern statement, return the pattern statement instead.
9788 Remove the corresponding (local) transformation from the vectype
9791 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
9793 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
9794 end and default to null.
9795 * tree-vect-loop.c (vect_create_epilog_for_reduction)
9796 (vectorizable_reduction): Update calls accordingly, dropping the
9797 gimple ** argument if the passed-back statement isn't needed.
9798 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
9799 (vect_recog_rotate_pattern): Likewise.
9800 (vect_recog_mask_conversion_pattern): Likewise.
9801 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
9802 (vect_mask_constant_operand_p): Likewise.
9803 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
9804 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
9805 (get_group_load_store_type, get_load_store_type): Likewise.
9806 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
9807 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
9808 (vectorizable_conversion, vectorizable_assignment): Likewise.
9809 (vectorizable_shift, vectorizable_operation): Likewise.
9810 (vectorizable_store, vect_is_simple_cond): Likewise.
9811 (vectorizable_condition, vectorizable_comparison): Likewise.
9812 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
9813 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
9814 and move it to the end. Cope with null def_stmt_outs.
9816 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
9818 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
9820 2018-06-29 Jeff Law <law@redhat.com>
9822 * config/v850/v850.c (v850_legitimate_address_p): Handle large
9823 displacements for TARGET_V850E2V3 and newer.
9824 (TARGET_LRA_P): Remove. Defaults to LRA now.
9825 * config/v850/v850.md (sign23byte_load): Remove.
9826 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
9827 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
9829 2018-06-29 Martin Liska <mliska@suse.cz>
9832 * coverage.c (coverage_init): Mangle full path name.
9833 * doc/invoke.texi: Document the change.
9834 * gcov-io.c (mangle_path): New.
9835 * gcov-io.h (mangle_path): Likewise.
9836 * gcov.c (mangle_name): Use mangle_path for path mangling.
9838 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9840 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
9841 if starting source register is not even.
9843 2018-06-29 Martin Liska <mliska@suse.cz>
9845 PR tree-optimization/86263
9846 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
9847 Make edge redirection.
9849 2018-06-29 David Malcolm <dmalcolm@redhat.com>
9851 * dumpfile.c (dump_loc): Add indentation based on scope depth.
9852 (dump_scope_depth): New variable.
9853 (get_dump_scope_depth): New function.
9854 (dump_begin_scope): New function.
9855 (dump_end_scope): New function.
9856 * dumpfile.h (get_dump_scope_depth): New declaration.
9857 (dump_begin_scope): New declaration.
9858 (dump_end_scope): New declaration.
9859 (class auto_dump_scope): New class.
9860 (AUTO_DUMP_SCOPE): New macro.
9861 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
9864 2018-06-29 Richard Biener <rguenther@suse.de>
9866 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
9867 compute_all_dependences succeeds.
9868 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
9869 exceed --param loop-max-datarefs-for-datadeps.
9871 2018-06-29 Jakub Jelinek <jakub@redhat.com>
9873 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
9875 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
9878 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
9879 alternative 0 in preferred_for_speed attribute.
9881 2018-06-28 Paul Koning <ni1d@arrl.net>
9883 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
9884 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
9885 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
9886 * config/pdp11/pdp11.md: Correct "length" attribute calculation
9887 for shift insn patterns.
9889 2018-06-28 David Malcolm <dmalcolm@redhat.com>
9891 * cgraph.c (cgraph_node::get_body): Replace assignments to
9892 "dump_file" with calls to set_dump_file.
9893 * dumpfile.c (alt_dump_file): Make static, and group with...
9894 (alt_flags): ...this definition.
9895 (dumps_are_enabled): New variable.
9896 (refresh_dumps_are_enabled): New function.
9897 (set_dump_file): New function.
9898 (set_alt_dump_file): New function.
9899 (gcc::dump_manager::dump_start): Replace assignments to
9900 "dump_file" and "alt_dump_file" with calls to set_dump_file and
9902 (gcc::dump_manager::dump_finish): Likewise.
9903 * dumpfile.h (alt_dump_file): Delete decl.
9904 (dumps_are_enabled): New variable decl.
9905 (set_dump_file): New function decl.
9906 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
9908 * tree-nested.c (lower_nested_functions): Replace assignments to
9909 "dump_file" with calls to set_dump_file.
9911 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9913 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
9914 goto_locus of each outgoing edge of each basic block.
9916 2018-06-28 Richard Biener <rguenther@suse.de>
9918 * dwarf2out.c (decl_scope_table): Remove.
9919 (push_decl_scope): Likewise.
9920 (pop_decl_scope): Likewise.
9921 (gen_type_die_for_member): Do not call push/pop_decl_scope.
9922 (gen_struct_or_union_type_die): Likewise.
9923 (gen_tagged_type_die): Likewise.
9924 (dwarf2out_init): Do not initialize decl_scope_table.
9925 (dwarf2out_c_finalize): Do not free it.
9927 2018-06-28 Richard Biener <rguenther@suse.de>
9929 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
9930 deciding whether to not re-use a DIE.
9932 2018-06-28 Richard Biener <rguenther@suse.de>
9934 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
9935 DW_AT_abstract_origin attribute.
9937 2018-06-28 Martin Liska <mliska@suse.cz>
9939 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9940 Use newly introduced constants.
9941 * tree-switch-conversion.h (struct jump_table_cluster):
9942 Define max_ratio_for_size and max_ratio_for_speed.
9944 2018-06-28 Martin Liska <mliska@suse.cz>
9946 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9947 Add new checking assert to catch invalid state.
9948 (jump_table_cluster::can_be_handled): Handle single case
9950 (jump_table_cluster::is_beneficial): Bail out for such case.
9951 (bit_test_cluster::find_bit_tests):
9952 Add new checking assert to catch invalid state.
9953 (bit_test_cluster::can_be_handled): Handle single case
9955 (bit_test_cluster::is_beneficial): Bail out for such case.
9956 (switch_decision_tree::analyze_switch_statement):
9959 2018-06-28 Martin Liska <mliska@suse.cz>
9961 * common.opt: Introduce -completion option.
9962 * gcc.c (driver_handle_option): Handle it.
9963 (driver::main): Print completions if completion
9965 * opt-suggestions.c (option_proposer::get_completions):
9967 (option_proposer::suggest_completion): Likewise.
9968 (option_proposer::find_param_completions): Likewise.
9969 (verify_autocompletions): Likewise.
9970 (test_completion_valid_options): Likewise.
9971 (test_completion_valid_params): Likewise.
9972 (in_completion_p): Likewise.
9973 (empty_completion_p): Likewise.
9974 (test_completion_partial_match): Likewise.
9975 (test_completion_garbage): Likewise.
9976 (opt_proposer_c_tests): Likewise.
9977 * opt-suggestions.h: Declare new functions.
9978 * opts.c (common_handle_option): Handle OPT__completion_.
9979 * selftest-run-tests.c (selftest::run_tests): Add
9980 opt_proposer_c_tests.
9981 * selftest.c (assert_str_startswith): New.
9982 * selftest.h (assert_str_startswith): Likewise.
9983 (opt_proposer_c_tests): New.
9984 (ASSERT_STR_STARTSWITH): Likewise.
9986 2018-06-28 Martin Liska <mliska@suse.cz>
9988 * Makefile.in: Add opt-suggestions.o.
9989 * gcc-main.c: Include opt-suggestions.h.
9990 * gcc.c (driver::driver): Likewise.
9991 (driver::~driver): Remove m_option_suggestions.
9992 (driver::build_option_suggestions): Moved to option_proposer.
9993 (driver::suggest_option): Likewise.
9994 (driver::handle_unrecognized_options): Use option_proposer.
9995 * gcc.h (class driver): Add new memver m_option_proposer.
9996 * opt-suggestions.c: New file.
9997 * opt-suggestions.h: New file.
9999 2018-06-28 Martin Liska <mliska@suse.cz>
10001 * vec.h (class auto_string_vec): New (moved from auto_argvec).
10002 (auto_string_vec::~auto_string_vec): Likewise.
10004 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
10006 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
10007 prevent_decl_creation_for_types fields up and add reset_location field.
10008 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
10009 statement if id->reset_location is true.
10010 (copy_edges_for_bb): Do not set goto_locus on the new edges if
10011 id->reset_location is true.
10012 (copy_phis_for_bb): Force input_location on the arguments if
10013 id->reset_location is true.
10014 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
10015 is set on the function to be inlined.
10017 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
10019 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
10021 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
10023 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
10024 registers for Pmode.
10025 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
10026 hard registers for the clobbered pseudo.
10028 2018-06-27 Paul Koning <ni1d@arrl.net>
10030 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
10031 mutually exclusive options.
10032 * config/pdp11/constraints.md (h): New constraint.
10033 (O): Update definition to match shift code generation.
10034 (D): New constraint.
10035 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
10037 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
10039 (output_jump): Change arguments.
10040 (pdp11_fixed_cc_regs): New function.
10041 (pdp11_cc_mode): Ditto.
10042 (pdp11_expand_shift): Ditto.
10043 (pdp11_assemble_shift): Ditto.
10044 (pdp11_small_shift): Ditto.
10045 (pdp11_branch_cost): Remove.
10046 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
10048 (pdp11_register_move_cost): Update for CC registers.
10049 (pdp11_rtx_costs): Add case for LSHIFTRT.
10050 (pdp11_output_jump): Add CCNZ mode conditional branches.
10051 (notice_update_cc_on_set): Remove.
10052 (pdp11_cc_mode): New function.
10053 (simple_memory_operand): Correct pre/post decrement case.
10054 (no_side_effect_operand): New function.
10055 (pdp11_regno_reg_class): Add CC_REGS class.
10056 (pdp11_fixed_cc_regs): New function.
10057 (pdp11_small_shift): New function.
10058 (pdp11_expand_shift): New function to expand shift insns.
10059 (pdp11_assemble_shift): New function to output shifts.
10060 (pdp11_branch_cost): Remove.
10061 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
10062 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
10063 (WCHAR_TYPE): Ditto.
10064 (PTRDIFF_TYPE): Ditto.
10065 (ADJUST_INSN_LENGTH): New macro.
10066 (FIXED_REGISTERS): Add CC registers.
10067 (CALL_USED_REGISTERS): Ditto.
10068 (reg_class): Ditto.
10069 (REG_CLASS_NAMES): Ditto.
10070 (REG_CLASS_CONTENTS): Ditto.
10071 (SELECT_CC_MODE): Use new function.
10072 (TARGET_FLAGS_REGNUM): New macro.
10073 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
10074 (cc0_reg_rtx): Remove.
10075 (CC_STATUS_MDEP): Remove.
10076 (CC_STATUS_MDEFP_INIT): Remove.
10077 (CC_IN_FPU): Remove.
10078 (NOTICE_UPDATE_CC): Remove.
10079 (REGISTER_NAMES): Add CC registers.
10080 (BRANCH_COST): Change to constant 1.
10081 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
10083 * config/pdp11/pdp11.opt (mbcopy): Remove.
10084 (mbcopy-builtin): Remove.
10085 (mbranch-cheap): Remove.
10086 (mbranch-expensive): Remove.
10087 * config/pdp11/predicates.md (expand_shift_operand): Update to
10088 match shift code generation.
10089 (ccnz_operator): New predicate.
10090 * doc/invoke.texi (PDP-11 Options): Remove deleted options
10091 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
10092 Remove non-existent option -mabshi, -mno-abshi. Document mutually
10094 * doc/md.texi (PDP-11): Document new D and h constraints. Update
10095 description of O constraint.
10097 2018-06-27 Jeff Law <law@redhat.com>
10098 Austin Law <austinklaw@gmail.com>
10100 * config/v850/v850.md (addsi3_set_flags): New pattern.
10101 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
10102 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
10103 (zero_extendhisi2_v850_set_flags): Likewise.
10104 (zero_extendqisi2_v850_set_flags): Likewise.
10105 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
10106 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
10107 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
10109 * config/v850/v850-protos.h (notice_update_cc): Remove.
10110 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
10111 (v850_print_operand): Handle 'D' and "d".
10112 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
10113 Add handling of arithmetic/logical operations compared against zero.
10114 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
10115 Do not look at v850_compare_op, instead get mode from last argument.
10116 (v850_gen_compare): Remove
10117 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
10118 after reload for prologue insns.
10119 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
10121 (construct_save_jarl): Likewise.
10122 (TARGET_FLAGS_REGNUM): Define.
10123 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
10124 (NOTICE_UPDATE_CC): Remove.
10125 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
10126 than cc0. Conditionalize on reload_completed.
10127 (cmpsi_insn, setfcc_insn): Likewise.
10128 (tst1 splitter): Turn into define_and_split which sets the flags
10130 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
10131 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
10132 (cstoresf4, cstoredf4): Clobber the flags.
10133 (cmpsi, cmpsf, cmpdf): Remove expanders.
10134 (setf_insn): Remove pattern.
10135 (addsi3): Turn into define_and_split which clobbers the flags after
10136 reload and a suitable pattern (addsi3_clobber_flags) for use after
10138 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
10139 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
10140 (ashrsi3, ashrsi3_v850e2): Likewise.
10141 (bins): Clobber the flags.
10142 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
10143 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
10144 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
10145 (call_value_internal_short, call_value_internal_long): Likewise.
10146 (callt_save_interrupt, callt_return_interrupt): Likewise.
10147 (save_interrupt, return_interrupt): Likewise.
10148 (callt_save_all_interrupt, save_all_interrupt): Likewise.
10149 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
10150 (restore_all_interrupt, _restore_all_interrupt): Likewise.
10151 (All FP comparisons): Only allow after reload has completed.
10153 (divh, divhu): Tweak output template.
10154 (branch_z_normal, branch_z_invert): Remove
10155 (branch_nz_normal, branch_nz_invert): Likewise.
10156 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
10158 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
10159 * config/v850/v850.c (notice_update_cc): Remove.
10160 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
10161 (CC_NO_CARRY): Likewise.
10162 (NOTICE_UPDATE_CC): Define to nothing.
10163 * config/v850/v850.md: Remove block comment on cc0 handling
10164 Remove "cc" attribute from all patterns. Remove cc_status handling
10165 from all patterns. Minor formatting fixes.
10167 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10169 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
10170 (cortex-a76.cortex-a55): Likewise.
10171 * config/aarch64/aarch64-tune.md: Regenerate.
10172 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
10173 cortex-a76.cortex-a55.
10175 2018-06-27 Jeff Law <law@redhat.com>
10177 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
10178 (MULTILIB_DIRNAMES): Similarly.
10180 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
10182 * gimple.h (gimple_return_retbnd): Delete.
10183 (gimple_return_set_retbnd): Likewise.
10184 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
10185 gimple_return_set_retbnd.
10186 * gimple-pretty-print.c (dump_gimple_return): Remove call to
10187 gimple_return_retbnd and adjust.
10188 * tree-inline.h (struct copy_body_data): Remove retbnd field.
10189 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
10190 Explicitly return NULL in a couple more cases. Move assertion
10191 on debug statements and remove unreachable code.
10192 (reset_debug_binding): Do not test id->retbnd.
10193 (expand_call_inline): Do not set it.
10195 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
10197 * configure.ac: Add --disable-gcov option.
10198 * configure: Regenerate.
10199 * Makefile.in: Honour @enable_gcov@.
10200 * doc/install.texi: Document --disable-gcov.
10202 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10204 * config/arm/arm-cpus.in (cortex-a76): New entry.
10205 (cortex-a76.cortex-a55): Likewise.
10206 * config/arm/arm-tables.opt: Regenerate.
10207 * config/arm/arm-tune.md: Likewise.
10208 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
10209 * doc/invoke.texi (ARM Options): Document cortex-a76 and
10210 cortex-a76.cortex-a55.
10212 2018-06-27 Tamar Christina <tamar.christina@arm.com>
10215 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
10217 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
10219 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
10221 (EPILOGUE_USES): Likewise.
10223 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
10225 * tree-inline.c (remap_location): New function extracted from...
10226 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
10227 (copy_phis_for_bb): ...here. Call remap_location.
10228 (copy_cfg_body): Adjust call to copy_edges_for_bb.
10230 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
10232 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
10233 unaligned vsx for 16B memset.
10235 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
10238 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
10239 ieee128_float_type_node to long_double_type_node unless
10240 TARGET_LONG_DOUBLE_128 is set.
10242 2018-06-26 David Malcolm <dmalcolm@redhat.com>
10244 * cfgloop.c (get_loop_location): Convert return type from
10245 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
10246 by implicit construction from rtx_insn *, and using
10247 dump_user_location_t::from_function_decl for the fallback case.
10248 * cfgloop.h (get_loop_location): Convert return type from
10249 location_t to dump_user_location_t.
10250 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
10251 dump_printf_loc to pass in a dump_location_t rather than a
10252 location_t, via the gimple stmt.
10253 * coverage.c (get_coverage_counts): Update calls to
10254 dump_printf_loc to pass in dump_location_t rather than a
10256 * doc/optinfo.texi (Dump types): Convert example of
10257 dump_printf_loc from taking "locus" to taking "insn". Update
10258 description of the "_loc" calls to cover dump_location_t.
10259 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
10261 (dump_user_location_t::dump_user_location_t): New constructors,
10262 from gimple *stmt and rtx_insn *.
10263 (dump_user_location_t::from_function_decl): New function.
10264 (dump_loc): Make static.
10265 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
10266 const dump_location_t &.
10267 (dump_generic_expr_loc): Delete.
10268 (dump_printf_loc): Convert param "loc" from location_t to
10269 const dump_location_t &.
10270 (selftest::test_impl_location): New function.
10271 (selftest::dumpfile_c_tests): New function.
10272 * dumpfile.h: Include "profile-count.h".
10273 (class dump_user_location_t): New class.
10274 (struct dump_impl_location_t): New struct.
10275 (class dump_location_t): New class.
10276 (dump_printf_loc): Convert 2nd param from source_location to
10277 const dump_location_t &.
10278 (dump_generic_expr_loc): Delete.
10279 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
10280 const dump_location_t &.
10281 * gimple-fold.c (fold_gimple_assign): Update call to
10282 dump_printf_loc to pass in a dump_location_t rather than a
10283 location_t, via the gimple stmt.
10284 (gimple_fold_call): Likewise.
10285 * gimple-loop-interchange.cc
10286 (loop_cand::analyze_iloop_reduction_var): Update for change to
10287 check_reduction_path.
10288 (tree_loop_interchange::interchange): Update for change to
10289 find_loop_location.
10290 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
10291 change in return-type of find_loop_location.
10292 (graphite_regenerate_ast_isl): Likewise.
10293 * graphite-optimize-isl.c (optimize_isl): Likewise.
10294 * graphite.c (graphite_transform_loops): Likewise.
10295 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
10296 pass in a dump_location_t rather than a location_t, via the
10298 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10299 * ipa.c (walk_polymorphic_call_targets): Likewise.
10300 * loop-unroll.c (report_unroll): Convert "locus" param from
10301 location_t to dump_location_t.
10302 (decide_unrolling): Update for change to get_loop_location's
10304 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
10305 location_t to dump_user_location_t.
10306 (grid_find_single_omp_among_assignments_1): Updates calls to
10307 dump_printf_loc to pass in a dump_location_t rather than a
10308 location_t, via the gimple stmt.
10309 (grid_parallel_clauses_gridifiable): Convert "tloc" from
10310 location_t to dump_location_t. Updates calls to dump_printf_loc
10311 to pass in a dump_location_t rather than a location_t, via the
10313 (grid_inner_loop_gridifiable_p): Likewise.
10314 (grid_dist_follows_simple_pattern): Likewise.
10315 (grid_gfor_follows_tiling_pattern): Likewise.
10316 (grid_target_follows_gridifiable_pattern): Likewise.
10317 (grid_attempt_target_gridification): Convert initialization
10318 of local "grid" from memset to zero-initialization; FIXME: does
10319 this require C++11? Update call to dump_printf_loc to pass in a
10320 optinfo_location rather than a location_t, via the gimple stmt.
10321 * profile.c (read_profile_edge_counts): Updates call to
10322 dump_printf_loc to pass in a dump_location_t rather than a
10324 (compute_branch_probabilities): Likewise.
10325 * selftest-run-tests.c (selftest::run_tests): Call
10327 * selftest.h (dumpfile_c_tests): New decl.
10328 * tree-loop-distribution.c (pass_loop_distribution::execute):
10329 Update for change in return type of find_loop_location.
10330 * tree-parloops.c (parallelize_loops): Likewise.
10331 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
10332 "locus" from location_t to dump_user_location_t.
10333 (canonicalize_loop_induction_variables): Likewise.
10334 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
10335 for change in return type of find_loop_location.
10336 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
10337 to dump_printf_loc to pass in a dump_location_t rather than a
10338 location_t, via the stmt.
10339 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
10341 * tree-vect-loop-manip.c (find_loop_location): Convert return
10342 type from source_location to dump_user_location_t.
10343 (vect_do_peeling): Update for above change.
10344 (vect_loop_versioning): Update for change in type of
10346 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
10347 from location_t to dump_user_location_t.
10348 (vect_estimate_min_profitable_iters): Update for change in type
10350 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
10351 location_t to dump_location_t.
10352 (vect_slp_bb): Update for change in type of vect_location.
10353 * tree-vectorizer.c (vect_location): Convert from source_location
10354 to dump_user_location_t.
10355 (try_vectorize_loop_1): Update for change in vect_location's type.
10356 (vectorize_loops): Likewise.
10357 (increase_alignment): Likewise.
10358 * tree-vectorizer.h (vect_location): Convert from source_location
10359 to dump_user_location_t.
10360 (find_loop_location): Convert return type from source_location to
10361 dump_user_location_t.
10362 (check_reduction_path): Convert 1st param from location_t to
10363 dump_user_location_t.
10364 * value-prof.c (check_counter): Update call to dump_printf_loc to
10365 pass in a dump_user_location_t rather than a location_t; update
10366 call to error_at for change in type of "locus".
10367 (check_ic_target): Update call to dump_printf_loc to
10368 pass in a dump_user_location_t rather than a location_t, via the
10371 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
10373 * config/s390/s390.h (enum processor_flags): Do not use
10374 default tune parameter when -march was specified.
10376 2018-06-26 Jakub Jelinek <jakub@redhat.com>
10379 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
10380 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
10383 2018-06-26 Richard Biener <rguenther@suse.de>
10385 PR tree-optimization/86287
10387 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
10388 (vect_analyze_loop): Initialize n_stmts.
10390 2018-06-26 Richard Biener <rguenther@suse.de>
10392 PR middle-end/86271
10393 * fold-const.c (fold_convertible_p): Pointer extension
10396 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
10399 * dwarf2out.c (loc_list_has_views): Adjust comments.
10400 (dw_loc_list): Split single cross-partition range with
10403 2018-06-25 Jeff Law <law@redhat.com>
10405 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
10406 on -mbig-switch by default.
10408 * config/v850/predicates.md (const_float_1_operand): Fix match_code
10410 (const_float_0_operand): Remove unused predicate.
10411 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
10412 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
10413 (recipsf2): New expander. Original pattern now called
10415 (recipdf2, recipdf2_insn): Similarly.
10416 (rsqrtsf2, rsqrtsf2_insn): Similarly
10417 (rsqrtdf2, rsqrtdf2_insn): Similarly
10419 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
10421 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
10422 Simplify logic for FreeBSD (twice).
10424 2018-06-25 Martin Sebor <msebor@redhat.com>
10426 PR tree-optimization/86204
10427 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
10428 a strnlen result if it's less than the length of the string.
10430 2018-06-25 Martin Sebor <msebor@redhat.com>
10432 PR tree-optimization/85700
10433 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
10434 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
10435 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
10437 2018-06-25 Martin Sebor <msebor@redhat.com>
10439 * doc/extend.texi (Zero-length arrays): Update and clarify.
10441 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
10443 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
10444 added IEEE/IBM long double multilib support on PowerPC little
10445 endian Linux systems.
10446 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
10447 (MULTILIB_DEFAULTS): Likewise.
10448 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10450 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
10451 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
10452 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
10454 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
10456 PR middle-end/86311
10457 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
10458 (REORDER_45): Likewise.
10460 2018-06-25 Jeff Law <law@redhat.com>
10462 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
10463 dividend to 32 bits. Adjust length.
10464 (udivmodhi4): Cleanup output template. Fix length.
10466 2018-06-25 Carl Love <cel@us.ibm.com>
10468 * config/rs6000/vsx.md: Change word selector to prefered location.
10470 2018-06-25 Richard Biener <rguenther@suse.de>
10472 PR tree-optimization/86304
10473 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
10474 epilogue-if-converted loops as well.
10476 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
10478 * lto-section-out.c (lto_begin_section): Do not print section
10479 name for noaddr and unnumbered dumps.
10481 2018-06-25 Richard Biener <rguenther@suse.de>
10483 * tree-vectorizer.h (struct vec_info_shared): New structure
10484 with parts split out from struct vec_info and loop_nest from
10485 struct _loop_vec_info.
10486 (struct vec_info): Adjust accordingly.
10487 (struct _loop_vec_info): Likewise.
10488 (LOOP_VINFO_LOOP_NEST): Adjust.
10489 (LOOP_VINFO_DATAREFS): Likewise.
10490 (LOOP_VINFO_DDRS): Likewise.
10491 (struct _bb_vec_info): Likewise.
10492 (BB_VINFO_DATAREFS): Likewise.
10493 (BB_VINFO_DDRS): Likewise.
10494 (struct _stmt_vec_info): Add dr_aux member.
10495 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
10496 (DR_MISALIGNMENT_UNINITIALIZED): New.
10497 (set_dr_misalignment): Adjust.
10498 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
10499 (vect_analyze_loop): Adjust prototype.
10500 (vect_analyze_loop_form): Likewise.
10501 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10502 Compute dependences lazily.
10503 (vect_record_base_alignments): Use shared datarefs/ddrs.
10504 (vect_verify_datarefs_alignment): Likewise.
10505 (vect_analyze_data_refs_alignment): Likewise.
10506 (vect_analyze_data_ref_accesses): Likewise.
10507 (vect_analyze_data_refs): Likewise.
10508 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
10509 constructor parameter for shared part.
10510 (vect_analyze_loop_form): Pass in shared part and adjust.
10511 (vect_analyze_loop_2): Pass in storage for the number of
10512 stmts. Move loop nest finding to the caller. Compute
10514 (vect_analyze_loop): Pass in shared part.
10515 (vect_transform_loop): Verify shared datarefs are unchanged.
10516 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
10517 constructor parameter for shared part.
10518 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
10519 (vect_slp_bb): Verify shared datarefs are unchanged before
10521 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
10523 (new_stmt_vec_info): Initialize DR_AUX misalignment to
10524 DR_MISALIGNMENT_UNINITIALIZED.
10525 * tree-vectorizer.c (vec_info::vec_info): Add constructor
10526 parameter for shared part.
10527 (vec_info::~vec_info): Adjust.
10528 (vec_info_shared::vec_info_shared): New.
10529 (vec_info_shared::~vec_info_shared): Likewise.
10530 (vec_info_shared::save_datarefs): Likewise.
10531 (vec_info_shared::check_datarefs): Likewise.
10532 (try_vectorize_loop_1): Construct shared part live for analyses
10533 of a single loop for multiple vector sizes.
10534 * tree-parloops.c (gather_scalar_reductions): Adjust.
10536 2018-06-25 Richard Biener <rguenther@suse.de>
10538 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
10539 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
10540 (vect_analyze_data_refs): Remove similar code from here and
10541 simplify accordingly.
10543 2018-06-25 Richard Biener <rguenther@suse.de>
10545 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
10546 for reverse storage order accesses rather than asserting
10547 they cannot happen here.
10549 2018-06-25 Tom de Vries <tdevries@suse.de>
10552 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
10553 Use data16 instead of .byte for insn prefix.
10555 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
10558 * parser.c (make_char_string_pack): Pass this literal chars
10559 through cpp_interpret_string.
10560 (cp_parser_userdef_numeric_literal): Check the result of
10561 make_char_string_pack.
10563 2018-06-24 Maya Rashish <coypu@sdf.org>
10565 * ginclude/stddef.h: Simplify conditions around avoiding
10566 re-definition of __size_t.
10568 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10570 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
10571 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
10573 2018-06-22 Maya Rashish <coypu@sdf.org>
10575 * doc/invoke.texi (mno-fancy-math-387): Update for changes
10576 made to OpenBSD and NetBSD through the years.
10578 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10580 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10581 behavior of vec_pack (vector double, vector double) to match
10582 behavior of vec_float2 (vector double, vector double).
10584 2018-06-22 Olivier Hainque <hainque@adacore.com>
10586 * gimplify.c (gimplify_function_tree): Prevent creation
10587 of a trampoline for the address of the current function
10588 passed to entry/exit instrumentation hooks.
10590 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
10593 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
10596 2018-06-22 Martin Liska <mliska@suse.cz>
10598 PR tree-optimization/86263
10599 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10600 Bail out if is_enabled is false.
10601 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
10603 (jump_table_cluster::is_enabled): New function.
10605 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10607 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
10608 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
10609 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
10610 (lto_input_ts_binfo_tree_pointers): Likewise.
10611 * tree-streamer-out.c (streamer_write_tree_bitfields,
10612 write_ts_binfo_tree_pointers): Likewise.
10613 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
10615 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
10617 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
10619 2018-06-22 Martin Liska <mliska@suse.cz>
10621 * symbol-summary.h (get): Make it pure and inline move
10622 functionality from ::get function.
10623 (get): Remove and inline into ::get and ::get_create.
10624 (get_create): Move code from ::get function.
10626 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10629 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
10630 -x assembler-with-cpp.
10632 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10634 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
10635 _FILE_OFFSET_BITS=64 for C++.
10637 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
10639 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
10640 conversion insn that shows up when pr85657-3.c is compiled using
10641 IEEE 128-bit long double.
10642 (neg<mode>2_internal): Use the correct mode to check whether the
10643 mode is IBM extended.
10644 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
10645 multiply and divide external functions from being created more
10648 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10650 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
10652 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
10653 the edge can be forwarded.
10654 (cfg_layout_merge_blocks): Likewise.
10656 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10658 * except.c (finish_eh_generation): Commit edge insertions only after
10659 the EH edges have been redirected from post-landing to landing pads.
10661 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10663 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
10664 create_tmp_var_for to create the FRAME decl.
10665 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
10667 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
10669 * tree-inline.c (copy_edges_for_bb): Minor tweak.
10670 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
10671 debug statement when resetting its value.
10672 (expand_call_inline): Copy the locus of the call onto the assignment
10673 of the return value, if any. Use local variable in more cases.
10675 2018-06-21 Martin Liska <mliska@suse.cz>
10677 * ipa-pure-const.c (propagate_nothrow): Use
10678 funct_state_summaries->get.
10679 (dump_malloc_lattice): Likewise.
10680 (propagate_malloc): Likewise.
10682 2018-06-21 Richard Biener <rguenther@suse.de>
10684 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
10685 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
10686 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
10688 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
10689 BLOCK_ABSTRACT_ORIGIN unconditionally.
10691 2018-06-21 David Malcolm <dmalcolm@redhat.com>
10693 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
10695 * ipa-reference.c (ipa_reference_c_finalize): Delete
10696 ipa_ref_opt_sum_summaries and set it to NULL.
10698 2018-06-21 Tom de Vries <tdevries@suse.de>
10700 PR tree-optimization/85859
10701 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
10702 test with comment from bb_no_side_effects_p.
10704 2018-06-21 Richard Biener <rguenther@suse.de>
10706 PR tree-optimization/86232
10707 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
10708 max for constant niter.
10710 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10712 * config/aarch64/aarch64-simd.md
10713 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
10715 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
10717 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
10718 Make opernads of the unspec commutative.
10720 2018-06-21 Richard Biener <rguenther@suse.de>
10722 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
10723 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
10724 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
10725 (vect_analyze_data_ref_dependence): Re-order checks to deal with
10727 (vect_record_base_alignments): Do not record base alignment
10728 for gathers or scatters.
10729 (vect_compute_data_ref_alignment): Drop return value that is always
10730 true. Bail out early for gathers or scatters.
10731 (vect_enhance_data_refs_alignment): Bail out early for gathers
10733 (vect_find_same_alignment_drs): Likewise.
10734 (vect_analyze_data_refs_alignment): Remove dead code.
10735 (vect_slp_analyze_and_verify_node_alignment): Likewise.
10736 (vect_analyze_data_refs): For possible gathers or scatters do
10737 not create an alternate DR, just check their possible validity
10738 and mark them. Adjust DECL_NONALIASED handling to not rely
10739 on DR_BASE_ADDRESS.
10740 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
10741 update inits of gathers or scatters.
10742 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
10743 Also copy gather/scatter flag to pattern vinfo.
10745 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
10747 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
10748 behavior of vec_packsu (vector unsigned long long, vector unsigned
10749 long long) to match behavior of vec_packs with same signature.
10751 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
10752 Thomas Schwinge <thomas@codesourcery.com>
10753 Cesar Philippidis <cesar@codesourcery.com>
10755 * gimplify.c (gimplify_scan_omp_clauses): Add support for
10756 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10757 (gimplify_adjust_omp_clauses): Likewise.
10758 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
10759 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
10760 (gimplify_omp_target_update): Update handling of acc update and
10762 * omp-low.c (install_var_field): Remove unused parameter
10763 base_pointers_restrict.
10764 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
10765 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
10767 (omp_target_base_pointers_restrict_p): Delete.
10768 (scan_omp_target): Update call to scan_sharing_clauses.
10769 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
10771 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
10772 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
10773 (convert_local_omp_clauses): Likewise.
10774 * tree-pretty-print.c (dump_omp_clause): Likewise.
10775 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
10777 (omp_clause_code_name): Likewise.
10779 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10782 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
10785 PR tree-optimization/86231
10786 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
10787 anti-range don't overwrite *vr0min before using it to compute *vr0max.
10789 2018-06-20 Tom de Vries <tdevries@suse.de>
10791 PR tree-optimization/86097
10792 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
10793 iv type if signedness of iv type is not the same as that of *nit.
10795 2018-06-20 Jakub Jelinek <jakub@redhat.com>
10797 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
10798 EDGE_EH edges, verify they are all EDGE_EH.
10800 2018-06-20 Maya Rashish <coypu@sdf.org>
10802 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
10804 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10806 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
10807 * config/aarch64/aarch64.c (xgene1_tunings): Add
10808 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
10809 (aarch64_mode_valid_for_sched_fusion_p):
10810 Allow 16-byte modes.
10811 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
10812 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
10814 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
10816 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
10817 * config/aarch64/iterators.md (VQ2): New mode iterator.
10819 2018-06-20 Martin Liska <mliska@suse.cz>
10821 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
10822 Change default ratio from 10 to 8.
10824 2018-06-20 Martin Liska <mliska@suse.cz>
10826 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
10828 (bit_test_cluster::find_bit_tests): Likewise.
10829 (switch_decision_tree::analyze_switch_statement): Find clusters.
10830 * tree-switch-conversion.h (struct jump_table_cluster): Document
10833 2018-06-20 Martin Liska <mliska@suse.cz>
10835 * tree-switch-conversion.c (switch_conversion::collect):
10836 Record m_uniq property.
10837 (switch_conversion::expand): Bail out for special conditions.
10838 (group_cluster::~group_cluster): New.
10839 (group_cluster::group_cluster): Likewise.
10840 (group_cluster::dump): Likewise.
10841 (jump_table_cluster::emit): New.
10842 (switch_decision_tree::fix_phi_operands_for_edges): New.
10843 (struct case_node): Remove struct.
10844 (jump_table_cluster::can_be_handled): New.
10845 (case_values_threshold): Moved to header.
10846 (reset_out_edges_aux): Likewise.
10847 (jump_table_cluster::is_beneficial): New.
10848 (bit_test_cluster::can_be_handled): Likewise.
10849 (add_case_node): Remove.
10850 (bit_test_cluster::is_beneficial): New.
10851 (case_bit_test::cmp): New.
10852 (bit_test_cluster::emit): New.
10853 (expand_switch_as_decision_tree_p): Remove.
10854 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
10855 (fix_phi_operands_for_edge): Likewise.
10856 (switch_decision_tree::analyze_switch_statement): New.
10857 (compute_cases_per_edge): Move ...
10858 (switch_decision_tree::compute_cases_per_edge): ... here.
10859 (try_switch_expansion): Likewise.
10860 (switch_decision_tree::try_switch_expansion): Likewise.
10861 (record_phi_operand_mapping): Likewise.
10862 (switch_decision_tree::record_phi_operand_mapping): Likewise.
10863 (emit_case_decision_tree): Likewise.
10864 (switch_decision_tree::emit): Likewise.
10865 (balance_case_nodes): Likewise.
10866 (switch_decision_tree::balance_case_nodes): Likewise.
10867 (dump_case_nodes): Likewise.
10868 (switch_decision_tree::dump_case_nodes): Likewise.
10869 (emit_jump): Likewise.
10870 (switch_decision_tree::emit_jump): Likewise.
10871 (emit_cmp_and_jump_insns): Likewise.
10872 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
10873 (emit_case_nodes): Likewise.
10874 (switch_decision_tree::emit_case_nodes): Likewise.
10875 (conditional_probability): Remove.
10876 * tree-switch-conversion.h (enum cluster_type): New.
10878 (struct cluster): Likewise.
10879 (cluster::cluster): Likewise.
10880 (struct simple_cluster): Likewise.
10881 (simple_cluster::simple_cluster): Likewise.
10882 (struct group_cluster): Likewise.
10883 (struct jump_table_cluster): Likewise.
10884 (struct bit_test_cluster): Likewise.
10885 (struct min_cluster_item): Likewise.
10886 (struct case_tree_node): Likewise.
10887 (case_tree_node::case_tree_node): Likewise.
10888 (jump_table_cluster::case_values_threshold): Likewise.
10889 (struct case_bit_test): Likewise.
10890 (struct switch_decision_tree): Likewise.
10891 (struct switch_conversion): Likewise.
10892 (switch_decision_tree::reset_out_edges_aux): Likewise.
10894 2018-06-20 Martin Liska <mliska@suse.cz>
10896 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
10897 (hoist_edge_and_branch_if_true): Likewise.
10898 (expand_switch_using_bit_tests_p): Likewise.
10899 (struct case_bit_test): Likewise.
10900 (case_bit_test_cmp): Likewise.
10901 (emit_case_bit_tests): Likewise.
10902 (switch_conversion::switch_conversion): New class.
10903 (struct switch_conv_info): Remove old struct.
10904 (collect_switch_conv_info): More to ...
10905 (switch_conversion::collect): ... this.
10906 (check_range): Likewise.
10907 (switch_conversion::check_range): Likewise.
10908 (check_all_empty_except_final): Likewise.
10909 (switch_conversion::check_all_empty_except_final): Likewise.
10910 (check_final_bb): Likewise.
10911 (switch_conversion::check_final_bb): Likewise.
10912 (create_temp_arrays): Likewise.
10913 (switch_conversion::create_temp_arrays): Likewise.
10914 (free_temp_arrays): Likewise.
10915 (gather_default_values): Likewise.
10916 (switch_conversion::gather_default_values): Likewise.
10917 (build_constructors): Likewise.
10918 (switch_conversion::build_constructors): Likewise.
10919 (constructor_contains_same_values_p): Likewise.
10920 (switch_conversion::contains_same_values_p): Likewise.
10921 (array_value_type): Likewise.
10922 (switch_conversion::array_value_type): Likewise.
10923 (build_one_array): Likewise.
10924 (switch_conversion::build_one_array): Likewise.
10925 (build_arrays): Likewise.
10926 (switch_conversion::build_arrays): Likewise.
10927 (gen_def_assigns): Likewise.
10928 (switch_conversion::gen_def_assigns): Likewise.
10929 (prune_bbs): Likewise.
10930 (switch_conversion::prune_bbs): Likewise.
10931 (fix_phi_nodes): Likewise.
10932 (switch_conversion::fix_phi_nodes): Likewise.
10933 (gen_inbound_check): Likewise.
10934 (switch_conversion::gen_inbound_check): Likewise.
10935 (process_switch): Use the newly created class.
10936 (switch_conversion::expand): New.
10937 (switch_conversion::~switch_conversion): New.
10938 * tree-switch-conversion.h: New file.
10940 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10942 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
10943 tree-vect-patterns.c.
10944 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
10945 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
10946 (vect_recog_sad_pattern): Likewise.
10947 (vect_recog_widen_sum_pattern): Likewise.
10948 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
10949 (vect_recog_widen_shift_pattern): Remove the type_in argument.
10950 (vect_recog_rotate_pattern): Likewise.
10951 (vect_recog_mult_pattern): Likewise.
10952 (vect_recog_vector_vector_shift_pattern): Likewise.
10953 (vect_recog_divmod_pattern): Likewise.
10954 (vect_recog_mixed_size_cond_pattern): Likewise.
10955 (vect_recog_bool_pattern): Likewise.
10956 (vect_recog_mask_conversion_pattern): Likewise.
10957 (vect_try_gather_scatter_pattern): Likewise.
10958 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
10959 (vect_recog_over_widening_pattern): Likewise.
10960 (vect_recog_gather_scatter_pattern): Likewise.
10961 (vect_recog_func_ptr): Move from tree-vectorizer.h
10962 (vect_vect_recog_func_ptrs): Move further down the file.
10963 (vect_recog_func): Likewise. Remove the third argument.
10964 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10965 (vect_pattern_recog_1): Expect the pattern function to do any
10966 necessary target tests. Also expect it to provide a vector type.
10967 Remove the type_in handling.
10969 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10971 * tree-vect-patterns.c (vect_pattern_detected): New function.
10972 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10973 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10974 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10975 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10976 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10977 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10978 (vect_recog_mask_conversion_pattern)
10979 (vect_try_gather_scatter_pattern): Likewise.
10981 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10983 * tree-vect-patterns.c (vect_get_internal_def): New function.
10984 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10985 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10986 (search_type_for_mask_1): Use it.
10988 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10990 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10991 redundant WIDEN_SUM_EXPR handling.
10992 (vect_recog_sad_pattern): Likewise.
10994 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10996 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10997 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10998 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10999 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
11000 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
11002 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11004 * tree-vect-stmts.c (vectorizable_call): Make sure that we
11005 use the stmt_vec_info of the original bb statement for the
11006 new zero assignment, even if the call is part of a pattern.
11008 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11010 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
11011 that the sequence is attached to the original statement rather
11012 than the pattern statement.
11013 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
11014 PATTERN_DEF_SEQ from the original statement rather than
11015 the main pattern statement.
11016 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
11017 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
11018 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
11020 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
11022 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
11023 definition statements before the early exit for statements that aren't
11025 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
11027 (vect_transform_loop): ...here. Process pattern definition
11028 statements without first checking whether the main pattern
11029 statement is live or relevant.
11031 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11033 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
11034 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
11036 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
11038 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
11039 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
11040 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
11041 (expand_block_compare): Change select_block_compare_mode call.
11042 (expand_strncmp_align_check): Use new functions, fix comment.
11043 (emit_final_str_compare_gpr): New function.
11044 (expand_strn_compare): Refactor and clean up code.
11045 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
11047 2018-06-19 Tony Reix <tony.reix@atos.com>
11048 Damien Bergamini <damien.bergamini@atos.com>
11049 David Edelsohn <dje.gcc@gmail.com>
11051 * collect2.c (static_obj): New variable.
11052 (static_libs): New variable.
11053 (is_in_list): Uncomment declaration.
11054 (main): Track AIX libraries linked statically.
11055 (is_in_list): Uncomment definition.
11056 (scan_prog_file): Don't add AIX shared libraries initializer
11057 to constructor list if linking statically.
11059 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
11061 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
11063 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
11065 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
11067 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
11070 2018-06-19 Martin Liska <mliska@suse.cz>
11072 * config/i386/i386.c (ix86_can_inline_p): Do not use
11073 ipa_fn_summaries::get_create.
11074 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
11076 (devirtualization_time_bonus): Likewise.
11077 (ipcp_propagate_stage): Likewise.
11078 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11079 (edge_set_predicate): Likewise.
11080 (evaluate_conditions_for_known_args): Likewise.
11081 (evaluate_properties_for_edge): Likewise.
11082 (ipa_call_summary::reset): Tranform to ...
11083 (ipa_call_summary::~ipa_call_summary): ... this.
11084 (ipa_fn_summary::reset): Transform to ...
11085 (ipa_fn_summary::~ipa_fn_summary): ... this.
11086 (ipa_fn_summary_t::remove): Rename to ...
11087 (ipa_fn_summary_t::remove_callees): ... this.
11088 (ipa_fn_summary_t::duplicate): Use placement new
11089 instead of memory copy.
11090 (ipa_call_summary_t::duplicate): Likewise.
11091 (ipa_call_summary_t::remove): Remove.
11092 (dump_ipa_call_summary): Change get_create to get.
11093 (ipa_dump_fn_summary): Dump only when summary exists.
11094 (analyze_function_body): Use symbol_summary::get instead
11096 (compute_fn_summary): Likewise.
11097 (estimate_edge_devirt_benefit): Likewise.
11098 (estimate_edge_size_and_time): Likewise.
11099 (inline_update_callee_summaries): Likewise.
11100 (remap_edge_change_prob): Likewise.
11101 (remap_edge_summaries): Likewise.
11102 (ipa_merge_fn_summary_after_inlining): Likewise.
11103 (write_ipa_call_summary): Likewise.
11104 (ipa_fn_summary_write): Likewise.
11105 (ipa_free_fn_summary): Likewise.
11106 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
11107 (struct ipa_call_summary): Likewise.
11108 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
11110 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
11111 (estimate_size_after_inlining): Likewise.
11112 (estimate_growth): Likewise.
11113 (growth_likely_positive): Likewise.
11114 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11115 (inline_call): Likewise.
11116 * ipa-inline.c (caller_growth_limits): Likewise.
11117 (can_inline_edge_p): Likewise.
11118 (can_inline_edge_by_limits_p): Likewise.
11119 (compute_uninlined_call_time): Likewise.
11120 (compute_inlined_call_time): Likewise.
11121 (want_inline_small_function_p): Likewise.
11122 (edge_badness): Likewise.
11123 (update_caller_keys): Likewise.
11124 (update_callee_keys): Likewise.
11125 (inline_small_functions): Likewise.
11126 (inline_to_all_callers_1): Likewise.
11127 (dump_overall_stats): Likewise.
11128 (early_inline_small_functions): Likewise.
11129 (early_inliner): Likewise.
11130 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11131 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11132 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11133 * ipa-split.c (execute_split_functions): Likewise.
11134 * symbol-summary.h: Likewise.
11135 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11137 2018-06-19 Richard Biener <rguenther@suse.de>
11139 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
11140 (vectorize_loops): ... here. Fix dbgcnt handling.
11141 (try_vectorize_loop): Wrap try_vectorize_loop_1.
11143 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
11146 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
11147 ieee128 argument takes up only one (vector) register, not two (floating
11150 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
11152 * gimplify.c (gimplify_init_constructor): Really never clear for an
11153 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
11155 2018-06-19 Richard Biener <rguenther@suse.de>
11157 PR tree-optimization/86179
11158 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
11159 after failed recognition.
11161 2018-06-18 Martin Sebor <msebor@redhat.com>
11163 PR middle-end/85602
11164 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
11165 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
11166 Handle integer subtraction.
11167 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
11168 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
11170 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11172 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
11173 param from rtx to rtx_insn *.
11174 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
11176 (frv_ifcvt_modify_insn): Likwise.
11177 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
11178 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
11179 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
11180 as_a <rtx_insn *> cast to local "unprotected_region" once
11181 it's been established that it's not NULL or pc_rtx.
11182 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
11183 param "sethi" from rtx to rtx_insn *.
11184 (nds32_group_float_insns): Likewise for param "insn".
11185 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
11187 (vax_output_int_subtract): Likewise.
11188 * config/vax/vax.c (vax_output_int_add): Likewise for param
11190 (vax_output_int_subtract): Likewise.
11191 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
11192 (emit_pattern_after): Likewise for param "after".
11193 (emit_insn_after): Likewise.
11194 (emit_jump_insn_after): Likewise.
11195 (emit_call_insn_after): Likewise.
11196 (emit_debug_insn_after): Likewise.
11197 (emit_pattern_before): Likewise for param "before".
11198 (emit_insn_before): Likewise.
11199 (emit_jump_insn_before): Likewise.
11200 * final.c (get_insn_template): Likewise for param "insn", removing
11202 * output.h (get_insn_template): Likewise for 2nd param.
11203 * rtl.h (emit_insn_before): Likewise.
11204 (emit_jump_insn_before): Likewise.
11205 (emit_debug_insn_before_noloc): Likewise.
11206 (emit_insn_after): Likewise.
11207 (emit_jump_insn_after): Likewise.
11208 (emit_call_insn_after): Likewise.
11209 (emit_debug_insn_after): Likewise.
11210 (set_insn_deleted): Likewise for param.
11212 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
11215 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
11216 floating point modes, so that IFmode is numerically greater than
11217 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
11218 to declare the ordering. This prevents IFmode from being
11219 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
11220 machine. Include rs6000-modes.h to share the fractional values
11221 between genmodes* and the rest of the compiler.
11222 (IFmode): Likewise.
11223 (KFmode): Likewise.
11224 (TFmode): Likewise.
11225 * config/rs6000/rs6000-modes.h: New file.
11226 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
11227 meaning of rs6000_long_double_size so that 126..128 selects an
11228 appropriate 128-bit floating point type.
11229 (rs6000_option_override_internal): Likewise.
11230 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
11231 (TARGET_LONG_DOUBLE_128): Change the meaning of
11232 rs6000_long_double_size so that 126..128 selects an appropriate
11233 128-bit floating point type.
11234 (LONG_DOUBLE_TYPE_SIZE): Update comment.
11235 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
11236 source and destination to match the standard usage.
11237 (truncifkf2): Likewise.
11238 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
11239 ISA 2.07 to use an explicit clobber, instead of passing in a
11241 (copysign<mode>3_soft): Likewise.
11243 2018-06-18 David Malcolm <dmalcolm@redhat.com>
11245 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
11246 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
11247 (vect_slp_analyze_instance_dependence): Likewise.
11248 (vect_enhance_data_refs_alignment): Likewise.
11249 (vect_analyze_data_refs_alignment): Likewise.
11250 (vect_slp_analyze_and_verify_instance_alignment
11251 (vect_analyze_data_ref_accesses): Likewise.
11252 (vect_prune_runtime_alias_test_list): Likewise.
11253 (vect_analyze_data_refs): Likewise.
11254 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
11255 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
11256 (vect_analyze_scalar_cycles_1): Likewise.
11257 (vect_get_loop_niters): Likewise.
11258 (vect_analyze_loop_form_1): Likewise.
11259 (vect_update_vf_for_slp): Likewise.
11260 (vect_analyze_loop_operations): Likewise.
11261 (vect_analyze_loop): Likewise.
11262 (vectorizable_induction): Likewise.
11263 (vect_transform_loop): Likewise.
11264 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
11265 * tree-vect-slp.c (vect_analyze_slp): Likewise.
11266 (vect_make_slp_decision): Likewise.
11267 (vect_detect_hybrid_slp): Likewise.
11268 (vect_slp_analyze_operations): Likewise.
11269 (vect_slp_bb): Likewise.
11270 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
11271 (vectorizable_bswap): Likewise.
11272 (vectorizable_call): Likewise.
11273 (vectorizable_simd_clone_call): Likewise.
11274 (vectorizable_conversion): Likewise.
11275 (vectorizable_assignment): Likewise.
11276 (vectorizable_shift): Likewise.
11277 (vectorizable_operation): Likewise.
11278 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
11280 2018-06-18 Martin Sebor <msebor@redhat.com>
11282 PR tree-optimization/81384
11283 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
11284 * builtins.c (expand_builtin_strnlen): New function.
11285 (expand_builtin): Call it.
11286 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
11287 * builtins.def (BUILT_IN_STRNLEN): New.
11288 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
11289 Warn for bounds in excess of maximum object size.
11290 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
11291 single-value ranges. Handle strnlen.
11292 (handle_builtin_strlen): Handle strnlen.
11293 (strlen_check_and_optimize_stmt): Same.
11294 * doc/extend.texi (Other Builtins): Document strnlen.
11296 2018-06-18 Maya Rashish <coypu@sdf.org>
11298 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
11299 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
11300 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
11302 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
11304 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
11306 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
11308 * tree.c (escaped_string::escape): Replace cast to char * by
11309 const_cast<char *> (unescaped).
11311 2018-06-18 Nick Clifton <nickc@redhat.com>
11314 * tree.c (escaped_string): New class. Converts an unescaped
11315 string into its escaped equivalent.
11316 (warn_deprecated_use): Use the new class to convert the
11317 deprecation message, if present.
11318 (test_escaped_strings): New self test.
11319 (test_c_tests): Add test_escaped_strings.
11320 * doc/extend.texi (deprecated): Add a note that the
11321 deprecation message is affected by the -fmessage-length
11322 option, and that control characters will be escaped.
11323 (#pragma GCC error): Document this pragma.
11324 (#pragma GCC warning): Likewise.
11325 * doc/invoke.texi (-fmessage-length): Document this option's
11326 effect on the #warning and #error preprocessor directives and
11327 the deprecated attribute.
11329 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
11331 * tree.c (decl_value_expr_lookup): Revert latest change.
11332 (decl_value_expr_insert): Likewise.
11334 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
11336 * gimplify.c (nonlocal_vlas): Delete.
11337 (nonlocal_vla_vars): Likewise.
11338 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
11340 (gimplify_body): Do not create and destroy nonlocal_vlas.
11341 * tree-nested.c: Include diagnostic.h.
11342 (use_pointer_in_frame): Tweak.
11343 (lookup_field_for_decl): Add assertion and declare the transformation.
11344 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
11345 internal error when the reference is in a wrong context. Do not
11346 create a debug decl by default.
11347 (note_nonlocal_block_vlas): Delete.
11348 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
11349 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
11350 create a debug decl by default.
11351 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
11352 call is in a wrong context.
11353 (fixup_vla_decls): New function.
11354 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
11355 debug variables were created.
11356 * tree.c (decl_value_expr_lookup): Add checking assertion.
11357 (decl_value_expr_insert): Likewise.
11359 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11361 PR middle-end/82479
11362 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
11363 * tree-scalar-evolution.c (interpret_expr): Likewise.
11364 (expression_expensive_p): Likewise.
11365 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
11366 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
11367 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
11368 (ssa_defined_by_minus_one_stmt_p): New.
11370 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
11372 PR middle-end/64946
11373 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
11374 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
11375 * dojump.c (do_jump): Likewise.
11376 * expr.c (expand_expr_real_2): Check operand type's sign.
11377 * fold-const.c (const_unop): Handle ABSU_EXPR.
11378 (fold_abs_const): Likewise.
11379 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11380 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
11381 (strip_sign_op_1): Likesise.
11382 * match.pd: Add new pattern to generate ABSU_EXPR.
11383 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
11384 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
11385 * tree-eh.c (operation_could_trap_helper_p): Likewise.
11386 * tree-inline.c (estimate_operator_cost): Likewise.
11387 * tree-pretty-print.c (dump_generic_node): Likewise.
11388 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
11389 * tree.def (ABSU_EXPR): New.
11391 2018-06-16 Jakub Jelinek <jakub@redhat.com>
11393 PR middle-end/86095
11394 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
11395 documented as preserved for backward compatibility only.
11396 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
11398 PR rtl-optimization/86108
11399 * bb-reorder.c (create_forwarder_block): Renamed to ...
11400 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
11401 jump from new landing pad to the second part.
11402 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
11405 2018-06-15 Jakub Jelinek <jakub@redhat.com>
11407 PR middle-end/85878
11408 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
11409 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
11410 Only call store_expr for halves if the mode is the same.
11412 PR middle-end/86123
11413 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
11414 Fix up comment formatting.
11416 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
11418 * typed-splay-tree.h (typed_splay_tree::remove): New function.
11419 (typed_splay_tree::closure,
11420 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
11421 (typed_splay_tree::typed_splay_tree,
11422 typed_splay_tree::operator =): Declared private.
11423 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
11424 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
11425 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
11426 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
11427 typed_splay_tree::splay_tree_splay,
11428 typed_splay_tree::splay_tree_foreach_helper,
11429 typed_splay_tree::splay_tree_insert,
11430 typed_splay_tree::splay_tree_remove,
11431 typed_splay_tree::splay_tree_lookup,
11432 typed_splay_tree::splay_tree_predecessor,
11433 typed_splay_tree::splay_tree_successor,
11434 typed_splay_tree::splay_tree_min,
11435 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
11436 (typed_splay_tree::root, typed_splay_tree::comp,
11437 typed_splay_tree::delete_key,
11438 typed_splay_tree::delete_value): New data members.
11439 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
11440 typed_splay_tree::remove.
11442 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
11444 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
11445 -mginv and -mno-ginv to the assembler.
11446 * config/mips/mips.opt (-mcrc): New option.
11447 (-mginv): Likewise.
11448 * doc/invoke.text (-mcrc): Document.
11449 (-mginv): Likewise.
11451 2018-06-15 Nick Clifton <nickc@redhat.com>
11454 * tree.c (escaped_string): New class. Converts an unescaped
11455 string into its escaped equivalent.
11456 (warn_deprecated_use): Use the new class to convert the
11457 deprecation message, if present.
11458 (test_escaped_strings): New self test.
11459 (test_c_tests): Add test_escaped_strings.
11460 * doc/extend.texi (deprecated): Add a note that the
11461 deprecation message is affected by the -fmessage-length
11462 option, and that control characters will be escaped.
11463 (#pragma GCC error): Document this pragma.
11464 (#pragma GCC warning): Likewise.
11465 * doc/invoke.texi (-fmessage-length): Document this option's
11466 effect on the #warning and #error preprocessor directives and
11467 the deprecated attribute.
11469 2018-06-15 Richard Biener <rguenther@suse.de>
11471 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
11472 here, also noting vector size used.
11473 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
11474 size used in MSG_OPTIMIZED_LOCATIONS dump.
11475 (pass_slp_vectorize::execute): Adjust.
11477 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
11480 * config/arc/arc.c (arc_return_address_register): Fix
11483 2018-06-15 Richard Biener <rguenther@suse.de>
11485 PR middle-end/86159
11486 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
11487 leave useless conversion stripping to force_gimple_operand_gsi.
11488 (gimplify_build2): Likewise.
11489 (gimplify_build1): Likewise.
11491 2018-06-15 Richard Biener <rguenther@suse.de>
11493 PR middle-end/86076
11494 * tree-cfg.c (move_stmt_op): unshare invariant addresses
11495 before adjusting their block.
11497 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
11499 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
11500 multilibs for *-*-rtems*.
11501 * config/riscv/t-rtems: New file.
11503 2018-06-14 Jakub Jelinek <jakub@redhat.com>
11505 PR middle-end/86122
11506 * match.pd ((A +- CST1) +- CST2): Punt if last resort
11507 unsigned_type_for returns NULL.
11510 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
11511 subregs of multi-word pseudos unless the float mode has word size.
11513 2018-06-14 Richard Biener <rguenther@suse.de>
11515 PR middle-end/86139
11516 * tree-vect-generic.c (build_word_mode_vector_type): Remove
11517 duplicate and harmful type_hash_canon.
11518 * tree.c (type_hash_canon): Assert we didn't find ourselves.
11520 2018-06-14 Richard Biener <rguenther@suse.de>
11523 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
11526 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
11528 * config/rtems.h (STDINT_LONG32): Define.
11530 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
11531 Prachi Godbole <prachi.godbole@imgtec.com>
11533 * config/mips/mips-cpus.def: Define P6600.
11534 * config/mips/mips-tables.opt: Regenerate.
11535 * config/mips/mips.c (mips_ucbranch_type): New enum.
11536 (mips_rtx_cost_data): Add support for P6600.
11537 (mips_issue_rate): Likewise.
11538 (mips_multipass_dfa_lookahead): Likewise.
11539 (mips_avoid_hazard): Likewise.
11540 (mips_reorg_process_insns): Likewise.
11541 (mips_classify_branch_p6600): New function.
11542 * config/mips/mips.h (TUNE_P6600): New define.
11543 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
11544 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
11545 * config/mips/mips.md: Include p6600.md.
11546 (processor): Add p6600.
11547 * config/mips/p6600.md: New file.
11548 * doc/invoke.texi: Add p6600 to supported architectures.
11550 2018-06-13 Martin Sebor <msebor@redhat.com>
11552 PR tree-optimization/86114
11553 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
11555 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
11557 2018-06-13 Richard Biener <rguenther@suse.de>
11559 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
11560 Properly set vector type of the intermediate stmt.
11561 * tree-vect-stmts.c (vectorizable_operation): The destination
11562 var always has vectype_out type.
11564 2018-06-13 Jeff Law <law@redhat.com>
11566 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
11567 integer 0 for argument to print_rtl_with_bb.
11568 (rl78_reorg): Likewise.
11570 2018-06-13 David Malcolm <dmalcolm@redhat.com>
11572 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
11573 from rtx to rtx_insn *.
11574 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
11576 (add_sched_insns_for_speculation): Likewise for local "target",
11577 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11578 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
11579 from rtx_insn ** to rtx_code_label **.
11580 (reorg_emit_nops): Likewise.
11581 (c6x_reorg): Likewise for local "call_labels".
11582 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
11584 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
11585 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
11586 the loops over LABEL_REFS.
11587 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
11589 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
11590 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
11591 (split_branches): Strengthen local "olabel" from rtx to
11592 rtx_insn *, adding a safe_as_a cast.
11593 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
11595 (add_insn_after): Likewise for first two params.
11596 (add_insn_before): Likewise.
11597 (remove_insn): Likewise for param.
11598 (emit_pattern_before_noloc): Likewise for second and third params.
11599 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
11600 (emit_call_insn_before_noloc): Likewise.
11601 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
11603 (emit_barrier_before): Likewise.
11604 (emit_label_before): Strengthen "label" param from "rtx" to
11605 "rtx_code_label *". Strengthen "before" param from "rtx" to
11607 (emit_insn_after_1): Strengthen "after" param from "rtx" to
11609 (emit_pattern_after_noloc): Likewise.
11610 (emit_insn_after_noloc): Likewise.
11611 (emit_jump_insn_after_noloc): Likewise.
11612 (emit_call_insn_after_noloc): Likewise.
11613 (emit_debug_insn_after_noloc): Likewise.
11614 (emit_barrier_after): Likewise.
11615 (emit_label_after): Likewise for both params.
11616 (emit_pattern_after_setloc): Likewise for "after" param. Convert
11617 "loc" param from "int" to "location_t".
11618 (emit_insn_after_setloc): Likewise.
11619 (emit_jump_insn_after_setloc): Likewise.
11620 (emit_call_insn_after_setloc): Likewise.
11621 (emit_debug_insn_after_setloc): Likewise.
11622 (emit_pattern_before_setloc): Likewise for "before" param. Convert
11623 "loc" param from "int" to "location_t".
11624 (emit_pattern_before): Convert NULL_RTX to NULL.
11625 (emit_insn_before_setloc): Convert "loc" param from "int" to
11627 (emit_jump_insn_before_setloc): Likewise.
11628 (emit_call_insn_before_setloc): Likewise.
11629 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
11630 rtx_insn *. Convert "loc" param from "int" to "location_t".
11631 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
11632 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
11633 Convert 3rd param from "int" to "location_t".
11634 (emit_barrier_before, emit_barrier_after, next_real_insn):
11635 Strengthen param from rtx to rtx_insn *.
11636 (emit_label_before): Strengthen 1st param from "rtx" to
11637 "rtx_code_label *". Strengthen 2nd param from "rtx" to
11639 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
11640 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
11641 Strengthen 2nd param from "rtx" to "rtx_insn *".
11642 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
11643 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
11644 Likewise. Convert 3rd param from "int" to "location_t".
11645 (emit_label_after): Strengthen 1st param from "rtx" to
11646 "rtx_code_label *".
11647 (next_real_insn, remove_insn): Strengthen param from "rtx" to
11649 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
11650 from "rtx" to "rtx_insn *".
11652 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
11654 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
11655 bodies streamed in with -Q.
11656 * dumpfile.c (dump_files): Add lto-stream-out dump file.
11657 * dumpfile.h (tree_dump_index): Add lto_stream_out.
11658 * gimple-streamer-out.c: Include gimple-pretty-print.h
11659 (output_bb): Dump stmts streamed.
11660 * lto-section-out.c: Include print-tree.h
11661 (lto_begin_section): Dump sections created.
11662 (lto_output_decl_index): Dump decl encoded.
11663 * lto-streamer-out.c: Include print-tree.h
11664 (create_output_block): Dump output block created.
11665 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
11666 (output_function): Dump function output.
11667 (output_constructor): Dump constructor streamed.
11668 (write_global_stream): Output indexes encoded.
11669 (produce_asm_for_decls): Dump streams encoded.
11670 * lto-streamer.c (streamer_dump_file): New global var.
11671 * lto-streamer.h (streamer_dump_file): Declare.
11672 * passes.c (ipa_write_summaries): Initialize streamer dump.
11673 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
11676 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
11679 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
11680 offsets for register save directives. Emit a second batch of save
11681 directives, if need be, when the function accesses prior frames.
11683 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11685 * config/arc/fpu.md (fmasf4): Force operand to register.
11686 (fnmasf4): Likewise.
11688 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11690 * config/arc/arc-protos.h (arc_pad_return): Remove.
11691 * config/arc/arc.c (machine_function): Remove force_short_suffix
11693 (arc_print_operand): Adjust printing of '&'.
11694 (arc_verify_short): Remove conditional printing of short suffix.
11695 (arc_final_prescan_insn): Remove reference to size_reason.
11696 (pad_return): New function.
11697 (arc_reorg): Call pad_return.
11698 (arc_pad_return): Remove.
11699 (arc_init_machine_status): Remove reference to force_short_suffix.
11700 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
11701 (attr length): When attribute iscompact is true force to 2
11702 regardless; in the case of maybe check if we want to force the
11703 instruction to have 4 bytes length.
11704 (nopv): Change it to generate 4 byte long nop as well.
11705 (blockage): New pattern.
11706 (simple_return): Remove call to arc_pad_return.
11707 (p_return_i): Likewise.
11709 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11711 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
11713 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11715 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
11718 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11720 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
11721 for ARC700 and ARCv2.
11723 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
11726 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
11727 operands[2] instead of operands[1].
11730 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11732 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
11733 case, check whether the outer register overlaps an unallocatable
11734 register, not just whether it fits the required class.
11736 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
11738 * poly-int.h (can_div_trunc_p): Add new overload in which all values
11740 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
11741 (memrefs_conflict_p): Likewise.
11742 (init_alias_analysis): Likewise.
11743 * cfgexpand.c (expand_debug_expr): Likewise.
11744 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
11745 * cse.c (fold_rtx): Likewise.
11746 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
11747 * expr.c (emit_block_move_hints): Likewise.
11748 (clear_storage_hints, push_block, emit_push_insn): Likewise.
11749 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
11750 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
11751 (emit_group_store): Likewise.
11752 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
11753 to read the PRE/POST_MODIFY increment.
11754 * calls.c (store_one_arg): Use strip_offset.
11755 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
11757 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
11759 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
11760 (simplify_binary_operation_1): Extend CONST_INT handling to
11762 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
11763 than a HOST_WIDE_INT.
11764 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
11766 (adjust_mems, add_stores): Update accodingly.
11767 (vt_canonicalize_addr): Track polynomial offsets.
11768 (emit_note_insn_var_location): Likewise.
11769 (vt_add_function_parameter): Likewise.
11770 (vt_initialize): Likewise.
11772 2018-06-12 Jeff Law <law@redhat.com>
11774 * config.gcc (alpha*-*-freebsd*): Remove.
11775 * config/alpha/freebsd.h: Remove.
11777 2018-06-12 David Malcolm <dmalcolm@redhat.com>
11780 * spellcheck-tree.c (levenshtein_distance): Rename to...
11781 (get_edit_distance): ...this, and update for underlying renaming.
11782 * spellcheck-tree.h (levenshtein_distance): Rename to...
11783 (get_edit_distance): ...this.
11784 * spellcheck.c (levenshtein_distance): Rename to...
11785 (get_edit_distance): ...this. Convert from Levenshtein distance
11786 to Damerau-Levenshtein distance by supporting transpositions of
11787 adjacent characters. Rename "v1" to "v_next" and "v0" to
11789 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
11790 (selftest::test_edit_distance_unit_test_oneway): ...this, and
11791 update for underlying renaming.
11792 (selftest::levenshtein_distance_unit_test): Rename to...
11793 (selftest::test_get_edit_distance_unit): ...this, and update for
11794 underlying renaming.
11795 (selftest::test_find_closest_string): Add example from PR 69968
11796 where transposition helps
11797 (selftest::test_metric_conditions): Update for renaming.
11798 (selftest::test_metric_conditions): Likewise.
11799 (selftest::spellcheck_c_tests): Likewise.
11800 * spellcheck.h (levenshtein_distance): Rename both overloads to...
11801 (get_edit_distance): ...this.
11802 (best_match::consider): Update for renaming.
11804 2018-06-12 Martin Sebor <msebor@redhat.com>
11806 PR tree-optimization/85259
11807 * builtins.c (compute_objsize): Handle constant offsets.
11808 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
11809 true iff a warning has been issued.
11810 * gimple.h (gimple_nonartificial_location): New function.
11811 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
11812 gimple_nonartificial_location and handle -Wno-system-headers.
11813 (handle_builtin_stxncpy): Same.
11815 2018-06-12 Martin Sebor <msebor@redhat.com>
11818 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
11820 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
11822 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
11823 BUILTIN_VEC_XST entries for pointer to double and long long.
11825 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
11828 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11830 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
11833 2018-06-12 Martin Liska <mliska@suse.cz>
11835 * doc/options.texi: Document IntegerRange.
11837 2018-06-12 Martin Liska <mliska@suse.cz>
11839 * config/i386/i386.opt: Make MPX-related options as Deprecated.
11840 * opt-functions.awk: Handle Deprecated flag.
11841 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
11843 (read_cmdline_option): Report warning for a deprecated option.
11844 * opts.h (struct cl_option): Add new field cl_deprecated.
11845 (CL_ERR_DEPRECATED): New.
11847 2018-06-12 Martin Liska <mliska@suse.cz>
11849 * doc/options.texi: Document Deprecated option flag.
11851 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11853 * config/arc/arc-arch.h (arc_extras): New enum.
11854 (arc_cpu_t):Add field extra.
11855 (arc_cpu_types): Consider the extras.
11856 * config/arc/arc-cpus.def: Add extras info.
11857 * config/arc/arc-opts.h (processor_type): Consider extra field.
11858 * config/arc/arc.c (arc_override_options): Handle extra field.
11860 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
11862 * config/arc/arc-arch.h: Update ARC_OPTX macro.
11863 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
11865 * config/arc/arc.c (arc_init): Update pic warning.
11866 (irq_range): Update irq range parsing warnings.
11867 (arc_override_options): Update various warning messages.
11868 (arc_handle_aux_attribute): Likewise.
11870 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
11872 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
11874 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11876 * doc/sourcebuild.texi: Document usage of line number 0 in verify
11877 compiler messages directives.
11879 2018-06-12 Matthew Fortune <mfortune@gmail.com>
11881 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
11882 * config/mips/mips-tables.opt: Regenerate.
11883 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
11885 * doc/invoke.texi: Document -march=i6500.
11887 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
11889 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
11890 (i6400_gpmul): Add cpu_unit.
11891 (i6400_gpdiv): Likewise.
11892 (i6400_msa_add_d): Update reservations.
11893 (i6400_msa_int_add) Likewise.
11894 (i6400_msa_short_logic3) Likewise.
11895 (i6400_msa_short_logic2) Likewise.
11896 (i6400_msa_short_logic) Likewise.
11897 (i6400_msa_move) Likewise.
11898 (i6400_msa_cmp) Likewise.
11899 (i6400_msa_short_float2) Likewise.
11900 (i6400_msa_div_d) Likewise.
11901 (i6400_msa_long_logic1) Likewise.
11902 (i6400_msa_long_logic2) Likewise.
11903 (i6400_msa_mult) Likewise.
11904 (i6400_msa_long_float2) Likewise.
11905 (i6400_msa_long_float4) Likewise.
11906 (i6400_msa_long_float5) Likewise.
11907 (i6400_msa_long_float8) Likewise.
11908 (i6400_fpu_fadd): Include frint type.
11909 (i6400_fpu_store): New define_insn_reservation.
11910 (i6400_fpu_load): Likewise.
11911 (i6400_fpu_move): Likewise.
11912 (i6400_fpu_fcmp): Likewise.
11913 (i6400_fpu_fmadd): Likewise.
11914 (i6400_int_mult): Include imul3nc type and update reservation.
11915 (i6400_int_div): Include idiv3 type and update reservation.
11916 (i6400_int_load): Update to check type not move_type.
11917 (i6400_int_store): Likewise.
11918 (i6400_int_prefetch): Set zero latency.
11920 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
11922 * gcc.c: Document new %@{...} sequence.
11923 (LINK_COMMAND_SPEC): Use it for the -L switches.
11924 (cpp_unique_options): Use it for the -I switches.
11925 (at_file_argbuf): New global variable.
11926 (in_at_file): Likewise.
11927 (alloc_args): Create at_file_argbuf.
11928 (clear_args): Truncate at_file_argbuf.
11929 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
11930 (open_at_file): New function.
11931 (close_at_file): Likewise.
11932 (create_at_file): Delete.
11933 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
11936 (validate_switches_from_spec): Deal with %@{...} sequence.
11937 (validate_switches): Likewise.
11938 (driver::finalize): Call clear_args.
11940 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
11942 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
11944 2018-06-11 Martin Sebor <msebor@redhat.com>
11946 * doc/invoke.texi (-Wall): List -Wc++17-compat.
11947 (Wno-class-memaccess): Add @opindex.
11948 (Wno-templates, Wno-multiple-inheritance): Same.
11949 (Wno-virtual-inheritance, Wno-namespaces): Same.
11950 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
11951 (Wno-format-overflow, Wno-format-truncation): Same.
11952 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
11953 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
11954 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
11955 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
11956 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
11957 (Wno-misspelled-isr): Same.
11959 2018-06-11 Martin Sebor <msebor@redhat.com>
11961 * PR tree-optimization/86083
11962 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11964 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
11966 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11968 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
11971 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11972 on the correct operand.
11973 (*movdi_internal64): Ditto.
11975 2018-06-11 Martin Liska <mliska@suse.cz>
11977 PR tree-optimization/86089
11978 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11980 2018-06-11 Julia Koval <julia.koval@intel.com>
11982 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11983 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11984 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11986 2018-06-11 Olivier Hainque <hainque@adacore.com>
11988 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11989 for Ada with strict dwarf2.
11991 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
11994 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11997 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
11999 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
12001 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
12003 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
12006 2018-06-08 Martin Liska <mliska@suse.cz>
12008 * tree-cfg.h (debug_function): Fix argument type to match
12011 2018-06-08 Martin Liska <mliska@suse.cz>
12013 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
12014 Remove usage of MPX-related (and removed) fields.
12015 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
12017 2018-06-08 David Malcolm <dmalcolm@redhat.com>
12019 * cfg.c (debug): Use TDF_NONE rather than 0.
12020 * cfghooks.c (debug): Likewise.
12021 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
12022 (struct dump_option_value_info): Convert to...
12023 (struct kv_pair): ...this template type.
12024 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
12026 (optinfo_verbosity_options): Likewise.
12027 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
12029 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
12030 than int for "optgroup_flags" param.
12031 (dump_generic_expr_loc): Use dump_flags_t rather than int for
12033 (dump_dec): Likewise.
12034 (dump_finish): Use TDF_NONE rather than 0.
12035 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
12036 rather than int for "optgroup_flags" param. Use TDF_NONE rather
12037 than 0. Update for change to option_ptr.
12038 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
12039 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
12040 0. Update for changes to optinfo_verbosity_options and
12042 (opt_info_switch_p): Convert optgroup_flags from int to
12044 (dump_basic_block): Use dump_flags_t rather than int
12045 for "dump_kind" param.
12046 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
12047 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
12048 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
12049 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
12050 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
12051 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
12052 TDF_NONE): Convert from macros to...
12053 (enum dump_flag): ...this new enum.
12054 (dump_flags_t): Update to use enum.
12055 (operator|, operator&, operator~, operator|=, operator&=):
12056 Implement for dump_flags_t.
12057 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
12058 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
12059 Convert from macros to...
12060 (enum optgroup_flag): ...this new enum.
12061 (optgroup_flags_t): New typedef.
12062 (operator|, operator|=): Implement for optgroup_flags_t.
12063 (struct dump_file_info): Convert field "alt_flags" to
12064 dump_flags_t. Convert field "optgroup_flags" to
12066 (dump_basic_block): Use dump_flags_t rather than int for param.
12067 (dump_generic_expr_loc): Likewise.
12068 (dump_dec): Likewise.
12069 (dump_register): Convert param "optgroup_flags" to
12071 (opt_info_enable_passes): Likewise.
12072 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
12074 * gimple-pretty-print.c (debug): Likewise.
12075 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
12076 (merged_store_group::apply_stores): Likewise.
12077 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
12078 * gimple.c (verify_gimple_pp): Likewise.
12079 * graphite-poly.c (print_pbb_body): Likewise.
12080 * passes.c (pass_manager::register_one_dump_file): Convert
12081 local "optgroup_flags" to optgroup_flags_t.
12082 * print-tree.c (print_node): Use TDF_NONE rather than 0.
12084 (debug_body): Likewise.
12085 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
12086 to optgroup_flags_t.
12087 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
12089 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
12090 (convert_mult_to_fma): Likewise.
12091 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
12092 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
12093 * tree-vect-data-refs.c (dump_lower_bound): Convert param
12094 "dump_kind" to dump_flags_t.
12096 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12098 * config/rs6000/rs6000.c (min, max): Delete.
12100 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
12102 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
12103 -mabi=spe and -mabi=no-spe.
12105 2018-06-08 Martin Liska <mliska@suse.cz>
12107 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
12108 where we expect an existing summary.
12110 2018-06-08 Martin Liska <mliska@suse.cz>
12112 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
12113 * ipa-inline.h (estimate_edge_growth): Likewise.
12115 2018-06-08 Martin Liska <mliska@suse.cz>
12117 * cgraph.c (function_version_hasher::hash): Use
12118 cgraph_node::get_uid ().
12119 (function_version_hasher::equal):
12120 * cgraph.h (cgraph_node::get_uid): New method.
12121 * ipa-inline.c (update_caller_keys): Use
12122 cgraph_node::get_uid ().
12123 (update_callee_keys): Likewise.
12124 * ipa-utils.c (searchc): Likewise.
12125 (ipa_reduced_postorder): Likewise.
12126 * lto-cgraph.c (input_node): Likewise.
12127 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
12128 * symbol-summary.h (symtab_insertion): Likewise.
12129 (symtab_removal): Likewise.
12130 (symtab_duplication): Likewise.
12131 * tree-pretty-print.c (dump_function_header): Likewise.
12132 * tree-sra.c (convert_callers_for_node): Likewise.
12134 2018-06-08 Martin Liska <mliska@suse.cz>
12136 * cgraph.c (symbol_table::create_edge): Always assign a new
12138 (symbol_table::free_edge): Do not recycle numbers.
12139 * cgraph.h (cgraph_edge::get): New method.
12140 * symbol-summary.h (symtab_removal): Use it.
12141 (symtab_duplication): Likewise.
12142 (call_summary::hashable_uid): Remove.
12144 2018-06-08 Martin Liska <mliska@suse.cz>
12146 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
12147 (initialize_growth_caches): Remove.
12148 (free_growth_caches): Likewise.
12149 (do_estimate_edge_time): Use edge_growth_cache.
12150 (do_estimate_edge_size): Likewise.
12151 (do_estimate_edge_hints): Likewise.
12152 * ipa-inline.c (reset_edge_caches): Likewise.
12153 (recursive_inlining): Likewise.
12154 (inline_small_functions): Likewise.
12155 * ipa-inline.h (initialize_growth_caches): Remove.
12156 (estimate_edge_size): Likewise.
12157 (estimate_edge_time): Likewise.
12158 (estimate_edge_hints): Likewise.
12159 (reset_edge_growth_cache): Likewise.
12160 * symbol-summary.h (call_summary::remove): New method.
12162 2018-06-08 Martin Liska <mliska@suse.cz>
12164 * ipa-cp.c (class edge_clone_summary): New summary.
12165 (grow_edge_clone_vectors): Remove.
12166 (ipcp_edge_duplication_hook): Remove.
12167 (class edge_clone_summary_t): New call_summary class.
12168 (ipcp_edge_removal_hook): Remove.
12169 (edge_clone_summary_t::duplicate): New function.
12170 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
12171 (create_specialized_node): Likewise.
12172 (ipcp_driver): Initialize edge_clone_summaries and do not
12175 2018-06-08 Martin Liska <mliska@suse.cz>
12177 * symbol-summary.h (get): New function.
12178 (call_summary::m_initialize_when_cloning): New class member.
12180 2018-06-08 Martin Liska <mliska@suse.cz>
12182 * cgraph.c (cgraph_node::remove): Do not recycle uid.
12183 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
12184 (symbol_table::allocate_cgraph_symbol): Do not set uid.
12185 * passes.c (uid_hash_t): Record removed_nodes by their uids.
12186 (remove_cgraph_node_from_order): Use the removed_nodes set.
12187 (do_per_function_toporder): Likwise.
12188 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
12189 instead of summary_uid.
12190 (symtab_removal): Likewise.
12191 (symtab_duplication): Likewise.
12193 2018-06-08 Martin Liska <mliska@suse.cz>
12195 * ipa-cp.c (ipcp_store_bits_results): Use
12196 ipcp_transformation_sum.
12197 (ipcp_store_vr_results): Likewise.
12198 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
12200 (ipcp_transformation_initialize): ... this.
12201 (ipa_set_node_agg_value_chain):
12202 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
12203 (write_ipcp_transformation_info): Likewise.
12204 (read_ipcp_transformation_info): Likewise.
12205 (ipcp_update_bits): Likewise.
12206 (ipcp_update_vr): Likewise.
12207 (ipcp_transform_function): Likewise.
12208 * ipa-prop.h: Rename ipcp_transformation_summary to
12209 ipcp_transformation.
12210 (class ipcp_transformation_t): New function summary.
12211 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
12212 (ipa_get_agg_replacements_for_node): Likewise.
12214 2018-06-08 Martin Liska <mliska@suse.cz>
12216 * ipa-pure-const.c (struct funct_state_d): Do it class instead
12218 (class funct_state_summary_t): New function_summary class.
12219 (has_function_state): Remove.
12220 (get_function_state): Likewise.
12221 (set_function_state): Likewise.
12222 (add_new_function): Likewise.
12223 (funct_state_summary_t::insert): New function.
12224 (duplicate_node_data): Remove.
12225 (remove_node_data): Remove.
12226 (funct_state_summary_t::duplicate): New function.
12227 (register_hooks): Create new funct_state_summaries.
12228 (pure_const_generate_summary): Use it.
12229 (pure_const_write_summary): Likewise.
12230 (pure_const_read_summary): Likewise.
12231 (propagate_pure_const): Likewise.
12232 (propagate_nothrow): Likewise.
12233 (dump_malloc_lattice): Likewise.
12234 (propagate_malloc): Likewise.
12235 (execute): Do not register hooks, just remove summary
12237 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
12240 2018-06-08 Martin Liska <mliska@suse.cz>
12242 * ipa-reference.c (remove_node_data): Remove.
12243 (duplicate_node_data): Likewise.
12244 (class ipa_ref_var_info_summary_t): New class.
12245 (class ipa_ref_opt_summary_t): Likewise.
12246 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
12247 (get_reference_optimization_summary): Use
12248 ipa_ref_opt_sum_summaries.
12249 (set_reference_vars_info): Remove.
12250 (set_reference_optimization_summary): Likewise.
12251 (ipa_init): Create summaries.
12252 (init_function_info): Use function summary.
12253 (ipa_ref_opt_summary_t::duplicate): New function.
12254 (ipa_ref_opt_summary_t::remove): New function.
12255 (get_read_write_all_from_node): Fix GNU coding style.
12256 (propagate): Use function summary.
12257 (write_node_summary_p): Fix GNU coding style.
12258 (stream_out_bitmap): Likewise.
12259 (ipa_reference_read_optimization_summary): Use function summary.
12260 (ipa_reference_c_finalize): Do not release hooks.
12262 2018-06-08 Martin Liska <mliska@suse.cz>
12264 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
12265 (analyze_function_body): Extract multiple calls of get_create.
12266 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12267 * ipa-inline.c (recursive_inlining): Use ::get method.
12268 * ipa-inline.h (estimate_edge_growth): Likewise.
12270 2018-06-08 Martin Liska <mliska@suse.cz>
12272 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
12274 (hsa_function_summary::hsa_function_summary): Use the new enum
12276 (hsa_gpu_implementation_p): Use hsa_summaries::get.
12277 * hsa-gen.c (hsa_get_host_function): Likewise.
12278 (get_brig_function_name): Likewise.
12279 * ipa-hsa.c (process_hsa_functions): Likewise.
12280 (ipa_hsa_write_summary): Likewise.
12281 * symbol-summary.h (symtab_duplication): Use ::get function/
12282 (get): New function.
12284 2018-06-08 Martin Liska <mliska@suse.cz>
12286 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
12288 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
12289 (hsa_register_kernel): Likewise.
12290 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
12291 * hsa-gen.c (hsa_get_host_function): Likewise.
12292 (get_brig_function_name): Likewise.
12293 (generate_hsa): Likewise.
12294 (pass_gen_hsail::execute): Likewise.
12295 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
12296 (devirtualization_time_bonus): Likewise.
12297 (ipcp_propagate_stage): Likewise.
12298 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
12299 (edge_set_predicate): Likewise.
12300 (evaluate_conditions_for_known_args): Likewise.
12301 (evaluate_properties_for_edge): Likewise.
12302 (ipa_fn_summary::reset): Likewise.
12303 (ipa_fn_summary_t::duplicate): Likewise.
12304 (dump_ipa_call_summary): Likewise.
12305 (ipa_dump_fn_summary): Likewise.
12306 (analyze_function_body): Likewise.
12307 (compute_fn_summary): Likewise.
12308 (estimate_edge_devirt_benefit): Likewise.
12309 (estimate_edge_size_and_time): Likewise.
12310 (estimate_calls_size_and_time): Likewise.
12311 (estimate_node_size_and_time): Likewise.
12312 (inline_update_callee_summaries): Likewise.
12313 (remap_edge_change_prob): Likewise.
12314 (remap_edge_summaries): Likewise.
12315 (ipa_merge_fn_summary_after_inlining): Likewise.
12316 (ipa_update_overall_fn_summary): Likewise.
12317 (read_ipa_call_summary): Likewise.
12318 (inline_read_section): Likewise.
12319 (write_ipa_call_summary): Likewise.
12320 (ipa_fn_summary_write): Likewise.
12321 (ipa_free_fn_summary): Likewise.
12322 * ipa-hsa.c (process_hsa_functions): Likewise.
12323 (ipa_hsa_write_summary): Likewise.
12324 (ipa_hsa_read_section): Likewise.
12325 * ipa-icf.c (sem_function::merge): Likewise.
12326 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
12327 (do_estimate_edge_time): Likewise.
12328 (estimate_size_after_inlining): Likewise.
12329 (estimate_growth): Likewise.
12330 (growth_likely_positive): Likewise.
12331 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
12332 (inline_call): Likewise.
12333 * ipa-inline.c (caller_growth_limits): Likewise.
12334 (can_inline_edge_p): Likewise.
12335 (can_inline_edge_by_limits_p): Likewise.
12336 (compute_uninlined_call_time): Likewise.
12337 (compute_inlined_call_time): Likewise.
12338 (want_inline_small_function_p): Likewise.
12339 (edge_badness): Likewise.
12340 (update_caller_keys): Likewise.
12341 (update_callee_keys): Likewise.
12342 (recursive_inlining): Likewise.
12343 (inline_small_functions): Likewise.
12344 (inline_to_all_callers_1): Likewise.
12345 (dump_overall_stats): Likewise.
12346 (early_inline_small_functions): Likewise.
12347 (early_inliner): Likewise.
12348 * ipa-inline.h (estimate_edge_growth): Likewise.
12349 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
12350 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
12351 * ipa-prop.h (IPA_NODE_REF): Likewise.
12352 (IPA_EDGE_REF): Likewise.
12353 * ipa-pure-const.c (malloc_candidate_p): Likewise.
12354 (propagate_malloc): Likewise.
12355 * ipa-split.c (execute_split_functions): Likewise.
12356 * symbol-summary.h: Rename get to get_create.
12358 (get_create): Likewise.
12359 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
12361 2018-06-08 Martin Liska <mliska@suse.cz>
12363 * symbol-summary.h (release): Move definition out of class
12365 (symtab_removal): Likewise.
12366 (symtab_duplication): Likewise.
12368 2018-06-08 Martin Liska <mliska@suse.cz>
12370 * symbol-summary.h (function_summary): Move constructor
12371 implementation out of class declaration.
12372 (release): Likewise.
12373 (symtab_insertion): Likewise.
12374 (symtab_removal): Likewise.
12375 (symtab_duplication): Likewise.
12378 2018-06-08 Martin Liska <mliska@suse.cz>
12380 * Makefile.in: Remove support for MPX (macros, related functions,
12381 fields in cgraph_node, ...).
12382 * builtin-types.def (BT_BND): Likewise.
12383 (BT_FN_BND_CONST_PTR): Likewise.
12384 (BT_FN_CONST_PTR_BND): Likewise.
12385 (BT_FN_VOID_PTR_BND): Likewise.
12386 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
12387 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
12388 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
12389 (expand_builtin_mempcpy_with_bounds): Likewise.
12390 (expand_builtin_memset_with_bounds): Likewise.
12391 (expand_builtin_memset_args): Likewise.
12392 (std_expand_builtin_va_start): Likewise.
12393 (expand_builtin): Likewise.
12394 (expand_builtin_with_bounds): Likewise.
12395 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
12396 (DEF_LIB_BUILTIN_CHKP): Likewise.
12397 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
12398 (DEF_CHKP_BUILTIN): Likewise.
12399 (BUILT_IN_MEMCPY): Likewise.
12400 (BUILT_IN_MEMMOVE): Likewise.
12401 (BUILT_IN_MEMPCPY): Likewise.
12402 (BUILT_IN_MEMSET): Likewise.
12403 (BUILT_IN_STPCPY): Likewise.
12404 (BUILT_IN_STRCAT): Likewise.
12405 (BUILT_IN_STRCHR): Likewise.
12406 (BUILT_IN_STRCPY): Likewise.
12407 (BUILT_IN_STRLEN): Likewise.
12408 (BUILT_IN_MEMCPY_CHK): Likewise.
12409 (BUILT_IN_MEMMOVE_CHK): Likewise.
12410 (BUILT_IN_MEMPCPY_CHK): Likewise.
12411 (BUILT_IN_MEMSET_CHK): Likewise.
12412 (BUILT_IN_STPCPY_CHK): Likewise.
12413 (BUILT_IN_STRCAT_CHK): Likewise.
12414 (BUILT_IN_STRCPY_CHK): Likewise.
12415 * calls.c (store_bounds): Likewise.
12416 (emit_call_1): Likewise.
12417 (special_function_p): Likewise.
12418 (maybe_warn_nonstring_arg): Likewise.
12419 (initialize_argument_information): Likewise.
12420 (finalize_must_preallocate): Likewise.
12421 (compute_argument_addresses): Likewise.
12422 (expand_call): Likewise.
12423 * cfgexpand.c (expand_call_stmt): Likewise.
12424 (expand_return): Likewise.
12425 (expand_gimple_stmt_1): Likewise.
12426 (pass_expand::execute): Likewise.
12427 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
12428 (cgraph_node::remove): Likewise.
12429 (cgraph_node::dump): Likewise.
12430 (cgraph_node::verify_node): Likewise.
12431 * cgraph.h (chkp_function_instrumented_p): Likewise.
12432 (symtab_node::get_alias_target): Likewise.
12433 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
12434 (cgraph_local_p): Likewise.
12435 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
12436 (cgraph_edge::rebuild_references): Likewise.
12437 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
12438 (walk_polymorphic_call_targets): Likewise.
12439 (cgraph_node::expand_thunk): Likewise.
12440 (symbol_table::output_weakrefs): Likewise.
12441 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
12442 (ix86_handle_option): Likewise.
12443 * config/i386/constraints.md: Likewise.
12444 * config/i386/i386-builtin-types.def (BND): Likewise.
12448 * config/i386/i386-builtin.def (BDESC_END): Likewise.
12449 (BDESC_FIRST): Likewise.
12451 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
12452 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
12453 * config/i386/i386.c (enum reg_class): Likewise.
12454 (ix86_target_string): Likewise.
12455 (ix86_option_override_internal): Likewise.
12456 (ix86_conditional_register_usage): Likewise.
12457 (ix86_valid_target_attribute_inner_p): Likewise.
12458 (ix86_set_indirect_branch_type): Likewise.
12459 (ix86_set_current_function): Likewise.
12460 (ix86_function_arg_regno_p): Likewise.
12461 (init_cumulative_args): Likewise.
12462 (ix86_function_arg_advance): Likewise.
12463 (ix86_function_arg): Likewise.
12464 (ix86_pass_by_reference): Likewise.
12465 (ix86_function_value_regno_p): Likewise.
12466 (ix86_function_value_1): Likewise.
12467 (ix86_function_value_bounds): Likewise.
12468 (ix86_return_in_memory): Likewise.
12469 (ix86_setup_incoming_vararg_bounds): Likewise.
12470 (ix86_va_start): Likewise.
12471 (indirect_thunk_need_prefix): Likewise.
12472 (print_reg): Likewise.
12473 (ix86_print_operand): Likewise.
12474 (ix86_expand_call): Likewise.
12475 (ix86_output_function_return): Likewise.
12476 (reg_encoded_number): Likewise.
12477 (BDESC_VERIFYS): Likewise.
12478 (ix86_init_mpx_builtins): Likewise.
12479 (ix86_init_builtins): Likewise.
12480 (ix86_emit_cmove): Likewise.
12481 (ix86_emit_move_max): Likewise.
12482 (ix86_expand_builtin): Likewise.
12483 (ix86_builtin_mpx_function): Likewise.
12484 (ix86_get_arg_address_for_bt): Likewise.
12485 (ix86_load_bounds): Likewise.
12486 (ix86_store_bounds): Likewise.
12487 (ix86_load_returned_bounds): Likewise.
12488 (ix86_store_returned_bounds): Likewise.
12489 (ix86_class_likely_spilled_p): Likewise.
12490 (ix86_hard_regno_mode_ok): Likewise.
12491 (x86_order_regs_for_local_alloc): Likewise.
12492 (ix86_mitigate_rop): Likewise.
12493 (ix86_bnd_prefixed_insn_p): Likewise.
12494 (ix86_mpx_bound_mode): Likewise.
12495 (ix86_make_bounds_constant): Likewise.
12496 (ix86_initialize_bounds): Likewise.
12497 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
12498 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
12499 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
12500 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
12501 (TARGET_CHKP_BOUND_MODE): Likewise.
12502 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
12503 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
12504 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
12505 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
12506 * config/i386/i386.h (TARGET_MPX): Likewise.
12507 (TARGET_MPX_P): Likewise.
12508 (VALID_BND_REG_MODE): Likewise.
12509 (FIRST_BND_REG): Likewise.
12510 (LAST_BND_REG): Likewise.
12511 (enum reg_class): Likewise.
12512 (BND_REG_P): Likewise.
12513 (BND_REGNO_P): Likewise.
12514 (BNDmode): Likewise.
12515 (ADJUST_INSN_LENGTH): Likewise.
12516 * config/i386/i386.md: Likewise.
12517 * config/i386/i386.opt: Likewise.
12518 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
12519 (defined): Likewise.
12520 (LINK_MPX): Likewise.
12521 (MPX_SPEC): Likewise.
12522 (LIBMPX_SPEC): Likewise.
12523 (LIBMPXWRAPPERS_SPEC): Likewise.
12524 (CHKP_SPEC): Likewise.
12525 * config/i386/predicates.md: Likewise.
12526 * dbxout.c (dbxout_type): Likewise.
12527 * doc/extend.texi: Likewise.
12528 * doc/invoke.texi: Likewise.
12529 * doc/md.texi: Likewise.
12530 * doc/tm.texi: Likewise.
12531 * doc/tm.texi.in: Likewise.
12532 * dwarf2out.c (is_base_type): Likewise.
12533 (gen_formal_types_die): Likewise.
12534 (gen_subprogram_die): Likewise.
12535 (gen_type_die_with_usage): Likewise.
12536 (gen_decl_die): Likewise.
12537 (dwarf2out_late_global_decl): Likewise.
12538 * expr.c (expand_assignment): Likewise.
12539 (emit_storent_insn): Likewise.
12540 (store_expr_with_bounds): Likewise.
12541 (store_expr): Likewise.
12542 (expand_expr_real_1): Likewise.
12543 * expr.h (store_expr_with_bounds): Likewise.
12544 * function.c (use_register_for_decl): Likewise.
12545 (struct bounds_parm_data): Likewise.
12546 (assign_parms_augmented_arg_list): Likewise.
12547 (assign_parm_find_entry_rtl): Likewise.
12548 (assign_parm_is_stack_parm): Likewise.
12549 (assign_parm_load_bounds): Likewise.
12550 (assign_bounds): Likewise.
12551 (assign_parms): Likewise.
12552 (expand_function_start): Likewise.
12553 * gcc.c (CHKP_SPEC): Likewise.
12554 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
12555 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
12556 (wrestrict_dom_walker::check_call): Likewise.
12557 * gimple.c (gimple_build_call_from_tree): Likewise.
12558 * gimple.h (enum gf_mask): Likewise.
12559 (gimple_call_with_bounds_p): Likewise.
12560 (gimple_call_set_with_bounds): Likewise.
12561 * gimplify.c (gimplify_init_constructor): Likewise.
12562 * ipa-cp.c (initialize_node_lattices): Likewise.
12563 (propagate_constants_across_call): Likewise.
12564 (find_more_scalar_values_for_callers_subset): Likewise.
12565 * ipa-hsa.c (process_hsa_functions): Likewise.
12566 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
12567 * ipa-icf.c (sem_function::merge): Likewise.
12568 * ipa-inline.c (early_inliner): Likewise.
12569 * ipa-pure-const.c (warn_function_noreturn): Likewise.
12570 (warn_function_cold): Likewise.
12571 (propagate_pure_const): Likewise.
12572 * ipa-ref.h (enum GTY): Likewise.
12573 * ipa-split.c (find_retbnd): Likewise.
12574 (consider_split): Likewise.
12575 (split_function): Likewise.
12576 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
12577 * ipa.c (walk_polymorphic_call_targets): Likewise.
12578 (symbol_table::remove_unreachable_nodes): Likewise.
12579 (process_references): Likewise.
12580 (cgraph_build_static_cdtor_1): Likewise.
12581 * lto-cgraph.c (lto_output_node): Likewise.
12582 (output_refs): Likewise.
12583 (compute_ltrans_boundary): Likewise.
12584 (input_overwrite_node): Likewise.
12585 (input_node): Likewise.
12586 (input_cgraph_1): Likewise.
12587 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
12588 * passes.c (pass_manager::execute_early_local_passes): Likewise.
12589 (class pass_chkp_instrumentation_passes): Likewise.
12590 (make_pass_chkp_instrumentation_passes): Likewise.
12591 * passes.def: Likewise.
12592 * rtl.h (struct GTY): Likewise.
12593 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
12594 * stor-layout.c (layout_type): Likewise.
12595 * symtab.c: Likewise.
12596 * target.def: Likewise.
12597 * targhooks.c (default_chkp_bound_type): Likewise.
12598 (default_chkp_bound_mode): Likewise.
12599 (default_builtin_chkp_function): Likewise.
12600 (default_chkp_function_value_bounds): Likewise.
12601 (default_chkp_make_bounds_constant): Likewise.
12602 (default_chkp_initialize_bounds): Likewise.
12603 * targhooks.h (default_chkp_bound_type): Likewise.
12604 (default_chkp_bound_mode): Likewise.
12605 (default_builtin_chkp_function): Likewise.
12606 (default_chkp_function_value_bounds): Likewise.
12607 (default_chkp_make_bounds_constant): Likewise.
12608 (default_chkp_initialize_bounds): Likewise.
12609 * toplev.c (compile_file): Likewise.
12610 (process_options): Likewise.
12611 * tree-core.h (DEF_BUILTIN): Likewise.
12612 (DEF_BUILTIN_CHKP): Likewise.
12613 * tree-inline.c (declare_return_variable): Likewise.
12614 (remap_gimple_stmt): Likewise.
12615 (copy_bb): Likewise.
12616 (initialize_inlined_parameters): Likewise.
12617 (expand_call_inline): Likewise.
12618 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
12619 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
12620 (make_pass_ipa_chkp_produce_thunks): Likewise.
12621 (make_pass_chkp): Likewise.
12622 (make_pass_chkp_opt): Likewise.
12623 (make_pass_chkp_instrumentation_passes): Likewise.
12624 * tree-pretty-print.c (dump_generic_node): Likewise.
12625 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
12626 * tree-ssa-dce.c (propagate_necessity): Likewise.
12627 (eliminate_unnecessary_stmts): Likewise.
12628 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
12629 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
12630 * tree-ssa-sccvn.h: Likewise.
12631 * tree-ssa-strlen.c (get_string_length): Likewise.
12632 (valid_builtin_call): Likewise.
12633 (adjust_last_stmt): Likewise.
12634 (handle_builtin_strchr): Likewise.
12635 (handle_builtin_strcpy): Likewise.
12636 (handle_builtin_stxncpy): Likewise.
12637 (handle_builtin_memcpy): Likewise.
12638 (handle_builtin_strcat): Likewise.
12639 (strlen_check_and_optimize_stmt): Likewise.
12640 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
12641 * tree-streamer-in.c: Likewise.
12642 * tree-streamer.c (record_common_node): Likewise.
12643 * tree.c (tree_code_size): Likewise.
12644 (wide_int_to_tree_1): Likewise.
12645 (type_contains_placeholder_1): Likewise.
12646 (build_common_tree_nodes): Likewise.
12647 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
12648 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
12649 (POINTER_BOUNDS_P): Likewise.
12650 (BOUNDED_TYPE_P): Likewise.
12651 (BOUNDED_P): Likewise.
12652 (CALL_WITH_BOUNDS_P): Likewise.
12653 (pointer_bounds_type_node): Likewise.
12654 * value-prof.c (gimple_ic): Likewise.
12655 * var-tracking.c (vt_add_function_parameters): Likewise.
12656 * varasm.c (make_decl_rtl): Likewise.
12657 (assemble_start_function): Likewise.
12658 (output_constant): Likewise.
12659 (maybe_assemble_visibility): Likewise.
12660 * varpool.c (ctor_for_folding): Likewise.
12661 * chkp-builtins.def: Remove.
12662 * ipa-chkp.c: Remove.
12663 * ipa-chkp.h: Remove.
12664 * rtl-chkp.c: Remove.
12665 * rtl-chkp.h: Remove.
12666 * tree-chkp-opt.c: Remove.
12667 * tree-chkp.c: Remove.
12668 * tree-chkp.h: Remove.
12670 2018-06-07 Carl Love <cel@us.ibm.com>
12672 * config/rs6000/vsx.md (vextract_fp_from_shorth,
12673 vextract_fp_from_shortl): Add BE support.
12675 2018-06-07 Paul Koning <ni1d@arrl.net>
12677 * compare-elim.c (try_merge_compare): Don't merge compare if
12678 address contains a side effect.
12679 (try_eliminate_compare): Likewise.
12681 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
12683 * config.gcc: Support "tremont".
12684 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
12685 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
12687 * config/i386/i386.c (m_TREMONT): Define.
12688 (processor_target_table): Add "tremont".
12689 (PTA_TREMONT): Define.
12690 (ix86_lea_outperforms): Add TARGET_TREMONT.
12691 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
12692 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
12693 and M_INTEL_GOLDMONT_PLUS.
12694 (fold_builtin_cpu): Add "tremont".
12695 (ix86_add_stmt_cost): Add TARGET_TREMONT.
12696 (ix86_option_override_internal): Add "tremont".
12697 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
12698 (processor_type): Add PROCESSOR_TREMONT.
12699 * config/i386/x86-tune.def: Add m_TREMONT.
12700 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
12702 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12704 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
12705 symbol defined for msp430i* devices to be lower case.
12707 2018-06-07 Richard Biener <rguenther@suse.de>
12709 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
12710 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
12711 Properly wrap signed arithmetic if overflow wraps.
12713 2018-06-07 Jakub Jelinek <jakub@redhat.com>
12715 PR tree-optimization/69615
12716 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
12717 of a cast from a same precision integral SSA_NAME in a bb dominated
12718 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
12719 cast to utype if rhs2 has already a compatible type.
12721 2018-06-07 Richard Biener <rguenther@suse.de>
12723 PR tree-optimization/85935
12724 * graphite-scop-detection.c (find_params_in_bb): Analyze
12725 condition operands with respect to the correct loop. Assert
12726 the analysis doesn't fail.
12728 2018-06-04 Carl Love <cel@us.ibm.com>
12730 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
12731 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
12732 as it is slightly cheaper.
12733 (first_match_or_eos_index_<mode>):
12734 Calculate index using natural element order.
12735 (first_match_index_<mode>):
12736 Calculate index using natural element order.
12737 (first_match_or_eos_index_<mode>):
12738 Calculate index using natural order.
12739 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
12740 for BE and LE modes.
12741 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
12742 P9V_BUILTIN_VCLZLSBB_V16QI.
12743 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
12746 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12748 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
12749 indentation and line wrap for many prototypes. Add missing
12750 @smallexample directives around block of prototypes for vec_xl and
12753 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
12755 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
12756 track if we pass or return IEEE 128-bit floating point.
12757 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
12758 C++ mangling that is compatible with GCC 8.1.
12759 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
12760 (init_cumulative_args): Note if we pass or return IEEE 128-bit
12761 floating point types.
12762 (rs6000_function_arg_advance_1): Likewise.
12763 (rs6000_mangle_type): Optionally generate mangled names that match
12764 what GCC 8.1 generated for IEEE 128-bit floating point types.
12765 (rs6000_globalize_decl_name): If we have an external function that
12766 passes or returns IEEE 128-bit types, generate a weak reference
12767 from the mangled name used in GCC 8.1 to the current mangled
12769 (rs6000_init_builtins): Make __ibm128 use the long double type if
12770 long double is IBM extended double. Make __float128 use the long
12771 double type if long double is IEEE 128-bit.
12774 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
12775 macro for __ibm128 built-in functions.
12776 (PACK_IF): Add __ibm128 pack/unpack functions.
12777 (UNPACK_IF): Likewise.
12778 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
12779 enable long double built-in functions if long double is IEEE
12780 128-bit floating point.
12781 (rs6000_invalid_builtin): Update long double built-in function
12783 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
12784 functions, adjust the built-in function to use the long double
12785 built-in function if __ibm128 and long double are the same type.
12786 * doc/extend.texi (PowerPC builtins): Update documention for
12787 __builtin_{,un}pack_longdouble. Add documentation for
12788 __builtin_{,un}pack_ibm128.
12790 2018-06-06 Jim Wilson <jimw@sifive.com>
12792 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
12793 (struct machine_function): New field interrupt_mode.
12794 (riscv_handle_type_attribute): New function. Add forward declaration.
12795 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
12796 (riscv_expand_epilogue): Check interrupt_mode field.
12797 (riscv_set_current_function): Check interrupt attribute args and
12798 set interrupt_mode field.
12799 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
12800 (riscv_sret, riscv_uret): New.
12801 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
12802 new arguments to interrupt attribute.
12804 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
12807 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
12808 Don't handle -mcpu=power8 if -mpower9-vector is also used.
12810 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
12812 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
12813 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
12814 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
12815 several redundant entries.
12817 2018-06-06 David Malcolm <dmalcolm@redhat.com>
12819 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
12820 type from "rtx" to "rtx_insn *".
12821 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
12822 for local "call_insn", removing cast.
12823 (ix86_expand_call): Likewise, introducing a "call_insn" local.
12825 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
12827 PR tree-optimization/86066
12828 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
12829 for BIT_INSERT_EXPR stores.
12831 2018-06-06 Richard Biener <rguenther@suse.de>
12833 PR tree-optimization/86062
12834 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
12835 component refs ontop
12836 of to be offsetted base.
12838 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12840 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
12841 to be static and remove check on interrupt attribute name.
12843 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
12845 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12846 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
12848 2018-06-05 Steve Ellcey <sellcey@cavium.com>
12851 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
12853 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
12854 Remove second argument, change how error is called.
12855 (aarch64_layout_arg): Remove second argument from
12856 aarch64_err_no_fpadvsimd call.
12857 (aarch64_init_cumulative_args): Ditto.
12858 (aarch64_gimplify_va_arg_expr): Ditto.
12859 * config/aarch64/aarch64.md (mov<mode>): Ditto.
12861 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
12863 * config/i386/i386.md (simple_return_indirect_internal): New expander.
12864 (*simple_return_indirect_internal<mode>): Rename from
12865 simple_return_indirect_internal. Use W mode iterator.
12866 (rstorssp): New expander.
12867 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
12868 (clrssbsy): New expander.
12869 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
12871 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
12873 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
12875 (cmse_check_pointed_object): Likewise.
12877 2018-06-05 Martin Liska <mliska@suse.cz>
12879 PR gcov-profile/47618
12880 * doc/invoke.texi: Document how -fprofile-dir format
12883 2018-06-05 Richard Biener <rguenther@suse.de>
12885 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
12886 removal pretend DOM info isn't available so we do not update
12887 it and only remove edges, not dominated blocks. Actually free
12888 DOM info in case we removed something. Remove unreachable blocks.
12889 (mfb_keep_latches): Work with either DOM info or marked backedges.
12890 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
12891 first. Mark backedges if DOM info isn't available.
12892 (Re-)compute DOM info after cleanup_control_flow_pre.
12894 2018-06-05 Richard Biener <rguenther@suse.de>
12896 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
12897 (locus_discrim_hasher::hash): Adjust.
12898 (locus_discrim_hasher::equal): Likewise.
12899 (next_discriminator_for_locus): Work on line directly.
12900 (same_line_p): Pass in expanded locus1 as well.
12901 (assign_discriminators): Avoid redundant location expansions.
12903 2018-06-05 Richard Biener <rguenther@suse.de>
12905 PR tree-optimization/86046
12906 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
12907 if required after clearing TREE_ADDRESSABLE.
12909 2018-06-05 Richard Biener <rguenther@suse.de>
12911 PR tree-optimization/86047
12912 * tree-ssa-loop.c (for_each_index): Glob handling of all
12913 decls and constants and really handle all of them.
12915 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12918 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
12919 qualifier_void_pointer and qualifier_const_void_pointer.
12920 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
12921 (arm_init_builtins): Handle the above.
12922 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
12923 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
12924 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
12927 2018-06-05 Martin Liska <mliska@suse.cz>
12929 * auto-profile.c (read_autofdo_file): Do not use
12930 gcov_ctr_summary struct.
12931 (afdo_callsite_hot_enough_for_early_inline): Likewise.
12932 * coverage.c (struct counts_entry): Likewise.
12933 (read_counts_file): Read just single summary entry.
12934 (get_coverage_counts): Use gcov_summary struct.
12935 * coverage.h (get_coverage_counts): Likewise.
12936 * gcov-dump.c (dump_working_sets): Likewise.
12937 (tag_summary): Dump just single summary.
12938 * gcov-io.c (gcov_write_summary): Write just histogram
12940 (gcov_read_summary): Read just single summary.
12941 (compute_working_sets): Use gcov_summary struct.
12942 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
12943 of GCOV_COUNTERS_SUMMABLE.
12944 (GCOV_COUNTERS_SUMMABLE): Remove.
12945 (GCOV_FIRST_VALUE_COUNTER): Replace with
12946 GCOV_COUNTER_V_INTERVAL.
12947 (struct gcov_ctr_summary): Remove.
12948 (struct gcov_summary): Directly use fields of former
12950 (compute_working_sets): Use gcov_summary struct.
12951 * gcov.c (read_count_file): Do not use ctrs fields.
12952 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
12954 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
12956 * profile.c: Likewise.
12957 * profile.h: Likewise.
12959 2018-06-05 Martin Liska <mliska@suse.cz>
12961 PR gcov-profile/84846
12962 * gcov.c (output_lines): Print working directory only
12963 in intermediate format.
12965 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
12967 * config/s390/s390-builtin-types.def: Add void function type.
12968 * config/s390/s390-builtins.def: Use the function type for the
12971 2018-06-04 Jim Wilson <jimw@sifive.com>
12973 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12975 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12976 and maybe_eh_return. Change regno to unsigned int. Use new args to
12977 handle EH_RETURN_DATA_REGNO registers properly.
12978 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12979 (riscv_expand_epilogue): Update comment. Change argument name and
12980 type. Update code to use new name and type. Pass new args to
12981 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
12983 * config/riscv/riscv.md (NORMAL_RETURN): New.
12984 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12985 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12986 (eh_return): Call gen_eh_return_internal and emit barrier.
12987 (eh_return_internal): Call riscv_expand_epilogue.
12989 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
12991 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12992 bit_insertion field and declare can_be_merged_into method.
12993 (merged_store_group::can_be_merged_into): New method.
12994 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12995 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
12996 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12998 2018-06-04 Richard Biener <rguenther@suse.de>
13000 PR tree-optimization/85955
13001 * builtins.c (fold_builtin_sincos): Convert pointers to
13002 destination to appropriate type before dereferencing.
13004 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
13006 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
13008 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
13010 * expr.c (expand_expr_real_1): Force the operand into memory if
13011 its TYPE_MODE is BLKmode and if there is no integer mode for
13012 the number of bits being extracted.
13014 2018-06-04 Jakub Jelinek <jakub@redhat.com>
13018 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13019 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
13021 2018-06-04 Richard Biener <rguenther@suse.de>
13023 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
13024 (cleanup_tree_cfg_noloop): ... single caller. Do
13025 start_recording_case_labels later.
13027 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
13029 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
13030 to _IMMINTRIN_H_INCLUDED.
13031 * config/i386/pconfigintrin.h: Ditto.
13032 * config/i386/waitpkgintrin.h: Ditto.
13033 * config/i386/immintrin.h: Add includes for sgxintrin.h,
13034 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
13035 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
13036 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
13037 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
13038 waitpkgintrin.h and cldemoteintrin.h.
13040 2018-06-04 Richard Biener <rguenther@suse.de>
13042 PR tree-optimization/86038
13043 * tracer.c (find_best_successor): Check probability for
13044 being initialized, bail out if not.
13046 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
13049 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
13050 of bits to ignore when comparing architectures.
13052 2018-06-04 Jakub Jelinek <jakub@redhat.com>
13054 PR tree-optimization/69615
13055 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
13056 maximum or minimum of the type, try to merge it also as if
13057 range1 is + [-, x - 1] or + [x + 1, -].
13060 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
13062 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
13064 PR tree-optimization/86034
13065 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
13066 the unsigned bitfield type in a bit insertion sequence if it does not
13067 have a larger precision than the bitfield size.
13068 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
13070 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
13072 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
13074 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13076 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
13077 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
13078 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
13079 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
13081 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
13083 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
13084 Disable -fdelete-null-pointer-checks for ELF toolchain.
13086 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13087 Kito Cheng <kito.cheng@gmail.com>
13089 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
13090 (nds32le-*-*, nds32be-*-*): Integrate checking process.
13091 (nds32*-*-*): Add glibc and uclibc conditions.
13092 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
13093 (TARGET_EXCEPT_UNWIND_INFO): Define.
13094 * config/nds32/elf.h: New file.
13095 * config/nds32/linux.h: New file.
13096 * config/nds32/nds32-elf.opt: New file.
13097 * config/nds32/nds32-linux.opt: New file.
13098 * config/nds32/nds32-fp-as-gp.c
13099 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
13100 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
13102 (nds32_asm_file_end): Ditto.
13103 (nds32_print_operand): Ditto.
13104 (nds32_insert_attributes): Ditto.
13105 (nds32_init_libfuncs): New function.
13106 (TARGET_HAVE_TLS): Define.
13107 (TARGET_INIT_LIBFUNCS): Define.
13108 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
13110 (TARGET_ELF): Apply different mcmodel setting.
13111 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
13112 been migrated into elf.h and linux.h files.
13113 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
13114 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
13115 (mcmodel): The content has been migrated into nds32-elf.opt and
13116 nds32-linux.opt files.
13117 * config/nds32/t-elf: New file.
13118 * config/nds32/t-linux: New file.
13120 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
13121 Shiva Chen <shiva0217@gmail.com>
13123 * config/nds32/constants.md (unspec_volatile_element): Add
13124 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
13125 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
13127 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
13128 (make_pass_nds32_fp_as_gp): Declare.
13129 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
13131 (nds32_asm_function_end_prologue): Remove unused asm output.
13132 (nds32_asm_function_begin_epilogue): Remove unused asm output.
13133 (nds32_asm_file_start): Output necessary fp_as_gp information.
13134 (nds32_option_override): Adjust register usage.
13135 (nds32_expand_prologue): Consider fp_as_gp situation.
13136 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
13137 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
13140 (simple_return): Ditto.
13141 (omit_fp_begin): Output special directive for fp_as_gp.
13142 (omit_fp_end): Output special directive for fp_as_gp.
13143 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
13144 mforbid-fp-as-gp): New options.
13146 2018-06-01 Mark Wielaard <mark@klomp.org>
13148 * dwarf2out.c (dwarf2out_finish): Remove generation of
13149 DW_AT_loclists_base.
13151 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
13153 * gimple-ssa-store-merging.c: Include gimple-fold.h.
13154 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
13155 (struct merged_store_group): Add bit_insertion field.
13156 (dump_char_array): Use standard hexadecimal format.
13157 (merged_store_group::merged_store_group): Set bit_insertion to false.
13158 (merged_store_group::apply_stores): Use optimal buffer size. Deal
13159 with BIT_INSERT_EXPR stores. Move up code updating the mask and
13160 also print the mask in the dump file.
13161 (pass_store_merging::gate): Minor tweak.
13162 (imm_store_chain_info::coalesce_immediate): Fix wrong association
13163 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
13164 stores with INTEGER_CST stores.
13165 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
13166 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
13167 and use it throughout. Generate bit insertion sequences if need be.
13168 (pass_store_merging::process_store): Remove redundant condition.
13169 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
13171 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
13173 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
13174 the 128-bit floating point types. Fix function comment.
13176 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13178 * config/aarch64/aarch64-simd.md
13179 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
13181 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
13184 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13186 PR tree-optimization/85989
13187 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
13189 (backprop::intersect_uses): Check it when deciding whether this
13190 is a backedge reference.
13191 (backprop::process_block): Add each phi to m_visited_phis
13192 after visiting it, then clear it at the end.
13194 2018-06-01 Richard Biener <rguenther@suse.de>
13196 * tree-vectorizer.h (vect_dr_stmt): New function.
13197 (vect_get_load_cost): Adjust.
13198 (vect_get_store_cost): Likewise.
13199 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
13200 Use vect_dr_stmt instead of DR_SMTT.
13201 (vect_record_base_alignments): Likewise.
13202 (vect_calculate_target_alignment): Likewise.
13203 (vect_compute_data_ref_alignment): Likewise and make static.
13204 (vect_update_misalignment_for_peel): Likewise.
13205 (vect_verify_datarefs_alignment): Likewise.
13206 (vector_alignment_reachable_p): Likewise.
13207 (vect_get_data_access_cost): Likewise. Pass down
13208 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
13209 (vect_get_peeling_costs_all_drs): Likewise.
13210 (vect_peeling_hash_get_lowest_cost): Likewise.
13211 (vect_enhance_data_refs_alignment): Likewise.
13212 (vect_find_same_alignment_drs): Likewise.
13213 (vect_analyze_data_refs_alignment): Likewise.
13214 (vect_analyze_group_access_1): Likewise.
13215 (vect_analyze_group_access): Likewise.
13216 (vect_analyze_data_ref_access): Likewise.
13217 (vect_analyze_data_ref_accesses): Likewise.
13218 (vect_vfa_segment_size): Likewise.
13219 (vect_small_gap_p): Likewise.
13220 (vectorizable_with_step_bound_p): Likewise.
13221 (vect_prune_runtime_alias_test_list): Likewise.
13222 (vect_analyze_data_refs): Likewise.
13223 (vect_supportable_dr_alignment): Likewise.
13224 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
13225 (vect_gen_prolog_loop_niters): Likewise.
13226 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
13227 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
13229 (vect_recog_mask_conversion_pattern): Likewise.
13230 (vect_try_gather_scatter_pattern): Likewise.
13231 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
13232 to vect_get_store_cost.
13233 (vect_get_store_cost): Get stmt_info instead of DR.
13234 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
13235 (vect_get_load_cost): Get stmt_info instead of DR.
13237 2018-06-01 Richard Biener <rguenther@suse.de>
13239 PR middle-end/86017
13240 * gimple-fold.c (var_decl_component_p): Also allow offsetted
13241 vars wrapped in MEM_REFs.
13243 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
13245 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13246 Fix subreg tests so that we only return a choice between
13247 GENERAL_REGS and FP_REGS if the original classes included both.
13249 2018-06-01 Richard Biener <rguenther@suse.de>
13252 * tree-ssa-structalias.c (get_function_part_constraint):
13253 Handle NULL fi->decl.
13254 (find_func_aliases_for_call): Properly handle indirect
13255 fi from direct call.
13256 (find_func_clobbers): Likewise.
13257 (ipa_pta_execute): Likewise.
13258 (create_variable_info_for): For functions that are ifunc_resolver
13259 resolve to a varinfo that contains the result of the resolver call.
13260 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
13263 2018-05-31 Michael Collison <michael.collison@arm.com>
13265 * config/aarch64/aarch64.md:
13266 (*fix_to_zero_extenddfdi2): New pattern.
13267 * gcc.target/aarch64/fix_extend1.c: New testcase.
13269 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
13271 PR middle-end/78809
13272 PR middle-end/83026
13273 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
13274 and BUILT_IN_STRNCMP_EQ.
13275 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
13276 BUILT_IN_STRNCMP_EQ.
13277 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
13278 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13279 (gimple_fold_builtin): Likewise.
13280 * tree-ssa-strlen.c (compute_string_length): New function.
13281 (determine_min_obsize): New function.
13282 (handle_builtin_string_cmp): New function to handle calls to
13283 string compare functions.
13284 (strlen_optimize_stmt): Add handling to builtin string compare
13286 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
13287 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
13288 * tree.c (build_common_builtin_nodes): Add new defines of
13289 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
13291 2018-05-31 Jakub Jelinek <jakub@redhat.com>
13294 * bb-reorder.c (pass_partition_blocks::gate): Return false for
13295 functions with naked attribute.
13297 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
13299 * config/i386/sse.md (avx_vec_concat<mode>):
13300 Substitute concat_tg_mode mode attribute with xtg_mode.
13301 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
13302 (concat_tg_mode): Remove mode attribute.
13304 2018-05-31 Martin Sebor <msebor@redhat.com>
13307 * calls.c (alloc_max_size): Correct a logic error/typo.
13308 Treat excessive arguments as infinite. Warn for invalid arguments.
13309 * doc/invoke.texi (-Walloc-size-larger-than): Update.
13311 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
13314 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
13315 and movx for Haswell.
13317 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
13318 Cesar Philippidis <cesar@codesourcery.com>
13320 PR middle-end/85879
13321 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
13322 when emitting error on private/firstprivate reductions.
13323 * omp-low.c (lower_omp_target): Avoid reference-type processing
13324 on pointers for firstprivate clause.
13326 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13328 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
13331 * config/aarch64/aarch64-simd.md
13332 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
13333 (aarch64_ld1_x3_<mode>): Likewise
13334 (aarch64_st1x2<VALLDIF:mode>): Likewise
13335 (aarch64_st1_x2_<mode>): Likewise
13336 (aarch64_st1x3<VALLDIF:mode>): Likewise
13337 (aarch64_st1_x3_<mode>): Likewise
13338 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
13339 (vld1_s8_x3): Likewise.
13340 (vld1_u16_x3): Likewise.
13341 (vld1_s16_x3): Likewise.
13342 (vld1_u32_x3): Likewise.
13343 (vld1_s32_x3): Likewise.
13344 (vld1_u64_x3): Likewise.
13345 (vld1_s64_x3): Likewise.
13346 (vld1_f16_x3): Likewise.
13347 (vld1_f32_x3): Likewise.
13348 (vld1_f64_x3): Likewise.
13349 (vld1_p8_x3): Likewise.
13350 (vld1_p16_x3): Likewise.
13351 (vld1_p64_x3): Likewise.
13352 (vld1q_u8_x3): Likewise.
13353 (vld1q_s8_x3): Likewise.
13354 (vld1q_u16_x3): Likewise.
13355 (vld1q_s16_x3): Likewise.
13356 (vld1q_u32_x3): Likewise.
13357 (vld1q_s32_x3): Likewise.
13358 (vld1q_u64_x3): Likewise.
13359 (vld1q_s64_x3): Likewise.
13360 (vld1q_f16_x3): Likewise.
13361 (vld1q_f32_x3): Likewise.
13362 (vld1q_f64_x3): Likewise.
13363 (vld1q_p8_x3): Likewise.
13364 (vld1q_p16_x3): Likewise.
13365 (vld1q_p64_x3): Likewise.
13366 (vst1_s64_x2): Likewise.
13367 (vst1_u64_x2): Likewise.
13368 (vst1_f64_x2): Likewise.
13369 (vst1_s8_x2): Likewise.
13370 (vst1_p8_x2): Likewise.
13371 (vst1_s16_x2): Likewise.
13372 (vst1_p16_x2): Likewise.
13373 (vst1_s32_x2): Likewise.
13374 (vst1_u8_x2): Likewise.
13375 (vst1_u16_x2): Likewise.
13376 (vst1_u32_x2): Likewise.
13377 (vst1_f16_x2): Likewise.
13378 (vst1_f32_x2): Likewise.
13379 (vst1_p64_x2): Likewise.
13380 (vst1q_s8_x2): Likewise.
13381 (vst1q_p8_x2): Likewise.
13382 (vst1q_s16_x2): Likewise.
13383 (vst1q_p16_x2): Likewise.
13384 (vst1q_s32_x2): Likewise.
13385 (vst1q_s64_x2): Likewise.
13386 (vst1q_u8_x2): Likewise.
13387 (vst1q_u16_x2): Likewise.
13388 (vst1q_u32_x2): Likewise.
13389 (vst1q_u64_x2): Likewise.
13390 (vst1q_f16_x2): Likewise.
13391 (vst1q_f32_x2): Likewise.
13392 (vst1q_f64_x2): Likewise.
13393 (vst1q_p64_x2): Likewise.
13394 (vst1_s64_x3): Likewise.
13395 (vst1_u64_x3): Likewise.
13396 (vst1_f64_x3): Likewise.
13397 (vst1_s8_x3): Likewise.
13398 (vst1_p8_x3): Likewise.
13399 (vst1_s16_x3): Likewise.
13400 (vst1_p16_x3): Likewise.
13401 (vst1_s32_x3): Likewise.
13402 (vst1_u8_x3): Likewise.
13403 (vst1_u16_x3): Likewise.
13404 (vst1_u32_x3): Likewise.
13405 (vst1_f16_x3): Likewise.
13406 (vst1_f32_x3): Likewise.
13407 (vst1_p64_x3): Likewise.
13408 (vst1q_s8_x3): Likewise.
13409 (vst1q_p8_x3): Likewise.
13410 (vst1q_s16_x3): Likewise.
13411 (vst1q_p16_x3): Likewise.
13412 (vst1q_s32_x3): Likewise.
13413 (vst1q_s64_x3): Likewise.
13414 (vst1q_u8_x3): Likewise.
13415 (vst1q_u16_x3): Likewise.
13416 (vst1q_u32_x3): Likewise.
13417 (vst1q_u64_x3): Likewise.
13418 (vst1q_f16_x3): Likewise.
13419 (vst1q_f32_x3): Likewise.
13420 (vst1q_f64_x3): Likewise.
13421 (vst1q_p64_x3): Likewise.
13423 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13425 * config/msp430/msp430.c (msp430_output_labelref): Prepend
13426 user_label_prefix to name.
13428 * tree-core.h: Update comment about the format of NAME string
13429 passed to handler in attribute_spec.
13431 * config/msp430/msp430.md: Remove erroneous subreg expression from
13432 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
13433 zero_extend{q,h}isi2.
13435 2018-05-30 Borislav Petkov <bp@suse.de>
13437 * doc/extend.texi: Document some architecture specific
13438 constraints and sort entries.
13440 2018-05-30 Martin Sebor <msebor@redhat.com>
13442 PR middle-end/85369
13443 * builtins.c (expand_builtin_stpcpy_1): New function.
13444 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
13445 only if the former succeeds.
13447 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
13449 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
13452 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13454 * doc/invoke.texi (-flinker-output): Document
13456 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13458 * passes.c (ipa_write_summaries): Only modify statements if body
13460 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
13461 incrementally linking.
13462 (ipa_passes): Likewise.
13463 * lto-cgraph.c (lto_output_node): When incrementally linking do not
13464 pass down resolution info.
13465 * common.opt (flag_incremental_link): Update info.
13466 * gcc.c (plugin specs): Turn flinker-output=* to
13467 -plugin-opt=-linker-output-known
13468 * toplev.c (compile_file): Also cut compilation when doing incremental
13470 * flag-types. (enum lto_partition_model): Add
13471 LTO_LINKER_OUTPUT_NOLTOREL.
13472 (invoke.texi): Add -flinker-output docs.
13473 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
13474 link same way as WPA; do not stream in dead initializers.
13476 * dwarf2out.c (dwarf2out_die_ref_for_decl,
13477 darf2out_register_external_decl): Support incremental link.
13479 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13481 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
13483 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
13485 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
13486 it down to simple_object_copy_lto_debug_sections.
13487 (run_gcc): Determine incremental LTO link time and configure
13488 lto1 into non-wpa mode, disable renaming of debug sections.
13490 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
13492 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
13493 descriptions of various incorrectly documented functions.
13495 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
13498 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13499 address check not strict.
13501 2018-05-30 Richard Biener <rguenther@suse.de>
13503 PR tree-optimization/85964
13504 * tracer.c (better_p): Drop initialized count check, we only
13505 call the function with initialized counts now.
13506 (find_best_successor): Do find a best edge if one
13507 has uninitialized count.
13508 (find_best_predecessor): Likewise. Do BB frequency check only
13509 if count is initialized.
13511 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
13513 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
13514 (aarch64_ldrstr_offset_compare): New.
13515 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
13516 load/store orderings.
13517 (aarch64_gen_adjusted_ldpstp): Likewise.
13519 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
13521 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
13522 Check for subset of GENERAL_REGS and FP_REGS.
13523 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
13526 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
13528 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
13529 and wi::to_poly_offset. Add the current offset and then check
13530 whether the sum fits, rather than using an unchecked addition of
13531 a checked term. Check for a shwi rather than a uhwi.
13532 * expr.c (get_bit_range): Use tree_to_poly_uint64.
13533 (store_constructor): Use poly_int_tree_p.
13534 (expand_expr_real_1): Likewise.
13535 * function.c (assign_temp): Likewise.
13536 * fold-const.c (const_binop): Use poly_int_tree_p and
13537 wi::to_poly_offset.
13538 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
13540 * ipa-icf-gimple.c (func_checker::compare_operand): Use
13541 to_poly_offset for MEM offsets.
13542 * ipa-icf.c (sem_variable::equals): Likewise.
13543 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
13544 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
13545 wi::to_poly_offset for BIT_FIELD_REF offsets.
13546 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
13547 wi::to_poly_offset.
13548 * var-tracking.c (emit_note_insn_var_location): Use
13549 tree_to_poly_uint64.
13551 2018-05-29 Jim Wilson <jimw@sifive.com>
13553 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
13555 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
13558 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
13559 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
13561 (sse4_1_round<mode>2): Use nonimmediate_operand
13562 for operand 1 predicate.
13564 2018-05-29 Martin Sebor <msebor@redhat.com>
13565 Richard Biener <rguenther@suse.de>
13568 * calls.c (get_size_range): Call determine_value_range instead
13569 of get_value_range..
13570 * tree-vrp.h (determine_value_range): Declared new function.
13571 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
13573 2018-05-29 Richard Biener <rguenther@suse.de>
13575 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
13576 sure to use non-pattern stmts for get_earlier_stmt arguments.
13577 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
13578 called on pattern stmts.
13579 (get_later_stmt): Likewise.
13581 2018-05-29 Martin Liska <mliska@suse.cz>
13583 PR gcov-profile/85759
13584 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
13587 2018-05-29 Jakub Jelinek <jakub@redhat.com>
13589 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
13591 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
13595 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13596 VEC_PACK_FLOAT_EXPR): New tree codes.
13597 * tree-pretty-print.c (op_code_prio): Handle
13598 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13599 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13600 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13601 * tree-inline.c (estimate_operator_cost): Likewise.
13602 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
13603 * fold-const.c (const_binop): Likewise.
13604 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
13605 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
13606 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
13607 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
13608 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
13609 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
13610 * expr.c (expand_expr_real_2): Likewise.
13611 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
13612 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
13613 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
13615 * optabs.c (expand_widen_pattern_expr): For
13616 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
13617 sign from result type rather than operand's type.
13618 (expand_binop_directly): For vec_packu_float_optab and
13619 vec_packs_float_optab allow result type to be different from operand's
13621 * optabs-tree.c (optab_for_tree_code): Handle
13622 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13623 VEC_PACK_FLOAT_EXPR. Formatting fixes.
13624 * tree-vect-generic.c (expand_vector_operations_1): Handle
13625 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
13626 VEC_PACK_FLOAT_EXPR.
13627 * tree-vect-stmts.c (supportable_widening_operation): Handle
13629 (supportable_narrowing_operation): Handle FLOAT_EXPR.
13630 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
13631 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
13632 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
13633 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
13635 (vec_pack<floatprefix>_float_<mode>): New expander.
13636 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
13638 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
13639 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
13640 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
13641 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
13642 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
13644 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
13645 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
13646 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
13647 VEC_PACK_FLOAT_EXPR): Document.
13649 2018-05-29 Richard Biener <rguenther@suse.de>
13651 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
13653 (stmt_vec_info_vec): Make pointer.
13654 (init_stmt_vec_info_vec): Remove.
13655 (free_stmt_vec_info_vec): Likewise.
13656 (set_stmt_vec_info_vec): New function.
13657 (free_stmt_vec_infos): Likewise.
13658 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
13659 (set_vinfo_for_stmt): Likewise.
13660 (get_earlier_stmt): Likewise.
13661 (get_later_stmt): Likewise.
13662 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
13663 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
13664 (vec_info::~vec_info): Free stmt_vec_infos.
13665 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
13666 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
13667 (pass_slp_vectorize::execute): Likewise.
13668 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
13669 (free_stmt_vec_info_vec): Likewise.
13670 (set_stmt_vec_info_vec): New function.
13671 (free_stmt_vec_infos): Likewise.
13672 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
13673 the global stmt_vec_info_vec.
13674 * tree-parloops.c (gather_scalar_reductions): Use
13675 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
13678 2018-05-29 Richard Biener <rguenther@suse.de>
13680 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
13682 2018-05-29 Martin Liska <mliska@suse.cz>
13683 David Malcolm <dmalcolm@redhat.com>
13685 * vec.c (test_reverse): New.
13686 (vec_c_tests): Add new test.
13687 * vec.h (vl_ptr>::reverse): New function.
13689 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
13691 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
13693 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
13696 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
13698 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
13700 2018-05-28 Richard Biener <rguenther@suse.de>
13702 PR tree-optimization/85933
13703 * tree-vect-data-refs.c (vect_record_base_alignments): Only
13704 look at stmts marked as vectorizable.
13706 2018-05-28 Richard Biener <rguenther@suse.de>
13708 PR tree-optimization/85934
13709 * tree-vect-generic.c (expand_vector_operations_1): Hoist
13710 vector boolean check before scalar optimization.
13712 2018-05-28 Jakub Jelinek <jakub@redhat.com>
13714 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
13717 2018-05-28 Mark Wielaard <mark@klomp.org>
13719 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
13720 if it is an expression containing a minus sign.
13722 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
13724 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
13726 2018-05-27 Paul Koning <ni1d@arrl.net>
13728 * config/pdp11/pdp11.md (truncsihi2): Remove.
13730 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
13731 Chung-Ju Wu <jasonwucj@gmail.com>
13733 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
13735 (unaligned_store_dw): Ditto.
13736 * config/nds32/nds32-memory-manipulation.c
13737 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
13738 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
13739 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
13740 (emit_setmem_word_loop): Rename to ...
13741 (emit_setmem_doubleword_loop): ... this.
13742 (nds32_gen_dup_4_byte_to_word_value): New function.
13743 (nds32_gen_dup_8_byte_to_double_word_value): New function.
13744 (nds32_expand_setmem_loop): Refine implementation.
13745 (nds32_expand_setmem_loop_v3m): Ditto.
13746 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
13749 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13751 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
13753 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
13755 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
13756 (nds32_init_machine_status): Initialize machine->attr_naked_p and
13757 machine->attr_no_prologue_p.
13758 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
13759 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
13760 (nds32_expand_epilogue): Consider attr_naked_p.
13761 (nds32_expand_epilogue_v3pop): Likewise.
13762 (nds32_can_use_return_insn): Likewise.
13763 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
13764 attr_no_prologue_p fields.
13765 * config/nds32/nds32.opt (mret-in-naked-func): New option.
13767 2018-05-27 Jakub Jelinek <jakub@redhat.com>
13770 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
13772 * config/i386/sse.md
13773 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
13775 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
13777 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
13779 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
13781 (*<floatsuffix>floatv2div2sf2): Rename to ...
13782 (*float<floatunssuffix>v2div2sf2): ... this.
13783 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
13784 (float<floatunssuffix>v2div2sf2_mask): ... this.
13785 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
13786 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
13787 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
13789 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
13791 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13793 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
13795 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13797 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
13799 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
13800 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
13801 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
13802 gen_ufix_truncv8dfv8si2.
13803 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
13804 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
13805 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
13806 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
13807 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
13808 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
13809 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
13810 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
13812 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13816 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
13818 2018-05-25 Jim Wilson <jimw@sifive.com>
13820 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
13821 * config/riscv/riscv.c (struct machine_function): Add
13822 interrupt_handler_p and attribute_checked_p fields.
13823 (riscv_attribute_table): Add interrupt.
13824 (riscv_interrupt_type_p): New.
13825 (riscv_save_reg_p): Save extra regs for interrupt handler.
13826 (riscv_use_save_libcall): Return false for interrupt handler.
13827 (riscv_first_stack_step): Add forward declaration.
13828 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
13829 for interrupt handler with large frame. Use it for saved reg list.
13830 (riscv_expand_prologue): Move flag_stack_usage_info support to
13831 eliminate duplication.
13832 (riscv_expand_epilogue): Generate mret for interrupt handler.
13833 (riscv_epilogue_uses): New.
13834 (riscv_can_use_return_insn): Return false for interrupt handler.
13835 (riscv_function_ok_for_sibcall): Likewise.
13836 (riscv_set_current_function): Add interrupt handler support.
13837 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
13838 * config/riscv/riscv.md (UNSPECV_MRET): New.
13840 (riscv_frflags, riscv_fsflags): Use tab after opcode.
13842 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
13844 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
13846 PR tree-optimization/85712
13847 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
13848 this candidate has already been replaced in-situ by a copy.
13850 2018-05-25 Jason Merrill <jason@redhat.com>
13852 PR c++/80485 - inline function non-zero address.
13853 * symtab.c (nonzero_address): Check DECL_COMDAT.
13855 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
13858 * config/alpha/alpha.md (ashlsi3): New insn pattern.
13859 (*ashlsi_se): Rename from *ashldi_se. Define as sign
13860 extension of SImode operation. Use const123_operand predicate.
13861 (*saddsi_1): Remove.
13862 (*saddl_se_1): Ditto.
13863 (*ssubsi_1): Ditto.
13864 (*ssubl_se_1): Ditto.
13865 * config/alpha/predicates.md (const123_operand): New predicate.
13866 * config/alpha/constraints.md (P): Use IN_RANGE.
13868 2018-05-25 Richard Biener <rguenther@suse.de>
13870 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
13872 (ref_maybe_used_by_stmt_p): Likewise.
13873 (stmt_may_clobber_ref_p): Likewise.
13874 (stmt_may_clobber_ref_p_1): Likewise.
13875 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
13877 (ref_maybe_used_by_stmt_p): Likewise.
13878 (stmt_may_clobber_ref_p): Likewise.
13879 (stmt_may_clobber_ref_p_1): Likewise.
13880 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
13881 the alias oracle to disambiguate DRs with stmts DR analysis
13883 (vect_analyze_data_refs): Do not give up on not analyzable
13884 DRs for BB vectorization. Remove code truncating the dataref
13887 2018-05-25 Jakub Jelinek <jakub@redhat.com>
13890 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
13891 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
13892 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
13894 2018-05-25 Richard Biener <rguenther@suse.de>
13896 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
13897 function, combining stmt data ref gathering and fatal analysis
13899 (vect_analyze_data_refs): Remove now redudnant code and simplify.
13900 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
13901 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
13902 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
13903 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
13905 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13907 PR tree-optimization/85720
13908 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
13909 SCC if all partitions are builtins.
13910 (version_loop_by_alias_check): New parameter. Generate cancelable
13911 runtime alias check if all partitions are builtins.
13912 (distribute_loop): Update call to above function.
13914 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13916 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
13917 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
13918 (parm_default_def_partition_arg): Ditto.
13919 (set_parm_default_def_partition): Ditto.
13920 (get_parm_default_def_partitions): Ditto and make it static.
13921 (get_undefined_value_partitions): Ditto and make it static.
13922 (remove_ssa_form): Refactor call to init_var_map here.
13923 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
13924 computation for loop region.
13925 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
13926 (register_default_def): Delete.
13927 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
13928 (parm_default_def_partition_arg): Ditto.
13929 (set_parm_default_def_partition): Ditto.
13930 (get_parm_default_def_partitions): Ditto and make it static.
13931 (get_undefined_value_partitions): Ditto and make it static.
13932 (coalesce_with_default, coalesce_with_default): Update comment.
13933 (create_coalesce_list_for_region): New func factored out from
13934 create_outofssa_var_map.
13935 (populate_coalesce_list_for_outofssa): New func factored out from
13936 create_outofssa_var_map and coalesce_ssa_name.
13937 (create_outofssa_var_map): Delete.
13938 (coalesce_ssa_name): Refactor to support live range computation.
13939 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
13940 (get_parm_default_def_partitions): Delete.
13941 (get_undefined_value_partitions): Ditto.
13942 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
13943 computation for loop region.
13944 (new_tree_live_info, loe_visit_block): Ditto.
13945 (live_worklist, set_var_live_on_entry): Ditto.
13946 (calculate_live_on_exit, verify_live_on_entry): Ditto.
13947 * tree-ssa-live.h (struct _var_map): New fields.
13948 (init_var_map): Change decl.
13949 (region_contains_p): New.
13951 2018-05-25 Bin Cheng <bin.cheng@arm.com>
13953 * tree-ssa-live.h (live_merge_and_clear): Delete.
13955 2018-05-25 Richard Biener <rguenther@suse.de>
13958 * tree-dump.c (dequeue_and_dump): Remove access to removed
13959 operand 2 of a SWITCH_EXPR.
13961 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13963 * doc/sourcebuild.texi (vect_double_cond_arith): Include
13964 multiplication and division.
13965 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13966 (cond_udiv@var{m}, cond_umod@var{m}): Document.
13967 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13968 (cond_udiv_optab, cond_umod_optab): New optabs.
13969 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13970 (IFN_COND_RDIV): New internal functions.
13971 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13972 TRUNC_MOD_EXPR and RDIV_EXPR.
13973 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13974 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13976 (SVE_INT_BINARY): Include mult.
13977 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13978 (optab, sve_int_op): Handle mult.
13979 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13981 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13982 for SVE_INT_BINARY_SD.
13984 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13986 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13987 (optab, sve_int_op): Handle div and udiv.
13988 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13989 for SVE_INT_BINARY_SD.
13990 (*<optab><mode>3): New insn for the same.
13992 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13994 * tree-vect-patterns.c: Include predict.h.
13995 (vect_recog_divmod_pattern): Restrict check for division support
13996 to when optimizing for size.
13998 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14000 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
14001 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
14002 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
14003 (gimple_match_op::set_op): Likewise.
14004 (gimple_resimplify4): Declare.
14005 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
14006 (expr::gen_transform): Likewise.
14007 (decision_tree::gen): Generate a simplification routine for 4 operands.
14008 * gimple-match-head.c (gimple_simplify): Add an overload for
14009 4 operands. In the top-level function, handle up to 4 call
14010 arguments and call gimple_resimplify4.
14011 (gimple_resimplify4): New function.
14012 (build_call_internal): Pass a fourth operand.
14013 (maybe_push_to_seq): Likewise.
14014 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
14015 Fold VEC_COND_EXPRs of an operation and a default value into
14016 an IFN_COND_* function if possible.
14017 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
14019 (SVE_COND_FP_BINARY): Include them.
14020 (optab, sve_fp_op): Handle them.
14021 (SVE_INT_BINARY_REV): New code iterator.
14022 (SVE_COND_FP_BINARY_REV): New int iterator.
14023 (commutative): New int attribute.
14024 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
14026 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
14028 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
14029 (*cond_<optab><mode>): New patterns for reversed operands.
14031 2018-05-25 Richard Biener <rguenther@suse.de>
14033 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
14034 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
14035 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
14036 (STMT_VINFO_GROUPED_ACCESS): Adjust.
14037 * tree-vect-data-refs.c (everywhere): Adjust users.
14038 * tree-vect-loop.c (everywhere): Likewise.
14039 * tree-vect-slp.c (everywhere): Likewise.
14040 * tree-vect-stmts.c (everywhere): Likewise.
14041 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
14043 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14045 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
14047 (gcc_cv_as_section_exclude): ... this.
14048 Try Solaris as #exclude syntax.
14049 * configure: Regenerate.
14050 * config.in: Regenerate.
14051 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
14053 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
14054 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
14056 * varasm.c (default_elf_asm_named_section): Don't check if
14057 HAVE_GAS_SECTION_EXCLUDE is defined.
14059 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14061 * doc/md.texi: Update the documentation of the cond_* optabs
14062 to mention the new final operand. Fix GET_MODE_NUNITS call.
14063 Describe the scalar case too.
14064 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
14065 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
14067 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
14068 (get_conditional_internal_fn): Update comment.
14069 * tree-vect-loop.c (vectorizable_reduction): Pass the original
14070 accumulator value as a final argument to conditional functions.
14071 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
14072 a define_expand and add an "else" operand. Assert for now that
14073 the else operand is equal to operand 2. Use SVE_INT_BINARY and
14074 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
14075 (*cond_<optab><mode>): New patterns.
14076 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
14077 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
14078 (UNSPEC_COND_EOR): Delete.
14079 (optab): Remove associated mappings.
14080 (SVE_INT_BINARY): New code iterator.
14081 (sve_int_op): Remove int attribute and add "minus" to the code
14083 (SVE_COND_INT_OP): Delete.
14084 (SVE_COND_FP_OP): Rename to...
14085 (SVE_COND_FP_BINARY): ...this.
14087 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
14089 * optabs.c (can_reuse_operands_p): New function.
14090 (maybe_legitimize_operands): Try to reuse the results for
14093 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14095 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
14096 Add {q} suffix to insn mnemonic.
14098 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
14100 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
14101 (msp430_warn_func_return): New.
14103 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
14105 * fold-const.c (tree_nonzero_bits): New function.
14106 * fold-const.h (tree_nonzero_bits): Likewise.
14107 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
14108 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
14110 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
14114 * varasm.c (assemble_alias): Check ifunc_resolver only on
14117 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
14120 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
14121 when memory input operand is handled.
14123 2018-05-24 Luis Machado <luis.machado@linaro.org>
14125 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
14127 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
14129 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14131 * match.pd: Delay FMA folds until after vectorization.
14133 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
14136 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
14137 address check not strict.
14139 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
14141 * gimple-match.h (gimple_match_op): New class.
14142 (mprts_hook): Replace parameters with a gimple_match_op *.
14143 (maybe_build_generic_op): Likewise.
14144 (gimple_simplified_result_is_gimple_val): Replace parameters with
14145 a const gimple_match_op *.
14146 (gimple_simplify): Replace code_helper * and tree * parameters with
14147 a gimple_match_op * parameter.
14148 (gimple_resimplify1): Replace code_helper *, tree and tree *
14149 parameters with a gimple_match_op * parameter.
14150 (gimple_resimplify2): Likewise.
14151 (gimple_resimplify3): Likewise.
14152 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
14153 parameters with a gimple_match_op * parameter.
14154 * gimple-match-head.c (gimple_simplify): Change prototypes of
14155 auto-generated functions to take a gimple_match_op * instead of
14156 separate code_helper * and tree * parameters. Make the same
14157 change in the top-level overload and update calls to the
14158 gimple_resimplify routines. Update calls to the auto-generated
14159 functions and to maybe_push_res_to_seq in the publicly-facing
14160 operation-specific gimple_simplify overloads.
14161 (gimple_match_op::MAX_NUM_OPS): Define.
14162 (gimple_resimplify1): Replace rcode and ops with a single res_op
14163 parameter. Update call to gimple_simplify.
14164 (gimple_resimplify2): Likewise.
14165 (gimple_resimplify3): Likewise.
14166 (mprts_hook): Replace parameters with a gimple_match_op *.
14167 (maybe_build_generic_op): Likewise.
14168 (build_call_internal): Replace type, nargs and ops with
14169 a gimple_match_op *.
14170 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
14171 with a single gimple_match_op *. Update calls to mprts_hook,
14172 build_call_internal and gimple_simplified_result_is_gimple_val.
14173 Factor out code that is common to the tree_code and combined_fn cases.
14174 * genmatch.c (expr::gen_transform): Replace tem_code and
14175 tem_ops with a gimple_match_op called tem_op. Update calls
14176 to the gimple_resimplify functions and maybe_push_res_to_seq.
14177 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
14178 res_ops. Update call to the gimple_resimplify functions.
14179 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
14180 (decision_tree::gen): Make the functions take a gimple_match_op *
14181 called res_op instead of separate res_code and res_ops parameters.
14182 Update call accordingly.
14183 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
14184 and ops with a single res_op parameter. Update calls to
14185 maybe_build_generic_op and maybe_push_res_to_seq.
14186 (fold_stmt_1): Update calls to gimple_simplify and
14187 replace_stmt_with_simplification.
14188 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
14189 and gimple_simplified_result_is_gimple_val.
14190 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
14192 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
14193 with a gimple_match_op *.
14194 (vn_nary_build_or_lookup): Likewise. Update call to
14195 vn_nary_build_or_lookup_1.
14196 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
14197 gimple_match_op *. Update calls to the gimple_resimplify routines
14198 and to gimple_simplified_result_is_gimple_val.
14199 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
14200 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
14201 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
14202 (visit_nary_op): Likewise.
14203 (visit_reference_op_load): Likewise.
14205 2018-05-23 Luis Machado <luis.machado@linaro.org>
14207 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
14208 modifier for printing the step amount.
14210 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
14213 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
14216 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
14218 * doc/sourcebuild.texi (Endianness): New subsubsection.
14220 2018-05-23 Luis Machado <luis.machado@linaro.org>
14222 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14223 <prefetch_dynamic_strides>: New const bool field.
14224 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14225 prefetch_dynamic_strides.
14226 (exynosm1_prefetch_tune): Likewise.
14227 (thunderxt88_prefetch_tune): Likewise.
14228 (thunderx_prefetch_tune): Likewise.
14229 (thunderx2t99_prefetch_tune): Likewise.
14230 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
14232 (aarch64_override_options_internal): Update to set
14233 PARAM_PREFETCH_DYNAMIC_STRIDES.
14234 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14235 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14236 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14237 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14238 prefetch-dynamic-strides setting.
14240 2018-05-23 Luis Machado <luis.machado@linaro.org>
14242 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14243 <minimum_stride>: New const int field.
14244 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14245 minimum_stride field defaulting to -1.
14246 (exynosm1_prefetch_tune): Likewise.
14247 (thunderxt88_prefetch_tune): Likewise.
14248 (thunderx_prefetch_tune): Likewise.
14249 (thunderx2t99_prefetch_tune): Likewise.
14250 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
14251 <default_opt_level>: Set to 3.
14252 (aarch64_override_options_internal): Update to set
14253 PARAM_PREFETCH_MINIMUM_STRIDE.
14254 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14255 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14256 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14257 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14258 stride is constant and is below the minimum stride threshold.
14260 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14262 * config/arm/arm-cpus.in (mode26): Delete.
14263 (armv4): Delete mode26 reference.
14264 * config/arm/arm.c (arm_configure_build_target): Delete use of
14267 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14269 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
14271 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
14272 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
14274 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
14275 (floatunsdidf2): Ditto.
14277 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
14279 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
14280 (fixuns_trunc<mode>si2_avx512f): Ditto.
14281 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
14282 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
14283 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
14285 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14287 PR rtl-optimization/79985
14288 * df-scan.c (df_insn_refs_collect): Remove special case for
14289 global registers and asm statements.
14291 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
14293 * extend.texi (Global Register Variables): Rewrite the bullet list.
14294 Note that the register is available for allocation. Note that access
14295 via inline asm must use constraints. Add note about async-signal
14296 handlers. Remove paragraph about automagic register selection.
14298 2018-05-23 Richard Biener <rguenther@suse.de>
14300 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
14301 of fixed offset from memset VN.
14303 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
14305 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
14306 first_interp field.
14307 (alloc_cand_and_find_basis): Initialize first_interp field.
14308 (slsr_process_mul): Modify first_interp field.
14309 (slsr_process_add): Likewise.
14310 (slsr_process_cast): Modify first_interp field for each new
14312 (slsr_process_copy): Likewise.
14313 (dump_candidate): Dump first_interp field.
14314 (replace_mult_candidate): Process all interpretations, not just
14316 (replace_rhs_if_not_dup): Likewise.
14317 (replace_one_candidate): Likewise.
14319 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
14321 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
14323 (aarch64_needs_frame_chain): New function.
14324 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
14326 2018-05-23 Sudakshina Das <sudi.das@arm.com>
14329 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
14330 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
14331 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
14332 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
14333 as true for strict-align.
14334 (aarch64_can_inline_p): Perform checks even when callee has no
14335 attributes to check for strict alignment.
14336 * doc/extend.texi (AArch64 Function Attributes): Document
14338 * doc/invoke.texi: (AArch64 Options): Likewise.
14340 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
14342 PR tree-optimization/85853
14343 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
14344 the handling of the root of the node to...
14345 (vect_slp_analyze_node_operations_1): ...this new function,
14346 and run the whole thing with the child nodes' def types
14347 set according to their SLP node's def type.
14349 2018-05-23 Richard Biener <rguenther@suse.de>
14351 PR middle-end/85874
14352 * tree-data-ref.c (create_runtime_alias_checks): Defer
14353 and ignore overflow warnings.
14355 2018-05-23 Yury Gribov <tetra2005@gmail.com>
14357 PR tree-optimization/85822
14358 * tree-vrp.c (is_masked_range_test): Fix handling of negative
14361 2018-05-23 Richard Biener <rguenther@suse.de>
14363 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
14364 memset constants via native_interpret_expr.
14366 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
14369 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
14371 (cgraph_node::create_alias): Likewise.
14372 (cgraph_node::get_availability): Check ifunc_resolver instead
14373 of looking up ifunc attribute.
14374 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
14375 * varasm.c (do_assemble_alias): Likewise.
14376 (assemble_alias): Likewise.
14377 (default_binds_local_p_3): Likewise.
14378 * cgraph.h (cgraph_node): Add ifunc_resolver.
14379 (cgraph_node::only_called_directly_or_aliased_p): Return false
14380 for IFUNC resolver.
14381 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
14383 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
14384 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
14385 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
14386 instead of looking up ifunc attribute.
14388 2018-05-22 Luis Machado <luis.machado@linaro.org>
14390 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
14392 2018-05-22 Martin Sebor <msebor@redhat.com>
14394 PR middle-end/85359
14395 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
14396 only when expasion succeeds.
14397 (expand_builtin_strcmp): Same.
14398 (expand_builtin_strncmp): Same.
14400 2018-05-22 Martin Sebor <msebor@redhat.com>
14402 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
14404 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14405 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14407 * config/aarch64/aarch64-ldpstp.md: Replace uses of
14408 aarch64_mem_pair_operand with memory_operand and delete operand swapping
14410 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14411 Add check for legitimate_address.
14412 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
14413 (aarch64_swap_ldrstr_operands): New.
14414 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
14417 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
14418 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14420 * config/aarch64/aarch64.md: New patterns to generate stp
14422 (store_pair_sw, store_pair_dw): New patterns to generate stp for
14423 single words and double words.
14424 (load_pair_sw, load_pair_dw): Likewise.
14425 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
14427 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
14429 * config/aarch64/aarch64-ldpstp.md: Modify peephole
14430 for different mode ldpstp and add peephole for merged zero stores.
14431 Likewise for loads.
14432 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
14434 (aarch64_gen_store_pair): Rename calls to match new patterns.
14435 (aarch64_gen_load_pair): Rename calls to match new patterns.
14436 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
14437 (load_pair<DREG:mode><DREG2:mode>): ... This.
14438 (store_pair<mode>): Rename to...
14439 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
14440 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
14441 New mode iterators.
14442 (V_INT_EQUIV): Handle SImode.
14443 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
14446 2018-05-22 Martin Sebor <msebor@redhat.com>
14449 * calls.c (maybe_warn_nonstring_arg): Use string length to set
14450 or ajust the presumed bound on an operation to avoid unnecessary
14453 2018-05-22 Martin Sebor <msebor@redhat.com>
14455 PR tree-optimization/85826
14456 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
14457 assuming that a DECL necesarily has a constant size.
14459 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
14461 PR middle-end/85862
14462 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
14464 2018-05-22 Richard Biener <rguenther@suse.de>
14466 PR tree-optimization/85834
14467 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
14468 non-constant and non-zero memset arguments.
14470 2018-05-22 Martin Liska <mliska@suse.cz>
14473 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
14475 2018-05-22 Richard Biener <rguenther@suse.de>
14477 PR tree-optimization/85863
14478 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
14479 comparisons when vectype is specified.
14480 (vectorizable_condition): Do not specify vectype for
14481 vect_is_simple_cond when SLP vectorizing.
14483 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
14486 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
14487 define __ibm128 as long double.
14488 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
14489 as a distinct type when IEEE 128-bit support is enabled.
14490 (init_float128_ieee): Fix up conversions between IFmode and IEEE
14491 128-bit types to use the correct functions.
14492 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
14493 convert between 128-bit floating point types that have different
14494 modes but the same representation, instead of using gen_lowpart to
14496 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
14498 (IFKF_reg): New attributes to give the register constraints for
14500 (extend<mode>tf2_internal): New insns to mark an explicit
14501 conversion between 128-bit floating point types that have a
14502 different mode but share the same representation.
14504 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
14506 PR tree-optimization/85814
14507 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
14508 a null return from get_strinfo when unsharing the next
14509 strinfo in the chain.
14511 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
14514 * varasm.c (weak_finish): Clean up weak_decls.
14516 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14518 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
14519 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
14520 UNSPEC_UADALP values.
14521 * config/aarch64/iterators.md (ABAL): New int iterator.
14524 (sur): Add mappings for the above.
14525 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
14527 (aarch64_<sur>abal<mode>_4): Likewise.
14528 (aarch64_<sur>adalp<mode>_3): Likewise.
14529 (<sur>sadv16qi): New define_expand.
14531 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
14533 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
14534 (*movdf_internal): Ditto.
14535 (*rcpsf2_sse): Ditto.
14536 (*rsqrtsf2_sse): Ditto.
14537 (*sqrt<mode>2_sse): Ditto.
14539 2018-05-21 Tamar Christina <tamar.christina@arm.com>
14541 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
14543 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
14544 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
14545 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
14546 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
14548 * config/aarch64/arm_neon.h: Likewise.
14549 * config/aarch64/iterators.md (VQ_I): New.
14551 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
14553 * config.gcc: Add arc/t-multilib-linux to tmake_file for
14555 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
14558 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14560 * config/nds32/constraints.md (S): New constraint.
14561 * config/nds32/nds32.md (call_internal): Use constraint S.
14562 (call_value_internal): Likewise.
14563 (sibcall_internal): Likewise.
14564 (sibcall_value_internal): Likewise.
14566 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14567 Chung-Ju Wu <jasonwucj@gmail.com>
14569 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
14570 into consideration.
14572 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
14573 Chung-Ju Wu <jasonwucj@gmail.com>
14575 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
14576 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
14577 (nds32_rtx_costs_impl): Simplify.
14578 (nds32_address_cost_impl): Simplify.
14579 (nds32_init_rtx_costs): New function.
14580 (nds32_rtx_costs_speed_prefer): Likewise.
14581 (nds32_rtx_costs_size_prefer): Likewise.
14582 (nds32_address_cost_speed_prefer): Likewise.
14583 (nds32_address_cost_speed_fwprop): Likewise.
14584 (nds32_address_cost_size_prefer): Likewise.
14585 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
14586 * config/nds32/nds32.c (nds32_option_override): Use
14587 nds32_init_rtx_costs function.
14589 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
14591 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
14592 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
14593 (TARGET_PIPELINE_N8): Likewise.
14594 (TARGET_PIPELINE_N10): Likewise.
14595 (TARGET_PIPELINE_N13): Likewise.
14596 (TARGET_PIPELINE_GRAYWOLF): Likewise.
14598 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
14600 * config/nds32/nds32-fpu.md: Update copyright year.
14602 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14604 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
14606 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14608 * config/nds32/nds32.c
14609 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
14610 * config/nds32/nds32.opt (minline-asm-r15): New option.
14612 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14614 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
14616 * config/nds32/nds32.md (abssi2): New pattern.
14618 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
14620 * config/i386/i386.md (rex64namesuffix): New mode attribute.
14621 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
14622 Merge insn pattern from sse_cvtsi2ss<round_name> and
14623 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
14624 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
14625 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
14626 using SWI48 mode iterator.
14627 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
14628 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
14629 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
14630 pattern from sse_cvttss2si<round_saeonly_name>
14631 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
14632 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
14633 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
14634 using SWI48 mode iterator.
14635 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
14636 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
14637 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
14638 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
14639 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
14640 using SWI48 mode iterator.
14641 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
14642 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
14643 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
14644 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
14645 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
14646 SWI48 mode iterator.
14647 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
14648 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
14649 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
14650 pattern from sse_cvttsd2si<round_saeonly_name>
14651 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
14653 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14655 * config/nds32/nds32-md-auxiliary.c
14656 (nds32_valid_smw_lwm_base_p): Refine.
14657 (nds32_output_smw_single_word): Refine.
14658 (nds32_output_smw_double_word): New.
14659 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
14661 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14663 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
14664 (nds32_output_stack_pop): Refine.
14665 (nds32_expand_unaligned_load): Refine.
14666 (nds32_expand_unaligned_store): Refine.
14668 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14669 Chung-Ju Wu <jasonwucj@gmail.com>
14671 * config/nds32/constants.md: Add TP_REGNUM constant.
14672 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
14673 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
14675 * config/nds32/nds32-doubleword.md: Consider flag_pic.
14676 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
14677 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
14678 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
14679 and PIC code generation.
14680 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
14682 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
14684 * config/nds32/nds32.md: Support TLS and PIC.
14685 * config/nds32/nds32.c: Support TLS and PIC.
14686 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
14687 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
14690 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
14692 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
14693 mode with E_ prefix.
14695 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
14696 Chung-Ju Wu <jasonwucj@gmail.com>
14698 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
14699 * config/nds32/nds32-md-auxiliary.c
14700 (symbolic_reference_mentioned_p): New.
14701 (nds32_legitimize_ict_address): New.
14702 (nds32_expand_ict_move): New.
14703 (nds32_indirect_call_referenced_p): New.
14704 (nds32_symbol_binds_local_p): Delete.
14705 (nds32_long_call_p): Modify.
14706 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
14707 * config/nds32/nds32-protos.h
14708 (symbolic_reference_mentioned_p): Declare.
14709 (nds32_legitimize_ict_address): Declare.
14710 (nds32_expand_ict_move): Declare.
14711 (nds32_indirect_call_referenced_p): Declare.
14712 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
14713 (nds32_relax_group): Use nds32_ict_const_p as condition.
14714 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
14715 (nds32_asm_file_start): Output ict_model directive in asm code.
14716 (nds32_legitimate_address_p): Consider indirect call.
14717 (nds32_print_operand): Consider indirect call.
14718 (nds32_print_operand_address): Consider indirect call.
14719 (nds32_insert_attributes): Handle "indirect_call" attribute.
14720 (TARGET_LEGITIMATE_ADDRESS_P): Define.
14721 (TARGET_LEGITIMATE_CONSTANT_P): Define.
14722 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
14723 (TARGET_DELEGITIMIZE_ADDRESS): Define.
14724 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
14725 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
14726 (TARGET_ICT_MODEL_SMALL): Define.
14727 (TARGET_ICT_MODEL_LARGE): Define.
14728 * config/nds32/nds32.md (movsi): Consider ict model.
14729 (call, call_value): Consider ict model.
14730 (sibcall, sibcall_value): Consider ict model.
14731 * config/nds32/nds32.opt (mict-model): New option.
14732 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
14735 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
14736 Monk Chiang <sh.chiang04@gmail.com>
14737 Jim Wilson <jimw@sifive.com>
14739 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
14740 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
14741 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
14742 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
14743 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
14744 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
14745 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
14746 compute save_libcall_adjustment properly.
14747 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
14748 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
14749 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
14750 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
14751 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
14752 (ABI_SPEC): Handle mabi=ilp32e.
14753 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
14754 (RVE): Add RVE mask.
14755 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
14756 <-march>: Add rv32e as an example.
14758 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
14761 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
14762 (intra_create_variable_infos): Handle C++ constructors.
14764 2018-05-18 Martin Liska <mliska@suse.cz>
14766 * passes.def: Remove a redundant pass.
14768 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
14771 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
14773 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14775 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
14777 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
14779 (armv2, armv2a, armv3, armv3m): Delete architectures.
14780 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
14781 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
14782 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
14784 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
14785 (*mulsidi3adddi): Likewise.
14786 (mulsidi3): Likewise.
14787 (*mulsidi3_nov6): Likewise.
14788 (umulsidi3): Likewise.
14789 (umulsidi3_nov6): Likewise.
14790 (umaddsidi4): Likewise.
14791 (*umulsidi3adddi): Likewise.
14792 (smulsi3_highpart): Likewise.
14793 (*smulsi3_highpart_nov6): Likewise.
14794 (umulsi3_highpart): Likewise.
14795 (*umulsi3_highpart_nov6): Likewise.
14796 * config/arm/arm.h (arm_arch3m): Delete.
14797 * config/arm/arm.c (arm_arch3m): Delete.
14798 (arm_option_override_internal): Update armv3-related comment.
14799 (arm_configure_build_target): Delete use of isa_bit_mode32.
14800 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
14801 (arm_rtx_costs_internal): Delete check of arm_arch3m.
14802 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
14803 (mulsa3): Likewise.
14804 (mulusa3): Likewise.
14805 * config/arm/arm-protos.h (arm_arch3m): Delete.
14806 * config/arm/arm-tables.opt: Regenerate.
14807 * config/arm/arm-tune.md: Likewise.
14808 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
14809 deleted architectures.
14811 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14813 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
14814 (armv5t, armv5te): New features.
14815 (ARMv5, ARMv5e): Delete fgroups.
14816 (ARMv5t, ARMv5te): Adjust for above changes.
14817 (ARMv6m): Likewise.
14818 (armv5, armv5e): Delete arches.
14819 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
14821 (*call_reg_arm): Likewise.
14822 (*call_value_reg_armv5): Likewise.
14823 (*call_value_reg_arm): Likewise.
14824 (*call_symbol): Likewise.
14825 (*call_value_symbol): Likewise.
14826 (*sibcall_insn): Likewise.
14827 (*sibcall_value_insn): Likewise.
14828 (clzsi2): Likewise.
14829 (prefetch): Likewise.
14830 (define_split and define_peephole2 dependent on arm_arch5):
14832 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
14834 (TARGET_ARM_QBIT): Likewise.
14835 (TARGET_DSP_MULTIPLY): Likewise.
14836 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
14837 (arm_arch5, arm_arch5e): Delete.
14838 (arm_arch5t, arm_arch5te): Declare.
14839 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
14840 (arm_arch5t): Declare.
14841 (arm_option_reconfigure_globals): Update for the above.
14842 (arm_options_perform_arch_sanity_checks): Update comment, replace
14843 use of arm_arch5 with arm_arch5t.
14844 (use_return_insn): Likewise.
14845 (arm_emit_call_insn): Likewise.
14846 (output_return_instruction): Likewise.
14847 (arm_final_prescan_insn): Likewise.
14848 (arm_coproc_builtin_available): Likewise.
14849 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
14850 arm_arch5e with arm_arch5t and arm_arch5te.
14851 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
14852 (arm_arch5t, arm_arch5te): Declare.
14853 * config/arm/arm-tables.opt: Regenerate.
14854 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
14855 * config/arm/t-multilib: Likewise.
14856 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
14857 instead of arm_arch5.
14858 (*call_reg_thumb1): Likewise.
14859 (*call_value_reg_thumb1_v5): Likewise.
14860 (*call_value_reg_thumb1): Likewise.
14861 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
14863 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
14865 2018-05-18 Martin Liska <mliska@suse.cz>
14867 PR gcov-profile/84846
14868 * doc/gcov.texi: Document -t option of gcov tool.
14870 2018-05-18 Martin Liska <mliska@suse.cz>
14872 PR gcov-profile/84846
14873 * gcov.c (print_usage): Add new -t option.
14874 (process_args): Handle the option.
14875 (generate_results): Use stdout as output when requested by
14878 2018-05-18 Martin Liska <mliska@suse.cz>
14880 PR gcov-profile/84846
14881 * coverage.c (coverage_init): Write PWD to .gcno file.
14882 * doc/gcov.texi: Document how working directory is printed.
14883 * gcov-dump.c (dump_gcov_file): Print PWD.
14884 * gcov.c (output_intermediate_file): Likewise.
14885 (read_graph_file): Read PWD string.
14886 (output_lines): Print PWD.
14888 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14890 PR middle-end/85817
14891 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
14892 for retval and return false if all args to phi are zero.
14894 2018-05-18 Richard Biener <rguenther@suse.de>
14896 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
14898 (evrp_dom_walker::before_dom_children): Call it.
14900 2018-05-18 Richard Biener <rguenther@suse.de>
14902 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
14903 results when processing array refs with variable index.
14905 2018-05-18 Toon Moene <toon@moene.org>
14907 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
14908 directly after that of -floop-interchange. Indicate that both
14909 options are enabled by default when specifying -O3.
14911 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14913 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
14914 iterator. Delete separate integer-mode vec_set<mode> expander.
14915 (aarch64_simd_vec_setv2di): Delete.
14916 (vec_setv2di): Delete.
14917 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
14918 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
14919 the "w, r" alternative.
14921 2018-05-18 Martin Liska <mliska@suse.cz>
14923 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
14924 * tree-pass.h (make_pass_lower_switch_O0): New function.
14925 * tree-switch-conversion.c (node_has_low_bound): Remove.
14926 (node_has_high_bound): Likewise.
14927 (node_is_bounded): Likewise.
14928 (class pass_lower_switch): Make it a template type and create
14930 (pass_lower_switch::execute): Add template argument.
14931 (make_pass_lower_switch): New function.
14932 (make_pass_lower_switch_O0): New function.
14933 (do_jump_if_equal): Remove.
14934 (emit_case_nodes): Simplify to just handle all 3 cases and leave
14935 all the hard work to tree optimization passes.
14937 2018-05-18 Martin Liska <mliska@suse.cz>
14939 * dbgcnt.c (limit_low): Renamed from limit.
14940 (limit_high): New variable.
14941 (dbg_cnt_is_enabled): Check for upper limit.
14942 (dbg_cnt): Adjust dumping.
14943 (dbg_cnt_set_limit_by_index): Add new argument for high
14945 (dbg_cnt_set_limit_by_name): Likewise.
14946 (dbg_cnt_process_single_pair): Parse new format.
14947 (dbg_cnt_process_opt): Use strtok.
14948 (dbg_cnt_list_all_counters): Remove 'value' and add
14950 * doc/invoke.texi: Document changes.
14952 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
14954 * doc/sourcebuild.texi (scalar_all_fma): Document.
14955 * tree.def (FMA_EXPR): Delete.
14956 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
14957 * internal-fn.c (ternary_direct): New macro.
14958 (expand_ternary_optab_fn): Likewise.
14959 (direct_ternary_optab_supported_p): Likewise.
14960 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
14961 * builtins.c (fold_builtin_fma): Delete.
14962 (fold_builtin_3): Don't call it.
14963 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14964 * expr.c (expand_expr_real_2): Likewise.
14965 * fold-const.c (operand_equal_p): Likewise.
14966 (fold_ternary_loc): Likewise.
14967 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14968 * gimple.c (DEFTREECODE): Likewise.
14969 * gimplify.c (gimplify_expr): Likewise.
14970 * optabs-tree.c (optab_for_tree_code): Likewise.
14971 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14972 * tree-eh.c (operation_could_trap_p): Likewise.
14973 (stmt_could_throw_1_p): Likewise.
14974 * tree-inline.c (estimate_operator_cost): Likewise.
14975 * tree-pretty-print.c (dump_generic_node): Likewise.
14976 (op_code_prio): Likewise.
14977 * tree-ssa-loop-im.c (stmt_cost): Likewise.
14978 * tree-ssa-operands.c (get_expr_operands): Likewise.
14979 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14980 * fold-const-call.h (fold_fma): Delete.
14981 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14982 CFN_FNMA and CFN_FNMS.
14983 (fold_fma): Delete.
14984 * genmatch.c (combined_fn): New enum.
14985 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14986 (commutative_op): New function.
14987 (commutate): Use it. Handle more than 2 operands.
14988 (dt_operand::gen_gimple_expr): Use commutative_op.
14989 (parser::parse_expr): Allow :c to be used with non-binary
14990 operators if the commutative operand is known.
14991 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14992 CFN_FMS, CFN_FNMA and CFN_FNMS.
14993 (backprop::process_assign_use): Remove FMA_EXPR handling.
14994 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14995 (gen_hsa_fma): New function.
14996 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14997 IFN_FNMA and IFN_FNMS.
14998 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14999 * gimple-fold.h (follow_all_ssa_edges): Declare.
15000 * gimple-fold.c (follow_all_ssa_edges): New function.
15001 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
15002 gimple_build interface and use follow_all_ssa_edges to fold the result.
15003 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
15004 instead of checking for optabs directly.
15005 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
15006 rather than FMA_EXPRs.
15007 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
15008 call to IFN_FMA instead of an FMA_EXPR.
15010 2018-05-17 Jim Wilson <jimw@sifive.com>
15012 * expr.c (do_tablejump): When converting index to Pmode, if we have a
15013 sign extended promoted subreg, and the range does not have the sign bit
15014 set, then do a sign extend.
15016 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
15017 test, check for sign extended subreg and/or constant operands, and
15018 do a sign extend in that case.
15020 2018-05-17 Steve Ellcey <sellcey@cavium.com>
15022 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
15023 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
15025 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
15026 Change logics_shift_reg to logics_shift_imm.
15027 (thunderx2t99_fp_loadpair_basic): Delete.
15028 (thunderx2t99_fp_storepair_basic): Delete.
15029 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
15030 (thunderx2t99_asimd_polynomial): Delete.
15031 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
15032 and neon_fp_mul_d_scalar_q.
15033 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
15034 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
15035 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
15036 (thunderx2t99_asimd_lut): Add missing tbl types.
15037 (thunderx2t99_asimd_ext): Delete.
15038 (thunderx2t99_asimd_load1_1_mult): Delete.
15039 (thunderx2t99_asimd_load1_2_mult): Delete.
15040 (thunderx2t99_asimd_load1_ldp): New.
15041 (thunderx2t99_asimd_load1): New.
15042 (thunderx2t99_asimd_load2): Add missing *load2* types.
15043 (thunderx2t99_asimd_load3): New.
15044 (thunderx2t99_asimd_load4): New.
15045 (thunderx2t99_asimd_store1_1_mult): Delete.
15046 (thunderx2t99_asimd_store1_2_mult): Delete.
15047 (thunderx2t99_asimd_store2_mult): Delete.
15048 (thunderx2t99_asimd_store2_onelane): Delete.
15049 (thunderx2t99_asimd_store_stp): New.
15050 (thunderx2t99_asimd_store1): New.
15051 (thunderx2t99_asimd_store2): New.
15052 (thunderx2t99_asimd_store3): New.
15053 (thunderx2t99_asimd_store4): New.
15055 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
15057 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
15058 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
15060 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
15061 Segher Boessenkool <segher@kernel.crashing.org>
15064 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
15067 2018-05-17 Richard Biener <rguenther@suse.de>
15069 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
15070 for pruning loop and prune defs feeding only already visited PHIs.
15072 2018-05-17 Richard Biener <rguenther@suse.de>
15074 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
15076 2018-05-17 Bin Cheng <bin.cheng@arm.com>
15077 Richard Biener <rguenther@suse.de>
15079 PR tree-optimization/85793
15080 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
15081 for VMAT_ELEMENTWISE.
15083 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15085 * internal-fn.h (lookup_internal_fn): Declare
15086 * internal-fn.c (lookup_internal_fn): New function.
15087 * gimple.c (gimple_build_call_from_tree): Handle calls to
15088 internal functions.
15089 * gimple-pretty-print.c (dump_gimple_call): Print "." before
15090 internal function names.
15091 * tree-pretty-print.c (dump_generic_node): Likewise.
15092 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
15094 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
15096 * gimple-fold.h (gimple_build): Make the function forms take
15097 combined_fn rather than built_in_function.
15098 (gimple_simplify): Likewise.
15099 * gimple-match-head.c (gimple_simplify): Likewise.
15100 * gimple-fold.c (gimple_build): Likewise.
15101 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
15102 rather than gimple_build_call_internal.
15103 (get_initial_defs_for_reduction): Likewise.
15104 (vect_create_epilog_for_reduction): Likewise.
15105 (vectorizable_live_operation): Likewise.
15107 2018-05-17 Martin Liska <mliska@suse.cz>
15109 * gimple-ssa-sprintf.c (format_directive): Do not use
15110 space in between 'G_' and '('.
15112 2018-05-17 Jakub Jelinek <jakub@redhat.com>
15115 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
15116 even if the mask is not all ones.
15119 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
15121 (ix86_gimple_fold_builtin): Likewise.
15124 * config/i386/i386.c: Include tree-vector-builder.h.
15125 (ix86_vector_shift_count): New function.
15126 (ix86_fold_builtin): Fold shift builtins by scalar count.
15127 (ix86_gimple_fold_builtin): Likewise.
15129 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
15130 _mm512_setzero): New intrinsics.
15132 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
15133 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15135 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
15136 code generation for cases where splatting a value is not useful.
15137 * simplify-rtx.c (simplify_ternary_operation): Simplify
15138 vec_merge across a vec_duplicate and a paradoxical subreg forming
15139 a vector mode to a vec_concat.
15141 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
15143 * config.gcc: Support "goldmont-plus".
15144 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15146 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15147 PROCESSOR_GOLDMONT_PLUS.
15148 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
15149 (processor_target_table): Add "goldmont-plus".
15150 (PTA_GOLDMONT_PLUS): Define.
15151 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
15152 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
15153 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
15154 (fold_builtin_cpu): Add "goldmont-plus".
15155 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
15156 (ix86_option_override_internal): Add "goldmont-plus".
15157 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
15158 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
15159 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
15160 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
15162 2018-05-17 Richard Biener <rguenther@suse.de>
15164 PR tree-optimization/85757
15165 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
15166 remove defs that only feed that PHI from further processing.
15168 2018-05-16 Jim Wilson <jimw@sifive.com>
15170 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
15172 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
15174 2018-05-16 Mark Wielaard <mark@klomp.org>
15176 * dwarf2out.c (count_index_strings): New function.
15177 (output_indirect_strings): Call count_index_strings and generate
15178 header for dwarf_version >= 5.
15180 2018-05-16 Mark Wielaard <mark@klomp.org>
15182 * dwarf2out.c (dwarf_FORM): New function.
15183 (set_indirect_string): Use dwarf_FORM.
15184 (reset_indirect_string): Likewise.
15185 (size_of_die): Likewise.
15186 (value_format): Likewise.
15187 (output_die): Likewise.
15188 (add_skeleton_AT_string): Likewise.
15189 (output_macinfo_op): Likewise.
15190 (index_string): Likewise.
15191 (output_index_string_offset): Likewise.
15192 (output_index_string): Likewise.
15193 (count_index_strings): Likewise.
15195 2018-05-16 Carl Love <cel@us.ibm.com>
15197 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15198 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
15200 2018-05-16 Martin Jambor <mjambor@suse.cz>
15202 * ipa-prop.c (ipa_free_all_edge_args): Remove.
15203 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
15205 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
15207 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
15208 (fnma<mode>4): Likewise.
15209 (fms<mode>4): Likewise.
15210 (fnms<mode>4): Likewise.
15211 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
15212 (aarch64_fnma<mode>4): Likewise.
15213 (aarch64_fms<mode>4): Likewise.
15214 (aarch64_fnms<mode>4): Likewise.
15215 (aarch64_fnmadd<mode>4): Likewise.
15217 2018-05-16 Jason Merrill <jason@redhat.com>
15219 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
15221 2018-05-16 Richard Biener <rguenther@suse.de>
15223 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
15224 (dump_stmt_cost): Declare.
15225 (add_stmt_cost): Dump cost we add.
15226 (add_stmt_costs): New function.
15227 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
15228 No longer exported.
15229 (vect_analyze_stmt): Adjust prototype.
15230 (vectorizable_condition): Likewise.
15231 (vectorizable_live_operation): Likewise.
15232 (vectorizable_reduction): Likewise.
15233 (vectorizable_induction): Likewise.
15234 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
15235 cost vector to pass to vectorizable_ and record afterwards.
15236 (vect_model_reduction_cost): Take cost vector argument and adjust.
15237 (vect_model_induction_cost): Likewise.
15238 (vectorizable_reduction): Likewise.
15239 (vectorizable_induction): Likewise.
15240 (vectorizable_live_operation): Likewise.
15241 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
15242 SLP_TREE_NUMBER_OF_VEC_STMTS.
15243 (vect_analyze_slp_cost_1): Remove.
15244 (vect_analyze_slp_cost): Likewise.
15245 (vect_slp_analyze_node_operations): Take visited args and
15246 a target cost vector. Avoid processing already visited stmt sets.
15247 (vect_slp_analyze_operations): Use a local cost vector to gather
15248 costs and register those of non-discarded instances.
15249 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
15250 (vect_schedule_slp_instance): Remove copying of
15251 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
15253 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
15254 adding cost. Record cost entry location.
15255 (vect_prologue_cost_for_slp_op): Function to compute cost of
15256 a constant or invariant generated for SLP vect in the prologue,
15257 split out from vect_analyze_slp_cost_1.
15258 (vect_model_simple_cost): Make static. Adjust for SLP costing.
15259 (vect_model_promotion_demotion_cost): Likewise.
15260 (vect_model_store_cost): Likewise, make static.
15261 (vect_model_load_cost): Likewise.
15262 (vectorizable_bswap): Add cost vector arg and adjust.
15263 (vectorizable_call): Likewise.
15264 (vectorizable_simd_clone_call): Likewise.
15265 (vectorizable_conversion): Likewise.
15266 (vectorizable_assignment): Likewise.
15267 (vectorizable_shift): Likewise.
15268 (vectorizable_operation): Likewise.
15269 (vectorizable_store): Likewise.
15270 (vectorizable_load): Likewise.
15271 (vectorizable_condition): Likewise.
15272 (vectorizable_comparison): Likewise.
15273 (can_vectorize_live_stmts): Likewise.
15274 (vect_analyze_stmt): Likewise.
15275 (vect_transform_stmt): Adjust calls to vectorizable_*.
15276 * tree-vectorizer.c: Include gimple-pretty-print.h.
15277 (dump_stmt_cost): New function.
15279 2018-05-16 Richard Biener <rguenther@suse.de>
15281 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
15282 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
15283 * tree-ssa-dse.c: Include tree-ssa-loop.h.
15284 (check_name): New callback.
15285 (dse_classify_store): Track cycles via a visited bitmap of PHI
15286 defs and simplify handling of in-loop and across loop dead stores
15287 and properly fail for loop-variant refs. Handle byte-tracking with
15288 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
15291 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
15293 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
15294 (vect_get_mask_type_for_stmt): Likewise.
15295 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
15297 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
15298 to determine the statement's vector type and the vector type that
15299 should be used for calculating nunits. Deal with cases in which
15300 the type has to be deferred.
15301 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
15302 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
15303 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
15304 (vect_determine_vf_for_stmt): New functions, split out from...
15305 (vect_determine_vectorization_factor): ...here.
15306 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
15307 (vect_get_mask_type_for_stmt): New functions, split out from
15308 vect_determine_vectorization_factor.
15310 2018-05-16 Richard Biener <rguenther@suse.de>
15312 * tree-cfg.c (verify_gimple_assign_ternary): Properly
15313 verify the [VEC_]COND_EXPR embedded comparison.
15315 2018-05-15 Martin Sebor <msebor@redhat.com>
15317 PR tree-optimization/85753
15318 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
15319 RECORD_TYPE in addition to ARRAY_TYPE.
15321 2018-05-15 Martin Sebor <msebor@redhat.com>
15323 PR middle-end/85643
15324 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
15326 2018-05-15 Richard Biener <rguenther@suse.de>
15328 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
15329 add by_clobber_p one. Change algorithm to collect all defs
15330 representing uses we need to walk and try reducing them to
15331 a single one before failing.
15332 (dse_dom_walker::dse_optimize_stmt): Adjust.
15334 2018-05-13 Mark Wielaard <mark@klomp.org>
15336 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
15337 (size_of_loc_descr): Likewise.
15338 (output_loc_operands): Likewise.
15339 (output_loc_operands_raw): Likewise.
15340 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
15341 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
15342 (hash_loc_operands): Likewise.
15343 (compare_loc_operands): Likewise.
15345 2018-05-14 Mark Wielaard <mark@klomp.org>
15347 * dwarf2out.c (count_index_addrs): New function.
15348 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
15350 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15352 PR tree-optimization/83648
15353 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
15354 return value as malloc candidate.
15356 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15359 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
15360 param as true in call to suggest_attribute.
15362 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
15364 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
15365 -mreadonly-in-sdata.
15367 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15369 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
15371 (aarch64_crypto_aesd_fused): Likewise.
15373 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
15375 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
15376 (movsi_aarch64): Likewise.
15377 (load_pairsi): Likewise.
15378 (load_pairdi): Likewise.
15379 (store_pairsi): Likewise.
15380 (store_pairdi): Likewise.
15381 (load_pairsf): Likewise.
15382 (load_pairdf): Likewise.
15383 (store_pairsf): Likewise.
15384 (store_pairdf): Likewise.
15385 (zero_extend): Likewise.
15386 (trunc): Swap alternatives.
15387 (fcvt_target): Add '?' to prefer w over r.
15389 2018-05-14 Jakub Jelinek <jakub@redhat.com>
15392 * config/i386/i386.md: Disallow non-commutative arithmetics in
15393 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
15394 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
15395 in the peephole2 before it.
15397 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
15399 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
15400 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
15401 (ix86_handle_option): Handle -mcldemote.
15402 * config.gcc: New header.
15403 * config/i386/cldemoteintrin.h: New file.
15404 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
15405 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
15407 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15408 OPTION_MASK_ISA_CLDEMOTE.
15409 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
15410 (ix86_valid_target_attribute_inner_p): Ditto.
15411 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
15412 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
15413 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
15414 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
15415 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
15417 * config/i386/i386.opt: Add -mcldemote.
15418 * config/i386/x86intrin.h: New header.
15419 * doc/invoke.texi: Add -mcldemote.
15421 2018-05-14 Richard Biener <rguenther@suse.de>
15423 * doc/match-and-simplify.texi: Adjust :s documentation.
15425 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
15427 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
15428 intended memcpy size.
15429 (REORDER_45): Likewise.
15431 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
15433 * sort.cc: New file.
15434 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
15435 * vec.c (qsort_chk): Use gcc_qsort.
15436 * Makefile.in (OBJS-libcommon): Add sort.o.
15437 (build/sort.o): New target. Use it...
15438 (BUILD_RTL): ... here, and...
15439 (build/gencfn-macros): ... here, and...
15440 (build/genmatch): ... here.
15442 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15443 Chung-Ju Wu <jasonwucj@gmail.com>
15445 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
15446 * config/nds32/nds32-graywolf.md: New file.
15447 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
15448 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
15450 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
15451 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
15452 * config/nds32/nds32.md (pipeline_model): Add graywolf.
15453 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
15454 * config/nds32/pipelines.md: Include n15 settings.
15456 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15457 Chung-Ju Wu <jasonwucj@gmail.com>
15459 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
15460 * config/nds32/nds32-n13.md: New file.
15461 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
15462 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
15464 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
15465 * config/nds32/nds32.md (pipeline_model): Add n13.
15466 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
15467 * config/nds32/pipelines.md: Include n13 settings.
15469 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
15470 Chung-Ju Wu <jasonwucj@gmail.com>
15472 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
15473 * config/nds32/nds32-n10.md: New file.
15474 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
15475 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
15477 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
15478 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
15479 * config/nds32/nds32.md (pipeline_model): Add n10.
15480 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
15481 * config/nds32/pipelines.md: Include n10 settings.
15483 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
15484 Kito Cheng <kito.cheng@gmail.com>
15485 Chung-Ju Wu <jasonwucj@gmail.com>
15487 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
15488 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
15489 Add enum values for DSP extension instructions.
15490 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
15492 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
15493 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
15494 New code iterators.
15495 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
15496 * config/nds32/nds32-dspext.md: New file for DSP implementation.
15497 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
15498 * config/nds32/nds32-intrinsic.md: Likewise.
15499 * config/nds32/nds32_intrinsic.h: Likewise.
15500 * config/nds32/nds32-md-auxiliary.c: Likewise.
15501 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
15502 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
15503 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
15504 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
15505 * config/nds32/nds32-protos.h: New declarations for DSP extension.
15506 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
15507 TYPE_DMAC in switch statement.
15508 * config/nds32/nds32.c: New checking and implementation for DSP
15509 extension instructions.
15510 * config/nds32/nds32.h: Likewise.
15511 * config/nds32/nds32.md: Likewise.
15512 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
15513 * config/nds32/predicates.md: Implement new predicates for DSP
15516 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
15518 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
15519 Reformat alternatives and attributes so it is easier to identify
15520 which constraints/attributes go with which instruction.
15521 (mov<mode>_hardfloat32, FMOVE64): Likewise.
15522 (mov<mode>_softfloat32, FMOVE64): Likewise.
15523 (mov<mode>_hardfloat64, FMOVE64): Likewise.
15524 (mov<mode>_softfloat64, FMOVE64): Likewise.
15526 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
15528 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15530 (Basic PowerPC Built-in Functions): The new name of the
15531 subsection previously known as "PowerPC Built-in Functions".
15532 (Basic PowerPC Built-in Functions Available on all Configurations):
15534 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
15535 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
15536 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
15537 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
15539 2018-05-11 Martin Jambor <mjambor@suse.cz>
15542 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
15545 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15548 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
15550 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
15552 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
15553 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
15554 (ix86_handle_option): Handle -mwaitpkg.
15555 * config.gcc: New header.
15556 * config/i386/cpuid.h (bit_WAITPKG): New bit.
15557 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
15558 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
15560 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15561 OPTION_MASK_ISA_WAITPKG.
15562 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
15563 (ix86_option_override_internal): Add PTA_WAITPKG.
15564 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
15565 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
15566 IX86_BUILTIN_TPAUSE.
15567 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
15568 __builtin_ia32_umwait and __builtin_ia32_tpause.
15569 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
15570 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
15571 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
15572 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
15573 UNSPECV_TPAUSE): New.
15574 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
15575 * config/i386/i386.opt: Add -mwaitpkg.
15576 * config/i386/waitpkgintrin.h: New file.
15577 * config/i386/x86intrin.h: New header.
15578 * doc/invoke.texi: Add -mwaitpkg.
15580 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
15583 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
15585 (cortex-m0): Use armv6s-m isa.
15586 (cortex-m0plus): Likewise.
15587 (cortex-m1): Likewise.
15588 (cortex-m0.small-multiply): Likewise.
15589 (cortex-m0plus.small-multiply): Likewise.
15590 (cortex-m1.small-multiply): Likewise.
15592 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
15593 Jakub Jelinek <jakub@redhat.com>
15595 PR tree-optimization/85692
15596 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
15597 source permute as well.
15599 2018-05-11 Martin Liska <mliska@suse.cz>
15602 * doc/extend.texi: Document LLVM style format for no_sanitize
15605 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
15607 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
15608 mode_supports_vsx_dform_quad to mode_supports_dq_form.
15609 (mode_supports_vsx_dform_quad): Likewise.
15610 (mode_supports_vmx_dform): Move these functions to be next to the
15611 other mode_supports functions.
15612 (mode_supports_dq_form): Likewise.
15613 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
15614 mode_supports_dq_form.
15615 (reg_offset_addressing_ok_p): Likewise.
15616 (offsettable_ok_by_alignment): Likewise.
15617 (rs6000_legitimate_offset_address_p): Likewise.
15618 (legitimate_lo_sum_address_p): Likewise.
15619 (rs6000_legitimize_address): Likewise.
15620 (rs6000_legitimize_reload_address): Likewise.
15621 (rs6000_secondary_reload_inner): Likewise.
15622 (rs6000_preferred_reload_class): Likewise.
15623 (rs6000_output_move_128bit): Likewise.
15625 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15627 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
15628 Generate SImode target register for null target.
15629 <case IX86_BUILTIN_XGETBV>: Ditto.
15630 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
15631 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
15633 2018-05-10 Carl Love <cel@us.ibm.com>
15635 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
15636 dcbtt and dcbtstt if operands[2] is 0.
15638 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
15641 * config/i386/sse.md (usadv64qi): New expander.
15643 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
15645 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
15646 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
15647 -maltivec=be support.
15648 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
15649 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
15650 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
15651 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
15652 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
15653 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
15654 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
15655 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
15656 altivec_vsumsws): Adjust.
15657 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
15658 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
15659 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
15661 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
15662 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
15663 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
15664 (altivec_lve<VI_char>x): Delete expand.
15665 (*altivec_lve<VI_char>x_internal): Rename to...
15666 (altivec_lve<VI_char>x): ... this.
15667 (altivec_lvxl_<mode>): Delete expand.
15668 (*altivec_lvxl_<mode>_internal): Rename to ...
15669 (altivec_lvxl_<mode>): ... this.
15670 (altivec_stvxl_<mode>): Delete expand.
15671 (*altivec_stvxl_<mode>_internal): Rename to ...
15672 (altivec_stvxl_<mode>): ... this.
15673 (altivec_stve<VI_char>x): Delete expand.
15674 (*altivec_stve<VI_char>x_internal): Rename to ...
15675 (altivec_stve<VI_char>x): ... this.
15676 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
15677 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
15678 reduc_plus_scal_<mode>): Adjust.
15679 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
15681 (rs6000_cpu_cpp_builtins): Adjust.
15682 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
15683 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
15684 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
15685 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
15686 -maltivec=be support.
15687 (rs6000_split_vec_extract_var): Adjust.
15688 (rs6000_split_v4si_init): Adjust.
15689 (swap_selector_for_mode): Delete.
15690 (altivec_expand_lvx_be, altivec_expand_stvx_be,
15691 altivec_expand_stvex_be): Delete.
15692 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
15693 -maltivec=be support.
15694 (rs6000_gimple_fold_builtin): Ditto.
15695 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
15697 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
15698 (TARGET_DIRECT_MOVE_64BIT): Adjust.
15699 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
15700 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
15701 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
15702 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
15703 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
15704 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
15705 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
15706 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
15707 anonymous split): Adjust.
15708 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
15709 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
15711 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
15713 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
15714 when --with-gxx-include-dir is also specified.
15715 * configure: Regenerate.
15717 2018-05-09 Jim Wilson <jimw@sifive.com>
15720 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
15721 * config/riscv/t-withmultilib: New.
15722 * config/riscv/withmultilib.h: New.
15723 * doc/install.texi: Document RISC-V --with-multilib-list support.
15725 2018-05-09 Richard Biener <rguenther@suse.de>
15727 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
15729 (vect_bb_vectorization_profitable_p): Adjust. Compute
15730 actual scalar cost using the cost vector and the add_stmt_cost
15733 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15735 PR rtl-optimization/85645
15736 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
15737 in the REG_CFA_REGISTER note for LR, don't leave it empty.
15739 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15741 PR rtl-optimization/85645
15742 * shrink-wrap.c (spread_components): Return a boolean saying if
15743 anything was changed.
15744 (try_shrink_wrapping_separate): Iterate spread_components until
15745 nothing changes anymore.
15747 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15749 PR rtl-optimization/85645
15750 * regrename.c (build_def_use): Also kill the chains that include the
15751 destination of a REG_CFA_REGISTER note.
15753 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
15755 PR rtl-optimization/85645
15756 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
15757 insn that has a REG_CFA_REGISTER note.
15759 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
15761 * cfgexpand.c (expand_clobber): New function.
15762 (expand_gimple_stmt_1): Use it.
15763 * tree-vect-stmts.c (vect_clobber_variable): New function,
15765 (vectorizable_simd_clone_call): ...here.
15766 (vectorizable_store): Emit a clobber either side of an
15767 IFN_STORE_LANES sequence.
15768 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
15770 2018-05-09 Tom de Vries <tom@codesourcery.com>
15773 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
15774 (define_insn "trap_if_false"): Add exit after trap.
15776 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
15778 PR rtl-optimization/85638
15779 * bb-reorder.c: Include common/common-target.h.
15780 (create_forwarder_block): New function extracted from...
15781 (fix_up_crossing_landing_pad): ...here. Rename into...
15782 (dw2_fix_up_crossing_landing_pad): ...this.
15783 (sjlj_fix_up_crossing_landing_pad): New function.
15784 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
15785 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
15786 from both partitions and exit the loop after one iteration.
15788 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15791 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15793 (Basic PowerPC Built-in Functions): The new name of the
15794 subsection previously known as "PowerPC Built-in Functions".
15795 (Basic PowerPC Built-in Functions Available on all Configurations):
15797 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15799 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15801 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15803 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15806 2018-05-08 Jim Wilson <jimw@sifive.com>
15808 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
15809 (LD_EMUL_SUFFIX): New.
15810 (LINK_SPEC): Use it.
15812 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
15814 * doc/extend.texi (PowerPC Built-in Functions): Rename this
15816 (Basic PowerPC Built-in Functions): The new name of the
15817 subsection previously known as "PowerPC Built-in Functions".
15818 (Basic PowerPC Built-in Functions Available on all Configurations):
15820 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
15822 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
15824 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
15826 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
15829 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15832 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
15833 after cmpelim optimization.
15835 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
15837 * config.gcc: Support "goldmont".
15838 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
15839 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15840 PROCESSOR_GOLDMONT.
15841 * config/i386/i386.c (m_GOLDMONT): Define.
15842 (processor_target_table): Add "goldmont".
15843 (PTA_GOLDMONT): Define.
15844 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
15845 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
15846 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
15847 (fold_builtin_cpu): Add "goldmont".
15848 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
15849 (ix86_option_override_internal): Add "goldmont".
15850 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
15851 (processor_type): Add PROCESSOR_GOLDMONT.
15852 * config/i386/i386.md: Add CPU "glm".
15853 * config/i386/glm.md: New file.
15854 * config/i386/x86-tune.def: Add m_GOLDMONT.
15855 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
15857 2018-05-08 Jakub Jelinek <jakub@redhat.com>
15860 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
15862 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
15863 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
15864 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
15867 * config/i386/i386.c (ix86_fold_builtin): Handle
15868 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
15871 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
15872 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
15874 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
15877 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
15878 (check_arch): Likewise.
15879 (check_fpu): Return the result rather than printing it.
15880 (end arch): Fix operator precedence.
15881 (end cpu): Likewise.
15882 (END): Print the result from check_fpu.
15884 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15885 Alan Hayward <alan.hayward@arm.com>
15886 David Sherwood <david.sherwood@arm.com>
15888 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
15889 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
15890 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
15891 (*fcmuo<mode>_and): New patterns.
15893 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
15895 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
15896 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
15897 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
15898 (cmp_op, sve_imm_con): New code attributes.
15899 (SVE_COND_INT_CMP, imm_con): Delete.
15900 (cmp_op): Remove above unspecs from int attribute.
15901 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
15903 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
15904 comparison-specific unspecs.
15905 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
15906 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
15907 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
15908 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
15909 (*vec_fcm<cmp_op><mode>): Rename to...
15910 (*fcm<cmp_op><mode>): ...this and adjust likewise.
15911 (*vec_fcmuo<mode>): Rename to...
15912 (*fcmuo<mode>): ...this and adjust likewise.
15913 (*pred_fcm<cmp_op><mode>): New pattern.
15914 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
15915 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
15917 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
15919 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
15920 (aarch64_emit_sve_predicated_cond): New function.
15921 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
15922 (aarch64_emit_unspec_cond_or): Replace with...
15923 (aarch64_emit_sve_or_conds): ...this new function. Use
15924 aarch64_emit_sve_ptrue_op for the individual comparisons and
15925 aarch64_emit_binop to OR them together.
15926 (aarch64_emit_inverted_unspec_cond): Replace with...
15927 (aarch64_emit_sve_inverted_cond): ...this new function. Use
15928 aarch64_emit_sve_ptrue_op for the comparison and
15929 aarch64_emit_unop to invert the result.
15930 (aarch64_expand_sve_vec_cmp_float): Update after the above
15931 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
15933 2018-05-07 Nathan Sidwell <nathan@acm.org>
15935 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
15936 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
15937 (Backwards Compatibility): Likewise.
15939 2018-05-07 Luis Machado <luis.machado@linaro.org>
15943 2018-05-07 Luis Machado <luis.machado@linaro.org>
15945 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15946 <prefetch_dynamic_strides>: New const bool field.
15947 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15948 prefetch_dynamic_strides.
15949 (exynosm1_prefetch_tune): Likewise.
15950 (thunderxt88_prefetch_tune): Likewise.
15951 (thunderx_prefetch_tune): Likewise.
15952 (thunderx2t99_prefetch_tune): Likewise.
15953 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15955 (aarch64_override_options_internal): Update to set
15956 PARAM_PREFETCH_DYNAMIC_STRIDES.
15957 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15958 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15959 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15960 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15961 prefetch-dynamic-strides setting.
15963 2018-05-07 Luis Machado <luis.machado@linaro.org>
15965 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15966 <minimum_stride>: New const int field.
15967 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15968 minimum_stride field.
15969 (exynosm1_prefetch_tune): Likewise.
15970 (thunderxt88_prefetch_tune): Likewise.
15971 (thunderx_prefetch_tune): Likewise.
15972 (thunderx2t99_prefetch_tune): Likewise.
15973 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15974 (aarch64_override_options_internal): Update to set
15975 PARAM_PREFETCH_MINIMUM_STRIDE.
15976 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15977 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15978 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15979 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15980 stride is constant and is below the minimum stride threshold.
15982 2018-05-07 Luis Machado <luis.machado@linaro.org>
15984 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15987 2018-05-07 Luis Machado <luis.machado@linaro.org>
15989 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15990 <prefetch_dynamic_strides>: New const bool field.
15991 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15992 prefetch_dynamic_strides.
15993 (exynosm1_prefetch_tune): Likewise.
15994 (thunderxt88_prefetch_tune): Likewise.
15995 (thunderx_prefetch_tune): Likewise.
15996 (thunderx2t99_prefetch_tune): Likewise.
15997 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15999 (aarch64_override_options_internal): Update to set
16000 PARAM_PREFETCH_DYNAMIC_STRIDES.
16001 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
16002 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
16003 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
16004 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
16005 prefetch-dynamic-strides setting.
16007 2018-05-07 Luis Machado <luis.machado@linaro.org>
16009 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
16010 <minimum_stride>: New const int field.
16011 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
16012 minimum_stride field.
16013 (exynosm1_prefetch_tune): Likewise.
16014 (thunderxt88_prefetch_tune): Likewise.
16015 (thunderx_prefetch_tune): Likewise.
16016 (thunderx2t99_prefetch_tune): Likewise.
16017 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
16018 (aarch64_override_options_internal): Update to set
16019 PARAM_PREFETCH_MINIMUM_STRIDE.
16020 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
16021 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
16022 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
16023 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
16024 stride is constant and is below the minimum stride threshold.
16026 2018-05-06 Jakub Jelinek <jakub@redhat.com>
16029 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
16030 the type is addressable. Don't force op into register if it has
16033 2018-05-05 Roland McGrath <mcgrathr@google.com>
16036 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
16037 any section for which we don't know a specific type it should have,
16038 regardless of name. Previously this was done only for the exact
16039 names ".init_array", ".fini_array", and ".preinit_array".
16040 (default_elf_asm_named_section): Add comment about
16041 relationship with default_section_type_flags and SECTION_NOTYPE.
16042 (get_section): Don't consider it a type conflict if one side has
16043 SECTION_NOTYPE and the other doesn't, as long as neither has the
16044 SECTION_BSS et al used in the default_section_type_flags logic.
16046 2018-05-05 Tom de Vries <tom@codesourcery.com>
16049 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
16050 (workaround_barsyncs): New function.
16051 (nvptx_reorg): Use workaround_barsyncs.
16052 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16053 (define_expand "nvptx_membar_cta"): New define_expand.
16054 (define_insn "*nvptx_membar_cta"): New insn.
16056 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
16058 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
16059 To improve optimization opportunities.
16060 * builtin-types.def: The new needed builtin types for the above.
16062 2018-05-04 Richard Biener <rguenther@suse.de>
16064 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
16065 * gimple-ssa-store-merging.c
16066 (imm_store_chain_info::output_merged_store): Remove redundant create,
16067 release split_store vector contents on failure.
16068 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
16069 scalar stmt vector on cache hit.
16071 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
16073 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
16075 * config.gcc (powerpc-xilinx-eabi*): Remove.
16076 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
16078 (fusion_addis_mem_combo_load): Ditto.
16079 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
16081 (rs6000_cpu_cpp_builtins): Ditto.
16082 * config/rs6000/rs6000-linux.c
16083 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
16084 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
16085 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
16087 (rs6000_setup_reg_addr_masks): Ditto.
16088 (rs6000_init_hard_regno_mode_ok): Ditto.
16089 (rs6000_option_override_internal): Ditto.
16090 (legitimate_lo_sum_address_p): Ditto.
16091 (rs6000_legitimize_address): Ditto.
16092 (rs6000_legitimize_reload_address): Ditto.
16093 (rs6000_legitimate_address_p): Ditto.
16094 (abi_v4_pass_in_fpr): Ditto.
16095 (setup_incoming_varargs): Ditto.
16096 (rs6000_gimplify_va_arg): Ditto.
16097 (rs6000_split_multireg_move): Ditto.
16098 (rs6000_savres_strategy): Ditto.
16099 (rs6000_emit_prologue_components): Ditto.
16100 (rs6000_emit_epilogue_components): Ditto.
16101 (rs6000_emit_prologue): Ditto.
16102 (rs6000_emit_epilogue): Ditto.
16103 (rs6000_elf_file_end): Ditto.
16104 (rs6000_function_value): Ditto.
16105 (rs6000_libcall_value): Ditto.
16106 * config/rs6000/rs6000.h: Ditto.
16107 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
16108 (TARGET_MINMAX): ... this. New.
16109 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
16110 * config/rs6000/rs6000.md: Remove Xilinx FP support.
16111 (*movsi_internal1_single): Delete.
16112 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
16113 mfpu=, mxilinx-fpu): Delete.
16114 * config/rs6000/singlefp.h: Delete.
16115 * config/rs6000/sysv4.h: Remove Xilinx FP support.
16116 * config/rs6000/t-rs6000: Ditto.
16117 * config/rs6000/t-xilinx: Delete.
16118 * config/rs6000/titan.md: Adjust for fp_type removal.
16119 * config/rs6000/vsx.md: Remove Xilinx FP support.
16120 (VStype_simple): Delete.
16121 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
16122 * config/rs6000/xfpu.h: Delete.
16123 * config/rs6000/xfpu.md: Delete.
16124 * config/rs6000/xilinx.h: Delete.
16125 * config/rs6000/xilinx.opt: Delete.
16126 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
16127 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
16129 2018-05-04 Tom de Vries <tom@codesourcery.com>
16132 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
16135 2018-05-04 Richard Biener <rguenther@suse.de>
16137 PR middle-end/85627
16138 * tree-complex.c (update_complex_assignment): We are always in SSA form.
16139 (expand_complex_div_wide): Likewise.
16140 (expand_complex_operations_1): Likewise.
16141 (expand_complex_libcall): Preserve EH info of the original stmt.
16142 (tree_lower_complex): Handle removed blocks.
16143 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
16144 on complex multiplication and division libcall builtins.
16146 2018-05-04 Richard Biener <rguenther@suse.de>
16148 PR middle-end/85574
16149 * fold-const.c (negate_expr_p): Restrict negation of operand
16150 zero of a division to when we know that can happen without
16152 (fold_negate_expr_1): Likewise.
16154 2018-05-04 Jakub Jelinek <jakub@redhat.com>
16157 * real.h (real_nextafter): Declare.
16158 * real.c (real_nextafter): New function.
16159 * fold-const-call.c (fold_const_nextafter): New function.
16160 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
16161 CASE_CFN_NEXTTOWARD.
16162 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
16163 even when arg1_mode is different from arg0_mode.
16165 2018-05-03 Nathan Sidwell <nathan@acm.org>
16167 * doc/extend.texi (Deprecated Features): Remove
16168 -ffriend-injection.
16169 (Backwards Compatibility): Likewise.
16170 * doc/invoke.texi (C++ Language Options): Likewise.
16171 (C++ Dialect Options): Likewise.
16173 2018-05-03 Jakub Jelinek <jakub@redhat.com>
16176 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
16177 _mm512_mask_mullox_epi64): New intrinsics.
16179 2018-05-03 Tom de Vries <tom@codesourcery.com>
16182 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16183 dump files): Add offload-tree.
16185 2018-05-03 Richard Biener <rguenther@suse.de>
16187 PR tree-optimization/85615
16188 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
16189 to loops not nested in BBs loop father to avoid creating multi-entry
16192 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16194 PR tree-optimization/70291
16195 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
16196 arguments. Change return type to tree. Emit libcall as a new
16197 statement rather than replacing existing one when inplace_p is true.
16198 (expand_complex_multiplication_components): New function.
16199 (expand_complex_multiplication): Expand floating-point complex
16200 multiplication using the above.
16201 (expand_complex_division): Rename inner_type parameter to type.
16202 Update expand_complex_libcall call-site.
16203 (expand_complex_operations_1): Update expand_complex_multiplication
16204 and expand_complex_division call-sites.
16206 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16209 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16210 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16211 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
16212 the highest significant bit of the shift count mask is clear. In
16213 check whether and[sq]i3 is needed verify that all significant bits
16214 of the shift count other than the highest are set.
16216 2018-05-02 Tom de Vries <tom@codesourcery.com>
16219 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
16220 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
16221 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
16222 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
16223 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
16224 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
16225 __builtin_goacc_parlevel_size.
16227 2018-05-02 Richard Biener <rguenther@suse.de>
16229 PR tree-optimization/85597
16230 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
16231 do not use split vect_get_vec_defs call but call vect_get_slp_defs
16234 2018-05-02 Tom de Vries <tom@codesourcery.com>
16237 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16238 dump files): Add ltrans-tree.
16240 2018-05-02 Tom de Vries <tom@codesourcery.com>
16243 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
16244 dump files): Add wpa-ipa.
16246 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
16248 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
16249 powerpc*-*-linux*paired* target.
16250 * config/rs6000/750cl.h: Delete.
16251 * config/rs6000/paired.h: Delete.
16252 * config/rs6000/paired.md: Delete.
16253 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
16255 * config/rs6000/rs6000-builtin.def: Remove paired float support.
16256 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
16257 comment. Remove paired float support.
16258 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
16259 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
16261 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
16262 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
16264 * config/rs6000/rs6000.c: Remove paired float support.
16265 (paired_expand_vector_init, paired_expand_vector_move,
16266 paired_emit_vector_compare, paired_emit_vector_cond_expr,
16267 (paired_expand_lv_builtin, paired_expand_stv_builtin,
16268 paired_expand_builtin, paired_expand_predicate_builtin,
16269 paired_init_builtins): Delete.
16270 * config/rs6000/rs6000.h: Remove paired float support.
16271 * config/rs6000/rs6000.md: Remove paired float support.
16272 (move_from_CR_ov_bit): Delete.
16273 * config/rs6000/rs6000.opt (mpaired): Delete.
16274 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
16275 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
16277 2018-05-02 Richard Biener <rguenther@suse.de>
16279 PR middle-end/85567
16280 * gimplify.c (gimplify_save_expr): When in SSA form allow
16281 SAVE_EXPRs to compute to SSA vars.
16283 2018-05-02 Jakub Jelinek <jakub@redhat.com>
16286 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
16287 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
16288 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
16289 clobber operands[2], instead use a new pseudo. Formatting fixes.
16291 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
16293 PR tree-optimization/85586
16294 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
16295 exit early for statements in the same group if the accesses are
16298 2018-05-02 Tom de Vries <tom@codesourcery.com>
16301 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
16304 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
16306 PR tree-optimization/85143
16307 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
16309 2018-05-01 Tom de Vries <tom@codesourcery.com>
16312 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
16313 not found" error message.
16315 2018-05-01 Tom de Vries <tom@codesourcery.com>
16318 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
16319 * vec.c (test_ordered_remove_if): New function.
16320 (vec_c_tests): Call test_ordered_remove_if.
16321 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
16322 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
16323 * tree-vect-patterns.c (vect_pattern_recog_1): Use
16324 VEC_ORDERED_REMOVE_IF.
16326 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
16328 PR tree-optimization/82665
16329 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
16330 pointer subtraction where arguments come from a memchr call.
16332 2018-05-01 Jakub Jelinek <jakub@redhat.com>
16334 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
16335 --push-state --as-needed and --pop-state instead of --as-needed and
16336 --no-as-needed if ld supports it.
16337 * configure: Regenerated.
16340 * doc/install.texi2html: Replace _002d with - and _002a with * in
16341 generated html files using sed.
16343 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16346 * gcc-rich-location.c (blank_line_before_p): New function.
16347 (use_new_line): New function.
16348 (gcc_rich_location::add_fixit_insert_formatted): New function.
16349 * gcc-rich-location.h
16350 (gcc_rich_location::add_fixit_insert_formatted): New function.
16352 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16354 * selftest.c (assert_streq): Rename "expected" and "actual" to
16355 "val1" and "val2". Extend NULL-handling to cover both inputs
16356 symmetrically, while still requiring both to be non-NULL for a pass.
16357 * selftest.h (assert_streq): Rename "expected" and "actual" to
16359 (ASSERT_EQ): Likewise.
16360 (ASSERT_EQ_AT): Likewise.
16361 (ASSERT_KNOWN_EQ): Likewise.
16362 (ASSERT_KNOWN_EQ_AT): Likewise.
16363 (ASSERT_NE): Likewise.
16364 (ASSERT_MAYBE_NE): Likewise.
16365 (ASSERT_MAYBE_NE_AT): Likewise.
16366 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
16367 the assertion to pass.
16368 (ASSERT_STREQ_AT): Likewise.
16370 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16372 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
16373 interaction with -pie.
16375 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16377 * selftest.h: Fix alphabetization of per-source-file selftest
16380 2018-04-30 Jason Merrill <jason@redhat.com>
16382 PR c++/61982 - dead stores to destroyed objects.
16383 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
16386 2018-04-30 Jason Merrill <jason@redhat.com>
16388 * tree.c (build_clobber): New.
16389 * tree.h: Declare it.
16390 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
16392 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16394 * diagnostic-show-locus.c (layout::layout): Update for
16395 location_get_source_line returning a char_span.
16396 (struct char_span): Move to input.h.
16397 (struct correction): Update for fields in char_span becoming
16399 (struct source_line): Update for location_get_source_line
16400 returning a char_span.
16401 (layout::print_line): Likewise.
16402 * edit-context.c (edited_file::print_content): Likewise.
16403 (edited_file::print_diff_hunk): Likewise.
16404 (edited_file::print_run_of_changed_lines): Likewise.
16405 (edited_file::get_num_lines): Likewise.
16406 (edited_line::edited_line): Likewise.
16407 * final.c (asm_show_source): Likewise.
16408 * input.c (location_get_source_line): Convert return type
16409 from const char * to char_span, losing the final "line_len"
16411 (dump_location_info): Update for the above.
16412 (get_substring_ranges_for_loc): Likewise. Use a char_span
16413 when handling the literal within the line.
16414 (test_reading_source_line): Update for location_get_source_line
16415 returning a char_span.
16416 * input.h (class char_span): Move here from
16417 diagnostic-show-locus.c, converting from a struct to a class.
16418 Make data members private.
16419 (char_span::operator bool): New.
16420 (char_span::length): New.
16421 (char_span::get_buffer): New.
16422 (char_span::operator[]): New.
16423 (char_span::subspan): Make const.
16424 (char_span::xstrdup): New.
16425 (location_get_source_line): Convert return type from const char *
16426 to char_span, losing the final "line_size" param.
16428 2018-04-30 Jan Hubicka <jh@suse.cz>
16430 * lto-wrapper.c (ltrans_priorities): New static var.
16431 (cmp_priority): New.
16432 (run_gcc): Read priorities and if doing parallel build order
16433 the Makefile by them.
16435 2018-04-30 David Malcolm <dmalcolm@redhat.com>
16437 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
16439 2018-04-30 Richard Biener <rguenther@suse.de>
16441 * tree-cfg.c (verify_address): Remove base argument, add
16442 flag whether to check TREE_ADDRESSABLE and do that.
16443 (verify_expr): Remove.
16444 (verify_types_in_gimple_reference): Add pieces from verify_expr.
16445 (verify_gimple_assign_single): Likewise.
16446 (verify_gimple_switch): Likewise.
16447 (verify_expr_location_1): Dereference tp once. Add (disabled)
16448 piece from verify_expr.
16449 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
16451 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16453 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
16455 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16457 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
16458 (small_data_pattern): Likewise.
16459 (arc_rewrite_small_data): Likewise.
16460 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
16461 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
16462 (get_symbol_alignment): New function.
16463 (legitimate_small_data_address_p): Likewise.
16464 (legitimate_scaled_address): Update, call
16465 legitimate_small_data_address_p.
16466 (output_sdata): New static variable.
16467 (arc_print_operand): Update how we handle small data operands.
16468 (arc_print_operand_address): Likewise.
16469 (arc_legitimate_address_p): Update, use
16470 legitimate_small_data_address_p.
16471 (arc_rewrite_small_data_p): Remove.
16472 (arc_rewrite_small_data_1): Likewise.
16473 (arc_rewrite_small_data): Likewise.
16474 (small_data_pattern): Likewise.
16475 (compact_sda_memory_operand): Update to use
16476 legitimate_small_data_address_p and get_symbol_alignment.
16477 (prepare_move_operands): Don't rewite sdata pattern.
16478 (prepare_extend_operands): Remove.
16479 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
16481 (zero_extendqisi2): Likewise.
16482 (zero_extendhisi2): Likewise.
16483 (extendqihi2): Likewise.
16484 (extendqisi2): Likewise.
16485 (extendhisi2): Likewise.
16486 (addsi3): Likewise.
16487 (subsi3): Likewise.
16488 (andsi3): Likewise.
16489 * config/arc/constraints.md (Usd): Change it to memory constraint.
16491 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
16493 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
16494 as source of std instructions.
16495 * config/arc/arc.md (movsi_insn): Update pattern predicate to
16496 allow 6-bit constants as source for store instructions.
16497 (movdi_insn): Update instruction pattern to allow 6-bit constants
16498 as source for store instructions.
16500 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
16502 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
16504 2018-04-30 Nathan Sidwell <nathan@acm.org>
16505 Sandra Loosemore <sandra@codesourcery.com>
16507 * dumpfile.c (dump_open): Allow '-' for stdout.
16508 * doc/invoke.texi (Developer Options): Document dump filename
16509 determination early. Document stdin/stdout selection.
16511 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
16513 Microblaze Target: PIC data text relative
16515 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
16516 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
16518 * config/microblaze/microblaze.h (microblaze_constant_address_p):
16519 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
16520 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
16521 New addressing mode for data-text relative position indepenedent code.
16522 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
16523 'ADDRESS_SYMBOLIC_TXT_REL'.
16524 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
16525 (microblaze_legitimate_pic_operand): Exclude function calls from
16526 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
16527 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
16529 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16530 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
16531 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
16532 for 'address + offset'.
16533 (microblaze_expand_prologue): Add new function prologue call for
16535 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
16536 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
16537 table in case of TARGET_PIC_DATA_TEXT_REL.
16538 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
16539 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
16540 Add new macros 'UNSPEC_TEXT',
16541 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
16542 + exclude function calls from 'UNSPEC_PLT' in case of data text
16544 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16545 new target hook for generating address diff vector tables in case of
16547 * doc/tm.texi : Regenerate.
16548 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
16549 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
16550 of addr diff vector generation.
16551 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
16552 target hook definition.
16553 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
16554 Add default function for generate_pic_addr_diff_vec -> flag_pic.
16555 * doc/invoke.texi (Add new pic option): Add new microblaze pic
16556 option for data text relative.
16558 2018-04-30 Richard Biener <rguenther@suse.de>
16560 * tree-chrec.h (evolution_function_is_constant_p): Remove
16562 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
16564 2018-04-30 Richard Biener <rguenther@suse.de>
16567 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
16569 2018-04-30 Richard Biener <rguenther@suse.de>
16571 PR tree-optimization/28364
16572 PR tree-optimization/85275
16573 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
16574 copying first exit test.
16576 2018-04-28 Mark Wielaard <mark@klomp.org>
16578 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
16579 dwarf_version >= 5.
16580 (dwarf_AT): Handle DW_AT_addr_base.
16581 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
16583 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
16586 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
16587 (*ashl<dwi>3_doubleword_mask_1): Ditto.
16588 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
16589 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
16591 2018-04-28 Richard Biener <rguenther@suse.de>
16593 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
16594 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
16595 to reflect use. Only add interesting stmts.
16597 2018-04-27 Martin Jambor <mjambor@suse.cz>
16600 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
16601 the jump function allows for passing through aggregate values.
16603 2018-04-27 David Malcolm <dmalcolm@redhat.com>
16605 * input.h (in_system_header_at): Convert from macro to inline
16607 (from_macro_expansion_at): Likewise.
16608 (from_macro_definition_at): Likewise.
16610 2018-04-27 Jeff Law <law@redhat.com>
16612 * config.gcc: Mark tile* targets as deprecated/obsolete.
16614 2018-04-27 Richard Biener <rguenther@suse.de>
16616 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
16619 2018-04-27 Richard Biener <rguenther@suse.de>
16621 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
16623 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
16625 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
16626 with Yd constraint. Set "preferred_for_speed" attribute from
16627 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
16628 with Yd constraint.
16629 (*movdi_internal): Ditto.
16630 (movti_interunit splitters): Remove
16631 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
16632 (movdi_interunit splitters): Ditto.
16633 * config/i386/constraints.md (Ye): Remove.
16634 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
16636 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16639 * config/aarch64/constraints.md (Usg): Limit to 31.
16640 (Usj): Limit to 63.
16642 2018-04-27 Jakub Jelinek <jakub@redhat.com>
16644 PR tree-optimization/85529
16645 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
16646 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
16647 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
16648 zero extension or masking of the MSB bit.
16649 (optimize_range_tests): Add FIRST_BB argument, pass it through
16650 to optimize_range_tests_var_bound.
16651 (maybe_optimize_range_tests, reassociate_bb): Adjust
16652 optimize_range_tests callers.
16654 2018-04-26 Richard Biener <rguenther@suse.de>
16655 Jakub Jelinek <jakub@redhat.com>
16657 * cgraph.h (symbol_table): Just declare debug method here.
16658 * symtab.c (symbol_table::debug): Define.
16660 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
16662 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
16664 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
16666 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
16667 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
16668 (*movdi_internal): Substitute Yi and Yj constraint with x
16669 and Ym and Yn constraint with y constraint. Update "isa"
16670 attribute and set "preferred_for_speed" attribute from
16671 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
16672 (*movsi_internal): Ditto.
16673 (*movdf_internal): Ditto.
16674 (*movsf_internal): Ditto.
16675 (*zero_extendsidi2): Ditto.
16676 * config/i386/sse.md (vec_set<mode>_0): Ditto.
16677 (sse2_loadld): Ditto.
16678 (*vec_extract<ssevecmodelower>_0): Ditto.
16679 (*vec_extractv4si_0_zext_sse4): Ditto.
16680 (vec_concatv2di): Ditto.
16681 (*vec_dup<mode>): Ditto.
16682 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
16683 * config/i386/constraints.md (Yi): Remove.
16688 2018-04-26 Nathan Sidwell <nathan@acm.org>
16690 * dumpfile.c (dump_open): New.
16691 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
16692 (dump_finish): Detect stdio/stderr by value not name.
16694 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
16696 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
16698 2018-04-26 Tom de Vries <tom@codesourcery.com>
16701 * config/nvptx/nvptx.c (verify_neutering_jumps)
16702 (verify_neutering_labels): New function
16703 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
16705 2018-04-26 Tom de Vries <tom@codesourcery.com>
16708 * config/nvptx/nvptx.c (needs_neutering_p): New function.
16709 (nvptx_single): Use needs_neutering_p to skip over insns that do not
16712 2018-04-26 Richard Biener <rguenther@suse.de>
16713 Tom de Vries <tom@codesourcery.com>
16716 * lto-streamer-out.c (output_function): Fixup loops if required to match
16717 discovery done in the reader.
16719 2018-04-26 Richard Biener <rguenther@suse.de>
16721 PR tree-optimization/85116
16722 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
16723 have a loop exit from the single latch predecessor. Remove
16724 case of header with just condition.
16725 (ch_base::copy_headers): Exclude infinite loops from any
16727 (pass_ch::execute): Record exits.
16729 2018-04-26 Richard Biener <rguenther@suse.de>
16731 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
16732 prologue cost vector and pass it to vect_get_load_cost.
16733 (vect_get_peeling_costs_all_drs): Likewise.
16734 (vect_peeling_hash_get_lowest_cost): Likewise.
16735 (vect_enhance_data_refs_alignment): Likewise.
16737 2018-04-26 Richard Biener <rguenther@suse.de>
16739 PR middle-end/85450
16740 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
16741 checking of integer<->pointer conversions.
16742 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
16743 sign-/zero-extending pointer types.
16744 (expand_omp_for_static_chunk): Likewise.
16746 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
16747 Jean Lee <xiaoyur347@gmail.com>
16749 * config/mips/mips.c (mips_asan_shadow_offset): New function.
16750 (TARGET_ASAN_SHADOW_OFFSET): Define.
16751 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
16752 true for -fsanitize=address.
16754 2018-04-25 Mark Wielaard <mark@klomp.org>
16756 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
16759 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16761 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
16762 than "alu", remove explicit "memory" and "imm_disp" attributes.
16763 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
16765 PR middle-end/85414
16766 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
16767 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
16768 gen_lowpart_no_emit.
16770 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
16773 * config/i386/i386.c (ix86_expand_builtin): Change memory
16774 operand to XI, extend p0 to Pmode.
16775 * config/i386/i386.md: Change unspec volatile and operand
16776 1 mode to XI, change operand 0 mode to P.
16778 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16780 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
16781 GET_MODE_MASK before any checking.
16782 (nds32_can_use_bset_p): Likewise.
16783 (nds32_can_use_btgl_p): Likewise.
16785 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16787 * config/nds32/nds32-doubleword.md: New define_split pattern for
16788 illegal register number.
16790 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16792 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
16794 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
16796 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
16798 2018-04-25 Richard Biener <rguenther@suse.de>
16800 * lto-streamer.h (LTO_major_version): Bump to 8.
16802 2018-04-25 Jakub Jelinek <jakub@redhat.com>
16804 * BASE-VER: Set to 9.0.0.
16806 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
16808 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
16809 in __abskf2 and __powikf2.
16811 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16814 * config/aarch64/constraints.md (Usg, Usj): New constraints.
16815 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
16816 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
16817 Use the above on operand 2. Reindent.
16818 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
16820 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16823 * common/config/i386/i386-common.c (ix86_handle_option): Don't
16825 * config/i386/i386.opt (mcet): Removed.
16826 * doc/install.texi: Remove -mcet documentation.
16827 * doc/invoke.texi: Likewise.
16829 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
16832 * doc/install.texi: Remove -mcet from bootstrap-cet.
16834 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16837 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
16838 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
16842 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
16843 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
16844 containing a CONST_VECTOR.
16846 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
16848 * doc/install.texi: Update newlib dependency for nvptx.
16850 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16853 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
16854 instead of INTVAL when shifting x left.
16856 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
16858 PR tree-optimization/85478
16859 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
16860 vect_grouped_store_supported for single element vectors.
16862 2018-04-24 Richard Biener <rguenther@suse.de>
16865 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
16866 load cost increase to the case of non-constant step.
16868 2018-04-24 Jakub Jelinek <jakub@redhat.com>
16871 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
16872 destination if any_malformed_asm.
16874 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
16876 PR middle-end/85496
16877 * expr.c (store_field): In the bitfield case, if the value comes from
16878 a function call and is returned in registers by means of a PARALLEL,
16879 do not change the mode of the temporary unless BLKmode and VOIDmode.
16881 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
16883 PR rtl-optimization/85423
16884 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
16885 dependencies to debug insns when the previous insn is non-debug.
16887 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16889 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
16890 enums into a single definition.
16891 (fls): Fix predicates and printing.
16894 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
16896 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
16897 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
16898 and short u6 immediate.
16899 (check_if_valid_sleep_operand): Remove.
16900 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
16902 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16904 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
16905 flag_always_save_lp condition.
16906 * config/nds32/nds32.opt (malways-save-lp): New option.
16908 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16910 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
16911 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
16912 * config/nds32/nds32.h
16913 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
16914 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
16916 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16918 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
16919 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
16921 2018-04-22 Shiva Chen <shiva0217@gmail.com>
16922 Chung-Ju Wu <jasonwucj@gmail.com>
16924 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
16926 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
16927 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
16929 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16931 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
16933 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16935 * config/nds32/nds32-protos.h (nds32_data_alignment,
16936 nds32_local_alignment): Declare.
16937 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
16938 nds32_local_alignment): New functions.
16939 (TARGET_CONSTANT_ALIGNMENT): Define.
16940 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
16942 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16944 * config/nds32/nds32.c
16945 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
16946 (TARGET_MODES_TIEABLE_P): Likewise.
16948 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
16950 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
16951 level Ofast and Og.
16953 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
16954 Chung-Ju Wu <jasonwucj@gmail.com>
16956 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16957 for unaligned access.
16958 * config/nds32/nds32-intrinsic.c: Implementation of expanding
16960 * config/nds32/nds32-intrinsic.md: Likewise.
16961 * config/nds32/nds32_intrinsic.h: Likewise.
16962 * config/nds32/nds32.h (nds32_builtins): Likewise.
16963 * config/nds32/nds32.opt (munaligned-access): New option.
16964 * config/nds32/nds32.c (nds32_asm_file_start): Display
16965 flag_unaligned_access status.
16967 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16969 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16970 -mno-relax is present.
16971 * config/riscv/linux.h (LINK_SPEC): Ditto.
16973 2018-04-20 Martin Sebor <msebor@redhat.com>
16976 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16978 (gimple_fold_builtin_stxcpy_chk): Same.
16979 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16981 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
16984 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16985 __powikf2 when long double is IEEE 128-bit.
16987 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16989 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16990 step to make sure stack always aligned.
16992 2018-04-20 Carl Love <cel@us.ibm.com>
16995 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16996 size check for arg0.
16998 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
16999 Tom de Vries <tom@codesourcery.com>
17002 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
17003 Emit insns for calls too.
17004 (nvptx_find_par): Always look for worker-level predecessor insn.
17005 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
17007 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
17008 (nvptx_process_pars): Propagate frames for calls.
17010 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
17013 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
17015 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
17016 (ix86_handle_option): Don't handle OPT_mibt.
17017 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
17019 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
17021 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
17022 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
17023 (ix86_target_macros): Define __CET__ with flag_cf_protection
17024 for -fcf-protection.
17025 * config/i386/i386.c (isa2_opts): Remove -mibt.
17026 * config/i386/i386.h (TARGET_IBT): Removed.
17027 (TARGET_IBT_P): Likewise.
17028 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
17029 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
17030 * config/i386/i386.opt (mcet): Update help message.
17031 (mshstk): Likewise.
17033 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
17034 -mcet as an alias for -mshstk.
17036 2018-04-20 Richard Biener <rguenther@suse.de>
17038 PR middle-end/85475
17039 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
17040 complexity by forcing a single use of the multiply operand.
17042 2018-04-20 Martin Jambor <mjambor@suse.cz>
17045 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
17046 recursion dependency to only apply to non-clones.
17048 2018-04-20 Martin Jambor <mjambor@suse.cz>
17051 * ipa-cp.c (create_specialized_node): Check that clones of
17052 self-recursive edges exist during IPA-CP.
17054 2018-04-19 Toon Moene <toon@moene.org>
17056 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
17059 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17061 PR tree-optimization/85467
17062 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
17063 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
17064 VECTOR_CST element to type.
17066 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17069 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
17070 * config/i386/i386.md (builtin_setjmp_setup): Removed.
17071 (builtin_longjmp): Likewise.
17072 (save_stack_nonlocal): New pattern.
17073 (restore_stack_nonlocal): Likewise.
17075 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17078 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17079 Replace ASM_OUTPUT_LABEL with fprintf.
17081 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
17084 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
17085 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
17086 * config/i386/i386-c.c (ix86_target_macros_internal): Also
17087 define __IBT__ and __SHSTK__ for -fcf-protection.
17088 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
17090 (ix86_trampoline_init): Likewise.
17091 (x86_output_mi_thunk): Likewise.
17092 (ix86_notrack_prefixed_insn_p): Likewise.
17093 (ix86_option_override_internal): Don't disallow -fcf-protection.
17094 * config/i386/i386.md (rdssp<mode>): Also enable for
17096 (incssp<mode>): Likewise.
17097 (nop_endbr): Likewise.
17098 * config/i386/i386.opt (mcet): Change help message to built-in
17101 (mshstk): Likewise.
17102 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
17103 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
17104 enable CET built-in functions.
17106 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
17108 * common/config/i386/i386-common.c
17109 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
17110 OPTION_MASK_ISA_MOVDIRI_UNSET,
17111 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
17112 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
17113 * config.gcc (movdirintrin.h): New header.
17114 * config/i386/cpuid.h (bit_MOVDIRI,
17115 bit_MOVDIR64B): New bits.
17116 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
17118 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
17119 (VOID, PVOID, PCVOID)): New function types.
17120 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
17121 __builtin_ia32_directstoreu_u64,
17122 __builtin_ia32_movdir64b): New builtins.
17123 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
17124 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
17126 (ix86_valid_target_attribute_inner_p): Ditto.
17127 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
17128 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
17129 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
17130 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
17131 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
17132 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
17133 (movdiri<mode>, movdir64b_<mode>): New.
17134 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
17135 * config/i386/immintrin.h: Include movdirintrin.h.
17136 * config/i386/movdirintrin.h: New file.
17137 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
17139 2018-04-19 Richard Biener <rguenther@suse.de>
17141 PR middle-end/85455
17142 * cfg.c (clear_bb_flags): When loop state says we have
17143 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
17145 2018-04-19 Richard Biener <rguenther@suse.de>
17147 PR tree-optimization/84737
17148 * tree-vect-data-refs.c (vect_copy_ref_info): New function
17149 copying restrict info.
17150 (vect_setup_realignment): Use it.
17151 * tree-vectorizer.h (vect_copy_ref_info): Declare.
17152 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
17153 the first DR to all generated stores.
17154 (vectorizable_load): Likewise for loads.
17156 2018-04-19 Jakub Jelinek <jakub@redhat.com>
17158 PR tree-optimization/85446
17159 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
17160 the integral and pointer types to have the same precision.
17162 * doc/install.texi: Document --disable-cet being the default and
17165 2018-04-18 Martin Liska <mliska@suse.cz>
17167 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
17170 2018-04-18 Martin Liska <mliska@suse.cz>
17173 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
17176 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
17177 arguments if they are comparable.
17179 2018-04-18 Martin Liska <mliska@suse.cz>
17182 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
17185 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
17188 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
17191 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
17192 ENDBR after calling __morestack.
17194 2018-04-18 David Malcolm <dmalcolm@redhat.com>
17197 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
17198 by using gcc_base_ver to generate a gcc_driver_version, and use
17199 it when generating GCC_DRIVER_NAME.
17200 * configure: Regenerate.
17202 2018-04-18 Jakub Jelinek <jakub@redhat.com>
17205 * config.gcc: Obsolete powerpc*-*-*spe*.
17207 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17210 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
17211 (stabstr_D): Change type of unum from unsigned int to
17212 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
17215 2018-04-17 Jim Wilson <jimw@sifive.com>
17218 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
17219 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
17220 Set arg_pointer_offset after using pretend_args_size.
17222 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17224 PR rtl-optimization/85431
17225 * dse.c (record_store): Ignore zero width stores.
17228 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
17229 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
17230 __builtin_stack_restore rather than after it.
17231 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
17232 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
17233 argument instead of virtual_dynamic_stack_rtx.
17235 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
17237 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
17239 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
17240 Add note to error message to explain internal mapping of overloaded
17241 built-in function name to non-overloaded built-in function name.
17242 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
17245 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
17248 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
17249 where the inputs overlap with the output.
17251 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17253 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
17254 (=v, v) alternative and explicit "memory" attribute.
17255 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
17256 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17258 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
17259 "sselog1" type instead of "sselog".
17260 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
17261 "sselog". Remove explicit "memory" attribute.
17262 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
17263 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17265 (vec_extract_hi_v32hi): Merge all alternatives into one, use
17266 "sselog1" type instead of "sselog". Remove explicit "memory"
17268 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
17269 use "sselog1" type instead of "sselog". Remove explicit "memory"
17271 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
17272 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
17274 (vec_extract_hi_v64qi): Merge all alternatives into one, use
17275 "sselog1" type instead of "sselog". Remove explicit "memory"
17277 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
17278 use "sselog1" type instead of "sselog". Remove explicit "memory"
17282 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
17284 PR middle-end/85414
17285 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
17288 2018-04-17 Martin Jambor <mjambor@suse.cz>
17291 * ipa-cp.c (create_specialized_node): Call
17292 expand_all_artificial_thunks if necessary.
17294 2018-04-17 Martin Liska <mliska@suse.cz>
17297 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
17298 in message, remote space in between '_G' and '('.
17300 2018-04-17 Jakub Jelinek <jakub@redhat.com>
17303 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
17304 avx512f_vmcmp<mode>3<round_saeonly_name>,
17305 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
17306 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
17307 avx512f_rndscale<mode><round_saeonly_name>,
17308 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
17309 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17310 Use %<iptr>2 instead of %2 for -masm=intel.
17311 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
17312 avx512f_vcvttss2usi<round_saeonly_name>,
17313 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
17315 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
17316 avx512f_vcvttsd2usi<round_saeonly_name>,
17317 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
17318 Use %q1 instead of %1 for -masm=intel.
17319 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
17320 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
17321 of %3 for -masm=intel.
17322 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
17324 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
17326 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
17328 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
17329 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
17331 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
17333 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
17334 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
17335 %g1 and one with %0 and %1.
17336 (avx512er_vmrcp28<mode><round_saeonly_name>,
17337 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
17338 %1 for -masm=intel.
17339 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
17340 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
17341 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
17342 of %0 and %{%4%} for -masm=intel.
17343 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
17344 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
17345 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
17346 order of %0 and %{%5%}%{z%} for -masm=intel.
17348 2018-04-17 Jan Hubicka <jh@suse.cz>
17351 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
17353 2018-04-17 Martin Liska <mliska@suse.cz>
17356 * multiple_target.c (create_dispatcher_calls): Set apostrophes
17357 for target_clone error message. Make default implementation
17358 clone to be a local declaration.
17359 (separate_attrs): Add new argument and check for an empty
17361 (expand_target_clones): Handle it.
17362 (ipa_target_clone): Make redirection just for target_clones
17365 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
17366 Tom de Vries <tom@codesourcery.com>
17368 PR middle-end/84955
17369 * omp-expand.c (expand_oacc_for): Add dummy false branch for
17370 tiled basic blocks without omp continue statements.
17372 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
17375 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
17376 vec_extract expression as having side effects to make sure it gets
17379 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
17382 * config/i386/i386.c (get_builtin_code_for_version): Check
17385 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
17388 * config.gcc: Support "skylake".
17389 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17391 * config/i386/i386.c (m_SKYLAKE): Define.
17392 (processor_target_table): Add "skylake".
17393 (ix86_option_override_internal): Add "skylake".
17394 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
17395 PROCESSOR_CANNONLAKE.
17396 (get_builtin_code_for_version): Fix priority for
17397 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
17398 PROCESSOR_SKYLAKE-AVX512.
17399 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
17400 (processor_type): Add PROCESSOR_SKYLAKE.
17402 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
17403 Jason Merrill <jason@redhat.com>
17406 * convert.c (convert_to_integer_1): Use direct recursion for
17407 enumeral types and types with a precision less than the number
17408 of bits in their mode.
17410 2018-04-16 Julia Koval <julia.koval@intel.com>
17413 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
17414 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
17416 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
17419 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
17420 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
17421 and -mno-direct-move.
17423 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
17426 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
17427 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
17428 Ensure negative shifts result in {0}.
17430 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
17432 PR rtl-optimization/79916
17433 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
17434 regs (if any) to define how to gnerate SD moves when LRA is in
17437 2018-04-13 Jakub Jelinek <jakub@redhat.com>
17439 PR rtl-optimization/85393
17440 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
17441 * except.c (expand_dw2_landing_pad_for_region): Make static.
17442 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
17443 a label and unconditional jump to old_bb, rather than
17444 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
17447 PR rtl-optimization/85376
17448 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
17449 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
17450 instead of a specific value.
17452 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17453 Bin Cheng <bin.cheng@arm.com>
17455 PR tree-optimization/82965
17456 PR tree-optimization/83991
17457 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
17458 by_profile_only parameter.
17459 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
17460 information if the loop was predicted to iterate too many times.
17461 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
17463 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
17466 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
17469 2018-04-13 Martin Liska <mliska@suse.cz>
17470 Jakub Jelinek <jakub@redhat.com>
17472 PR middle-end/81657
17473 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
17474 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
17475 * builtins.c (expand_builtin_memory_copy_args): Use
17476 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
17477 handle dest_addr == pc_rtx.
17479 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
17482 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
17483 asked to not generate direct moves.
17484 (fix_trunc<mode>si2_stfiwx): Similar.
17485 (fix_trunc<mode>si2_internal): Similar.
17487 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17490 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
17491 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
17492 lookup if dest in some wider mode is known to be const0_rtx and
17493 if so, record permanent equivalence for it to be ZERO_EXTEND of
17494 the narrower mode destination.
17496 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17498 * lto-streamer-out.c (output_function): Revert 259346.
17499 * omp-expand.c (expand_oacc_for): Likewise.
17501 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
17503 PR rtl-optimization/85354
17504 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
17505 * sel-sched.c (sel_global_init): ... here.
17507 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
17510 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
17511 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
17512 mode for PE-COFF targets.
17513 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
17514 (i386_pe_asm_lto_end): Likewise.
17515 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
17516 (TARGET_ASM_LTO_END): Likewise.
17517 * config/i386/winnt.c (saved_debug_info_level): New static variable.
17518 (i386_pe_asm_lto_start): New function.
17519 (i386_pe_asm_lto_end): Likewise.
17521 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
17522 Richard Biener <rguenther@suse.de>
17524 PR middle-end/84955
17525 * lto-streamer-out.c (output_function): Fix CFG loop state before
17527 * omp-expand.c (expand_oacc_for): Handle calls to internal
17528 functions like regular functions.
17530 2018-04-12 Richard Biener <rguenther@suse.de>
17533 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
17534 for the early LTO debug to properly generate references to it
17535 during DIE emission. Do not re-use that for the skeleton for
17537 (dwarf2out_early_finish): Likewise.
17539 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17542 * config/i386/sse.md
17543 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
17544 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
17545 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
17546 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
17547 and output is a reg, avoid creating invalid lowpart subreg, but
17548 instead split into a 512-bit move. Don't split if not AVX512VL,
17549 input is xmm16+ reg and output is a mem.
17550 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
17551 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
17552 xmm16+ reg and output is a mem.
17554 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17556 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
17557 also for flag_dwarf2_cfi_asm.
17559 2018-04-12 Jakub Jelinek <jakub@redhat.com>
17561 PR rtl-optimization/85342
17562 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
17563 a bool scalar var inside of the loop instead. Don't try to update
17564 recog_data.operand after failed apply_change_group.
17566 2018-04-12 Tom de Vries <tom@codesourcery.com>
17569 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
17570 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
17571 array with flexible array member as array without given dimension.
17572 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
17573 argument for undefined param to true.
17575 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
17578 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17579 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
17580 from PowerPC section.
17581 * config/rs6000/sysv4.opt (mcall-): Improve help text.
17582 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
17583 help text that is too long.
17584 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
17585 help text that is too long.
17586 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
17587 help text that is too long.
17589 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
17591 * config/alpha/alpha.md (stack_probe_internal): Rename
17592 from "probe_stack". Update all callers.
17594 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17596 PR rtl-optimization/84566
17597 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
17598 sched_macro_fuse_insns.
17600 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17603 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
17604 (compute_block_dependences): ... from here.
17606 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17608 PR tree-optimization/85331
17609 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
17610 from int to HOST_WIDE_INT.
17612 2018-04-11 Martin Jambor <mjambor@suse.cz>
17615 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
17616 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
17617 not the same as the source val.
17618 (cgraph_edge_brings_value_p): New parameter.
17619 (gather_edges_for_value): Pass destination value to
17620 cgraph_edge_brings_value_p.
17621 (perhaps_add_new_callers): Likewise.
17622 (get_info_about_necessary_edges): Likewise and exclude values brought
17623 only by self-recursive edges.
17624 (create_specialized_node): Redirect only clones of self-calling edges.
17625 (+self_recursive_pass_through_p): New function.
17626 (find_more_scalar_values_for_callers_subset): Use it.
17627 (find_aggregate_values_for_callers_subset): Likewise.
17628 (known_aggs_to_agg_replacement_list): Removed.
17629 (decide_whether_version_node): Re-calculate known constants for all
17630 remaining context clones.
17632 2018-04-11 Richard Biener <rguenther@suse.de>
17635 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
17636 from early DWARF output.
17637 (dwarf2out_early_finish): Output line info unconditionally into
17638 early DWARF and add reference to it.
17640 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17643 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
17644 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
17645 other than V2DFmode using iptr mode attribute.
17646 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
17648 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
17650 PR rtl-optimization/84659
17651 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
17653 2018-04-11 Jakub Jelinek <jakub@redhat.com>
17656 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
17658 (output_loc_list): Pass address of a dummy size variable even in the
17659 locview handling loop.
17660 (index_location_lists): Add comment on why skip_loc_list_entry can't
17663 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
17666 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
17669 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
17672 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
17673 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
17674 and -mstring-compare-inline-limit.
17676 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17679 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
17680 for stack clash protection in a register whenever we need it to be in
17683 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17685 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
17686 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
17688 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
17691 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
17693 (mlong-double-): Ditto.
17694 * config/rs6000/sysv4.opt (msdata=): Ditto.
17695 (mtls-size=): Ditto.
17697 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
17699 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17700 erroneous entries for
17701 "vector int vec_ldl (int, long int *)", and
17702 "vector unsigned int vec_ldl (int, unsigned long int *)".
17703 Add comments and entries for
17704 "vector bool char vec_ldl (int, bool char *)",
17705 "vector bool short vec_ldl (int, bool short *)",
17706 "vector bool int vec_ldl (int, bool int *)",
17707 "vector bool long long vec_ldl (int, bool long long *)",
17708 "vector pixel vec_ldl (int, pixel *)",
17709 "vector long long vec_ldl (int, long long *)",
17710 "vector unsigned long long vec_ldl (int, unsigned long long *)".
17711 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
17712 type tree bool_long_long_type_node and correct definition of
17713 bool_V2DI_type_node to make reference to this new type tree.
17714 (rs6000_mangle_type): Replace erroneous reference to
17715 bool_long_type_node with bool_long_long_type_node.
17716 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
17717 comments to emphasize sign distinctions for char and int types and
17718 replace RS6000_BTI_bool_long constant with
17719 RS6000_BTI_bool_long_long constant. Also add comment to restrict
17720 use of RS6000_BTI_pixel.
17721 (bool_long_type_node): Remove this macro definition.
17722 (bool_long_long_type_node): New macro definition
17724 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17726 PR rtl-optimization/85300
17727 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
17728 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
17729 simplify_unary_operation fails.
17731 2018-04-10 Martin Liska <mliska@suse.cz>
17733 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
17734 cgraph_edge and ipa_ref.
17736 2018-04-10 Jakub Jelinek <jakub@redhat.com>
17740 * config/i386/sse.md
17741 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
17742 computation of the VEC_MERGE selector from mask.
17743 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
17744 Fix decoding of the VEC_MERGE selector into mask.
17746 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17748 PR tree-optimization/85286
17749 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
17751 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
17753 * final.c (final_1): Set insn_last_address as well as
17754 insn_current_address.
17756 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17759 * explow.c (emit_stack_probe): Call validize_mem on memory location
17760 before passing it to gen_probe_stack. Create address operand and
17761 legitimize it for the probe_stack_address case.
17763 2018-04-09 Jan Hubicka <jh@suse.cz>
17766 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
17767 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
17768 * tree.c (free_lang_data_in_type): Fix handling of binfos;
17770 (free_lang_data): Rebuild type inheritance graph.
17772 2018-04-09 Martin Sebor <msebor@redhat.com>
17774 * invoke.texi (-finline-small-functions): Mention other optimization
17776 (-findirect-inlining, -fpartial-inlining): Same.
17777 (-finline-functions-called-once): Same.
17778 (-freorder-blocks-and-partition): Same.
17780 2018-04-09 Jan Hubicka <jh@suse.cz>
17783 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
17784 jumps; choose last target that matches the criteria (i.e.
17785 no partition changes for non-crossing jumps).
17786 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
17787 support for redirecting crossing jumps to non-crossing.
17789 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
17791 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
17792 also for naked functions.
17794 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
17796 * config/arc/arc.md (add_shift): New pattern.
17797 (add_shift2): Likewise.
17798 (sub_shift): Likewise.
17799 (sub_shift_cmp0_noout): Likewise.
17800 (compare_si_ashiftsi): Likewise.
17801 (xbfu_cmp0_noout): New combine pattern.
17802 (xbfu_cmp0"): Likewise.
17803 (movsi_set_cc_insn): Place the predicable variant first.
17804 (commutative_binary_cmp0_noout): Remove clobber.
17805 (commutative_binary_cmp0): New pattern.
17806 (noncommutative_binary_cmp0): Likewise.
17807 (noncommutative_binary_cmp0_noout): Likewise.
17808 (noncommutative_binary_comparison_result_used): Removed.
17809 (rsub_cmp0): New pattern.
17810 (rsub_cmp0_noout): Likewise.
17811 (extzvsi): Changed, keep only meaningful variants.
17812 (SQH, SEZ): New iterators.
17813 (SQH_postfix): New mode attribute.
17814 (SEZ_prefix): New code attribute.
17815 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
17816 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
17817 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
17818 of numerical value.
17819 (noncommutative_operator): Check the availability of barrel
17822 2018-04-09 Richard Biener <rguenther@suse.de>
17824 PR tree-optimization/85284
17825 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
17826 Only use the niter constraining form of simple_iv when the exit
17827 is always executed.
17829 2018-04-09 Tom de Vries <tom@codesourcery.com>
17832 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
17833 (define_expand "*memory_barrier"): New define_expand.
17834 (define_insn "memory_barrier"): New insn.
17836 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17838 PR rtl-optimization/80463
17839 PR rtl-optimization/83972
17840 PR rtl-optimization/83480
17842 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
17843 correct producer for the insn.
17844 (tidy_control_flow): Fixup seqnos in case of debug insns.
17846 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17848 PR rtl-optimization/83913
17850 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
17851 different sched-times when merging exprs.
17853 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17855 PR rtl-optimization/83962
17857 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
17858 tidy_fallthru_edge and tidy_control_flow.
17860 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
17862 PR rtl-optimization/83530
17864 * sel-sched.c (force_next_insn): New global variable.
17865 (remove_insn_for_debug): When force_next_insn is true, also leave only
17866 next insn in the ready list.
17867 (sel_sched_region): When the region wasn't scheduled, make another pass
17868 over it with force_next_insn set to 1.
17870 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17872 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
17874 * config/nds32/constants.md (unspec_volatile_element): Add enum values
17875 for interrupt control.
17876 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17877 functions for interrupt control.
17878 * config/nds32/nds32-intrinsic.md: Likewise.
17879 * config/nds32/nds32_intrinsic.h: Likewise.
17880 * config/nds32/nds32.h (nds32_builtins): Likewise.
17882 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17884 * config/nds32/nds32.c (nds32_init_machine_status,
17885 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
17886 strict_aligned_p field.
17887 (nds32_expand_to_rtl_hook): New function.
17888 (TARGET_EXPAND_TO_RTL_HOOK): Define.
17889 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
17891 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17892 Chung-Ju Wu <jasonwucj@gmail.com>
17894 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
17895 * config/nds32/nds32-n7.md: New file.
17896 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
17897 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
17899 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
17900 * config/nds32/nds32.md (pipeline_model): Add n7.
17901 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
17902 * config/nds32/pipelines.md: Include n7 settings.
17904 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17905 Chung-Ju Wu <jasonwucj@gmail.com>
17907 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
17908 * config/nds32/nds32-e8.md: New file.
17909 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
17910 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
17912 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
17913 * config/nds32/nds32.md (pipeline_model): Add e8.
17914 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
17915 * config/nds32/pipelines.md: Include e8 settings.
17917 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17918 Chung-Ju Wu <jasonwucj@gmail.com>
17920 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
17921 * config/nds32/nds32-n8.md: New file.
17922 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
17923 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
17925 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
17926 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
17927 * config/nds32/nds32.md (pipeline_model): Add n8.
17928 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
17929 * config/nds32/pipelines.md: Include n8 settings.
17931 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
17932 Chung-Ju Wu <jasonwucj@gmail.com>
17934 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
17935 * config/nds32/nds32-n9-2r1w.md: New file.
17936 * config/nds32/nds32-n9-3r2w.md: New file.
17937 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
17938 nds32_register_ports): New or modify for cpu n9.
17939 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
17941 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
17942 * config/nds32/nds32-utils.c: New file.
17943 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
17944 TARGET_MUL_SLOW): Define.
17945 * config/nds32/nds32.md (pipeline_model): New attribute.
17946 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
17947 New options that support cpu n9.
17948 * config/nds32/pipelines.md: Include n9 settings.
17949 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
17951 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
17953 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
17954 information if necessary.
17955 (output_cond_branch_compare_zero): Likewise.
17956 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
17957 (nds32_target_alignment): Refine for alignment.
17958 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
17959 (FUNCTION_BOUNDARY): Modify.
17960 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
17962 * config/nds32/nds32.opt (malways-align, malign-functions): New.
17964 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17966 * config/nds32/constants.md (unspec_volatile_element): Add values for
17967 TLB operation and data prefetch.
17968 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17969 functions for TLB operation and data prefetch.
17970 * config/nds32/nds32-intrinsic.md: Likewise.
17971 * config/nds32/nds32_intrinsic.h: Likewise.
17972 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17973 (nds32_print_operand): Likewise.
17974 * config/nds32/nds32.h (nds32_builtins): Likewise.
17976 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
17977 Andrew Pinski <pinsika@gcc.gnu.org>
17979 PR middle-end/82976
17980 * match.pd: Use constant_boolean_node of correct type instead of
17981 boolean_true_node or boolean_false_node for simplifying
17982 pointer comparisons to zero.
17984 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17986 PR tree-optimization/80021
17987 * tree.c (verify_type_variant): Make error call in verify_variant_match
17988 translatable and remove final full stop.
17990 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17992 * config/nds32/constants.md (unspec_volatile_element): Add
17993 UNSPEC_VOLATILE_EH_RETURN.
17994 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17995 nds32_output_stack_pop): Support dwarf exception handling process.
17996 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17997 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17998 exception handling process.
17999 (nds32_compute_stack_frame): Likewise.
18000 (nds32_return_addr_rtx): Likewise.
18001 (nds32_initial_elimination_offset): Likewise.
18002 (nds32_expand_prologue): Likewise.
18003 (nds32_expand_epilogue): Likewise.
18004 (nds32_dynamic_chain_address): New function.
18005 * config/nds32/nds32.h (machine_function): Add fields for dwarf
18006 exception handling.
18007 (DYNAMIC_CHAIN_ADDRESS): Define.
18008 (EH_RETURN_DATA_REGNO): Define.
18009 (EH_RETURN_STACKADJ_RTX): Define.
18010 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
18011 patterns for dwarf exception handling.
18013 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18015 * config/nds32/nds32.h: Clean up obsolete macros.
18017 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18019 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18020 Add enum values for particular instructions.
18021 * config/nds32/nds32-intrinsic.c: Implementation of expanding
18022 particular intrinsic functions.
18023 * config/nds32/nds32-intrinsic.md: Likewise.
18024 * config/nds32/nds32_intrinsic.h: Likewise.
18025 * config/nds32/nds32.h (nds32_builtins): Likewise.
18026 * config/nds32/nds32.md (type): Add pbsad and pbsada.
18027 (btst, ave): New patterns for particular instructions.
18029 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18031 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
18032 Add enum values for atomic load/store and memory sync.
18033 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
18035 * config/nds32/nds32-intrinsic.md: Likewise.
18036 * config/nds32/nds32_intrinsic.h: Likewise.
18037 * config/nds32/nds32.h (nds32_builtins): Likewise.
18039 2018-04-07 Jakub Jelinek <jakub@redhat.com>
18041 PR tree-optimization/85257
18042 * fold-const.c (native_encode_vector): If not all elts could fit
18043 and off is -1, return 0 rather than offset.
18044 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
18045 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
18046 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
18047 adjust buffer in native_interpret_expr call.
18049 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18051 * config/nds32/constants.md (unspec_volatile_element): Add cache
18052 control enum values.
18053 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
18054 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
18055 * config/nds32/nds32.c (nds32_cctl_names): New.
18056 (nds32_print_operand): Handle cache control register names.
18057 * config/nds32/nds32.h (nds32_builtins): New enum values.
18058 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
18060 * config/nds32/nds32.md (type): Add mmu.
18061 * config/nds32/pipelines.md (simple_insn): Add mmu.
18063 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18065 * config/nds32/nds32.md (type): Remove call.
18066 * config/nds32/pipelines.md (simple_insn): Likewise.
18068 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18070 * config/nds32/constants.md (unspec_volatile_element): Add
18071 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
18072 UNSPEC_VOLATILE_FMFCFG.
18073 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
18074 description for fmfcfg and fmfcsr.
18075 (bdesc_1arg): Add fmtcsr.
18076 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
18077 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
18078 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
18079 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
18080 unspec_fmfcfg): New patterns.
18081 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
18082 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
18083 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
18084 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
18085 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
18086 __nds32__fmfcfg): Define.
18088 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
18090 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
18091 intrinsic register names.
18092 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
18093 intrinsic register enum values and macros.
18095 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
18097 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
18098 for load/store addressing form.
18099 (nds32_print_operand_address): Likewise.
18101 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
18104 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
18105 based on LABEL_REF. Remove useless assertion.
18106 (pic_address_needs_scratch): Fix formatting.
18107 (sparc_legitimize_pic_address): Minor tweaks.
18108 (sparc_delegitimize_address): Adjust assertion accordingly.
18109 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
18110 into symbolic_operand.
18111 (movsi_high_pic_label_ref): Likewise.
18112 (movsi_lo_sum_pic_label_ref): Likewise.
18113 (movdi_pic_label_ref): Likewise.
18114 (movdi_high_pic_label_ref): Likewise.
18115 (movdi_lo_sum_pic_label_ref): Likewise.
18117 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
18119 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
18120 custom LIB_SPEC setup.
18122 2018-04-06 Ruslan Bukin <br@bsdpad.com>
18123 Kito Cheng <kito.cheng@gmail.com>
18125 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
18126 * config/riscv/freebsd.h: New.
18128 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18130 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
18131 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
18134 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18135 Kito Cheng <kito.cheng@gmail.com>
18137 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
18138 nds32_output_call, nds32_symbol_binds_local_p): New functions.
18139 * config/nds32/nds32-protos.h (nds32_output_call,
18140 nds32_output_return): Declare.
18141 * config/nds32/nds32.md: Refine all the call and return patterns.
18143 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18146 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
18147 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
18149 PR rtl-optimization/84872
18150 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
18151 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
18152 EDGE_CROSSING edge.
18154 2018-04-06 Tamar Christina <tamar.christina@arm.com>
18156 * expr.c (copy_blkmode_to_reg): Revert 254862.
18157 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
18159 2018-04-06 Richard Biener <rguenther@suse.de>
18161 PR middle-end/85244
18162 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
18163 after seeing a component reference with an adjacent field. Treat
18164 refs to arrays at struct end of external decls similar to
18165 refs to unconstrained commons.
18167 2018-04-06 Jakub Jelinek <jakub@redhat.com>
18170 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
18171 look through SAVE_EXPRs with non-side-effects argument. Adjust
18173 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
18176 2018-04-06 Richard Biener <rguenther@suse.de>
18178 PR middle-end/85180
18179 * alias.c (find_base_term): New wrapper around find_base_term
18180 unwinding CSELIB_VAL_PTR changes.
18181 (find_base_term): Do not restore CSELIB_VAL_PTR during the
18184 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18186 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
18188 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
18189 constant definitions.
18190 ("nop"): lr 0,0 -> nopr r0
18191 ("nop_lr0", "nop_lr1"): New insn definitions.
18193 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
18195 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
18196 NDS32_V3PUSH_AVAILABLE_P macro.
18198 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
18199 Chung-Ju Wu <jasonwucj@gmail.com>
18201 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
18202 (nds32*-*-*): Add float and fpu_config into supported_defaults.
18203 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
18204 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
18205 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
18206 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
18207 * config/nds32/constraints.md: New constraints and checking for hard
18208 float configuration.
18209 * config/nds32/iterators.md: New mode iterator and attribute for hard
18210 float configuration.
18211 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
18213 * config/nds32/nds32-fpu.md: New file.
18214 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
18215 deal with hard float code generation.
18216 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
18218 (abi_type, float_reg_number): New enum type.
18219 * config/nds32/nds32-predicates.c: New predicates for hard float.
18220 * config/nds32/nds32-protos.h: Declare functions for hard float.
18221 * config/nds32/nds32.c: Implementation for hard float configuration.
18222 * config/nds32/nds32.h: Definitions for hard float configuration.
18223 * config/nds32/nds32.md: Include hard float machine description and
18224 modify patterns for hard float configuration.
18225 * config/nds32/nds32.opt: New options for hard float configuration.
18226 * config/nds32/predicates.md: New predicates for hard float
18229 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18231 * common/config/nds32/nds32-common.c
18232 (nds32_option_optimization_table): Enable -mreleax-hint by default.
18234 2018-04-05 Jakub Jelinek <jakub@redhat.com>
18236 PR middle-end/85195
18237 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
18238 CONSTRUCTOR_ELT (ctor, ...)->value.
18240 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
18243 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
18245 2018-04-05 Tom de Vries <tom@codesourcery.com>
18248 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
18251 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18252 Kito Cheng <kito.cheng@gmail.com>
18254 * config/nds32/constraints.md (U33): Fine-tune checking condition.
18255 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
18256 * config/nds32/nds32.h (nds32_16bit_address_type): Add
18257 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
18259 2018-04-05 Shiva Chen <shiva0217@gmail.com>
18260 Kito Cheng <kito.cheng@gmail.com>
18262 * config/nds32/constraints.md (Ufe): New memory constraint.
18263 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
18264 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
18265 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
18267 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
18268 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
18270 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18272 * config/nds32/nds32.md: Use optimize_size in the condition for
18273 alu-shift instructions.
18275 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18277 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
18279 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18281 * config/nds32/nds32.md (negsi2): Refine pattern.
18283 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18284 Chung-Ju Wu <jasonwucj@gmail.com>
18286 * config/nds32/iterators.md (shift_rotate): New code iterator.
18287 (shift): New code attribute.
18288 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
18289 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
18290 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
18291 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
18292 bit-wise operations.
18293 (andsi3, *andsi3): Ditto.
18294 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
18295 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
18296 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
18297 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
18298 nds32_ior_operand, nds32_xor_operand): New predicates.
18300 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18302 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
18303 (addsi3, subsi3): ... this.
18305 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18307 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
18309 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18311 * config/nds32/nds32.md: Adjust indention.
18313 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
18315 * config/nds32/nds32.md (feature): New attribute.
18317 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
18319 * config/nds32/nds32.md (subtype): New attribute.
18321 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
18324 * config/arm/arm-builtins.c (arm_expand_builtin): Change
18325 expansion to perform a bitwise AND of the argument followed by a
18326 boolean negation of the result.
18328 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
18330 PR rtl-optimization/84878
18331 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
18332 the basic block. Assert the use reference is not artificial and that
18333 it has an associated insn.
18335 2018-04-04 Michael Matz <matz@suse.de>
18337 * builtins.c (compute_objsize): Pass correct operand
18338 to array_at_struct_end_p.
18340 2018-04-04 Richard Biener <rguenther@suse.de>
18343 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
18344 from contexts for DINFO_LEVEL_TERSE and below.
18346 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18348 * config/nds32/nds32-doubleword.md (move_<mode>): Require
18349 resiter_operand condition.
18350 * config/nds32/nds32.md (*move<mode>): Ditto.
18352 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18353 Monk Chiang <sh.chiang04@gmail.com>
18355 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
18357 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18359 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
18361 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18362 Kito Cheng <kito.cheng@gmail.com>
18364 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
18365 nds32_cond_code_str, output_cond_branch,
18366 output_cond_branch_compare_zero, nds32_expand_cbranch,
18367 nds32_expand_cstore, nds32_expand_movcc,
18368 nds32_output_cbranchsi4_equality_zero,
18369 nds32_output_cbranchsi4_equality_reg,
18370 nds32_output_cbranchsi4_equality_reg_or_const_int,
18371 nds32_output_cbranchsi4_greater_less_zero: New functions.
18372 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
18373 nds32_expand_cstore, nds32_expand_movcc,
18374 nds32_output_cbranchsi4_equality_zero,
18375 nds32_output_cbranchsi4_equality_reg,
18376 nds32_output_cbranchsi4_equality_reg_or_const_int,
18377 nds32_output_cbranchsi4_greater_less_zero): Declare.
18378 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
18379 nds32_rimm11s_operand): New predicates.
18380 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
18381 * config/nds32/nds32.md: Rewrite all the branch and conditional move
18384 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
18386 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
18387 * config/nds32/nds32.md: Ditto.
18388 * config/nds32/pipelines.md: Ditto.
18390 2018-04-04 Richard Biener <rguenther@suse.de>
18392 PR tree-optimization/85168
18393 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
18394 propagating abnormals.
18396 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18398 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
18400 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
18401 Kito Cheng <kito.cheng@gmail.com>
18403 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
18404 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
18405 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
18406 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
18407 * config/nds32/nds32.md (sibcall_internal): New.
18408 (sibcall_register): Remove.
18409 (sibcall_immediate): Remove.
18410 (sibcall_value_internal): New.
18411 (sibcall_value_register): Remove.
18412 (sibcall_value_immediate): Remove.
18413 * config/nds32/predicates.md (nds32_general_register_operand): New.
18414 (nds32_call_address_operand): New.
18416 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18418 PR rtl-optimization/85167
18419 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
18420 bb_defs if *split_p, instead preinitialize it to NULL.
18422 PR tree-optimization/85156
18423 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
18424 evaluating the argument multiple times.
18426 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
18428 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
18430 (_mm_cvtpd_ps): Likewise.
18431 (_mm_cvttpd_epi32): Likewise.
18432 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
18433 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
18434 vector, pixel, and bool following altivec.h include.
18436 2018-04-03 Martin Sebor <msebor@redhat.com>
18438 * doc/extend.texi (Common Function Attributes): Clarify.
18439 (const attribute): Likewise.
18440 (pure attribute): Likewise.
18442 2018-04-03 Jakub Jelinek <jakub@redhat.com>
18445 * config/i386/i386.c (ix86_expand_vector_set): Use
18446 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
18448 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
18450 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
18451 instructions when changing rounding bits to preserve precision bits
18452 in the x87 control word.
18454 2018-04-03 Martin Liska <mliska@suse.cz>
18456 PR tree-optimization/82491
18457 * rtl.h (strip_offset_and_add): Replace += suboffset with
18458 poly_uint64 () + suboffset.
18460 2018-03-29 Martin Liska <mliska@suse.cz>
18461 Martin Jambor <mjambor@suse.cz>
18464 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
18465 param_type is not an integral or pointer type.
18467 2018-04-03 Richard Biener <rguenther@suse.de>
18469 * sese.h (recompute_all_dominators): Remove.
18471 2018-04-02 Martin Sebor <msebor@redhat.com>
18473 * doc/invoke.texi (-Wrestrict): Fix typos.
18475 2018-04-02 Jim Wilson <jimw@sifive.com>
18477 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
18478 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
18479 (<optab>di3, <optab>si3_extend): Likewise.
18480 (<optab>si3_mask, <optab>si3_mask_1): New.
18481 (<optab>di3_mask, <optab>di3_mask_1): New.
18482 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
18483 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
18484 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
18486 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
18488 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
18491 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
18493 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
18494 (nds32_canonicalize_comparison): New function.
18496 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18497 Kito Cheng <kito.cheng@gmail.com>
18498 Kuan-Lin Chen <kuanlinchentw@gmail.com>
18500 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
18501 * config/nds32/constants.md (unspec_volatile_element): Add
18502 UNSPEC_VOLATILE_RELAX_GROUP.
18503 * config/nds32/nds32-relax-opt.c: New file.
18504 * config/nds32/nds32-predicates.c
18505 (nds32_symbol_load_store_p): New function.
18506 * config/nds32/nds32-protos.h
18507 (nds32_symbol_load_store_p): Declare function.
18508 (make_pass_nds32_relax_opt): Declare new rtl pass function.
18509 * config/nds32/nds32.c
18510 (nds32_register_pass): New function to register pass.
18511 (nds32_register_passes): New function to register passes.
18512 * config/nds32/nds32.md (relax_group): New pattern.
18513 * config/nds32/nds32.opt (mrelax-hint): New option.
18514 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
18516 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
18518 * config/nds32/t-nds32: Modify files dependency.
18520 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18522 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
18523 (PROFILE_HOOK): Define its implementation.
18525 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
18527 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
18528 type and 32-bit size.
18530 2018-04-01 Jakub Jelinek <jakub@redhat.com>
18532 PR middle-end/85090
18533 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
18534 (V_128_256): New mode iterator.
18535 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
18536 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
18537 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
18539 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
18540 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
18542 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
18545 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
18546 NaN inputs correctly.
18548 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
18551 * config/rs6000/vsx.md (??r): New mode attribute.
18552 (*vsx_mov<mode>_64bit): Use it.
18553 (*vsx_mov<mode>_32bit): Likewise.
18555 2018-03-30 Martin Sebor <msebor@redhat.com>
18557 PR tree-optimization/84818
18558 * builtins.c (check_access): Use warning_n.
18560 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18563 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
18565 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
18568 2018-03-30 Julia Koval <julia.koval@intel.com>
18571 * x86-tune.def (movx, partial_reg_dependency): Enable for
18574 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
18576 PR inline-asm/84985
18577 * lra-constraints.c (process_alt_operands): Move setting
18578 this_alternative_matches below.
18580 2018-03-29 Martin Liska <mliska@suse.cz>
18583 * doc/invoke.texi: Document how LTO works with debug info.
18584 Describe auto-load support of binutils. Mention 'x86-64'
18585 as valid option value of -march option.
18587 2018-03-29 Jakub Jelinek <jakub@redhat.com>
18589 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
18592 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
18593 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
18594 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
18597 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18600 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
18601 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
18602 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
18603 for RS6000_BTM_POWERPC64.
18604 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
18605 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
18606 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
18609 (DIVDEU): Likewise.
18611 2018-03-28 Carl Love <cel@us.ibm.com>
18614 2017-09-27 Carl Love <cel@us.ibm.com>
18616 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
18617 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
18618 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
18621 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18623 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
18624 instead of __vector bool.
18625 (_mm_max_pu8): Likewise.
18626 (_mm_min_pi16): Likewise.
18628 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
18631 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
18632 (DIVWEUO): Likewise.
18633 (DIVDEO): Likewise.
18634 (DIVDEUO): Likewise.
18635 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
18636 DIVWEUO and DIVDEUO.
18637 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
18638 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
18639 (div_extend): Likewise.
18640 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
18642 (__builtin_divweuo): Likewise.
18643 (__builtin_divdeo): Likewise.
18644 (__builtin_divdeuo): Likewise.
18646 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18649 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
18650 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
18652 PR tree-optimization/82004
18653 * gimple-match-head.c (optimize_pow_to_exp): New function.
18654 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
18655 Don't fold to exp if optimize_pow_to_exp is false.
18657 2018-03-28 Martin Liska <mliska@suse.cz>
18660 * calls.c (initialize_argument_information): Fix trailing space.
18661 * common.opt: Fix typo and provide better explanation for
18662 -fsanitize-coverage option.
18663 * config/i386/i386.opt: Fix typo.
18665 2018-03-28 Jakub Jelinek <jakub@redhat.com>
18666 Martin Liska <mliska@suse.cz>
18669 * gimplify.c (asan_poison_variable): Don't do the check for
18670 gimplify_omp_ctxp here.
18671 (gimplify_decl_expr): Do it here.
18672 (gimplify_target_expr): Likewise.
18674 2018-03-28 Martin Liska <mliska@suse.cz>
18677 * config/i386/i386.c (ix86_function_arg_advance): Do not call
18678 chkp_type_bounds_count if MPX is not enabled.
18680 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
18682 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
18684 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
18687 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
18688 function to create the function decl for complex long double
18689 multiply and divide for -mabi=ieeelongdouble.
18690 (init_float128_ieee): Call it.
18692 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
18695 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
18696 -fcf-protection=branch -mibt.
18697 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
18699 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18702 * config/arm/arm.c (arm_valid_symbolic_address): Handle
18703 arm_word_relocations.
18705 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
18708 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
18709 extern array declarations.
18711 2018-03-27 Richard Biener <rguenther@suse.de>
18713 PR middle-end/84067
18714 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
18715 explicit single_use checks.
18717 2018-03-27 Richard Biener <rguenther@suse.de>
18719 PR tree-optimization/85082
18720 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
18723 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
18725 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
18726 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
18727 Turn on fasynchronous-unwind-tables and funwind-tables.
18729 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
18732 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
18733 (*bmi_blsr_<mode>_ccz): Ditto.
18735 2018-03-26 Tom de Vries <tom@codesourcery.com>
18737 PR tree-optimization/85063
18738 * omp-general.c (offloading_function_p): New function. Factor out
18740 * omp-offload.c (pass_omp_target_link::gate): ... here.
18741 * omp-general.h (offloading_function_p): Declare.
18742 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
18743 with attribute omp declare target for offloading functions.
18745 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
18747 PR tree-optimization/84005
18748 * tree-data-ref.h (get_base_for_alignment): Declare.
18749 * tree-data-ref.c (get_base_for_alignment_1): New function.
18750 (get_base_for_alignment): Likewise.
18751 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
18752 get_base_for_alignment to find a suitable base object, instead
18753 of always using drb->base_address.
18755 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18757 PR inline-asm/85022
18758 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
18759 known size by default.
18761 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
18763 PR inline-asm/85030
18764 * lra-constraints.c (process_alt_operands): Don't match BLKmode
18765 and non BLKmode operands.
18767 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18770 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
18771 Clean up attributes.
18773 2018-03-23 Richard Biener <rguenther@suse.de>
18776 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
18777 we are going to emit early debug for LTO.
18779 2018-03-23 Jakub Jelinek <jakub@redhat.com>
18781 PR inline-asm/85034
18782 * function.c (match_asm_constraints_1): Don't optimize if input
18783 doesn't satisfy general_operand predicate for output's mode.
18785 PR inline-asm/85022
18786 * alias.c (write_dependence_p): Don't require for x_canonicalized
18787 non-VOIDmode if x has VOIDmode.
18790 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
18791 just don't try to optimize it rather than assert it never happens.
18793 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18795 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
18796 macro expansions for definition of ST_INTERNAL_<mode> and
18797 LD_INTERNAL_<mode> builtins.
18798 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
18800 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
18802 (altivec_expand_st_builtin): Likewise.
18803 (altivec_expand_builtin): Remove calls to deleted functions.
18804 (rs6000_address_for_altivec): Delete this function.
18805 * config/rs6000/vector.md: Remove expands for
18806 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
18808 2018-03-22 Sudakshina Das <sudi.das@arm.com>
18811 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
18812 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
18813 re-computing once computed.
18814 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
18815 (arm_init_machine_status): Initialize
18816 machine->static_chain_stack_bytes.
18818 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
18821 * doc/extend.texi: Add four new prototypes for vec_ld.
18822 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
18823 definitions for more logical presentation.
18824 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
18825 entries for V1TI variants of __builtin_altivec_ld builtin.
18826 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
18827 handling of V1TI variant of LVX icode pattern.
18828 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
18829 (rs6000_gimple_fold_builtin): Likewise.
18830 (altivec_init_builtins): Add code to define
18831 __builtin_altivec_lvx_v1ti function.
18833 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18835 PR inline-asm/84941
18836 * function.c (match_asm_constraints_1): Don't do the optimization
18837 if input isn't a REG, SUBREG, MEM or constant.
18839 2018-03-22 Tom de Vries <tom@codesourcery.com>
18841 PR tree-optimization/84956
18842 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
18843 bb_has_abnormal_pred.
18845 2018-03-22 Jakub Jelinek <jakub@redhat.com>
18848 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
18849 DECL_INITIAL (decl) to decl at the end.
18850 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
18851 adjust the comment.
18853 2018-03-21 Joseph Myers <joseph@codesourcery.com>
18855 * doc/extend.texi (__builtin_tgmath): Document when complex
18856 integer types are treated as _Complex _Float64.
18858 2018-03-21 Tom de Vries <tom@codesourcery.com>
18860 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
18862 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18864 PR tree-optimization/84960
18865 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
18866 if it is ENTRY block, move them into single succ of ENTRY in that case.
18868 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
18870 PR tree-optimization/84811
18871 * poly-int.h (poly_span_traits): Remove the T3 parameter and
18872 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
18873 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
18874 (known_subrange_p): Update accordingly. Cast each value involved
18875 in the size comparison, rather than casting the result of the
18878 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18880 PR tree-optimization/84982
18881 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
18882 by flipping the least significant bit rather than all bits from
18883 bitpos to bitpos + bitsize - 1.
18885 2018-03-21 Nathan Sidwell <nathan@acm.org>
18887 * doc/extend.texi (Deprecated Features): Remove mention of
18888 long-deleted deprecations.
18890 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18893 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
18894 * configure: Regenerate.
18896 2018-03-21 Tom de Vries <tom@codesourcery.com>
18898 PR tree-optimization/83126
18899 * tree-parloops.c (num_phis): New function.
18900 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
18902 2018-03-21 Nathan Sidwell <nathan@acm.org>
18904 * doc/extend.texi (Deprecated Features): Update deprecated flags,
18905 mention anon-struct/union members and trailing attributes.
18907 2018-03-21 Bin Cheng <bin.cheng@arm.com>
18909 PR tree-optimization/84969
18910 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
18911 builtin memset partitions if they set different rhs values.
18913 2018-03-21 Jakub Jelinek <jakub@redhat.com>
18915 PR rtl-optimization/84989
18916 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
18917 VEC_DUPLICATE with scalar result mode.
18919 2018-03-21 Martin Liska <mliska@suse.cz>
18922 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
18923 not intended return statement.
18925 2018-03-21 Martin Liska <mliska@suse.cz>
18928 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
18929 (chkp_find_bound_slots_1): Limit number of iterations.
18931 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
18934 * Minor grammar fixes for x86 options.
18936 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18939 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
18940 holding REG_CFA_RESTORE notes, instead turn them into a USE.
18942 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
18945 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
18946 (altivec_lvx_<mode>_1op): Likewise.
18947 (altivec_stvx_<mode>_2op): Likewise.
18948 (altivec_stvx_<mode>_1op): Likewise.
18949 (altivec_lvx_<VM2:mode>): New define_expand.
18950 (altivec_stvx_<VM2:mode>): Likewise.
18951 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
18952 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18953 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
18954 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
18955 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
18956 (rs6000_gen_lvx): Likewise.
18957 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
18958 (altivec_expand_stv_builtin): Likewise.
18959 (altivec_expand_builtin): Likewise.
18960 * config/rs6000/vector.md: Likewise.
18962 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18965 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18968 2018-03-20 Richard Biener <rguenther@suse.de>
18971 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18972 sign-conversions as zero, fall back to standard scalar_stmt
18975 2018-03-20 Martin Liska <mliska@suse.cz>
18978 * predict.c (rebuild_frequencies): Handle case when we have
18979 PROFILE_ABSENT, but flag_guess_branch_prob is false.
18981 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18984 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18985 flag_section_anchors.
18986 * varasm.c (use_blocks_for_decl_p): Remove hack for
18987 dw2_force_const_mem.
18990 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18992 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
18993 be created, use lowpart_subreg of operands[0] rather than operands[0]
18995 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18996 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18997 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18998 and n constraint instead of aarch64_shift_imm_di and Usd.
18999 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
19000 (*aarch64_<optab>_reg_minus<mode>3): ... this.
19002 2018-03-20 Sudakshina Das <sudi.das@arm.com>
19005 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
19006 to favor GPR over NEON registers.
19007 (<shift>di3_neon): Likewise.
19009 2018-03-20 Tom de Vries <tom@codesourcery.com>
19012 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
19013 (nvptx_process_pars): Emit bar.sync asap and alap.
19015 2018-03-20 Tom de Vries <tom@codesourcery.com>
19018 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
19019 seen_label if seen_label is already set.
19021 2018-03-20 Jakub Jelinek <jakub@redhat.com>
19024 * config/i386/i386.c (fold_builtin_cpu): For features above 31
19025 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
19026 Use 1U instead of 1. Formatting fixes.
19029 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
19030 instead of TREE_TYPE (s1) for the return value.
19032 2018-03-19 Jakub Jelinek <jakub@redhat.com>
19034 PR tree-optimization/84946
19035 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
19036 bitsize + bitsize in poly_uint64 rather than poly_int64.
19039 * dwarf2asm.c: Include fold-const.c.
19040 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
19041 of decl rather than decl itself.
19043 PR rtl-optimization/84643
19044 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
19046 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
19049 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
19050 calling assemble_variable.
19052 2018-03-19 Sudakshina Das <sudi.das@arm.com>
19055 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
19056 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
19058 2018-03-19 Jim Wilson <jimw@sifive.com>
19061 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
19062 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
19063 (riscv_first_stack_step): Likewise.
19064 (riscv_option_override): Use STACK_BOUNDARY instead of
19065 MIN_STACK_BOUNDARY.
19066 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
19067 MIN_STACK_BOUNDARY.
19068 (BIGGEST_ALIGNMENT): Set to 128.
19069 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
19070 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
19073 2018-03-19 Richard Biener <rguenther@suse.de>
19075 PR tree-optimization/84933
19076 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
19077 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
19079 2018-03-19 Richard Biener <rguenther@suse.de>
19081 PR tree-optimization/84859
19082 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
19083 (cond_if_else_store_replacement): Perform sinking operation on
19084 single-store BBs regardless of MAX_STORES_TO_SINK setting.
19085 Generalize what a BB with a single eligible store is.
19087 2018-03-19 Richard Biener <rguenther@suse.de>
19089 PR tree-optimization/84929
19090 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
19091 chrec_is_positive against non-chrec arg.
19093 2018-03-19 Tamar Christina <tamar.christina@arm.com>
19096 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
19098 2018-03-18 Martin Liska <mliska@suse.cz>
19100 PR rtl-optimization/84635
19101 * regrename.c (build_def_use): Use matches_mode only when
19104 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
19106 PR tree-optimization/84913
19107 * tree-vect-loop.c (vectorizable_reduction): Don't try to
19108 vectorize chains of COND_EXPRs.
19110 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19112 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
19114 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19116 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
19118 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
19120 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
19122 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19123 Kito Cheng <kito.cheng@gmail.com>
19125 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
19126 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
19127 (nds32_adjust_reg_alloc_order): New function.
19128 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
19130 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
19132 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
19133 nds32_print_operand, nds32_print_operand_address): Use
19134 HOST_WIDE_INT_PRINT_DEC instead.
19136 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
19138 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
19140 2018-03-17 Jakub Jelinek <jakub@redhat.com>
19143 * config/i386/i386.c (initial_ix86_tune_features,
19144 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
19145 unsigned long long.
19146 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
19147 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
19148 rather than 1u << ix86_tune. Formatting fix.
19149 (ix86_option_override_internal): Change ix86_arch_mask from
19150 unsigned int to unsigned HOST_WIDE_INT, initialize to
19151 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
19152 (ix86_function_specific_restore): Likewise.
19154 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19157 * postreload.c (reload_combine_recognize_pattern): Perform
19158 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
19159 truncate_int_for_mode the result for the destination's mode.
19162 * hsa-gen.c (mem_type_for_type): Fix comment typo.
19163 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
19165 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19168 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
19171 * lra-assigns.c (lra_split_hard_reg_for): Don't use
19172 regno_allocno_class_array and sorted_pseudos.
19173 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
19174 insns where regno is used.
19176 2018-03-16 Martin Liska <mliska@suse.cz>
19179 * multiple_target.c (create_dispatcher_calls): Redirect
19180 reference in the symbol table.
19182 2018-03-16 Martin Liska <mliska@suse.cz>
19185 * multiple_target.c (create_dispatcher_calls): Redirect also
19188 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19192 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
19193 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
19194 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
19196 2018-03-16 Julia Koval <julia.koval@intel.com>
19198 * doc/invoke.texi (Skylake Server): Add CLWB.
19199 Cannonlake): Remove CLWB.
19201 2018-03-16 Jakub Jelinek <jakub@redhat.com>
19203 PR tree-optimization/84841
19204 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
19206 (FLOAT_ONE_CONST_TYPE): Define.
19207 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
19208 (sort_by_operand_rank): Put entries with higher constant_type last
19209 rather than first to match comments.
19211 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
19213 * config/nios2/nios2.md (movsi_internal): Fix thinko in
19216 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19219 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
19220 check and use address of target always.
19222 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
19225 * config/i386/i386.c (indirect_thunk_needed): Update comments.
19226 (indirect_thunk_bnd_needed): Likewise.
19227 (indirect_thunks_used): Likewise.
19228 (indirect_thunks_bnd_used): Likewise.
19229 (indirect_return_needed): New.
19230 (indirect_return_bnd_needed): Likewise.
19231 (output_indirect_thunk_function): Add a bool argument for
19233 (output_indirect_thunk_function): Don't generate alias for
19234 function return thunk.
19235 (ix86_code_end): Call output_indirect_thunk_function to generate
19236 function return thunks.
19237 (ix86_output_function_return): Set indirect_return_bnd_needed
19238 and indirect_return_needed instead of indirect_thunk_bnd_needed
19239 and indirect_thunk_needed.
19241 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
19243 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
19244 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
19245 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
19247 2018-03-15 David Malcolm <dmalcolm@redhat.com>
19248 Paul Hua <paul.hua.gm@gmail.com>
19251 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
19253 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
19255 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
19256 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
19257 resp. SFmode cases.
19259 2018-03-15 Tamar Christina <tamar.christina@arm.com>
19262 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
19263 instead of GET_MODE_SIZE when comparing Units.
19265 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19268 * varasm.c (hash_section): Return an unchangeble hash value
19269 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
19270 Return !aarch64_can_use_per_function_literal_pools_p ().
19272 2018-03-15 Jakub Jelinek <jakub@redhat.com>
19275 * optabs.c (emit_conditional_move): Pass address of cmode's copy
19276 rather than address of cmode as last argument to prepare_cmp_insn.
19278 2018-03-15 Julia Koval <julia.koval@intel.com>
19280 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
19281 F_AVX512VNNI, F_AVX512BITALG): New.
19283 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
19286 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
19287 insn for floating-point loads and stores.
19289 2018-03-14 Carl Love <cel@us.ibm.com>
19291 * config/rs6000/rs6000-c.c: Add macro definitions for
19292 ALTIVEC_BUILTIN_VEC_PERMXOR.
19293 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
19294 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
19295 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
19296 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
19298 * config/doc/extend.texi: Add prototypes for vec_permxor.
19300 2018-03-14 David Malcolm <dmalcolm@redhat.com>
19303 * diagnostic-show-locus.c (class layout_point): Convert m_line
19304 from int to linenum_type.
19305 (line_span::comparator): Use linenum "compare" function when
19306 comparing line numbers.
19307 (test_line_span): New function.
19308 (layout_range::contains_point): Convert param "row" from int to
19310 (layout_range::intersects_line_p): Likewise.
19311 (layout::will_show_line_p): Likewise.
19312 (layout::print_source_line): Likewise.
19313 (layout::should_print_annotation_line_p): Likewise.
19314 (layout::print_annotation_line): Likewise.
19315 (layout::print_leading_fixits): Likewise.
19316 (layout::annotation_line_showed_range_p): Likewise.
19317 (struct line_corrections): Likewise for field m_row.
19318 (line_corrections::line_corrections): Likewise for param "row".
19319 (layout::print_trailing_fixits): Likewise.
19320 (layout::get_state_at_point): Likewise.
19321 (layout::get_x_bound_for_row): Likewise.
19322 (layout::print_line): Likewise.
19323 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
19324 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
19325 * input.c (selftest::test_linenum_comparisons): New function.
19326 (selftest::input_c_tests): Call it.
19327 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
19328 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
19329 * selftest.h (ASSERT_GT): New macro.
19330 (ASSERT_GT_AT): New macro.
19331 (ASSERT_LT): New macro.
19332 (ASSERT_LT_AT): New macro.
19334 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
19336 PR rtl-optimization/84780
19337 * combine.c (distribute_links): Don't make a link based on pc_rtx.
19339 2018-03-14 Martin Liska <mliska@suse.cz>
19341 * tree.c (record_node_allocation_statistics): Use
19342 get_stats_node_kind.
19343 (get_stats_node_kind): New function extracted from
19344 record_node_allocation_statistics.
19345 (free_node): Use get_stats_node_kind.
19347 2018-03-14 Richard Biener <rguenther@suse.de>
19349 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
19350 that the value-set of ANTIC_IN doesn't grow.
19353 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19355 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19356 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19358 2018-03-14 Julia Koval <julia.koval@intel.com>
19360 * config.gcc (icelake-client, icelake-server): New.
19362 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
19363 (initial_ix86_arch_features): Ditto.
19364 (PTA_SKYLAKE): Add SGX.
19365 (PTA_ICELAKE): Remove.
19366 (PTA_ICELAKE_CLIENT): New.
19367 (PTA_ICELAKE_SERVER): New.
19368 (ix86_option_override_internal): Split up icelake on icelake client and
19370 (get_builtin_code_for_version): Ditto.
19371 (fold_builtin_cpu): Ditto.
19372 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
19373 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
19374 * config/i386/i386.h (processor_type): Ditto.
19375 * doc/invoke.texi: Ditto.
19377 2018-03-14 Jakub Jelinek <jakub@redhat.com>
19380 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
19381 INTEGER_CST offset, add it together with bitpos / 8 and
19382 sign extend based on POINTER_SIZE.
19386 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
19389 * config/i386/constraints.md (Yc): New register constraint.
19390 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
19391 Use Yc constraint for alternative 2 of operand 0. Remove
19392 preferred_for_speed attribute.
19394 2018-03-14 Richard Biener <rguenther@suse.de>
19396 PR tree-optimization/84830
19397 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
19398 with the old one to avoid oscillations.
19400 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
19403 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
19405 (assign_by_spills): Return a flag of reload assignment failure.
19406 Do not process the reload assignment failures. Do not spill other
19407 reload pseudos if they has the same reg class. Update n if
19409 (lra_assign): Add a return arg. Set up from the result of
19410 assign_by_spills call.
19411 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19412 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19413 usage_insns if it is not NULL.
19414 (spill_hard_reg_in_range): New function.
19415 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19416 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19417 function prototypes.
19418 (lra_assign): Change prototype.
19419 * lra.c (lra): Add code to deal with fails by splitting hard reg
19422 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
19424 * config/riscv/riscv.opt (mrelax): New option.
19425 * config/riscv/riscv.c (riscv_file_start): Emit ".option
19426 "norelax" when riscv_mrelax is disabled.
19427 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
19429 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19432 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
19433 reassociation for int modes.
19435 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19437 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
19438 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
19440 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
19441 * config/aarch64/aarch64-sve.md
19442 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
19443 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
19444 (*extend<mode><Vwide>2): Rename to...
19445 (aarch64_sve_extend<mode><Vwide>2): ...this.
19446 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
19447 renaming the old pattern to...
19448 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
19450 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
19451 define_expand, renaming the old pattern to...
19452 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
19453 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
19454 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
19455 account when deciding which SVE instruction the optab should use.
19456 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
19458 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19460 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
19461 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
19462 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
19463 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
19464 (tlsdesc_small_<mode>): Turn a define_expand and use
19465 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
19466 (tlsdesc_small_advsimd_<mode>): ...this.
19467 (tlsdesc_small_sve_<mode>): New pattern.
19469 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
19471 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
19472 (UNSPEC_UMUL_HIGHPART): New constants.
19473 (MUL_HIGHPART): New int iteraor.
19474 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
19475 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
19477 (*<su>mul<mode>3_highpart): New define_insn.
19479 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
19482 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
19485 2018-03-13 Martin Liska <mliska@suse.cz>
19488 * (sem_item_optimizer::sem_item_optimizer): Initialize new
19490 (sem_item_optimizer::~sem_item_optimizer): Release it.
19491 (sem_item_optimizer::merge_classes): Register variable aliases.
19492 (sem_item_optimizer::fixup_pt_set): New function.
19493 (sem_item_optimizer::fixup_points_to_sets): Likewise.
19494 * ipa-icf.h: Declare new variables and functions.
19496 2018-03-13 Jakub Jelinek <jakub@redhat.com>
19498 PR middle-end/84834
19499 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
19500 integer_pow2p@2 and test integer_pow2p in condition.
19501 (A < 0 ? C : 0): Similarly for @1.
19503 PR middle-end/84831
19504 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
19505 characters starting at p contain '\0' character, don't look beyond
19509 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
19510 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
19513 * reg-stack.c (change_stack): Change update_end var from int to
19514 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
19515 also call set_block_for_insn on the newly added insns and rescan.
19518 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
19519 on the last operand.
19522 * tree.c (stabilize_reference_1): Return save_expr (e) for
19523 STATEMENT_LIST even if it doesn't have side-effects.
19525 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
19527 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
19529 2018-03-12 Renlin Li <renlin.li@arm.com>
19531 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
19532 aarch64_output_scalar_simd_mov_immediate.
19534 2018-03-12 Martin Sebor <msebor@redhat.com>
19536 PR tree-optimization/83456
19537 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
19538 for perfectly overlapping calls to memcpy.
19539 (gimple_fold_builtin_memory_chk): Same.
19540 (gimple_fold_builtin_strcpy): Handle no-warning.
19541 (gimple_fold_builtin_stxcpy_chk): Same.
19542 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
19544 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19546 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
19547 parameter. Use it for SFmode.
19548 (rs6000_function_arg_advance_1): Adjust.
19549 (rs6000_function_arg): Adjust.
19550 (rs6000_gimplify_va_arg): Pass false for that new parameter.
19552 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
19554 PR rtl-optimization/84169
19555 PR rtl-optimization/84780
19556 * combine.c (can_combine_p): Check for a 2-insn combination whether
19557 the destination register is used between the two insns, too.
19559 2018-03-12 Richard Biener <rguenther@suse.de>
19561 PR tree-optimization/84803
19562 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
19563 for refs DR analysis didn't process.
19565 2018-03-12 Richard Biener <rguenther@suse.de>
19567 PR tree-optimization/84777
19568 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
19569 force-vectorize loops ignore whether we are optimizing for size.
19571 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
19573 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
19574 (TARGET_MD_ASM_ADJUST): Define.
19576 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
19577 Kito Cheng <kito.cheng@gmail.com>
19578 Chung-Ju Wu <jasonwucj@gmail.com>
19580 * config/nds32/nds32.c (nds32_compute_stack_frame,
19581 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
19582 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
19583 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
19584 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
19585 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
19586 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
19587 * config/nds32/nds32.md (prologue, epilogue): Use macro
19588 NDS32_V3PUSH_AVAILABLE_P to do checking.
19590 2018-03-11 Jakub Jelinek <jakub@redhat.com>
19593 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
19594 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
19595 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
19596 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
19597 addition of most attributes on !orig_type_die or the attribute not
19598 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
19600 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19601 Chung-Ju Wu <jasonwucj@gmail.com>
19603 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
19604 __NDS32_VH__ macro.
19605 * config/nds32/nds32.opt (mvh): New option.
19607 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19608 Chung-Ju Wu <jasonwucj@gmail.com>
19610 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
19612 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
19613 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
19616 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
19617 Chung-Ju Wu <jasonwucj@gmail.com>
19619 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
19621 * config/nds32/nds32-multiple.md (strlensi): New pattern.
19622 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
19624 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
19625 Kito Cheng <kito.cheng@gmail.com>
19626 Chung-Ju Wu <jasonwucj@gmail.com>
19628 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
19629 UNSPEC_FFMISM and UNSPEC_FLMISM.
19630 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
19631 for ffb, ffmism and flmism.
19632 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
19633 (unspec_ffmism): Ditto.
19634 (unspec_flmism): Ditto.
19635 (nds32_expand_builtin_impl): Check if string extension is available.
19636 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
19637 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
19639 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
19642 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19645 * lra-assigns.c (assign_by_spills): Return a flag of reload
19646 assignment failure. Do not process the reload assignment
19647 failures. Do not spill other reload pseudos if they has the same
19649 (lra_assign): Add a return arg. Set up from the result of
19650 assign_by_spills call.
19651 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19652 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19653 usage_insns if it is not NULL.
19654 (spill_hard_reg_in_range): New function.
19655 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19656 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19657 function prototypes.
19658 (lra_assign): Change prototype.
19659 * lra.c (lra): Add code to deal with fails by splitting hard reg
19662 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
19665 * config/i386/i386.opt: Replace Enforcment with Enforcement.
19667 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
19670 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
19671 (dw_val_node): Add val_symbolic_view.
19672 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
19673 (symview_upper_bound): New.
19674 (new_line_info_table): Initialize symviews_since_reset.
19675 (dwarf2out_source_line): Count symviews_since_reset and set
19676 symview_upper_bound.
19677 (dw_val_equal_p): Handle symview.
19678 (add_AT_symview): New.
19679 (print_dw_val): Handle symview.
19680 (attr_checksum, attr_checksum_ordered): Likewise.
19681 (same_dw_val_p, size_of_die): Likewise.
19682 (value_format, output_die): Likewise.
19683 (add_high_low_attributes): Use add_AT_symview for entry_view.
19684 (dwarf2out_finish): Reset symview_upper_bound, clear
19687 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
19690 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
19691 Add strict argument and use it.
19692 (rs6000_split_multireg_move): Update for new strict argument.
19693 (mem_operand_gpr): Disallow all non-offsettable addresses.
19694 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
19696 2018-03-09 Jakub Jelinek <jakub@redhat.com>
19699 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
19700 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
19701 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
19704 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
19705 decl, use remap_type if we want to use the type.
19707 2018-03-09 Martin Sebor <msebor@redhat.com>
19709 PR tree-optimization/84526
19710 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
19712 (builtin_access::generic_overlap): Be prepared to handle non-array
19715 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
19717 PR rtl-optimization/84682
19718 * lra-constraints.c (process_address_1): Check is_address flag
19719 for address constraints.
19720 (process_alt_operands): Likewise.
19721 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
19722 preprocess_constraints.
19723 * recog.h (preprocess_constraints): Add oploc parameter.
19725 * recog.c (preprocess_constraints): Test address_operand for
19726 CT_ADDRESS constraints.
19728 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
19731 * lra-assigns.c (assign_by_spills): Return a flag of reload
19732 assignment failure. Do not process the reload assignment
19733 failures. Do not spill other reload pseudos if they has the same
19735 (lra_assign): Add a return arg. Set up from the result of
19736 assign_by_spills call.
19737 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
19738 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
19739 usage_insns if it is not NULL.
19740 (spill_hard_reg_in_range): New function.
19741 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
19742 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
19743 function prototypes.
19744 (lra_assign): Change prototype.
19745 * lra.c (lra): Add code to deal with fails by splitting hard reg
19748 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19751 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
19752 Accept complain bool parameter. Only emit errors if it is true.
19753 (arm_parse_cpu_option_name): Likewise.
19754 (arm_target_thumb_only): Adjust callers of the above.
19755 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
19756 prototype to take a default true bool parameter.
19757 (arm_parse_arch_option_name): Likewise.
19759 2018-03-09 David Malcolm <dmalcolm@redhat.com>
19760 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
19764 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
19765 * configure: Regenerate.
19766 * configure.ac ("linker --version-script option"): New.
19767 ("linker soname option"): New.
19769 2018-03-09 Richard Biener <rguenther@suse.de>
19771 PR tree-optimization/84775
19772 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
19773 immediate uses of predicate stmts and mark them modified.
19776 PR tree-optimization/84178
19777 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19779 (version_loop_for_if_conversion): Delay update_ssa call.
19780 (tree_if_conversion): Delay update_ssa until after predicate
19783 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
19786 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
19787 when the function accesses prior frames.
19789 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19792 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
19793 gen_llsym, otherwise call maybe_gen_llsym.
19795 PR inline-asm/84742
19796 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
19797 has ',' character inside of it.
19799 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19802 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
19803 as clobbering CC_REGNUM.
19805 2018-03-08 Richard Biener <rguenther@suse.de>
19807 PR middle-end/84552
19808 * tree-scalar-evolution.c: Include tree-into-ssa.h.
19809 (follow_copies_to_constant): Do not follow SSA names registered
19812 2018-03-08 Richard Biener <rguenther@suse.de>
19814 PR tree-optimization/84178
19815 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
19817 (version_loop_for_if_conversion): Delay update_ssa call.
19818 (tree_if_conversion): Delay update_ssa until after predicate
19821 2018-03-08 David Malcolm <dmalcolm@redhat.com>
19823 PR tree-optimization/84178
19824 * tree-if-conv.c (release_bb_predicate): Remove the
19825 the assertion that the stmts have NULL use_ops.
19826 Discard the statements, asserting that they haven't
19827 yet been added to a BB.
19829 2018-03-08 Richard Biener <rguenther@suse.de>
19831 PR tree-optimization/84746
19832 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
19833 (phi_translate): Pass in destination ANTIC_OUT set.
19834 (phi_translate_1): Likewise. For a simplified result lookup
19835 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
19836 (phi_translate_set): Adjust.
19837 (do_pre_regular_insertion): Likewise.
19838 (do_pre_partial_partial_insertion): Likewise.
19840 2018-03-08 Martin Liska <mliska@suse.cz>
19842 PR gcov-profile/84735
19843 * doc/gcov.texi: Document usage of profile files.
19844 * gcov-io.h: Document changes in the format.
19846 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
19850 * dwarf2out.c (struct dw_line_info_table): Update comments for
19852 (FORCE_RESET_NEXT_VIEW): New.
19853 (FORCE_RESETTING_VIEW_P): New.
19854 (RESETTING_VIEW_P): Check for -1 too.
19855 (ZERO_VIEW_P): Likewise.
19856 (new_line_info_table): Force-reset next view.
19857 (dwarf2out_begin_function): Likewise.
19858 (dwarf2out_source_line): Simplify zero_view_p initialization.
19859 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
19860 view directly. Omit view when omitting .loc at line 0.
19862 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19864 PR tree-optimization/84740
19865 * tree-switch-conversion.c (process_switch): Call build_constructors
19866 only if info.phi_count is non-zero.
19868 PR tree-optimization/84739
19869 * tree-tailcall.c (find_tail_calls): Check call arguments against
19870 DECL_ARGUMENTS (current_function_decl) rather than
19871 DECL_ARGUMENTS (func) when checking for tail recursion.
19873 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19875 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
19876 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
19877 Volker Reichelt's entry and add entries for people that perform
19878 GCC fuzzy testing and report numerous bugs.
19880 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
19883 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
19884 readonly data in sdata, if that is disabled.
19885 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
19886 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
19887 -mreadonly-in-sdata option.
19889 2018-03-07 Martin Sebor <msebor@redhat.com>
19891 PR tree-optimization/84468
19892 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
19893 basic block when looking for nul assignment.
19895 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
19898 * except.h (output_function_exception_table): Adjust prototype.
19899 * except.c (output_function_exception_table): Remove FNNAME parameter
19900 and add SECTION parameter. Ouput one part of the table at a time.
19901 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
19902 the first part of the exception table and emit unwind directives.
19903 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
19904 (i386_pe_seh_cold_init): Likewise.
19905 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
19906 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
19907 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
19908 (ix86_output_call_insn): Emit a nop in one more case for SEH.
19909 * config/i386/winnt.c: Include except.h.
19910 (struct seh_frame_state): Add reg_offset, after_prologue and
19911 in_cold_section fields.
19912 (i386_pe_seh_end_prologue): Set seh->after_prologue.
19913 (i386_pe_seh_cold_init): New function.
19914 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
19915 to seh->in_cold_section.
19916 (seh_emit_push): Record the offset of the push.
19917 (seh_emit_save): Record the offet of the save.
19918 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
19919 Test seh->after_prologue to disregard the epilogue.
19920 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
19921 (i386_pe_end_cold_function): New function.
19923 2018-03-07 Jakub Jelinek <jakub@redhat.com>
19926 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
19927 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
19930 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
19932 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
19933 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
19935 PR middle-end/84723
19936 * multiple_target.c: Include tree-inline.h and intl.h.
19937 (expand_target_clones): Diagnose and fail if node->definition and
19938 !tree_versionable_function_p (node->decl).
19940 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
19942 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
19944 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19945 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19946 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
19948 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19951 * combine.c (try_combine): Use reg_or_subregno instead of handling
19952 just paradoxical SUBREGs and REGs.
19954 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
19956 * config/arc/arc.c (arc_finalize_pic): Remove function.
19957 (arc_must_save_register): We use single base PIC register, remove
19958 checks to save/restore the PIC register.
19959 (arc_expand_prologue): Likewise.
19960 * config/arc/arc-protos.h (arc_set_default_type_attributes):
19962 (arc_verify_short): Likewise.
19963 (arc_attr_type): Likewise.
19964 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19965 (walk_stores): Likewise.
19966 (arc_address_cost): Make it static.
19967 (arc_verify_short): Likewise.
19968 (branch_dest): Likewise.
19969 (arc_attr_type): Likewise.
19970 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19971 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19972 (arc_final_prescan_insn): Remove inserting the nops due to
19973 hardware hazards. It is done in reorg step.
19974 (insn_length_variant_t): Remove.
19975 (insn_length_parameters_t): Likewise.
19976 (arc_insn_length_parameters): Likewise.
19977 (arc_get_insn_variants): Likewise.
19978 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19980 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19982 PR inline-asm/84683
19983 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19986 PR tree-optimization/84687
19987 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19989 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19991 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19993 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19994 Rename to ppc_speculation_barrier.
19995 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19996 __builtin_ppc_speculation_barrier.
19998 2018-03-05 Jakub Jelinek <jakub@redhat.com>
20001 * combine.c (combine_simplify_rtx): Don't try to simplify if
20002 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
20005 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
20007 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
20008 to 32 bytes when compiling for POWER9.
20010 2018-03-05 Jakub Jelinek <jakub@redhat.com>
20013 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
20014 regparm >= 3 with no arg reg available also for calls with
20015 flag_force_indirect_call. Pass decl to ix86_function_regparm.
20018 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
20020 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
20022 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
20025 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
20027 2018-03-05 Richard Biener <rguenther@suse.de>
20029 PR tree-optimization/84486
20030 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
20031 When inserting a __builtin_assume_aligned call set the LHS
20032 SSA name alignment info accordingly.
20034 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
20036 PR tree-optimization/84114
20037 * config/aarch64/aarch64.c (aarch64_reassociation_width)
20038 Avoid reassociation of FLOAT_MODE addition.
20040 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
20042 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
20043 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
20044 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
20045 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
20046 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
20047 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
20048 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
20050 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
20051 __builtin_ia32_wbinvd): New builtins.
20052 (SPECIAL_ARGS2): New.
20053 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
20054 (SPECIAL_ARGS2): New.
20055 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
20056 (ix86_valid_target_attribute_inner_p): Ditto.
20057 (ix86_init_mmx_sse_builtins): Add special_args2.
20058 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
20059 TARGET_WBNOINVD_P): New.
20060 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
20061 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
20062 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
20063 * config/i386/immintrin.h (_wbinvd): New intrinsic.
20064 * config/i386/pconfigintrin.h: New file.
20065 * config/i386/wbnoinvdintrin.h: Ditto.
20066 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
20068 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
20070 2018-03-05 Richard Biener <rguenther@suse.de>
20072 PR tree-optimization/84670
20073 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
20075 (BB_VISITED_WITH_VISITED_SUCCS): New define.
20076 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
20077 (compute_antic_aux): Only assert the number of values in ANTIC_IN
20078 doesn't grow if all successors (recursively) were visited at least
20081 2018-03-05 Richard Biener <rguenther@suse.de>
20083 PR tree-optimization/84650
20084 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
20085 if executed in the loop pipeline.
20087 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
20089 * doc/configfiles.texi (Configuration Files): Move info about
20090 conditionalizing $target-protos.h to...
20091 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
20092 differs from $target-protos.h.
20094 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
20095 Chung-Ju Wu <jasonwucj@gmail.com>
20097 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
20098 * config/nds32/nds32-multiple.md (setmemsi): Define.
20099 * config/nds32/nds32-memory-manipulation.c
20100 (nds32_gen_dup_4_byte_to_word_value): New.
20101 (emit_setmem_word_loop): New.
20102 (emit_setmem_byte_loop): New.
20103 (nds32_expand_setmem_loop): New.
20104 (nds32_expand_setmem_loop_v3m): New.
20105 (nds32_expand_setmem_unroll): New.
20106 (nds32_expand_setmem): New.
20108 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20109 Chung-Ju Wu <jasonwucj@gmail.com>
20111 * config/nds32/nds32-memory-manipulation.c
20112 (nds32_emit_load_store): New.
20113 (nds32_emit_post_inc_load_store): New.
20114 (nds32_emit_mem_move): New.
20115 (nds32_emit_mem_move_block): New.
20116 (nds32_expand_movmemsi_loop_unknown_size): New.
20117 (nds32_expand_movmemsi_loop_known_size): New.
20118 (nds32_expand_movmemsi_loop): New.
20119 (nds32_expand_movmemsi_unroll): New.
20120 (nds32_expand_movmemqi): Rename ...
20121 (nds32_expand_movmemsi): ... to this.
20122 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
20123 (movmemsi): ... to this.
20124 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
20125 (nds32_expand_movmemsi): ... to this.
20127 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20128 Monk Chiang <sh.chiang04@gmail.com>
20129 Chung-Ju Wu <jasonwucj@gmail.com>
20131 * config/nds32/nds32-protos.h
20132 (nds32_expand_load_multiple): New arguments.
20133 (nds32_expand_store_multiple): Ditto.
20134 (nds32_valid_multiple_load_store): Rename ...
20135 (nds32_valid_multiple_load_store_p): ... to this.
20136 * config/nds32/nds32-memory-manipulation.c
20137 (nds32_expand_load_multiple): Refine implementation.
20138 (nds32_expand_store_multiple): Ditto.
20139 * config/nds32/nds32-multiple.md
20140 (load_multiple): Update nds32_expand_load_multiple interface.
20141 (store_multiple): Update nds32_expand_store_multiple interface.
20142 * config/nds32/nds32-predicates.c
20143 (nds32_valid_multiple_load_store): Rename ...
20144 (nds32_valid_multiple_load_store_p): ... to this and refine
20146 * config/nds32/predicates.md
20147 (nds32_load_multiple_and_update_address_operation): New predicate.
20148 (nds32_store_multiple_and_update_address_operation): New predicate.
20150 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
20151 Chung-Ju Wu <jasonwucj@gmail.com>
20153 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
20154 (combo): New attribute.
20155 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
20157 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
20159 * config/nds32/nds32.opt: Change -mcmodel= default value.
20161 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
20162 Monk Chiang <sh.chiang04@gmail.com>
20163 Chung-Ju Wu <jasonwucj@gmail.com>
20165 * config/nds32/constants.md (unspec_element): New enum.
20166 * config/nds32/constraints.md (Umw): New constraint.
20167 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
20168 * config/nds32/nds32-intrinsic.md: Likewise.
20169 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
20170 (nds32_valid_smw_lwm_base_p): New.
20171 (nds32_output_smw_single_word): New.
20172 (nds32_output_lmw_single_word): New.
20173 (nds32_expand_unaligned_load): New.
20174 (nds32_expand_unaligned_store): New.
20175 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
20176 (nds32_output_smw_single_word): Declare.
20177 (nds32_output_lmw_single_word): Declare.
20178 (nds32_expand_unaligned_load): Declare.
20179 (nds32_expand_unaligned_store): Declare.
20180 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
20181 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
20182 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
20183 NDS32_BUILTIN_UASTORE_DW.
20184 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
20187 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
20188 Kito Cheng <kito.cheng@gmail.com>
20189 Chung-Ju Wu <jasonwucj@gmail.com>
20191 * config/nds32/nds32-intrinsic.c
20192 (nds32_expand_builtin_null_ftype_reg): Delete.
20193 (nds32_expand_builtin_reg_ftype_imm): Ditto.
20194 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
20195 (nds32_read_argument): New.
20196 (nds32_legitimize_target): Ditto.
20197 (nds32_legitimize_argument): Ditto.
20198 (nds32_check_constant_argument): Ditto.
20199 (nds32_expand_unop_builtin): Ditto.
20200 (nds32_expand_unopimm_builtin): Ditto.
20201 (nds32_expand_binop_builtin): Ditto.
20202 (nds32_builtin_decl_impl): Ditto.
20203 (builtin_description): Ditto.
20204 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
20205 (nds32_init_builtins_impl): Ditto.
20206 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
20207 (nds32_builtin_decl): New.
20208 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
20209 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
20211 2018-03-02 Jeff Law <law@redhat.com>
20213 * reorg.c (stop_search_p): Handle DEBUG_INSN.
20214 (redundant_insn, fill_simple_delay_slots): Likewise.
20215 (fill_slots_from_thread): Likewise.
20216 * resource.c (mark_referenced_resources): Likewise.
20217 (mark_set_resources, find_dead_or_set_registers): Likewise.
20219 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20221 * substring-locations.h (format_warning_va): Formatting fix for
20222 ATTRIBUTE_GCC_DIAG.
20223 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
20225 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
20226 * substring-locations.c: Include intl.h.
20227 (format_warning_va): Turned into small wrapper around
20228 format_warning_n_va, renamed to ...
20229 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
20230 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
20232 (format_warning_at_substring_n): New function.
20233 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
20234 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
20235 format_warning_at_substring with just a shorter name instead of
20236 const function pointer.
20237 (fmtwarn_n): New function.
20238 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
20239 appropriate, get rid of all the fmtstr temporaries, move conditionals
20240 with G_() wrapped string literals directly into fmtwarn arguments,
20241 cast dir.len to (int), formatting fixes.
20243 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
20245 * doc/invoke.texi: Remove "Cilk Plus" references.
20247 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20248 Richard Biener <rguenther@suse.de>
20251 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
20252 for error or warning attributes if CALL_FROM_THUNK_P is set.
20255 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20258 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
20259 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
20262 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
20263 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
20265 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
20266 instead of -1U in last predictors element's probability member.
20268 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
20271 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
20272 arguments if they are comparable.
20274 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20276 PR tree-optimization/84634
20277 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
20278 masks and masked_loop_p with a single loop_masks, making sure it's
20279 null for bb vectorization.
20281 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
20283 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
20284 (vect_analyze_data_ref_access): Use loop->safe_len rather than
20285 loop->force_vectorize to check whether there is no alias.
20287 2018-03-02 Jakub Jelinek <jakub@redhat.com>
20290 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
20292 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
20294 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
20295 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
20296 instead of a loop around prev_real_insn.
20297 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
20300 PR inline-asm/84625
20301 * config/i386/i386.c (ix86_print_operand): Use conditional
20302 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
20305 2018-03-02 Richard Biener <rguenther@suse.de>
20307 PR tree-optimization/84427
20308 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
20309 (bitmap_set_subtract_values): Rewrite to handle multiple
20312 (prune_clobbered_mems): Likewise.
20313 (phi_translate): Take edge instead of pred/phiblock.
20314 (phi_translate_1): Likewise.
20315 (phi_translate_set): Likewise. Insert all translated
20316 exprs for a value into the set, keeping possibly multiple
20317 expressions per value.
20318 (compute_antic_aux): Adjust for phi_translate changes.
20319 When intersecting union the expressions and prune those
20320 not in the final value set, keeping possibly multiple
20321 expressions per value. Do not use value-insertion
20322 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
20323 all expressions. Add verification that the value-sets
20324 only shrink during iteration.
20325 (compute_partial_antic_aux): Adjust for the phi_translate changes.
20326 (do_pre_regular_insertion): Likewise.
20327 (do_pre_partial_partial_insertion): Likewise.
20329 2018-03-02 Richard Biener <rguenther@suse.de>
20332 * config/darwin.c (saved_debug_info_level): New static global.
20333 (darwin_asm_lto_start): Disable debug info generation for LTO out.
20334 (darwin_asm_lto_end): Restore debug info generation settings.
20336 2018-03-01 Martin Liska <mliska@suse.cz>
20339 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
20340 volatile arguments.
20342 2018-03-01 Richard Biener <rguenther@suse.de>
20345 * dwarf2out.c (gen_variable_die): Properly handle late VLA
20346 type annotation with LTO when debug was disabled at compile-time.
20348 2018-03-01 Matthew Fortune <mfortune@gmail.com>
20350 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
20352 (mips_final_postscan_insn): Likewise.
20354 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
20356 PR rtl-optimization/84528
20357 * alias.c (init_alias_target): Add commentary.
20358 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
20359 a unique base value if the frame pointer is not eliminated
20360 to the stack pointer.
20362 2018-03-01 Tom de Vries <tom@codesourcery.com>
20364 PR rtl-optimization/83327
20365 * lra-int.h (hard_regs_spilled_into): Declare.
20366 * lra.c (hard_regs_spilled_into): Define.
20367 (init_reg_info): Init hard_regs_spilled_into.
20368 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
20369 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
20370 (process_bb_lives): Handle hard_regs_spilled_into.
20371 (lra_create_live_ranges_1): Before doing liveness propagation, clear
20372 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
20374 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
20376 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
20377 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
20378 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
20379 * config/rs6000/aix72.h: New file.
20381 2018-02-28 Jakub Jelinek <jakub@redhat.com>
20383 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
20384 instead of warning_at with conditional singular and plural messages
20388 * stor-layout.c (update_alignment_for_field): For
20389 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
20390 && !DECL_PACKED (field), do the alignment update, just use
20391 only desired_align instead of MAX (type_align, desired_align)
20393 (place_field): Don't do known_align < desired_align handling
20394 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
20395 is non-NULL, instead do it after rli->prev_field handling and
20396 only if not within a bitfield word. For DECL_PACKED (field)
20397 use type_align of BITS_PER_UNIT.
20399 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
20401 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
20402 superfluous parentheses and trailing spaces.
20404 2018-02-28 Richard Biener <rguenther@suse.de>
20406 PR tree-optimization/84584
20407 * graphite-scop-detection.c (scop_detection::add_scop): Discard
20408 SCoPs with fake exit edge.
20410 2018-02-28 Martin Liska <mliska@suse.cz>
20413 * timevar.c (timer::print): Fix format to properly print 100%
20416 2018-02-28 Richard Biener <rguenther@suse.de>
20418 PR middle-end/84607
20419 * genmatch.c (capture_info::walk_match): Do not mark
20420 captured expressions without operands as expr_p given
20421 they act more like predicates and should be subject to
20422 "lost tail" side-effect preserving.
20424 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
20426 PR rtl-optimization/81611
20427 * auto-inc-dec.c (attempt_change): Move dead note from
20428 mem_insn if it's the next use of regno
20429 (find_address): Take address use of reg holding
20430 non-incremented value. Add parm to limit search to the named
20432 (merge_in_block): Attempt to use a mem insn that is the next
20433 use of the original regno.
20435 2018-02-27 Martin Sebor <msebor@redhat.com>
20438 * doc/invoke.texi (-Wmissing-attributes): New option.
20439 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
20441 2018-02-27 Martin Sebor <msebor@redhat.com>
20443 PR translation/84207
20444 * diagnostic-core.h (warning_n, error_n, inform_n): Change
20445 n argument to unsigned HOST_WIDE_INT.
20446 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
20447 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
20448 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
20449 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
20451 2018-02-27 Richard Biener <rguenther@suse.de>
20453 PR tree-optimization/84512
20454 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
20455 Do not use the estimate returned from record_stmt_cost for
20456 the scalar iteration cost but sum properly using add_stmt_cost.
20458 2018-02-27 Richard Biener <rguenther@suse.de>
20460 PR tree-optimization/84466
20461 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20462 Adjust last change to less strictly validate use operands.
20464 2018-02-27 Martin Liska <mliska@suse.cz>
20466 PR gcov-profile/84548
20467 * gcov.c (process_file): Allow partial overlap and consider it
20468 also as group functions.
20469 (output_lines): Properly calculate range of lines for a group.
20471 2018-02-27 Martin Liska <mliska@suse.cz>
20473 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
20474 'ggc' suffixes. Change first column width.
20475 (timer::print): Fix formatting of the column.
20477 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
20479 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
20480 preserve inline entry blocks for the sake of debug inline
20481 entry point markers alone.
20482 (remove_unused_locals): Suggest in comments a better place to
20483 force the preservation of inline entry blocks that are
20484 otherwise unused, but do not preserve them.
20486 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20488 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
20490 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20493 * config/i386/constraints.md (Bs): Replace
20494 ix86_indirect_branch_register with
20495 TARGET_INDIRECT_BRANCH_REGISTER.
20497 * config/i386/i386.md (indirect_jump): Likewise.
20498 (tablejump): Likewise.
20499 (*sibcall_memory): Likewise.
20500 (*sibcall_value_memory): Likewise.
20501 Peepholes of indirect call and jump via memory: Likewise.
20502 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
20503 (*sibcall_value_GOT_32): Likewise.
20504 * config/i386/predicates.md (indirect_branch_operand): Likewise.
20505 (GOT_memory_operand): Likewise.
20506 (call_insn_operand): Likewise.
20507 (sibcall_insn_operand): Likewise.
20508 (GOT32_symbol_operand): Likewise.
20509 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
20511 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20513 PR rtl-optimization/83496
20514 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
20515 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
20516 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
20517 redundant insn, if any.
20518 (relax_delay_slots): Likewise.
20519 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
20521 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
20523 PR tree-optimization/83965
20524 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
20525 that grouped statements are part of a reduction chain. Return
20526 true if the statement is not marked as a reduction itself but
20527 is part of a group.
20528 (vect_recog_dot_prod_pattern): Don't check whether the statement
20529 is part of a group here.
20530 (vect_recog_sad_pattern): Likewise.
20531 (vect_recog_widen_sum_pattern): Likewise.
20533 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
20536 * final.c (rest_of_clean_state): Also look for calls inside sequences.
20538 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
20541 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
20543 (ix86_output_indirect_function_return): New prototype.
20544 (ix86_split_simple_return_pop_internal): Likewise.
20545 * config/i386/i386.c (indirect_return_via_cx): New.
20546 (indirect_return_via_cx_bnd): Likewise.
20547 (indirect_thunk_name): Handle return va CX_REG.
20548 (output_indirect_thunk_function): Create alias for
20549 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
20550 (ix86_output_indirect_jmp): Remove the bool argument.
20551 (ix86_output_indirect_function_return): New function.
20552 (ix86_split_simple_return_pop_internal): Likewise.
20553 * config/i386/i386.md (*indirect_jump): Don't pass false
20554 to ix86_output_indirect_jmp.
20555 (*tablejump_1): Likewise.
20556 (simple_return_pop_internal): Change it to define_insn_and_split.
20557 Call ix86_split_simple_return_pop_internal to split it for
20558 -mfunction-return=.
20559 (simple_return_indirect_internal): Call
20560 ix86_output_indirect_function_return instead of
20561 ix86_output_indirect_jmp.
20563 2018-02-26 Jakub Jelinek <jakub@redhat.com>
20566 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
20567 memset and value initialization afterwards.
20569 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
20571 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
20573 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
20576 * common/config/aarch64/aarch64-common.c
20577 (aarch_option_optimization_table[]): Switch
20578 off fomit-frame-pointer
20580 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20581 Chung-Ju Wu <jasonwucj@gmail.com>
20583 * config/nds32/nds32-multiple.md (load_multiple): Disallow
20585 (store_multiple): Ditto.
20587 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
20589 * config.gcc: Add --with-cpu support for nds32 target.
20590 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
20591 * config/nds32/nds32.opt: Add -mcpu= option.
20593 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
20595 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
20596 isel=yes): Warn for these deprecated options.
20598 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
20600 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
20601 ISA_2_5_MASKS_EMBEDDED.
20603 2018-02-23 Jakub Jelinek <jakub@redhat.com>
20605 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
20606 p->max as pointers rather than using iterative_hash_expr.
20608 2018-02-23 Carl Love <cel@us.ibm.com>
20610 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
20611 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
20613 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
20614 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
20615 P8V_BUILTIN_VEC_VUNSIGNED2.
20617 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
20620 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
20621 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
20623 * lra-constraints.c (curr_insn_transform): Set up
20624 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
20626 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
20627 * lra-eliminations.c (spill_pseudos): Ditto.
20628 (process_insn_for_elimination): Ditto.
20629 * lra-lives.c (reg_early_clobber_p): Use the new macros.
20630 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
20631 LRA_NON_CLOBBERED_ALT.
20633 2018-02-22 Martin Sebor <msebor@redhat.com>
20635 PR tree-optimization/84480
20636 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
20637 to maybe_diag_stxncpy_trunc. Call it.
20638 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
20639 from gimple_fold_builtin_strcpy. Print inlining stack.
20640 (handle_builtin_stxncpy): Print inlining stack.
20641 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
20643 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
20646 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
20647 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
20648 and -fcheck-pointer-bounds are used together.
20649 (indirect_thunk_prefix): New enum.
20650 (indirect_thunk_need_prefix): New function.
20651 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
20652 "_nt" instead of "_bnd" for NOTRACK prefix.
20653 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
20654 (output_indirect_thunk_function): Likewise.
20656 (ix86_code_end): Update output_indirect_thunk_function calls.
20657 (ix86_output_indirect_branch_via_reg): Replace
20658 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
20659 (ix86_output_indirect_branch_via_push): Likewise.
20660 (ix86_output_function_return): Likewise.
20661 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
20662 incompatible with -fcf-protection=branch and
20663 -fcheck-pointer-bounds.
20665 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20668 * config/aarch64/aarch64.c (aarch64_print_address_internal):
20669 Change gcc_assert call to output_operand_lossage.
20671 2018-02-22 Steve Ellcey <sellcey@cavium.com>
20673 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
20675 2018-02-22 DJ Delorie <dj@redhat.com>
20676 Sebastian Perta <sebastian.perta@renesas.com>
20677 Oleg Endo <olegendo@gcc.gnu.org>
20679 * config/rx/rx.c (rx_rtx_costs): New function.
20680 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
20682 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
20684 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
20686 2018-02-22 Martin Liska <mliska@suse.cz>
20689 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
20690 Add "native" as a possible value.
20692 2018-02-22 Martin Liska <mliska@suse.cz>
20695 * config/i386/i386.c (ix86_option_override_internal):
20696 Add "native" as a possible value for -march and -mtune.
20698 2018-02-22 Jakub Jelinek <jakub@redhat.com>
20701 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
20702 to all type variants.
20704 PR tree-optimization/84503
20705 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
20706 width as info->bitpos + info->bitsize - start.
20707 (merged_store_group::merge_overlapping): Simplify width computation.
20708 (check_no_overlap): New function.
20709 (imm_store_chain_info::try_coalesce_bswap): Compute expected
20710 start + width and last_order of the group, fail if check_no_overlap
20712 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
20713 to group if check_no_overlap fails.
20715 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20717 * config/rs6000/altivec.md: Delete contraint arguments to
20718 define_expand, define_split, and define_peephole2, and in
20719 define_insn_and_split if always unused.
20720 * config/rs6000/darwin.md: Ditto.
20721 * config/rs6000/dfp.md: Ditto.
20722 * config/rs6000/rs6000.md: Ditto.
20723 * config/rs6000/sync.md: Ditto.
20724 * config/rs6000/vector.md: Ditto.
20725 * config/rs6000/vsx.md: Ditto.
20727 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
20729 * config/rs6000/altivec.md: Write output control strings as braced
20730 blocks instead of double-quoted strings.
20731 * config/rs6000/darwin.md: Ditto.
20732 * config/rs6000/rs6000.md: Ditto.
20733 * config/rs6000/vector.md: Ditto.
20734 * config/rs6000/vsx.md: Ditto.
20736 2018-02-21 Jason Merrill <jason@redhat.com>
20738 PR c++/84314 - ICE with templates and fastcall attribute.
20739 * attribs.c (build_type_attribute_qual_variant): Remove assert.
20741 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20743 * ipa-cp.c (determine_versionability): Fix comment typos.
20745 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
20748 * ipa-cp.c (determine_versionability): Do not version functions caling
20751 2018-02-21 Martin Liska <mliska@suse.cz>
20754 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
20755 Add "native" as a possible value.
20756 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
20757 the macro when native cpu detection is available.
20759 2018-02-21 Martin Liska <mliska@suse.cz>
20762 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
20763 Add "native" as a possible value.
20764 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
20765 when native cpu detection is available.
20767 2018-02-21 Jakub Jelinek <jakub@redhat.com>
20768 Martin Sebor <msebor@redhat.com>
20770 PR tree-optimization/84478
20771 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
20773 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
20774 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
20775 support which is conservatively correct, for 2 only stay conservative
20776 for maxlen. Formatting and comment capitalization fixes. Add STRICT
20777 argument to the 2 argument get_range_strlen, adjust 6 arg
20778 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
20780 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
20781 (gimple_fold_builtin_strlen): Pass true as last argument to
20784 2018-02-20 Martin Sebor <msebor@redhat.com>
20786 PR middle-end/84095
20787 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
20788 (builtin_memref::set_base_and_offset): Same. Handle inner references.
20789 (builtin_memref::builtin_memref): Factor out parts into
20790 set_base_and_offset and call it.
20792 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20794 PR middle-end/84406
20795 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
20796 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
20797 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
20798 search at the associated MODE_INT.
20800 2018-02-20 Jeff Law <law@redhat.com>
20802 PR middle-end/82123
20803 PR tree-optimization/81592
20804 PR middle-end/79257
20805 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
20806 for range data rather than using global data.
20807 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
20808 range data rather than using global data.
20809 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
20810 pass it to children as needed.
20811 (struct directive::fmtresult): Similarly.
20812 (struct directive::set_width): Similarly.
20813 (struct directive::set_precision): Similarly.
20814 (format_integer, format_directive, parse_directive): Similarly.
20815 (format_none): Accept unnamed vr_values parameter.
20816 (format_percent, format_floating, format_character): Similarly.
20817 (format_string, format_plain): Similarly.
20818 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
20819 the EVRP range analyzer for range data rather than using global data.
20820 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
20821 gimple-ssa-evrp-analyze.h
20822 (class sprintf_dom_walker): Add after_dom_children member function.
20823 Add evrp_range_analyzer member.
20824 (sprintf_dom_walker::before_dom_children): Call into the EVRP
20825 range analyzer as needed.
20826 (sprintf_dom_walker::after_dom_children): New member function.
20827 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
20829 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
20830 (evrp_range_analyzer::pop_to_marker): Likewise.
20832 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
20834 PR tree-optimization/84419
20835 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
20836 with the required type if its current type is compatible but
20839 2018-02-20 Jakub Jelinek <jakub@redhat.com>
20841 PR middle-end/82004
20842 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
20843 after vectorization.
20845 2018-02-20 Martin Liska <mliska@suse.cz>
20848 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
20849 possible values if we don't have a hint.
20851 2018-02-20 Martin Liska <mliska@suse.cz>
20855 * final.c (shorten_branches): Build align_tab array with one
20857 * opts.c (finish_options): Add alignment option limit check.
20858 (MAX_CODE_ALIGN): Likewise.
20859 (MAX_CODE_ALIGN_VALUE): Likewise.
20860 * doc/invoke.texi: Document maximum allowed option value for
20861 all -falign-* options.
20863 2018-02-19 Jakub Jelinek <jakub@redhat.com>
20866 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
20867 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
20868 * var-tracking.c (emit_note_insn_var_location): Remove all references
20869 to NOTE_INSN_CALL_ARG_LOCATION.
20870 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
20871 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
20872 Use copy_rtx_if_shared.
20873 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
20874 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
20875 (dwarf2out_var_location): Remove handling of
20876 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
20878 * final.c (final_scan_insn): Remove all references to
20879 NOTE_INSN_CALL_ARG_LOCATION.
20880 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
20881 before dumping final insns.
20882 * except.c (emit_note_eh_region_end): Remove all references to
20883 NOTE_INSN_CALL_ARG_LOCATION.
20884 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
20885 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
20886 * config/arc/arc.c (hwloop_optimize): Likewise.
20887 * config/arm/arm.c (create_fix_barrier): Likewise.
20888 * config/s390/s390.c (s390_chunkify_start): Likewise.
20889 * config/sh/sh.c (find_barrier): Likewise.
20890 * config/i386/i386.c (rest_of_insert_endbranch,
20891 ix86_seh_fixup_eh_fallthru): Likewise.
20892 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
20893 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
20894 * config/frv/frv.c (frv_function_prologue): Likewise.
20895 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
20897 (note_outside_basic_block_p): Remove all references to
20898 NOTE_INSN_CALL_ARG_LOCATION.
20899 * gengtype.c (adjust_field_rtx_def): Likewise.
20900 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
20902 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
20903 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
20906 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
20909 PR tree-optimization/84452
20910 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
20911 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
20914 2018-02-19 Martin Liska <mliska@suse.cz>
20917 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
20919 2018-02-19 Martin Liska <mliska@suse.cz>
20920 Richard Sandiford <richard.sandiford@linaro.org>
20922 PR tree-optimization/82491
20923 * gimple-fold.c (get_base_constructor): Make earlier bail out
20926 2018-02-19 Carl Love <cel@us.ibm.com>
20928 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
20929 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
20931 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
20932 P8V_BUILTIN_VEC_NEG.
20934 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
20936 * config/rl78/rl78.md (movdf): New define expand.
20938 2018-02-19 Martin Liska <mliska@suse.cz>
20941 * doc/invoke.texi: Fix typo.
20942 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
20944 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
20946 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
20947 handle rs6000_single_float and rs6000_double_float specially for
20950 2018-02-16 Jeff Law <law@redhat.com>
20952 * config/rx/rx.c (add_pop_cfi_notes): New function.;
20953 (pop_regs): Use it.
20955 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20958 * ipa-inline.c (inline_small_functions): Fix a typo.
20960 2018-02-16 Nathan Sidwell <nathan@acm.org>
20962 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20964 2018-02-16 Carl Love <cel@us.ibm.com>
20966 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20967 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20968 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20969 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20970 expansion to P8V_BUILTIN_VEC_FLOAT2.
20972 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
20974 PR rtl-optimization/70023
20975 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20976 src_regno into account.
20978 2018-02-16 Carl Love <cel@us.ibm.com>
20980 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20981 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20982 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20983 * config/rs6000/rs6000.c: Remove case statements for
20984 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20985 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20986 and P9V_BUILTIN_VEC_VINSERT4B.
20987 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20988 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20989 * config/rs6000/vsx.md:
20990 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20993 2018-02-16 Carl Love <cel@us.ibm.com>
20995 * config/rs6000/altivec.h: Add builtin names vec_extract4b
20997 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20999 * config/rs6000/rs6000-c.c: Add the definitions for
21000 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
21001 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
21002 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
21003 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
21004 definition for insert4b and define insn *insert3b_internal.
21005 * doc/extend.texi: Add documentation for vec_extract4b.
21007 2018-02-16 Nathan Sidwell <nathan@acm.org>
21009 * doc/extend.texi (Backwards Compatibility): Mention friend
21010 injection. Note for-scope is deprecated.
21011 * doc/invoke.texi (-ffriend-injection): Deprecate.
21013 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
21015 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
21016 that moved to I2, also allow destinations that are a paradoxical
21017 subreg (instead of a normal reg).
21019 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
21022 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
21025 2018-02-16 Richard Biener <rguenther@suse.de>
21027 PR tree-optimization/84037
21028 PR tree-optimization/84016
21030 * config/i386/i386.c (ix86_builtin_vectorization_cost):
21031 Adjust vec_construct for the fact we need additional higher latency
21032 128bit inserts for AVX256 and AVX512 vector builds.
21033 (ix86_add_stmt_cost): Scale vector construction cost for
21036 2018-02-16 Richard Biener <rguenther@suse.de>
21038 PR tree-optimization/84417
21039 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
21040 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
21041 (non_rewritable_lvalue_p): Likewise, use poly-ints.
21043 2018-02-16 Martin Liska <mliska@suse.cz>
21046 * internal-fn.def (ASAN_CHECK): Set proper flags.
21047 (ASAN_MARK): Likewise.
21049 2018-02-16 Julia Koval <julia.koval@intel.com>
21051 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
21052 from PTA_CANNONLAKE.
21054 2018-02-16 Jakub Jelinek <jakub@redhat.com>
21057 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
21058 Use ++iter rather than iter++ for std::list iterators.
21059 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
21060 defer deleting them until all nodes in the forest are processed. Do
21061 free even leaf nodes. Change to_process into auto_vec.
21064 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
21065 * vec.h (vec_default_construct): Use memset instead of placement new
21066 if BROKEN_VALUE_INITIALIZATION is defined.
21067 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
21068 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
21071 PR rtl-optimization/83723
21072 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
21073 * lra.c (lra_substitute_pseudo): Likewise. If true, use
21074 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
21076 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
21078 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
21080 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
21082 PR rtl-optimization/81443
21083 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
21084 from inner REGs to paradoxical SUBREGs.
21086 2018-02-16 Richard Biener <rguenther@suse.de>
21088 PR tree-optimization/84399
21089 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
21090 For operands we can analyze at their definition make sure we can
21091 analyze them at each use as well.
21093 2018-02-16 Richard Biener <rguenther@suse.de>
21095 PR tree-optimization/84190
21096 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
21097 volatile accesses if the decl isn't volatile.
21099 2018-02-15 Jason Merrill <jason@redhat.com>
21101 PR c++/84314 - ICE with templates and fastcall attribute.
21102 * attribs.c (build_type_attribute_qual_variant): Don't clobber
21103 TYPE_CANONICAL on an existing type.
21105 2018-02-15 Jakub Jelinek <jakub@redhat.com>
21107 PR tree-optimization/84383
21108 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
21109 dstoff nor call operand_equal_p if dstbase is NULL.
21111 PR tree-optimization/84334
21112 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
21113 also a CONSTANT_CLASS_P, punt.
21115 2018-02-14 Jim Wilson <jimw@sifive.com>
21117 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
21118 first SMALL_OPERAND check. New local min_second_step. Move assert
21119 to where locals are set. Add TARGET_RVC support.
21120 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
21122 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
21124 * doc/invoke.texi: Correct -Wformat-overflow code sample.
21126 2018-02-14 Martin Sebor <msebor@redhat.com>
21128 PR tree-optimization/83698
21129 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
21130 arrays constrain the offset range to their bounds.
21131 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
21132 (builtin_access::overlap): Avoid setting the size of overlap if it's
21134 (maybe_diag_overlap): Also consider arrays when deciding what values
21135 of offsets to include in diagnostics.
21137 2018-02-14 Martin Sebor <msebor@redhat.com>
21140 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
21141 that correspond to the kind of a declaration.
21143 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
21146 * config/pa/pa.md: Load address of PIC label using the linkage table
21147 if the label is nonlocal.
21149 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
21151 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
21152 warning message if user requests -maltivec=be.
21153 * doc/invoke.texi: Document deprecation of -maltivec=be.
21155 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
21158 * config/rs6000/rs6000-c.c: Update definitions for
21159 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
21160 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
21162 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21165 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
21166 add _get_ssp intrinsics. Remove argument from
21167 __builtin_ia32_rdssp[d|q].
21168 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
21169 * config/i386/i386-builtin.def: Remove argument from
21170 __builtin_ia32_rdssp[d|q].
21171 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
21172 ix86_expand_special_args_builtin for _rdssp[d|q].
21173 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
21174 Clear register before usage.
21175 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
21176 Add documentation for new _get_ssp and _inc_ssp intrinsics.
21178 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
21180 PR tree-optimization/84357
21181 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
21182 operand 1 of an ARRAY_REF too.
21184 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
21187 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
21188 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21190 (set_of_reg): New struct.
21191 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
21192 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
21193 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
21195 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
21196 Split into bitclr, bitset, bitinvert patterns if appropriate.
21197 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
21198 use rx_fuse_in_memory_bitop.
21199 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
21200 to named insn, correct maximum insn length.
21202 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
21205 * machmode.def: Define a complex mode for PARTIAL_INT.
21206 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
21208 * doc/rtl.texi: Document CSPImode.
21209 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
21211 (msp430_hard_regno_nregs_with_padding): Likewise.
21213 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
21216 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
21218 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
21220 PR rtl-optimization/84169
21221 * combine.c (try_combine): New variable split_i2i3. Set it to true if
21222 we generated a parallel as new i3 and we split that to new i2 and i3
21223 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
21224 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
21225 those to i2, not i1. Partially rewrite this scan code.
21227 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21230 * stor-layout.c (place_field): For variable length fields, adjust
21231 offset_align afterwards not just based on the field's alignment,
21232 but also on the size.
21234 PR middle-end/84309
21235 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
21236 of exps and logs in the use_exp2 case.
21238 2018-02-13 Jeff Law <law@redhat.com>
21240 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
21241 entry for "vector".
21243 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
21246 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
21250 * common.opt (gas-loc-support, gas-locview-support): New.
21251 (ginline-points, ginternal-reset-location-views): New.
21252 * doc/invoke.texi: Document them. Use @itemx where intended.
21253 (gvariable-location-views): Adjust.
21254 * target.def (reset_location_view): New.
21255 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
21256 (TARGET_RESET_LOCATION_VIEW): New.
21257 * doc/tm.texi: Rebuilt.
21258 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
21259 (dwarf2out_default_as_locview_support): New.
21260 (output_asm_line_debug_info): Use option variables.
21261 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
21262 (output_loc_list): Likewise.
21263 (add_high_low_attributes): Check option variables.
21264 Don't output entry view attribute in strict mode.
21265 (gen_inlined_subroutine_die): Check option variables.
21266 (dwarf2out_inline_entry): Likewise.
21267 (init_sections_and_labels): Likewise.
21268 (dwarf2out_early_finish): Likewise.
21269 (maybe_reset_location_view): New, from...
21270 (dwarf2out_var_location): ... here. Call it.
21271 * debug.h (dwarf2out_default_as_loc_support): Declare.
21272 (dwarf2out_default_as_locview_support): Declare.
21273 * hooks.c (hook_int_rtx_insn_0): New.
21274 * hooks.h (hook_int_rtx_insn_0): Declare.
21275 * toplev.c (process_options): Take -gas-loc-support and
21276 -gas-locview-support from dwarf2out. Enable
21277 -gvariable-location-views by default only with locview
21278 assembler support. Enable -ginternal-reset-location-views by
21279 default only if the target defines the corresponding hook.
21280 Enable -ginline-points by default if location views are
21281 enabled; force it disabled if statement frontiers are
21283 * tree-inline.c (expand_call_inline): Check option variables.
21284 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
21286 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
21288 PR tree-optimization/84321
21289 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
21290 handling. Also check whether the anti-range contains any values
21291 that satisfy the mask; switch to a VR_RANGE if not.
21293 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
21296 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
21298 2018-02-13 Martin Jambor <mjambor@suse.cz>
21301 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
21302 of call statements, also set location of a load to a temporary.
21304 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
21306 * config/rl78/rl78.c (add_vector_labels): New function.
21307 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
21308 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
21309 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
21310 which checks that no arguments are passed.
21311 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
21312 * doc/extend.texi: Documentation for the new attribute.
21314 2018-02-13 Andreas Schwab <schwab@suse.de>
21316 * config/riscv/linux.h (CPP_SPEC): Define.
21318 2018-02-13 Jakub Jelinek <jakub@redhat.com>
21321 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
21322 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
21323 OPTION_MASK_ISA_AES as first argument to def_builtin_const
21324 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
21325 instead of OPTION_MASK_ISA_PCLMUL as first argument to
21326 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
21327 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
21328 temporarily for AES and PCLMUL builtins.
21330 PR tree-optimization/84339
21331 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
21332 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
21335 PR middle-end/84309
21336 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
21337 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
21338 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
21340 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
21342 * omp-simd-clone.h: New file.
21343 * omp-simd-clone.c: Include omp-simd-clone.h.
21344 (expand_simd_clones): No longer static.
21345 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
21346 cgraph.h and omp-simd-clone.h.
21347 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
21348 (vect_recog_widen_shift_pattern): Formatting fix.
21349 (vect_pattern_recog_1): Don't check optab for calls.
21352 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
21353 operands[2] into a REG before using gen_lowpart on it.
21355 2018-02-12 Jeff Law <law@redhat.com>
21358 * config/sh/sh.c (find_barrier): Consider a sibling call
21361 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
21362 successfully back substituting a reg.
21364 2018-02-12 Richard Biener <rguenther@suse.de>
21366 PR tree-optimization/84037
21367 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
21368 parameter, move visited init to caller.
21369 (vect_slp_analyze_operations): Separate cost from validity
21370 check, initialize visited once for all instances.
21371 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
21373 * tree-vect-stmts.c (vect_model_simple_cost): Make early
21375 (vect_model_promotion_demotion_cost): Likewise.
21376 (vectorizable_bswap): Guard cost modeling with !slp_node
21377 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
21379 (vectorizable_call): Likewise.
21380 (vectorizable_conversion): Likewise.
21381 (vectorizable_assignment): Likewise.
21382 (vectorizable_shift): Likewise.
21383 (vectorizable_operation): Likewise.
21384 (vectorizable_store): Likewise.
21385 (vectorizable_load): Likewise.
21386 (vectorizable_condition): Likewise.
21387 (vectorizable_comparison): Likewise.
21389 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
21392 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
21393 (ASAN_MARK): Fix fnspec to account for return value, change pointer
21394 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
21396 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21398 PR middle-end/83665
21399 * params.def (inline-min-speedup): Increase from 8 to 15.
21400 (max-inline-insns-auto): Decrease from 40 to 30.
21401 * ipa-split.c (consider_split): Add some buffer for function to
21402 be considered inlining candidate.
21403 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
21406 2018-02-12 Richard Biener <rguenther@suse.de>
21408 PR tree-optimization/84037
21409 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
21410 matched stmts if we cannot swap the non-matched ones.
21412 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21414 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
21415 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
21416 _mm_maskz_scalef_round_ss): New intrinsics.
21417 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
21418 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
21419 __builtin_ia32_scalefss_round): Remove.
21420 (__builtin_ia32_scalefsd_mask_round,
21421 __builtin_ia32_scalefss_mask_round): New intrinsics.
21422 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
21423 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
21424 ((match_operand:VF_128 2 "<round_nimm_predicate>"
21425 "<round_constraint>")): Changed to ...
21426 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
21427 "<round_scalar_constraint>")): ... this.
21428 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
21429 %0, %1, %2<round_op3>}"): Changed to ...
21430 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
21431 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
21432 %2<round_scalar_mask_op3>}"): ... this.
21433 * config/i386/subst.md (round_scalar_nimm_predicate): New.
21435 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
21437 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
21438 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
21439 (_mm_maskz_sqrt_round_ss): New intrinsics.
21440 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
21441 (__builtin_ia32_sqrtsd_mask_round)
21442 (__builtin_ia32_sqrtss_mask_round): New builtins.
21443 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
21444 (__builtin_ia32_sqrtss_round): Remove.
21445 (__builtin_ia32_sqrtsd_mask_round)
21446 (__builtin_ia32_sqrtss_mask_round): New builtins.
21447 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
21448 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
21449 ((match_operand:VF_128 1 "vector_operand"
21450 "xBm,<round_constraint>")): Changed to ...
21451 ((match_operand:VF_128 1 "vector_operand"
21452 "xBm,<round_scalar_constraint>")): ... this.
21453 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
21454 %0, %2, %<iptr>1<round_op3>}): Changed to ...
21455 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
21456 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
21457 %<iptr>1<round_scalar_mask_op3>}): ... this.
21458 ((set_attr "prefix" "<round_prefix>")): Changed to ...
21459 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
21461 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
21464 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
21465 Cast vec_cmpeq result to correct type.
21466 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
21467 Cast vec_cmpgt result to correct type.
21469 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
21471 * final.c (final_scan_insn_1): Renamed from...
21472 (final_scan_insn): ... this. New wrapper, to recover
21473 seen from the outermost call in recursive ones.
21474 * config/sparc/sparc.c (output_return): Drop seen from call.
21475 (output_sibcall): Likewise.
21476 * config/visium/visium.c (output_branch): Likewise.
21478 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
21480 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
21483 2018-02-10 Alan Modra <amodra@gmail.com>
21486 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
21487 Specify LR as an input.
21489 2018-02-10 Jakub Jelinek <jakub@redhat.com>
21492 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
21493 remove_member_access_dummy_vars): New functions.
21494 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
21495 lower_omp_1, execute_lower_omp): Use them.
21497 PR rtl-optimization/84308
21498 * shrink-wrap.c (spread_components): Release todo vector.
21500 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
21502 PR rtl-optimization/57193
21503 * ira-color.c (struct allocno_color_data): Add member
21504 conflict_allocno_hard_prefs.
21505 (update_conflict_allocno_hard_prefs): New.
21506 (bucket_allocno_compare_func): Add a preference based on
21507 conflict_allocno_hard_prefs.
21508 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
21509 (color_allocnos): Remove a dead code. Initiate
21510 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
21512 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21515 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
21516 constraint from =wa to wa. Avoid a subreg on the output operand,
21517 instead use a pseudo and subreg it in a move.
21518 (p9_xxbrd_<mode>): Changed to ...
21519 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
21520 (p9_xxbrd_v2df): New expander.
21521 (p9_xxbrw_<mode>): Changed to ...
21522 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
21523 (p9_xxbrw_v4sf): New expander.
21525 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21527 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
21529 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
21532 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
21533 multiply in 32-bit mode.
21534 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
21535 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
21538 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
21540 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
21541 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
21542 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
21543 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
21545 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
21548 * dwarf2out.c (is_trivial_indirect_ref): New function.
21549 (dwarf2out_late_global_decl): Do not generate a location
21550 attribute for variables that have a non-trivial DECL_VALUE_EXPR
21551 and that are not defined in the current unit.
21553 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21555 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
21556 instead of a libcall for UNORDERED.
21558 2018-02-09 Tamar Christina <tamar.christina@arm.com>
21561 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
21562 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
21564 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21567 * config/s390/s390.c (s390_set_current_function): Invoke
21568 s390_indirect_branch_settings also if fndecl didn't change.
21570 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21572 * config/rs6000/rs6000.md (blockage): Set length to zero.
21574 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
21576 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
21578 2018-02-09 Jakub Jelinek <jakub@redhat.com>
21581 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
21582 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
21586 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
21587 PARALLEL incoming that failed vt_get_decl_and_offset check.
21589 PR middle-end/84237
21590 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
21591 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
21593 (get_variable_section): For decls in named .bss* sections pass true as
21594 second argument to bss_initializer_p.
21596 2018-02-09 Marek Polacek <polacek@redhat.com>
21597 Jakub Jelinek <jakub@redhat.com>
21600 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
21601 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
21602 Sync some changes from cxx_fold_indirect_ref.
21604 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
21606 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
21608 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
21609 (BLOCK_INLINE_ENTRY_LABEL): New.
21610 (dwarf2out_var_location): Disregard inline entry markers.
21611 (inline_entry_data): New struct.
21612 (inline_entry_data_hasher): New hashtable type.
21613 (inline_entry_data_hasher::hash): New.
21614 (inline_entry_data_hasher::equal): New.
21615 (inline_entry_data_table): New variable.
21616 (add_high_low_attributes): Add DW_AT_entry_pc and
21617 DW_AT_GNU_entry_view attributes if a pending entry is found
21618 in inline_entry_data_table. Add old entry_pc attribute only
21619 if debug nonbinding markers are disabled.
21620 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
21621 markers are enabled.
21622 (block_within_block_p, dwarf2out_inline_entry): New.
21623 (dwarf2out_finish): Check that no entries remained in
21624 inline_entry_data_table.
21625 * final.c (reemit_insn_block_notes): Handle inline entry notes.
21626 (final_scan_insn, notice_source_line): Likewise.
21627 (rest_of_clean_state): Skip inline entry markers.
21628 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
21630 * gimple.c (gimple_build_debug_inline_entry): New.
21631 * gimple.h (enum gimple_debug_subcode): Add
21632 GIMPLE_DEBUG_INLINE_ENTRY.
21633 (gimple_build_debug_inline_entry): Declare.
21634 (gimple_debug_inline_entry_p): New.
21635 (gimple_debug_nonbind_marker_p): Adjust.
21636 * insn-notes.def (INLINE_ENTRY): New.
21637 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
21638 inline entry marker notes.
21639 (print_insn): Likewise.
21640 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
21641 (INSN_DEBUG_MARKER_KIND): Likewise.
21642 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
21643 * tree-inline.c (expand_call_inline): Build and insert
21644 debug_inline_entry stmt.
21645 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
21646 inline entry blocks early, if nonbind markers are enabled.
21647 (dump_scope_block): Dump fragment info.
21648 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
21649 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
21650 (gimple_build_debug_inline_entry): New.
21651 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
21652 Enable/disable inline entry points too.
21653 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
21654 (DEBUG_INSN): Describe inline entry markers.
21656 * common.opt (gvariable-location-views): New.
21657 (gvariable-location-views=incompat5): New.
21658 * config.in: Rebuilt.
21659 * configure: Rebuilt.
21660 * configure.ac: Test assembler for view support.
21661 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
21662 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
21663 * dwarf2out.c (var_loc_view): New typedef.
21664 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
21665 (dwarf2out_locviews_in_attribute): New.
21666 (dwarf2out_locviews_in_loclist): New.
21667 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
21668 (enum dw_line_info_opcode): Add LI_adv_address.
21669 (struct dw_line_info_table): Add view.
21670 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
21671 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
21672 (zero_view_p): New variable.
21673 (ZERO_VIEW_P): New macro.
21674 (output_asm_line_debug_info): New.
21675 (struct var_loc_node): Add view.
21676 (add_AT_view_list, AT_loc_list): New.
21677 (add_var_loc_to_decl): Add view param. Test it against last.
21678 (new_loc_list): Add view params. Record them.
21679 (AT_loc_list_ptr): Handle loc and view lists.
21680 (view_list_to_loc_list_val_node): New.
21681 (print_dw_val): Handle dw_val_class_view_list.
21682 (size_of_die): Likewise.
21683 (value_format): Likewise.
21684 (loc_list_has_views): New.
21685 (gen_llsym): Set vl_symbol too.
21686 (maybe_gen_llsym, skip_loc_list_entry): New.
21687 (dwarf2out_maybe_output_loclist_view_pair): New.
21688 (output_loc_list): Output view list or entries too.
21689 (output_view_list_offset): New.
21690 (output_die): Handle dw_val_class_view_list.
21691 (output_dwarf_version): New.
21692 (output_compilation_unit_header): Use it.
21693 (output_skeleton_debug_sections): Likewise.
21694 (output_rnglists, output_line_info): Likewise.
21695 (output_pubnames, output_aranges): Update version comments.
21696 (output_one_line_info_table): Output view numbers in asm comments.
21697 (dw_loc_list): Determine current endview, pass it to new_loc_list.
21698 Call maybe_gen_llsym.
21699 (loc_list_from_tree_1): Adjust.
21700 (add_AT_location_description): Create view list attribute if
21701 needed, check it's absent otherwise.
21702 (convert_cfa_to_fb_loc_list): Adjust.
21703 (maybe_emit_file): Call output_asm_line_debug_info for test.
21704 (dwarf2out_var_location): Reset views as needed. Precompute
21705 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
21706 attribute. Set view.
21707 (new_line_info_table): Reset next view.
21708 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
21709 (dwarf2out_source_line): Likewise. Output view resets and labels to
21710 the assembler, or select appropriate line info opcodes.
21711 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
21712 (optimize_string_length): Catch it. Adjust.
21713 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
21714 dw_val_class_view_list, and remove it if no longer needed.
21715 (hash_loc_list): Hash view numbers.
21716 (loc_list_hasher::equal): Compare them.
21717 (optimize_location_lists): Check whether a view list symbol is
21718 needed, and whether the locview attribute is present, and
21719 whether they match. Remove the locview attribute if no longer
21721 (index_location_lists): Call skip_loc_list_entry for test.
21722 (dwarf2out_finish): Call output_asm_line_debug_info for test.
21723 Use output_dwarf_version.
21724 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
21725 (struct dw_val_node): Add val_view_list.
21726 * final.c (SEEN_NEXT_VIEW): New.
21727 (set_next_view_needed): New.
21728 (clear_next_view_needed): New.
21729 (maybe_output_next_view): New.
21730 (final_start_function): Rename to...
21731 (final_start_function_1): ... this. Take pointer to FIRST,
21732 add SEEN parameter. Emit param bindings in the initial view.
21733 (final_start_function): Reintroduce SEEN-less interface.
21734 (final): Rename to...
21735 (final_1): ... this. Take SEEN parameter. Output final pending
21736 next view at the end.
21737 (final): Reintroduce seen-less interface.
21738 (final_scan_insn): Output pending next view before switching
21739 sections or ending a block. Mark the next view as needed when
21740 outputting variable locations. Notify debug backend of section
21741 changes, and of location view changes.
21742 (rest_of_handle_final): Adjust.
21743 * toplev.c (process_options): Autodetect value for debug variable
21744 location views option. Warn on incompat5 without -gdwarf-5.
21745 * doc/invoke.texi (gvariable-location-views): New.
21746 (gvariable-location-views=incompat5): New.
21747 (gno-variable-location-views): New.
21749 2018-02-08 David Malcolm <dmalcolm@redhat.com>
21751 PR tree-optimization/84136
21752 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
21753 that the result of find_edge is non-NULL.
21755 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21758 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
21759 storing integer register in SImode. Fix cost of 256 and 512
21760 byte aligned SSE register store.
21762 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
21764 * config/i386/i386.c (ix86_multiplication_cost): Fix
21765 multiplication cost for TARGET_AVX512DQ.
21767 2018-02-08 Marek Polacek <polacek@redhat.com>
21769 PR tree-optimization/84238
21770 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
21773 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21775 PR tree-optimization/84265
21776 * tree-vect-stmts.c (vectorizable_store): Don't treat
21777 VMAT_CONTIGUOUS accesses as grouped.
21778 (vectorizable_load): Likewise.
21780 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21782 PR tree-optimization/81635
21783 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
21784 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
21785 (test_round_for_mask): New functions.
21786 (wide_int_cc_tests): Call test_round_for_mask.
21787 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
21788 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
21789 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
21790 range returned by get_range_info.
21792 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
21795 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
21796 * symtab.c: Include builtins.h
21797 (symtab_node::output_to_lto_symbol_table_p): Move here
21798 from lto-streamer-out.c:output_symbol_p.
21799 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
21800 (output_symbol_p): Move all logic to symtab.c
21801 (produce_symtab): Update.
21803 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
21805 * config/s390/s390-opts.h (enum indirect_branch): Define.
21806 * config/s390/s390-protos.h (s390_return_addr_from_memory)
21807 (s390_indirect_branch_via_thunk)
21808 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
21809 (enum s390_indirect_branch_type): Define.
21810 * config/s390/s390.c (struct s390_frame_layout, struct
21811 machine_function): Remove.
21812 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
21813 (indirect_branch_table_label_no, indirect_branch_table_name):
21815 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
21816 (enum s390_indirect_branch_option): Define.
21817 (s390_return_addr_from_memory): New function.
21818 (s390_handle_string_attribute): New function.
21819 (s390_attribute_table): Add new attribute handler.
21820 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
21821 (s390_indirect_branch_via_thunk): New function.
21822 (s390_indirect_branch_via_inline_thunk): New function.
21823 (s390_function_ok_for_sibcall): When jumping via thunk disallow
21824 sibling call optimization for non z10 compiles.
21825 (s390_emit_call): Force indirect branch target to be a single
21826 register. Add r1 clobber for non-z10 compiles.
21827 (s390_emit_epilogue): Emit return jump via return_use expander.
21828 (s390_reorg): Handle JUMP_INSNs as execute targets.
21829 (s390_option_override_internal): Perform validity checks for the
21830 new command line options.
21831 (s390_indirect_branch_attrvalue): New function.
21832 (s390_indirect_branch_settings): New function.
21833 (s390_set_current_function): Invoke s390_indirect_branch_settings.
21834 (s390_output_indirect_thunk_function): New function.
21835 (s390_code_end): Implement target hook.
21836 (s390_case_values_threshold): Implement target hook.
21837 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
21839 * config/s390/s390.h (struct s390_frame_layout)
21840 (struct machine_function): Move here from s390.c.
21841 (TARGET_INDIRECT_BRANCH_NOBP_RET)
21842 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
21843 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
21844 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
21845 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
21846 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
21847 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
21848 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
21849 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
21850 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
21851 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
21852 (mnemonic attribute): Add values which aren't recognized
21854 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
21855 pattern for branch conversion. Fix mnemonic attribute.
21856 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
21857 indirect branch via thunk if requested.
21858 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
21859 ("*indirect_jump"): Disable for branch conversion using out of
21861 ("indirect_jump_via_thunk<mode>_z10")
21862 ("indirect_jump_via_thunk<mode>")
21863 ("indirect_jump_via_inlinethunk<mode>_z10")
21864 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
21865 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
21866 ("casesi_jump_via_inlinethunk<mode>_z10")
21867 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
21868 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
21869 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
21870 ("*indirect2_jump"): Disable for branch conversion.
21871 ("casesi_jump"): Turn into expander and expand patterns for branch
21873 ("return_use"): New expander.
21874 ("*return"): Emit return via thunk and rename it to ...
21875 ("*return<mode>"): ... this one.
21876 * config/s390/s390.opt: Add new options and and enum for the
21879 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
21881 * lra-constraints.c (match_reload): Unconditionally use
21882 gen_lowpart_SUBREG, rather than selecting between that
21883 and equivalent gen_rtx_SUBREG code.
21885 2018-02-08 Richard Biener <rguenther@suse.de>
21887 PR tree-optimization/84233
21888 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
21889 changed flag instead of boguously re-using phi_inserted.
21891 2018-02-08 Martin Jambor <mjambor@suse.cz>
21893 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
21894 static local variables.
21896 2018-02-08 Richard Biener <rguenther@suse.de>
21898 PR tree-optimization/84278
21899 * tree-vect-stmts.c (vectorizable_store): When looking for
21900 smaller vector types to perform grouped strided loads/stores
21901 make sure the mode is supported by the target.
21902 (vectorizable_load): Likewise.
21904 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21906 * config/aarch64/aarch64.c (aarch64_components_for_bb):
21907 Increase LDP/STP opportunities by adding adjacent callee-saves.
21909 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
21911 PR rtl-optimization/84068
21912 PR rtl-optimization/83459
21913 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
21915 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
21917 PR tree-optimization/84224
21918 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
21919 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
21920 non-zero arguments.
21922 2018-02-07 Iain Sandoe <iain@codesourcery.com>
21925 * config/rs6000/altivec.md (*restore_world): Remove LR use.
21926 * config/rs6000/predicates.md (restore_world_operation): Adjust op
21927 count, remove one USE.
21929 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
21931 * doc/install.texi (Configuration): Document the
21932 --with-long-double-format={ibm,ieee} PowerPC configuration
21936 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
21937 Convert from define_expand to be define_insn_and_split. Rework
21938 float/double/_Float128 conversions to QI/HI/SImode to work with
21939 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
21940 conversions to QI/HImode types did a store and then a load to
21941 truncate the value. For conversions to VSX registers, don't split
21942 the insn, instead emit the code directly. Use the code iterator
21943 any_fix to combine signed and unsigned conversions.
21944 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
21945 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
21946 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21947 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
21948 (fix_<mode>di2_hw): Likewise.
21949 (fixuns_<mode>di2_hw): Likewise.
21950 (fix_<mode>si2_hw): Likewise.
21951 (fixuns_<mode>si2_hw): Likewise.
21952 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
21953 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
21954 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
21955 fix<uns>_trunc<SFDF:mode>si2_p8.
21956 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
21957 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
21958 (fix<uns>_<mode>_mem): Likewise.
21959 (fctiw<u>z_<mode>_mem): Likewise.
21960 (fix<uns>_<mode>_mem): Likewise.
21961 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
21962 the register allocator from doing a direct move to the GPRs to do
21963 a store, and instead use the ISA 3.0 store byte/half-word from
21964 vector register instruction. For IEEE 128-bit floating point,
21965 also optimize stores of 32-bit ints.
21966 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21968 2018-02-07 Alan Hayward <alan.hayward@arm.com>
21970 * genextract.c (push_pathstr_operand): New function to support
21972 (walk_rtx): Call push_pathstr_operand.
21973 (print_path): Support [a-zA-Z].
21975 2018-02-07 Richard Biener <rguenther@suse.de>
21977 PR tree-optimization/84037
21978 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21979 (cse_and_gimplify_to_preheader): Declare.
21980 (vect_get_place_in_interleaving_chain): Likewise.
21981 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21983 (_loop_vec_info::~_loop_vec_info): Delete it.
21984 (cse_and_gimplify_to_preheader): New function.
21985 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21986 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21987 (vectorizable_load): Likewise. For grouped stores always base
21988 the IV on the first element.
21989 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21990 condition before gimplifying.
21992 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21994 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21995 *DIV_EXPR and *MOD_EXPR.
21997 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
22000 * config/i386/i386.c (ix86_option_override_internal): Mask out
22001 the CF_SET bit when checking -fcf-protection.
22003 2018-02-07 Tom de Vries <tom@codesourcery.com>
22006 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
22009 2018-02-07 Richard Biener <rguenther@suse.de>
22011 PR tree-optimization/84204
22012 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
22015 PR tree-optimization/84205
22016 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
22017 special-case isl_ast_op_zdiv_r.
22019 PR tree-optimization/84223
22020 * graphite-scop-detection.c (gather_bbs::before_dom_children):
22021 Only add conditions from within the region.
22022 (gather_bbs::after_dom_children): Adjust.
22024 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
22027 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
22028 * config/avr/avr.md: Only post-reload split REG-REG moves if
22029 either register is GENERAL_REG_P.
22031 2018-02-07 Jakub Jelinek <jakub@redhat.com>
22033 PR tree-optimization/84235
22034 * tree-ssa-scopedtables.c
22035 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
22036 if the subtraction is performed in floating point type where NaNs are
22037 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
22038 build 1. Formatting fix.
22040 2018-02-06 Jakub Jelinek <jakub@redhat.com>
22043 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
22044 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
22045 and skip it regardless of bb boundaries. Use CALL_P macro,
22046 don't test INSN_P (insn) together with CALL_P or JUMP_P check
22047 unnecessarily, formatting fix.
22049 2018-02-06 Michael Collison <michael.collison@arm.com>
22051 * config/arm/thumb2.md:
22052 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
22053 (*thumb_mov_notscc): Ditto.
22055 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
22058 * config/rs6000/rs6000.md (su code attribute): Use "u" for
22059 unsigned_fix, not "s".
22061 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22063 * configure.ac (gcc_fn_eh_frame_ro): New function.
22064 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
22065 correct .eh_frame permissions.
22066 * configure: Regenerate.
22068 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
22070 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
22071 irrelevant options.
22073 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22075 * config/rs6000/rs6000.c (rs6000_option_override_internal):
22076 Display warning message for -mno-speculate-indirect-jumps.
22078 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
22080 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
22082 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
22084 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
22086 PR tree-optimization/84225
22087 * tree-eh.c (find_trapping_overflow): Only call
22088 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
22090 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22093 * config/i386/i386.c: Reimplement the check of possible options
22094 -mibt/-mshstk conbination. Change error messages.
22095 * doc/invoke.texi: Fix a typo: remove extra '='.
22097 2018-02-06 Marek Polacek <polacek@redhat.com>
22099 PR tree-optimization/84228
22100 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
22102 2018-02-06 Tamar Christina <tamar.christina@arm.com>
22105 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
22106 emitted arch directives.
22107 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
22108 __ARM_FEATURE_COPROC before changing architectures.
22110 2018-02-06 Richard Biener <rguenther@suse.de>
22112 * config/i386/i386.c (print_reg): Fix typo.
22113 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
22115 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
22117 * configure: Regenerate.
22119 2018-02-05 Martin Sebor <msebor@redhat.com>
22121 PR tree-optimization/83369
22122 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
22125 2018-02-05 Martin Liska <mliska@suse.cz>
22127 * doc/invoke.texi: Cherry-pick upstream r323995.
22129 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
22131 * ira.c (ira_init_register_move_cost): Adjust comment.
22133 2018-02-05 Martin Liska <mliska@suse.cz>
22135 PR gcov-profile/84137
22136 * doc/gcov.texi: Fix typo in documentation.
22138 2018-02-05 Martin Liska <mliska@suse.cz>
22140 PR gcov-profile/83879
22141 * doc/gcov.texi: Document necessity of --dynamic-list-data when
22142 using dlopen functionality.
22144 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
22146 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
22147 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
22148 _mm_maskz_range_ss, _mm_mask_range_round_ss,
22149 _mm_maskz_range_round_ss): New intrinsics.
22150 (__builtin_ia32_rangesd128_round)
22151 (__builtin_ia32_rangess128_round): Remove.
22152 (__builtin_ia32_rangesd128_mask_round,
22153 __builtin_ia32_rangess128_mask_round): New builtins.
22154 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
22155 __builtin_ia32_rangess128_round): Remove.
22156 (__builtin_ia32_rangesd128_mask_round,
22157 __builtin_ia32_rangess128_mask_round): New builtins.
22158 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
22159 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
22160 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
22161 "<round_saeonly_constraint>")): Changed to ...
22162 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
22163 "<round_saeonly_scalar_constraint>")): ... this.
22164 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
22165 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
22166 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
22167 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
22168 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
22170 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
22172 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
22174 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
22175 Remove all values except native, 8540 and 8548.
22177 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
22179 * config/i386/i386.c (ix86_output_function_return): Pass
22180 INVALID_REGNUM, instead of -1, as invalid register number to
22181 indirect_thunk_name and output_indirect_thunk.
22183 2018-02-02 Julia Koval <julia.koval@intel.com>
22185 * config.gcc: Add -march=icelake.
22186 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
22187 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
22188 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
22189 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
22190 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
22191 (processor_target_table): Add icelake.
22192 (ix86_option_override_internal): Handle new PTAs.
22193 (get_builtin_code_for_version): Handle icelake.
22194 (M_INTEL_COREI7_ICELAKE): New.
22195 (fold_builtin_cpu): Handle icelake.
22196 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
22197 * doc/invoke.texi: Add -march=icelake.
22199 2018-02-02 Julia Koval <julia.koval@intel.com>
22201 * config/i386/i386.c (ix86_option_override_internal): Change flags type
22202 to wide_int_bitmask.
22203 * wide-int-bitmask.h: New.
22205 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
22208 * config/i386/i386.md: Replace Pmode with word_mode in
22209 builtin_setjmp_setup and builtin_longjmp to support x32.
22211 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
22215 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
22217 (rs6000_supported_cpu_names): New static variable.
22218 (linux_cpu_translation_table): Likewise.
22219 (elf_platform) <cpu>: Define new static variable and use it.
22220 Translate kernel AT_PLATFORM name to canonical name if needed.
22221 Error if platform name is unknown.
22223 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
22226 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
22228 2018-02-01 Jeff Law <law@redhat.com>
22231 * config/i386/i386.c (release_scratch_register_on_entry): Add new
22232 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
22233 the scratch if RELEASE_VIA_POP is false.
22234 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
22235 If we have to save a temporary register, decrement SIZE appropriately.
22236 Pass new arguments to release_scratch_register_on_entry.
22237 (ix86_adjust_stack_and_probe): Likewise.
22238 (ix86_emit_probe_stack_range): Pass new arguments to
22239 release_scratch_register_on_entry.
22241 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
22243 PR rtl-optimization/84157
22244 * combine.c (change_zero_ext): Use REG_P predicate in
22245 front of HARD_REGISTER_P predicate.
22247 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
22249 * config/avr/avr.c (avr_option_override): Move disabling of
22250 -fdelete-null-pointer-checks to...
22251 * common/config/avr/avr-common.c (avr_option_optimization_table):
22254 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22256 PR tree-optimization/81635
22257 * tree-data-ref.c (split_constant_offset_1): For types that
22258 wrap on overflow, try to use range info to prove that wrapping
22261 2018-02-01 Renlin Li <renlin.li@arm.com>
22264 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
22265 TAILCALL_ADDR_REGS.
22266 (aarch64_register_move_cost): Likewise.
22267 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
22268 TAILCALL_ADDR_REGS.
22269 (REG_CLASS_NAMES): Likewise.
22270 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
22271 TAILCALL_ADDR_REGS. Remove IP registers.
22272 * config/aarch64/aarch64.md (Ucs): Update register constraint.
22274 2018-02-01 Richard Biener <rguenther@suse.de>
22276 * domwalk.h (dom_walker::dom_walker): Add additional constructor
22277 for specifying RPO order and allow NULL for that.
22278 * domwalk.c (dom_walker::dom_walker): Likewise.
22279 (dom_walker::walk): Handle NULL RPO order.
22280 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
22282 (rewrite_update_dom_walker): Likewise.
22283 (mark_def_dom_walker): Likewise.
22285 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22287 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
22288 (aarch64_maybe_expand_sve_subreg_move): Declare.
22289 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
22290 * config/aarch64/predicates.md (aarch64_any_register_operand): New
22292 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
22293 that are semantically a reverse operation.
22294 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
22295 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
22296 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
22298 (aarch64_can_change_mode_class): For big-endian, forbid changes
22299 between two SVE modes if they have different element sizes.
22301 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22303 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
22304 the TImode handling for big-endian targets.
22306 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22308 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
22309 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
22311 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
22312 Remove BSWAP handing for big-endian targets and use the form of
22313 LD1RQ appropariate for the mode.
22315 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22317 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
22318 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
22319 duplicated element.
22321 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
22324 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
22325 check for operands that need to go through aarch64_sve_reload_be.
22327 2018-02-01 Jakub Jelinek <jakub@redhat.com>
22329 PR tree-optimization/81661
22330 PR tree-optimization/84117
22331 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
22332 * tree-eh.c: Include gimplify.h.
22333 (find_trapping_overflow, replace_trapping_overflow,
22334 rewrite_to_non_trapping_overflow): New functions.
22335 * tree-vect-loop.c: Include tree-eh.h.
22336 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
22337 * tree-data-ref.c: Include tree-eh.h.
22338 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
22340 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
22342 PR rtl-optimization/84123
22343 * combine.c (change_zero_ext): Check if hard register satisfies
22344 can_change_dest_mode before calling gen_lowpart_SUBREG.
22346 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
22349 * ira.c (ira_init_register_move_cost): Remove assert.
22351 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22353 PR rtl-optimization/84071
22354 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
22355 * doc/tm.texi: Regenerate.
22357 2018-01-31 Richard Biener <rguenther@suse.de>
22359 PR tree-optimization/84132
22360 * tree-data-ref.c (analyze_miv_subscript): Properly
22361 check whether evolution_function_is_affine_multivariate_p
22362 before calling gcd_of_steps_may_divide_p.
22364 2018-01-31 Julia Koval <julia.koval@intel.com>
22367 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
22368 * config/i386/i386.md (rdpid_rex64) New.
22369 (rdpid): Make 32bit only.
22371 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
22374 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
22375 an IDENTIFIER_NODE for FUNCTION_TYPE's.
22377 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22380 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
22382 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
22384 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
22386 PR rtl-optimization/84071
22387 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
22388 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
22390 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22392 * config/arc/arc.c (arc_handle_aux_attribute): New function.
22393 (arc_attribute_table): Add 'aux' attribute.
22394 (arc_in_small_data_p): Consider aux like variables.
22395 (arc_is_aux_reg_p): New function.
22396 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
22397 (arc_get_aux_arg): New function.
22398 (prepare_move_operands): Handle aux-register access.
22399 (arc_handle_aux_attribute): New function.
22400 * doc/extend.texi (ARC Variable attributes): Add subsection.
22402 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
22404 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
22405 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
22406 (arc_attribute_table): Add 'uncached' attribute.
22407 (arc_print_operand): Print '.di' flag for uncached memory
22409 (arc_in_small_data_p): Do not consider for small data the uncached
22411 (arc_is_uncached_mem_p): New function.
22412 * config/arc/predicates.md (compact_store_memory_operand): Check
22413 for uncached memory accesses.
22414 (nonvol_nonimm_operand): Likewise.
22415 * doc/extend.texi (ARC Type Attribute): New subsection.
22417 2018-01-31 Jakub Jelinek <jakub@redhat.com>
22420 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
22421 falign-loops=): Add Optimization flag.
22423 2018-01-30 Jeff Law <law@redhat.com>
22426 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
22427 INT_REGISTERS_SAVED. Check it prior to calling
22428 get_scratch_register_on_entry.
22429 (ix86_adjust_stack_and_probe): Similarly.
22430 (ix86_emit_probe_stack_range): Similarly.
22431 (ix86_expand_prologue): Corresponding changes.
22433 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22436 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
22437 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
22439 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
22442 * lra-constraints.c (curr_insn_transform): Process AND in the
22445 2018-01-30 Jakub Jelinek <jakub@redhat.com>
22447 PR rtl-optimization/83986
22448 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
22449 dependence against last_pending_memory_flush in addition to
22450 pending_jump_insns.
22452 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
22454 PR tree-optimization/81611
22455 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
22458 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
22461 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
22464 2018-01-30 Richard Biener <rguenther@suse.de>
22465 Jakub Jelinek <jakub@redhat.com>
22467 PR tree-optimization/84111
22468 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
22469 inner loops added during recursion, as they don't have up-to-date
22472 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22475 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
22476 (can_inline_edge_by_limits_p): ... here.
22477 (can_early_inline_edge_p, check_callers,
22478 update_caller_keys, update_callee_keys, recursive_inlining,
22479 add_new_edges_to_heap, speculation_useful_p,
22480 inline_small_functions,
22481 inline_small_functions, flatten_function,
22482 inline_to_all_callers_1): Update.
22484 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
22486 * profile-count.c (profile_count::combine_with_ipa_count): Handle
22489 2018-01-30 Richard Biener <rguenther@suse.de>
22491 PR tree-optimization/83008
22492 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
22493 invariant and constant vector uses in stmts when they need
22494 more than one stmt.
22496 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22499 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
22500 * configure: Regenerate.
22502 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22504 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
22506 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
22507 Use gen_rtx_REG rather than gen_lowpart.
22509 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22511 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
22512 rather than 0 when creating partial subregs.
22514 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
22516 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
22519 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
22522 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
22523 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
22524 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
22525 flags. This restores the settings used before the 2017-07-24.
22526 Turning off pre increment/decrement/modify allows IVOPTS to
22527 optimize DF/SF loops where the index is an int.
22529 2018-01-29 Richard Biener <rguenther@suse.de>
22530 Kelvin Nilsen <kelvin@gcc.gnu.org>
22533 * tree-vect-stmts.c (vectorizable_call): Don't call
22534 targetm.vectorize_builtin_md_vectorized_function if callee is
22537 2018-01-22 Carl Love <cel@us.ibm.com>
22539 * doc/extend.tex: Fix typo in second arg in
22540 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
22542 2018-01-29 Richard Biener <rguenther@suse.de>
22544 PR tree-optimization/84086
22545 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
22546 (flush_ssaname_freelist): When SSA names were released reset
22547 the SCEV hash table.
22549 2018-01-29 Richard Biener <rguenther@suse.de>
22551 PR tree-optimization/84057
22552 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
22553 removed paths when removing edges.
22555 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
22557 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
22558 -mfunction-return=@var{choice}.
22560 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
22562 PR diagnostic/84034
22563 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
22564 Handle CR like TAB.
22565 (layout::print_source_line): Likewise.
22566 (test_get_line_width_without_trailing_whitespace): Add test cases.
22568 2018-01-27 Jakub Jelinek <jakub@redhat.com>
22570 PR middle-end/84040
22571 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
22574 2018-01-26 Jim Wilson <jimw@sifive.com>
22576 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
22578 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
22581 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22583 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
22584 and CMP + SUB-immediate -> SUBS.
22586 2018-01-26 Martin Sebor <msebor@redhat.com>
22588 PR tree-optimization/83896
22589 * tree-ssa-strlen.c (get_string_len): Rename...
22590 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
22591 Avoid assuming length is constant.
22592 (handle_char_store): Use HOST_WIDE_INT for string length.
22594 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
22597 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
22598 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
22600 2018-01-26 Richard Biener <rguenther@suse.de>
22602 PR rtl-optimization/84003
22603 * dse.c (record_store): Only record redundant stores when
22604 the earlier store aliases at least all accesses the later one does.
22606 2018-01-26 Jakub Jelinek <jakub@redhat.com>
22608 PR rtl-optimization/83985
22609 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
22610 REG_CFA_RESTORE insns.
22611 (delete_unmarked_insns): Don't ignore separate shrink wrapping
22612 REG_CFA_RESTORE insns here.
22615 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
22616 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
22618 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22620 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
22621 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
22622 (arc_init): Likewise.
22623 (arc_override_options): Likewise.
22624 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
22626 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
22628 * config/arc/arc.h (TARGET_DBNZ): Define.
22629 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
22630 properly set the tune attribute.
22631 (dbnz): Use TARGET_DBNZ guard.
22632 * config/arc/arc.opt (mtune): Add core3 option.
22634 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22636 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
22637 recognize new pic like addresses.
22638 (arc_delegitimize_address): Clean up.
22640 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22642 * config/arc/arc-arches.def: Option mrf16 valid for all
22644 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
22645 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
22646 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
22647 * config/arc/arc-tables.opt: Regenerate.
22648 * config/arc/arc.c (arc_conditional_register_usage): Handle
22649 reduced register file case.
22650 (arc_file_start): Set must have build attributes.
22651 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
22652 mrf16 option value.
22653 * config/arc/arc.opt (mrf16): Add new option.
22654 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
22655 * config/arc/genmultilib.awk: Handle new mrf16 option.
22656 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
22657 * config/arc/t-multilib: Regenerate.
22658 * doc/invoke.texi (ARC Options): Document mrf16 option.
22660 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22662 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
22663 * config/arc/arc.c (arc_handle_secure_attribute): New function.
22664 (arc_attribute_table): Add 'secure_call' attribute.
22665 (arc_print_operand): Print secure call operand.
22666 (arc_function_ok_for_sibcall): Don't optimize tail calls when
22668 (arc_is_secure_call_p): New function. * config/arc/arc.md
22669 (call_i): Add support for sjli instruction.
22670 (call_value_i): Likewise.
22671 * config/arc/constraints.md (Csc): New constraint.
22673 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
22674 John Eric Martin <John.Martin@emmicro-us.com>
22676 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
22677 * config/arc/arc.c (_arc_jli_section): New struct.
22678 (arc_jli_section): New type.
22679 (rc_jli_sections): New static variable.
22680 (arc_handle_jli_attribute): New function.
22681 (arc_attribute_table): Add jli_always and jli_fixed attribute.
22682 (arc_file_end): New function.
22683 (TARGET_ASM_FILE_END): Define.
22684 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
22685 (arc_add_jli_section): New function.
22686 (jli_call_scan): Likewise.
22687 (arc_reorg): Call jli_call_scan.
22688 (arc_output_addsi): Remove 'S' from printing asm operand.
22689 (arc_is_jli_call_p): New function.
22690 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
22692 (movhi_insn): Likewise.
22693 (movsi_insn): Likewise.
22694 (movsi_set_cc_insn): Likewise.
22695 (loadqi_update): Likewise.
22696 (load_zeroextendqisi_update): Likewise.
22697 (load_signextendqisi_update): Likewise.
22698 (loadhi_update): Likewise.
22699 (load_zeroextendhisi_update): Likewise.
22700 (load_signextendhisi_update): Likewise.
22701 (loadsi_update): Likewise.
22702 (loadsf_update): Likewise.
22703 (movsicc_insn): Likewise.
22704 (bset_insn): Likewise.
22705 (bxor_insn): Likewise.
22706 (bclr_insn): Likewise.
22707 (bmsk_insn): Likewise.
22708 (bicsi3_insn): Likewise.
22709 (cmpsi_cc_c_insn): Likewise.
22710 (movsi_ne): Likewise.
22711 (movsi_cond_exec): Likewise.
22712 (clrsbsi2): Likewise.
22713 (norm_f): Likewise.
22723 (call_i): Remove 'S' asm letter, add jli instruction.
22724 (call_value_i): Likewise.
22725 * config/arc/arc.op (mjli-always): New option.
22726 * config/arc/constraints.md (Cji): New constraint.
22727 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
22729 (subsf3_fpx): Likewise.
22730 (mulsf3_fpx): Likewise.
22731 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
22733 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
22734 function attrbutes.
22735 * doc/invoke.texi (ARC): Document mjli-always option.
22737 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
22739 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
22740 avoid addition with 0 and use incw and decw where possible.
22742 2018-01-26 Richard Biener <rguenther@suse.de>
22744 PR tree-optimization/81082
22745 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
22746 association if it requires casting to unsigned.
22747 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
22748 from fold_plusminus_mult_expr to catch important cases late when
22749 range info is available.
22751 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
22753 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
22754 * configure.ac (hidden_linkonce): New test.
22755 * configure: Regenerate.
22756 * config.in: Regenerate.
22758 2018-01-26 Julia Koval <julia.koval@intel.com>
22760 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
22761 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
22762 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
22763 _mm_mask_bitshuffle_epi64_mask): Fix type.
22764 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
22765 USI_FTYPE_V4DI_V4DI_USI): Remove.
22766 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
22767 __builtin_ia32_vpshufbitqmb256_mask,
22768 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
22769 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
22770 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
22772 2018-01-26 Alan Modra <amodra@gmail.com>
22775 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
22776 UNSPEC_VBPERMQ. Sort other unspecs.
22778 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
22780 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
22782 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
22784 PR middle-end/83055
22785 * predict.c (drop_profile): Do not push/pop cfun; update also
22787 (handle_missing_profiles): Fix logic looking for zero profiles.
22789 2018-01-25 Jakub Jelinek <jakub@redhat.com>
22791 PR middle-end/83977
22792 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
22793 on functions with #pragma omp declare simd or functions with simd
22795 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
22796 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
22797 Remove trailing \n from warning_at calls.
22799 2018-01-25 Tom de Vries <tom@codesourcery.com>
22802 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22803 for neutered workers.
22805 2018-01-24 Joseph Myers <joseph@codesourcery.com>
22808 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
22809 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
22811 2018-01-24 Jeff Law <law@redhat.com>
22814 * i386.c (get_probe_interval): Move to earlier point.
22815 (ix86_compute_frame_layout): If -fstack-clash-protection and
22816 the frame is larger than the probe interval, then use pushes
22817 to save registers rather than reg->mem moves.
22818 (ix86_expand_prologue): Remove conditional for int_registers_saved
22821 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
22824 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
22825 min/max for never referenced object.
22827 2018-01-24 Jakub Jelinek <jakub@redhat.com>
22829 PR middle-end/83977
22830 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
22832 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
22833 attributes from DECL_ATTRIBUTES (decl) without affecting
22834 DECL_ATTRIBUTES (current_function_decl).
22835 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
22836 functions with non-NULL DECL_ABSTRACT_ORIGIN.
22838 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
22840 PR tree-optimization/83979
22841 * fold-const.c (fold_comparison): Use constant_boolean_node
22842 instead of boolean_{true,false}_node.
22844 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
22846 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
22849 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22851 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22852 Simplify the clause that sets the length attribute.
22853 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22854 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
22855 clause that sets the length attribute.
22856 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22858 2018-01-24 Tom de Vries <tom@codesourcery.com>
22861 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
22862 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
22863 Add strict parameter.
22864 (prevent_branch_around_nothing): Insert dummy insn between branch to
22865 label and label with no ptx insn inbetween.
22866 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
22868 2018-01-24 Tom de Vries <tom@codesourcery.com>
22871 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
22872 for neutered threads in warp.
22873 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
22875 2018-01-24 Richard Biener <rguenther@suse.de>
22877 PR tree-optimization/83176
22878 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
22881 2018-01-24 Richard Biener <rguenther@suse.de>
22883 PR tree-optimization/82819
22884 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
22885 code generating pluses that are no-ops in the target precision.
22887 2018-01-24 Richard Biener <rguenther@suse.de>
22889 PR middle-end/84000
22890 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
22892 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22894 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
22895 to merge probabilities.
22896 * predict.c (probably_never_executed): Also mark as cold functions
22897 with global 0 profile and guessed local profile.
22898 * profile-count.c (profile_probability::combine_with_count): New
22900 * profile-count.h (profile_probability::operator*,
22901 profile_probability::operator*=, profile_probability::operator/,
22902 profile_probability::operator/=): Reduce precision to adjusted
22903 and set value to guessed on contradictory divisions.
22904 (profile_probability::combine_with_freq): Remove.
22905 (profile_probability::combine_wiht_count): Declare.
22906 (profile_count::force_nonzero):: Set to adjusted.
22907 (profile_count::probability_in):: Set quality to adjusted.
22908 * tree-ssa-tail-merge.c (replace_block_by): Use
22909 combine_with_count.
22911 2018-01-23 Andrew Waterman <andrew@sifive.com>
22912 Jim Wilson <jimw@sifive.com>
22914 * config/riscv/riscv.c (riscv_stack_boundary): New.
22915 (riscv_option_override): Set riscv_stack_boundary. Handle
22916 riscv_preferred_stack_boundary_arg.
22917 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
22918 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
22919 (STACK_BOUNDARY): Set to riscv_stack_boundary.
22920 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
22921 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
22922 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
22924 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
22927 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
22928 of struct ix86_frame.
22929 (ix86_expand_epilogue): Likewise. Add a local variable for
22930 the reg_save_offset field in struct ix86_frame.
22932 2018-01-23 Bin Cheng <bin.cheng@arm.com>
22934 PR tree-optimization/82604
22935 * tree-loop-distribution.c (enum partition_kind): New enum item
22936 PKIND_PARTIAL_MEMSET.
22937 (partition_builtin_p): Support above new enum item.
22938 (generate_code_for_partition): Ditto.
22939 (compute_access_range): Differentiate cases that equality can be
22940 proven at all loops, the innermost loops or no loops.
22941 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
22942 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
22943 (finalize_partitions, distribute_loop): Don't fuse partition of
22944 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
22945 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
22946 parloop is enabled.
22948 2018-01-23 Martin Liska <mliska@suse.cz>
22950 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
22951 order to ignore the predictor.
22952 (PRED_POLYMORPHIC_CALL): Likewise.
22953 (PRED_RECURSIVE_CALL): Likewise.
22955 2018-01-23 Martin Liska <mliska@suse.cz>
22957 * tree-profile.c (tree_profiling): Print function header to
22958 aware reader which function we are working on.
22959 * value-prof.c (gimple_find_values_to_profile): Do not print
22960 not interesting value histograms.
22962 2018-01-23 Martin Liska <mliska@suse.cz>
22964 * profile-count.h (enum profile_quality): Add
22965 profile_uninitialized as the first value. Do not number values
22966 as they are zero based.
22967 (profile_count::verify): Update sanity check.
22968 (profile_probability::verify): Likewise.
22970 2018-01-23 Nathan Sidwell <nathan@acm.org>
22972 * doc/invoke.texi (ffor-scope): Deprecate.
22974 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22976 PR tree-optimization/83510
22977 * domwalk.c (set_all_edges_as_executable): New function.
22978 (dom_walker::dom_walker): Convert bool param
22979 "skip_unreachable_blocks" to enum reachability. Move setup of
22980 edge flags to set_all_edges_as_executable and only do it when
22981 reachability is REACHABLE_BLOCKS.
22982 * domwalk.h (enum dom_walker::reachability): New enum.
22983 (dom_walker::dom_walker): Convert bool param
22984 "skip_unreachable_blocks" to enum reachability.
22985 (set_all_edges_as_executable): New decl.
22986 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
22987 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22989 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22990 but converting true to REACHABLE_BLOCKS.
22991 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22993 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22994 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22995 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22997 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22998 if check_all_array_refs will be called.
23000 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23002 * tree.c (selftest::test_location_wrappers): Add more test
23005 2018-01-23 David Malcolm <dmalcolm@redhat.com>
23007 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
23008 (selftest::test_bit_in_range): Likewise.
23010 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
23013 * doc/sourcebuild.texi (vect_float): Say that the selector
23014 only describes the situation when -funsafe-math-optimizations is on.
23015 (vect_float_strict): Document.
23017 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
23019 PR tree-optimization/83965
23020 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
23021 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
23022 instead of checking only for a reduction.
23023 (vect_recog_widen_sum_pattern): Likewise.
23025 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23027 * predict.c (probably_never_executed): Only use precise profile info.
23028 (compute_function_frequency): Skip after inlining hack since we now
23029 have quality checking.
23031 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
23033 * profile-count.h (profile_probability::very_unlikely,
23034 profile_probability::unlikely, profile_probability::even): Set
23035 precision to guessed.
23037 2018-01-23 Richard Biener <rguenther@suse.de>
23039 PR tree-optimization/83963
23040 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
23041 Properly terminate dominator walk when crossing the exit edge not
23042 when visiting its source block.
23044 2018-01-23 Jakub Jelinek <jakub@redhat.com>
23047 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
23048 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
23050 2018-01-22 Jakub Jelinek <jakub@redhat.com>
23052 PR tree-optimization/83957
23053 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
23054 semicolon after for body surrounded by braces.
23056 PR tree-optimization/83081
23057 * profile-count.h (profile_probability::split): New method.
23058 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
23059 Use profile_probability::split.
23060 (do_compare_rtx_and_jump): Fix adjustment of probabilities
23061 when splitting a single conditional jump into 2.
23063 2018-01-22 David Malcolm <dmalcolm@redhat.com>
23065 PR tree-optimization/69452
23066 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
23069 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23071 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
23072 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
23073 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
23075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23077 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
23079 * config/rl78/rl78.md (movdi): New define_expand.
23080 * config/rl78/rl78.c (rl78_split_movdi): New function.
23082 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
23085 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
23087 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
23088 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
23089 128-bit to produce an UNSPEC move to get the double word with the
23090 signbit and then a shift directly to do signbit.
23091 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
23092 implementation with a new version that just does either a direct
23093 move or a regular move. Move memory interface to separate insns.
23094 Move insns so they are next to the expander.
23095 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
23096 with signbit move. Split big and little endian case.
23097 (signbit<mode>2_dm_mem_le): Likewise.
23098 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
23099 (signbit<mode>2_dm2): Likewise.
23101 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23103 * config/rl78/rl78.md (anddi3): New define_expand.
23105 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23107 * config/rl78/rl78.md (umindi3): New define_expand.
23109 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23111 * config/rl78/rl78.md (smindi3): New define_expand.
23113 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23115 * config/rl78/rl78.md (smaxdi3): New define_expand.
23117 2018-01-22 Carl Love <cel@us.ibm.com>
23119 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
23120 LVX_V1TI): Add macro expansion.
23121 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
23122 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
23123 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
23124 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
23125 Change check to determine if the instruction is a byte reversing
23126 entry. Fix typo in comment.
23127 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
23128 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
23129 Add def_builtin calls for new builtins.
23130 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
23131 Add define_insn expansion.
23133 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23135 * config/rl78/rl78.md (umaxdi3): New define_expand.
23137 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
23139 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
23140 for non-QImode registers.
23142 2018-01-22 Richard Biener <rguenther@suse.de>
23144 PR tree-optimization/83963
23145 * graphite-scop-detection.c (scop_detection::get_sese): Delay
23146 including the loop exit block.
23147 (scop_detection::merge_sese): Likewise.
23148 (scop_detection::add_scop): Do it here instead.
23150 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23152 * doc/sourcebuild.texi (arm_softfloat): Document.
23154 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
23157 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
23158 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
23159 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
23161 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23162 David Edelsohn <dje.gcc@gmail.com>
23165 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23166 Change "crset eq" to "crset 2".
23167 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23168 (*call_indirect_aix<mode>_nospec): Likewise.
23169 (*call_value_indirect_aix<mode>_nospec): Likewise.
23170 (*call_indirect_elfv2<mode>_nospec): Likewise.
23171 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
23172 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
23173 change assembly output from . to $.
23174 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23175 (indirect_jump<mode>_nospec): Change assembly output from . to $.
23176 (*tablejump<mode>_internal1_nospec): Likewise.
23178 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
23181 * config/sh/sh_optimize_sett_clrt.cc:
23182 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
23184 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
23186 PR tree-optimization/83940
23187 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
23188 offset_dt to vect_constant_def rather than vect_unknown_def_type.
23189 (vect_check_load_store_mask): Add a mask_dt_out parameter and
23190 use it to pass back the definition type.
23191 (vect_check_store_rhs): Likewise rhs_dt_out.
23192 (vect_build_gather_load_calls): Add a mask_dt argument and use
23193 it instead of a call to vect_is_simple_use.
23194 (vectorizable_store): Update calls to vect_check_load_store_mask
23195 and vect_check_store_rhs. Use the dt returned by the latter instead
23196 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
23197 instead of calls to vect_is_simple_use. Pass the scalar rather
23198 than the vector operand to vect_is_simple_use when handling
23199 second and subsequent copies of an rhs value.
23200 (vectorizable_load): Update calls to vect_check_load_store_mask
23201 and vect_build_gather_load_calls. Use the cached mask_dt and
23202 gs_info.offset_dt instead of calls to vect_is_simple_use.
23204 2018-01-20 Jakub Jelinek <jakub@redhat.com>
23206 PR middle-end/83945
23207 * tree-emutls.c: Include gimplify.h.
23208 (lower_emutls_2): New function.
23209 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
23210 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
23211 it before further processing.
23214 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
23215 UINTVAL (trueop1) instead of INTVAL (op1).
23217 2018-01-19 Jakub Jelinek <jakub@redhat.com>
23221 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
23222 INCOMING_FRAME_SP_OFFSET if not defined.
23223 (scan_trace): Add ENTRY argument. If true and
23224 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
23225 emit a note to adjust the CFA offset.
23226 (create_cfi_notes): Adjust scan_trace callers.
23227 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
23228 INCOMING_FRAME_SP_OFFSET in the CIE.
23229 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
23230 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
23232 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
23233 * doc/tm.texi: Regenerated.
23235 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
23237 PR rtl-optimization/83147
23238 * lra-constraints.c (remove_inheritance_pseudos): Use
23239 lra_substitute_pseudo_within_insn.
23241 2018-01-19 Tom de Vries <tom@codesourcery.com>
23242 Cesar Philippidis <cesar@codesourcery.com>
23245 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
23247 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
23250 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
23251 spaces for function labels.
23253 2018-01-19 Martin Liska <mliska@suse.cz>
23255 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
23256 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
23257 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
23258 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
23259 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
23260 (PRED_CONST_RETURN): Change from 69 to 65.
23261 (PRED_NULL_RETURN): Change from 91 to 71.
23262 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
23263 (PRED_LOOP_GUARD): Change from 66 to 73.
23265 2018-01-19 Martin Liska <mliska@suse.cz>
23267 * predict.c (predict_insn_def): Add new assert.
23268 (struct branch_predictor): Change type to signed integer.
23269 (test_prediction_value_range): Amend test to cover
23270 PROB_UNINITIALIZED.
23271 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
23272 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
23273 (PRED_LOOP_ITERATIONS_MAX): Likewise.
23274 (PRED_LOOP_IV_COMPARE): Likewise.
23275 * predict.h (PROB_UNINITIALIZED): Define new constant.
23277 2018-01-19 Martin Liska <mliska@suse.cz>
23279 * predict.c (dump_prediction): Add new format for
23280 analyze_brprob.py script which is enabled with -details
23282 * profile-count.h (precise_p): New function.
23284 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23286 PR tree-optimization/83922
23287 * tree-vect-loop.c (vect_verify_full_masking): Return false if
23288 there are no statements that need masking.
23289 (vect_active_double_reduction_p): New function.
23290 (vect_analyze_loop_operations): Use it when handling phis that
23291 are not in the loop header.
23293 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
23295 PR tree-optimization/83914
23296 * tree-vect-loop.c (vectorizable_induction): Don't convert
23297 init_expr or apply the peeling adjustment for inductions
23298 that are nested within the vectorized loop.
23300 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23302 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
23305 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23309 * function.h (gimplify_parameters): Add gimple_seq * argument.
23310 * function.c: Include gimple.h and options.h.
23311 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
23312 for the added local temporaries if needed.
23313 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
23314 if there are any parameter cleanups, wrap whole body into a
23315 try/finally with the cleanups.
23317 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
23320 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
23321 Use GET_MODE_CLASS for scalar floating point.
23323 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23327 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
23328 Fix call of call_cgraph_insertion_hooks.
23330 2018-01-18 Martin Sebor <msebor@redhat.com>
23332 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
23334 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
23337 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
23340 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
23343 * common.opt: (-ffile-prefix-map): New option.
23344 * opts.c (common_handle_option): Defer it.
23345 * opts-global.c (handle_common_deferred_options): Handle it.
23346 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
23347 * file-prefix-map.h: New file.
23348 (remap_debug_filename, add_debug_prefix_map): ...here.
23349 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
23350 * final.c (debug_prefix_map, add_debug_prefix_map
23351 remap_debug_filename): Move to...
23352 * file-prefix-map.c: New file.
23353 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
23354 generalize, get rid of alloca(), use strrchr() instead of strchr().
23355 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
23356 Implement in terms of add_prefix_map().
23357 (remap_macro_filename, remap_debug_filename): Implement in term of
23359 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
23360 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
23361 * dbxout.c: Include file-prefix-map.h.
23362 * varasm.c: Likewise.
23363 * vmsdbgout.c: Likewise.
23364 * xcoffout.c: Likewise.
23365 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
23366 * doc/cppopts.texi (-fmacro-prefix-map): Document.
23367 * doc/invoke.texi (-ffile-prefix-map): Document.
23368 (-fdebug-prefix-map): Update description.
23370 2018-01-18 Martin Liska <mliska@suse.cz>
23372 * config/i386/i386.c (indirect_thunk_name): Document that also
23374 (output_indirect_thunk): Document why both instructions
23375 (pause and lfence) are generated.
23377 2018-01-18 Richard Biener <rguenther@suse.de>
23379 PR tree-optimization/83887
23380 * graphite-scop-detection.c
23381 (scop_detection::get_nearest_dom_with_single_entry): Remove.
23382 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
23383 (scop_detection::merge_sese): Re-implement with a flood-fill
23384 algorithm that properly finds a SESE region if it exists.
23386 2018-01-18 Jakub Jelinek <jakub@redhat.com>
23389 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
23390 pointer_diff optimizations use view_convert instead of convert.
23392 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23394 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
23395 Generate different code for -mno-speculate-indirect-jumps.
23396 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
23397 (*call_indirect_aix<mode>): Disable for
23398 -mno-speculate-indirect-jumps.
23399 (*call_indirect_aix<mode>_nospec): New define_insn.
23400 (*call_value_indirect_aix<mode>): Disable for
23401 -mno-speculate-indirect-jumps.
23402 (*call_value_indirect_aix<mode>_nospec): New define_insn.
23403 (*sibcall_nonlocal_sysv<mode>): Generate different code for
23404 -mno-speculate-indirect-jumps.
23405 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
23407 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
23409 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
23410 long double type, set the flags for noting the default long double
23411 type, even if we don't pass or return a long double type.
23413 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
23416 * ipa-inline.c (flatten_function): Do not overwrite final inlining
23419 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
23421 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
23422 support for merge[hl].
23423 (fold_mergehl_helper): New helper function.
23424 (tree-vector-builder.h): New #include for tree_vector_builder usage.
23425 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
23426 (altivec_vmrglw_direct): Add xxmrglw insn.
23428 2018-01-17 Andrew Waterman <andrew@sifive.com>
23430 * config/riscv/riscv.c (riscv_conditional_register_usage): If
23431 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
23433 2018-01-17 David Malcolm <dmalcolm@redhat.com>
23436 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
23437 call the lto_location_cache before reading the
23438 DECL_SOURCE_LOCATION of the types.
23440 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
23441 Richard Sandiford <richard.sandiford@linaro.org>
23443 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
23444 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
23445 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
23446 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
23447 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
23449 * config/aarch64/constraints.md (aarch64_movti_operand):
23451 * config/aarch64/predicates.md (Uti): Add new constraint.
23453 2018-01-17 Carl Love <cel@us.ibm.com>
23455 * config/rs6000/vsx.md (define_expand xl_len_r,
23456 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
23457 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
23459 (define_expand, define_insn): Move the shift left from the
23460 define_insn to the define_expand for lxvl and stxvl instructions.
23461 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
23462 and XL_LEN_R definitions to PURE.
23464 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
23466 * config/i386/i386.c (indirect_thunk_name): Declare regno
23467 as unsigned int. Compare regno with INVALID_REGNUM.
23468 (output_indirect_thunk): Ditto.
23469 (output_indirect_thunk_function): Ditto.
23470 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
23471 in the call to output_indirect_thunk_function.
23473 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
23475 PR middle-end/83884
23476 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
23477 rather than the size of inner_type to determine the stack slot size
23478 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
23480 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
23483 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
23486 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
23488 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
23489 endian Linux systems to optionally enable multilibs for selecting
23490 the long double type if the user configured an explicit type.
23491 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
23492 have no long double multilibs if not defined.
23493 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
23494 warn if the user used -mabi={ieee,ibm}longdouble and we built
23495 multilibs for long double.
23496 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
23497 appropriate multilib option.
23498 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
23500 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
23501 for building long double multilibs.
23502 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
23504 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
23506 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
23509 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
23511 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
23514 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
23517 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
23520 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23522 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
23523 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
23525 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
23527 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
23528 different rtl trees depending on TARGET_64BIT.
23529 (rs6000_gen_lvx): Likewise.
23531 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
23533 * config/visium/visium.md (nop): Tweak comment.
23534 (hazard_nop): Likewise.
23536 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23538 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
23539 -mspeculate-indirect-jumps.
23540 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
23541 for -mno-speculate-indirect-jumps.
23542 (*call_indirect_elfv2<mode>_nospec): New define_insn.
23543 (*call_value_indirect_elfv2<mode>): Disable for
23544 -mno-speculate-indirect-jumps.
23545 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
23546 (indirect_jump): Emit different RTL for
23547 -mno-speculate-indirect-jumps.
23548 (*indirect_jump<mode>): Disable for
23549 -mno-speculate-indirect-jumps.
23550 (*indirect_jump<mode>_nospec): New define_insn.
23551 (tablejump): Emit different RTL for
23552 -mno-speculate-indirect-jumps.
23553 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
23554 (tablejumpsi_nospec): New define_expand.
23555 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
23556 (tablejumpdi_nospec): New define_expand.
23557 (*tablejump<mode>_internal1): Disable for
23558 -mno-speculate-indirect-jumps.
23559 (*tablejump<mode>_internal1_nospec): New define_insn.
23560 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
23563 2018-01-16 Artyom Skrobov tyomitch@gmail.com
23565 * caller-save.c (insert_save): Drop unnecessary parameter. All
23568 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23569 Richard Biener <rguenth@suse.de>
23572 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
23573 return early, inline manually is_gimple_sizepos. Make sure if we
23574 call gimplify_expr we don't end up with a gimple constant.
23575 * tree.c (variably_modified_type_p): Don't return true for
23576 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
23577 * gimplify.h (is_gimple_sizepos): Remove.
23579 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23581 PR tree-optimization/83857
23582 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
23583 vectorizable_live_operation for pure SLP statements.
23584 (vectorizable_live_operation): Handle PHIs.
23586 2018-01-16 Richard Biener <rguenther@suse.de>
23588 PR tree-optimization/83867
23589 * tree-vect-stmts.c (vect_transform_stmt): Precompute
23590 nested_in_vect_loop_p since the scalar stmt may get invalidated.
23592 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23595 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
23596 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
23597 If off is not INTEGER_CST, issue a may not be aligned warning
23598 rather than isn't aligned. Use isn%'t rather than isn't.
23599 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
23601 <case MULT_EXPR>: Improve the case when bottom and one of the
23602 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
23603 operand, in that case check if the other operand is multiple of
23604 bottom divided by the INTEGER_CST operand.
23606 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23609 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
23610 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
23611 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
23612 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
23613 * config/pa/pa.c (pa_function_arg_advance): Likewise.
23614 (pa_function_arg, pa_arg_partial_bytes): Likewise.
23615 (pa_function_arg_size): New function.
23617 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23619 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
23620 in a separate statement.
23622 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
23624 PR tree-optimization/83847
23625 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
23626 group gathers and scatters.
23628 2018-01-16 Jakub Jelinek <jakub@redhat.com>
23630 PR rtl-optimization/86620
23631 * params.def (max-sched-ready-insns): Bump minimum value to 1.
23633 PR rtl-optimization/83213
23634 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
23635 to last if both are JUMP_INSNs.
23637 PR tree-optimization/83843
23638 * gimple-ssa-store-merging.c
23639 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
23640 store_immediate_info for bswap/nop orig_stores.
23642 2018-01-15 Andrew Waterman <andrew@sifive.com>
23644 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
23646 <UDIV>: Increase cost if !TARGET_DIV.
23648 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
23650 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
23651 (define_attr "cr_logical_3op"): New.
23652 (cceq_ior_compare): Adjust.
23653 (cceq_ior_compare_complement): Adjust.
23654 (*cceq_rev_compare): Adjust.
23655 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
23656 (is_cracked_insn): Adjust.
23657 (insn_must_be_first_in_group): Adjust.
23658 * config/rs6000/40x.md: Adjust.
23659 * config/rs6000/440.md: Adjust.
23660 * config/rs6000/476.md: Adjust.
23661 * config/rs6000/601.md: Adjust.
23662 * config/rs6000/603.md: Adjust.
23663 * config/rs6000/6xx.md: Adjust.
23664 * config/rs6000/7450.md: Adjust.
23665 * config/rs6000/7xx.md: Adjust.
23666 * config/rs6000/8540.md: Adjust.
23667 * config/rs6000/cell.md: Adjust.
23668 * config/rs6000/e300c2c3.md: Adjust.
23669 * config/rs6000/e500mc.md: Adjust.
23670 * config/rs6000/e500mc64.md: Adjust.
23671 * config/rs6000/e5500.md: Adjust.
23672 * config/rs6000/e6500.md: Adjust.
23673 * config/rs6000/mpc.md: Adjust.
23674 * config/rs6000/power4.md: Adjust.
23675 * config/rs6000/power5.md: Adjust.
23676 * config/rs6000/power6.md: Adjust.
23677 * config/rs6000/power7.md: Adjust.
23678 * config/rs6000/power8.md: Adjust.
23679 * config/rs6000/power9.md: Adjust.
23680 * config/rs6000/rs64.md: Adjust.
23681 * config/rs6000/titan.md: Adjust.
23683 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23685 * config/i386/predicates.md (indirect_branch_operand): Rewrite
23686 ix86_indirect_branch_register logic.
23688 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23690 * config/i386/constraints.md (Bs): Update
23691 ix86_indirect_branch_register check. Don't check
23692 ix86_indirect_branch_register with GOT_memory_operand.
23694 * config/i386/predicates.md (GOT_memory_operand): Don't check
23695 ix86_indirect_branch_register here.
23696 (GOT32_symbol_operand): Likewise.
23698 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23700 * config/i386/predicates.md (constant_call_address_operand):
23701 Rewrite ix86_indirect_branch_register logic.
23702 (sibcall_insn_operand): Likewise.
23704 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23706 * config/i386/constraints.md (Bs): Replace
23707 ix86_indirect_branch_thunk_register with
23708 ix86_indirect_branch_register.
23710 * config/i386/i386.md (indirect_jump): Likewise.
23711 (tablejump): Likewise.
23712 (*sibcall_memory): Likewise.
23713 (*sibcall_value_memory): Likewise.
23714 Peepholes of indirect call and jump via memory: Likewise.
23715 * config/i386/i386.opt: Likewise.
23716 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23717 (GOT_memory_operand): Likewise.
23718 (call_insn_operand): Likewise.
23719 (sibcall_insn_operand): Likewise.
23720 (GOT32_symbol_operand): Likewise.
23722 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23724 PR middle-end/83837
23725 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
23726 type rather than type addr's type points to.
23727 (expand_omp_atomic_mutex): Likewise.
23728 (expand_omp_atomic): Likewise.
23730 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
23733 * config/i386/i386.c (output_indirect_thunk_function): Use
23734 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
23735 for __x86_return_thunk.
23737 2018-01-15 Richard Biener <rguenther@suse.de>
23739 PR middle-end/83850
23740 * expmed.c (extract_bit_field_1): Fix typo.
23742 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23745 * config/arm/iterators.md (VF): New mode iterator.
23746 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
23747 Remove integer-related logic from pattern.
23748 (neon_vabd<mode>_3): Likewise.
23750 2018-01-15 Jakub Jelinek <jakub@redhat.com>
23752 PR middle-end/82694
23753 * common.opt (fstrict-overflow): No longer an alias.
23754 (fwrapv-pointer): New option.
23755 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
23756 also for pointer types based on flag_wrapv_pointer.
23757 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
23758 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
23759 opts->x_flag_wrapv got set.
23760 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
23761 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
23762 POINTER_TYPE_OVERFLOW_UNDEFINED.
23763 * match.pd: Likewise in address comparison pattern.
23764 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
23766 2018-01-15 Richard Biener <rguenther@suse.de>
23769 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
23770 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
23771 Reset type names to their identifier if their TYPE_DECL doesn't
23772 have linkage (and thus is used for ODR and devirt).
23773 (save_debug_info_for_decl): Remove.
23774 (save_debug_info_for_type): Likewise.
23775 (add_tree_to_fld_list): Adjust.
23776 * tree-pretty-print.c (dump_generic_node): Make dumping of
23777 type names more robust.
23779 2018-01-15 Richard Biener <rguenther@suse.de>
23781 * BASE-VER: Bump to 8.0.1.
23783 2018-01-14 Martin Sebor <msebor@redhat.com>
23786 * builtins.c (check_access): Avoid warning when the no-warning bit
23789 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
23791 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
23792 * ira-color (allocno_hard_regs_compare): Likewise.
23794 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
23797 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
23798 Use .pushsection/.popsection.
23800 2018-01-14 Martin Sebor <msebor@redhat.com>
23803 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
23805 2018-01-14 Jakub Jelinek <jakub@redhat.com>
23807 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
23808 entry from extra_headers.
23809 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
23810 extra_headers, make the list bitwise identical to the i?86-*-* one.
23812 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23814 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
23815 -mcmodel=large with -mindirect-branch=thunk,
23816 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
23817 -mfunction-return=thunk-extern.
23818 * doc/invoke.texi: Document -mcmodel=large is incompatible with
23819 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
23820 -mfunction-return=thunk and -mfunction-return=thunk-extern.
23822 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23824 * config/i386/i386.c (print_reg): Print the name of the full
23825 integer register without '%'.
23826 (ix86_print_operand): Handle 'V'.
23827 * doc/extend.texi: Document 'V' modifier.
23829 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23831 * config/i386/constraints.md (Bs): Disallow memory operand for
23832 -mindirect-branch-register.
23834 * config/i386/predicates.md (indirect_branch_operand): Likewise.
23835 (GOT_memory_operand): Likewise.
23836 (call_insn_operand): Likewise.
23837 (sibcall_insn_operand): Likewise.
23838 (GOT32_symbol_operand): Likewise.
23839 * config/i386/i386.md (indirect_jump): Call convert_memory_address
23840 for -mindirect-branch-register.
23841 (tablejump): Likewise.
23842 (*sibcall_memory): Likewise.
23843 (*sibcall_value_memory): Likewise.
23844 Disallow peepholes of indirect call and jump via memory for
23845 -mindirect-branch-register.
23846 (*call_pop): Replace m with Bw.
23847 (*call_value_pop): Likewise.
23848 (*sibcall_pop_memory): Replace m with Bs.
23849 * config/i386/i386.opt (mindirect-branch-register): New option.
23850 * doc/invoke.texi: Document -mindirect-branch-register option.
23852 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23854 * config/i386/i386-protos.h (ix86_output_function_return): New.
23855 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
23856 set function_return_type.
23857 (indirect_thunk_name): Add ret_p to indicate thunk for function
23859 (output_indirect_thunk_function): Pass false to
23860 indirect_thunk_name.
23861 (ix86_output_indirect_branch_via_reg): Likewise.
23862 (ix86_output_indirect_branch_via_push): Likewise.
23863 (output_indirect_thunk_function): Create alias for function
23864 return thunk if regno < 0.
23865 (ix86_output_function_return): New function.
23866 (ix86_handle_fndecl_attribute): Handle function_return.
23867 (ix86_attribute_table): Add function_return.
23868 * config/i386/i386.h (machine_function): Add
23869 function_return_type.
23870 * config/i386/i386.md (simple_return_internal): Use
23871 ix86_output_function_return.
23872 (simple_return_internal_long): Likewise.
23873 * config/i386/i386.opt (mfunction-return=): New option.
23874 (indirect_branch): Mention -mfunction-return=.
23875 * doc/extend.texi: Document function_return function attribute.
23876 * doc/invoke.texi: Document -mfunction-return= option.
23878 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
23880 * config/i386/i386-opts.h (indirect_branch): New.
23881 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
23882 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
23883 with local indirect jump when converting indirect call and jump.
23884 (ix86_set_indirect_branch_type): New.
23885 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
23886 (indirectlabelno): New.
23887 (indirect_thunk_needed): Likewise.
23888 (indirect_thunk_bnd_needed): Likewise.
23889 (indirect_thunks_used): Likewise.
23890 (indirect_thunks_bnd_used): Likewise.
23891 (INDIRECT_LABEL): Likewise.
23892 (indirect_thunk_name): Likewise.
23893 (output_indirect_thunk): Likewise.
23894 (output_indirect_thunk_function): Likewise.
23895 (ix86_output_indirect_branch_via_reg): Likewise.
23896 (ix86_output_indirect_branch_via_push): Likewise.
23897 (ix86_output_indirect_branch): Likewise.
23898 (ix86_output_indirect_jmp): Likewise.
23899 (ix86_code_end): Call output_indirect_thunk_function if needed.
23900 (ix86_output_call_insn): Call ix86_output_indirect_branch if
23902 (ix86_handle_fndecl_attribute): Handle indirect_branch.
23903 (ix86_attribute_table): Add indirect_branch.
23904 * config/i386/i386.h (machine_function): Add indirect_branch_type
23905 and has_local_indirect_jump.
23906 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
23908 (tablejump): Likewise.
23909 (*indirect_jump): Use ix86_output_indirect_jmp.
23910 (*tablejump_1): Likewise.
23911 (simple_return_indirect_internal): Likewise.
23912 * config/i386/i386.opt (mindirect-branch=): New option.
23913 (indirect_branch): New.
23916 (thunk-inline): Likewise.
23917 (thunk-extern): Likewise.
23918 * doc/extend.texi: Document indirect_branch function attribute.
23919 * doc/invoke.texi: Document -mindirect-branch= option.
23921 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
23924 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
23926 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
23928 * ipa-inline.c (want_inline_small_function_p): Return false if
23929 inlining has already failed with CIF_FINAL_ERROR.
23930 (update_caller_keys): Call want_inline_small_function_p before
23932 (update_callee_keys): Likewise.
23934 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
23936 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
23938 (rs6000_quadword_masked_address_p): Likewise.
23939 (quad_aligned_load_p): Likewise.
23940 (quad_aligned_store_p): Likewise.
23941 (const_load_sequence_p): Add comment to describe the outer-most loop.
23942 (mimic_memory_attributes_and_flags): New function.
23943 (rs6000_gen_stvx): Likewise.
23944 (replace_swapped_aligned_store): Likewise.
23945 (rs6000_gen_lvx): Likewise.
23946 (replace_swapped_aligned_load): Likewise.
23947 (replace_swapped_load_constant): Capitalize argument name in
23948 comment describing this function.
23949 (rs6000_analyze_swaps): Add a third pass to search for vector loads
23950 and stores that access quad-word aligned addresses and replace
23951 with stvx or lvx instructions when appropriate.
23952 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
23953 New function prototype.
23954 (rs6000_quadword_masked_address_p): Likewise.
23955 (rs6000_gen_lvx): Likewise.
23956 (rs6000_gen_stvx): Likewise.
23957 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
23958 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
23959 when memory address is aligned.
23960 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
23961 this split to select lvx instruction when memory address is aligned.
23962 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
23963 instruction when memory address is aligned.
23964 (*vsx_le_perm_load_v16qi): Likewise.
23965 (four unnamed splitters): Modify to select the stvx instruction
23966 when memory is aligned.
23968 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
23970 * predict.c (determine_unlikely_bbs): Handle correctly BBs
23971 which appears in the queue multiple times.
23973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23974 Alan Hayward <alan.hayward@arm.com>
23975 David Sherwood <david.sherwood@arm.com>
23977 * tree-vectorizer.h (vec_lower_bound): New structure.
23978 (_loop_vec_info): Add check_nonzero and lower_bounds.
23979 (LOOP_VINFO_CHECK_NONZERO): New macro.
23980 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23981 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23982 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23983 fields. Make seg_len the distance travelled, not including the
23985 (dr_direction_indicator): Declare.
23986 (dr_zero_step_indicator): Likewise.
23987 (dr_known_forward_stride_p): Likewise.
23988 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23990 (runtime_alias_check_p): Allow runtime alias checks with
23992 (operator ==): Compare access_size and align.
23993 (prune_runtime_alias_test_list): Rework for new distinction between
23994 the access_size and seg_len.
23995 (create_intersect_range_checks_index): Likewise. Cope with polynomial
23997 (get_segment_min_max): New function.
23998 (create_intersect_range_checks): Use it.
23999 (dr_step_indicator): New function.
24000 (dr_direction_indicator): Likewise.
24001 (dr_zero_step_indicator): Likewise.
24002 (dr_known_forward_stride_p): Likewise.
24003 * tree-loop-distribution.c (data_ref_segment_size): Return
24004 DR_STEP * (niters - 1).
24005 (compute_alias_check_pairs): Update call to the dr_with_seg_len
24007 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
24008 (vect_preserves_scalar_order_p): New function, split out from...
24009 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
24010 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
24011 (vect_vfa_access_size): New function.
24012 (vect_vfa_align): Likewise.
24013 (vect_compile_time_alias): Take access_size_a and access_b arguments.
24014 (dump_lower_bound): New function.
24015 (vect_check_lower_bound): Likewise.
24016 (vect_small_gap_p): Likewise.
24017 (vectorizable_with_step_bound_p): Likewise.
24018 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
24019 depencies if the vectorization factor is 1. Convert the checks
24020 for nonzero steps into checks on the bounds of DR_STEP. Try using
24021 a bunds check for variable steps if the minimum required step is
24022 relatively small. Update calls to the dr_with_seg_len
24023 constructor and to vect_compile_time_alias.
24024 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
24026 (vect_loop_versioning): Call it.
24027 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
24029 (vect_estimate_min_profitable_iters): Account for any bounds checks.
24031 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24032 Alan Hayward <alan.hayward@arm.com>
24033 David Sherwood <david.sherwood@arm.com>
24035 * doc/sourcebuild.texi (vect_scatter_store): Document.
24036 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
24038 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
24040 * genopinit.c (main): Add supports_vec_scatter_store and
24041 supports_vec_scatter_store_cached to target_optabs.
24042 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
24043 IFN_MASK_SCATTER_STORE.
24044 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
24046 * internal-fn.h (internal_store_fn_p): Declare.
24047 (internal_fn_stored_value_index): Likewise.
24048 * internal-fn.c (scatter_store_direct): New macro.
24049 (expand_scatter_store_optab_fn): New function.
24050 (direct_scatter_store_optab_supported_p): New macro.
24051 (internal_store_fn_p): New function.
24052 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
24053 IFN_MASK_SCATTER_STORE.
24054 (internal_fn_mask_index): Likewise.
24055 (internal_fn_stored_value_index): New function.
24056 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
24057 for scatter stores.
24058 * optabs-query.h (supports_vec_scatter_store_p): Declare.
24059 * optabs-query.c (supports_vec_scatter_store_p): New function.
24060 * tree-vectorizer.h (vect_get_store_rhs): Declare.
24061 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
24062 true for scatter stores.
24063 (vect_gather_scatter_fn_p): Handle scatter stores too.
24064 (vect_check_gather_scatter): Consider using scatter stores if
24065 supports_vec_scatter_store_p.
24066 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
24067 scatter stores too.
24068 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24069 internal_fn_stored_value_index.
24070 (check_load_store_masking): Handle scatter stores too.
24071 (vect_get_store_rhs): Make public.
24072 (vectorizable_call): Use internal_store_fn_p.
24073 (vectorizable_store): Handle scatter store internal functions.
24074 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
24075 when deciding whether the end of the group has been reached.
24076 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
24077 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
24078 (mask_scatter_store<mode>): New insns.
24080 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24081 Alan Hayward <alan.hayward@arm.com>
24082 David Sherwood <david.sherwood@arm.com>
24084 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
24085 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
24086 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
24088 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
24089 Use vect_truncate_gather_scatter_offset if we can't treat the
24090 operation as a normal gather load or scatter store.
24091 (get_group_load_store_type): Take the gather_scatter_info
24092 as argument. Try using a gather load or scatter store for
24093 single-element groups.
24094 (get_load_store_type): Update calls to get_group_load_store_type
24095 and vect_use_strided_gather_scatters_p.
24097 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24098 Alan Hayward <alan.hayward@arm.com>
24099 David Sherwood <david.sherwood@arm.com>
24101 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
24102 optional tree argument.
24103 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
24105 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
24106 but continue to use the current value as a fallback.
24107 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
24108 to compare the updates.
24109 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
24110 (get_load_store_type): Use it when handling a strided access.
24111 (vect_get_strided_load_store_ops): New function.
24112 (vect_get_data_ptr_increment): Likewise.
24113 (vectorizable_load): Handle strided gather loads. Always pass
24114 a step to vect_create_data_ref_ptr and bump_vector_ptr.
24116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24117 Alan Hayward <alan.hayward@arm.com>
24118 David Sherwood <david.sherwood@arm.com>
24120 * doc/md.texi (gather_load@var{m}): Document.
24121 (mask_gather_load@var{m}): Likewise.
24122 * genopinit.c (main): Add supports_vec_gather_load and
24123 supports_vec_gather_load_cached to target_optabs.
24124 * optabs-tree.c (init_tree_optimization_optabs): Use
24125 ggc_cleared_alloc to allocate target_optabs.
24126 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
24127 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
24129 * internal-fn.h (internal_load_fn_p): Declare.
24130 (internal_gather_scatter_fn_p): Likewise.
24131 (internal_fn_mask_index): Likewise.
24132 (internal_gather_scatter_fn_supported_p): Likewise.
24133 * internal-fn.c (gather_load_direct): New macro.
24134 (expand_gather_load_optab_fn): New function.
24135 (direct_gather_load_optab_supported_p): New macro.
24136 (direct_internal_fn_optab): New function.
24137 (internal_load_fn_p): Likewise.
24138 (internal_gather_scatter_fn_p): Likewise.
24139 (internal_fn_mask_index): Likewise.
24140 (internal_gather_scatter_fn_supported_p): Likewise.
24141 * optabs-query.c (supports_at_least_one_mode_p): New function.
24142 (supports_vec_gather_load_p): Likewise.
24143 * optabs-query.h (supports_vec_gather_load_p): Declare.
24144 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
24145 and memory_type field.
24146 (NUM_PATTERNS): Bump to 15.
24147 * tree-vect-data-refs.c: Include internal-fn.h.
24148 (vect_gather_scatter_fn_p): New function.
24149 (vect_describe_gather_scatter_call): Likewise.
24150 (vect_check_gather_scatter): Try using internal functions for
24151 gather loads. Recognize existing calls to a gather load function.
24152 (vect_analyze_data_refs): Consider using gather loads if
24153 supports_vec_gather_load_p.
24154 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
24155 (vect_get_gather_scatter_offset_type): Likewise.
24156 (vect_convert_mask_for_vectype): Likewise.
24157 (vect_add_conversion_to_patterm): Likewise.
24158 (vect_try_gather_scatter_pattern): Likewise.
24159 (vect_recog_gather_scatter_pattern): New pattern recognizer.
24160 (vect_vect_recog_func_ptrs): Add it.
24161 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
24162 internal_fn_mask_index and internal_gather_scatter_fn_p.
24163 (check_load_store_masking): Take the gather_scatter_info as an
24164 argument and handle gather loads.
24165 (vect_get_gather_scatter_ops): New function.
24166 (vectorizable_call): Check internal_load_fn_p.
24167 (vectorizable_load): Likewise. Handle gather load internal
24169 (vectorizable_store): Update call to check_load_store_masking.
24170 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
24171 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
24172 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
24173 (aarch64_gather_scale_operand_d): New predicates.
24174 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
24175 (mask_gather_load<mode>): New insns.
24177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24178 Alan Hayward <alan.hayward@arm.com>
24179 David Sherwood <david.sherwood@arm.com>
24181 * optabs.def (fold_left_plus_optab): New optab.
24182 * doc/md.texi (fold_left_plus_@var{m}): Document.
24183 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
24184 * internal-fn.c (fold_left_direct): Define.
24185 (expand_fold_left_optab_fn): Likewise.
24186 (direct_fold_left_optab_supported_p): Likewise.
24187 * fold-const-call.c (fold_const_fold_left): New function.
24188 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
24189 * tree-parloops.c (valid_reduction_p): New function.
24190 (gather_scalar_reductions): Use it.
24191 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
24192 (vect_finish_replace_stmt): Declare.
24193 * tree-vect-loop.c (fold_left_reduction_fn): New function.
24194 (needs_fold_left_reduction_p): New function, split out from...
24195 (vect_is_simple_reduction): ...here. Accept reductions that
24196 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
24197 (vect_force_simple_reduction): Also store the reduction type in
24198 the assignment's STMT_VINFO_REDUC_TYPE.
24199 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
24200 (merge_with_identity): New function.
24201 (vect_expand_fold_left): Likewise.
24202 (vectorize_fold_left_reduction): Likewise.
24203 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
24204 scalar phi in place for it. Check for target support and reject
24205 cases that would reassociate the operation. Defer the transform
24206 phase to vectorize_fold_left_reduction.
24207 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
24208 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
24209 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
24211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24213 * tree-if-conv.c (predicate_mem_writes): Remove redundant
24214 call to ifc_temp_var.
24216 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24217 Alan Hayward <alan.hayward@arm.com>
24218 David Sherwood <david.sherwood@arm.com>
24220 * target.def (legitimize_address_displacement): Take the original
24221 offset as a poly_int.
24222 * targhooks.h (default_legitimize_address_displacement): Update
24224 * targhooks.c (default_legitimize_address_displacement): Likewise.
24225 * doc/tm.texi: Regenerate.
24226 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
24227 as an argument, moving assert of ad->disp == ad->disp_term to...
24228 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
24229 Try calling targetm.legitimize_address_displacement before expanding
24230 the address rather than afterwards, and adjust for the new interface.
24231 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
24232 Match the new hook interface. Handle SVE addresses.
24233 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
24234 new hook interface.
24236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24238 * Makefile.in (OBJS): Add early-remat.o.
24239 * target.def (select_early_remat_modes): New hook.
24240 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
24241 * doc/tm.texi: Regenerate.
24242 * targhooks.h (default_select_early_remat_modes): Declare.
24243 * targhooks.c (default_select_early_remat_modes): New function.
24244 * timevar.def (TV_EARLY_REMAT): New timevar.
24245 * passes.def (pass_early_remat): New pass.
24246 * tree-pass.h (make_pass_early_remat): Declare.
24247 * early-remat.c: New file.
24248 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
24250 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
24252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24253 Alan Hayward <alan.hayward@arm.com>
24254 David Sherwood <david.sherwood@arm.com>
24256 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
24257 vfm1 with a bound_epilog parameter.
24258 (vect_do_peeling): Update calls accordingly, and move the prologue
24259 call earlier in the function. Treat the base bound_epilog as 0 for
24260 fully-masked loops and retain vf - 1 for other loops. Add 1 to
24261 this base when peeling for gaps.
24262 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
24263 with fully-masked loops.
24264 (vect_estimate_min_profitable_iters): Handle the single peeled
24265 iteration in that case.
24267 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24268 Alan Hayward <alan.hayward@arm.com>
24269 David Sherwood <david.sherwood@arm.com>
24271 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
24272 single-element interleaving even if the size is not a power of 2.
24273 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
24274 accesses for single-element interleaving if the group size is
24277 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24278 Alan Hayward <alan.hayward@arm.com>
24279 David Sherwood <david.sherwood@arm.com>
24281 * doc/md.texi (fold_extract_last_@var{m}): Document.
24282 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
24283 * optabs.def (fold_extract_last_optab): New optab.
24284 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
24285 * internal-fn.c (fold_extract_direct): New macro.
24286 (expand_fold_extract_optab_fn): Likewise.
24287 (direct_fold_extract_optab_supported_p): Likewise.
24288 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
24289 * tree-vect-loop.c (vect_model_reduction_cost): Handle
24290 EXTRACT_LAST_REDUCTION.
24291 (get_initial_def_for_reduction): Do not create an initial vector
24292 for EXTRACT_LAST_REDUCTION reductions.
24293 (vectorizable_reduction): Leave the scalar phi in place for
24294 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
24295 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
24296 epilogue code for EXTRACT_LAST_REDUCTION and defer the
24297 transform phase to vectorizable_condition.
24298 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
24300 (vect_finish_stmt_generation): ...here.
24301 (vect_finish_replace_stmt): New function.
24302 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
24303 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
24305 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
24307 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24308 Alan Hayward <alan.hayward@arm.com>
24309 David Sherwood <david.sherwood@arm.com>
24311 * doc/md.texi (extract_last_@var{m}): Document.
24312 * optabs.def (extract_last_optab): New optab.
24313 * internal-fn.def (EXTRACT_LAST): New internal function.
24314 * internal-fn.c (cond_unary_direct): New macro.
24315 (expand_cond_unary_optab_fn): Likewise.
24316 (direct_cond_unary_optab_supported_p): Likewise.
24317 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
24318 loops using EXTRACT_LAST.
24319 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
24320 (extract_last_<mode>): ...this optab.
24321 (vec_extract<mode><Vel>): Update accordingly.
24323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24324 Alan Hayward <alan.hayward@arm.com>
24325 David Sherwood <david.sherwood@arm.com>
24327 * target.def (empty_mask_is_expensive): New hook.
24328 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
24329 * doc/tm.texi: Regenerate.
24330 * targhooks.h (default_empty_mask_is_expensive): Declare.
24331 * targhooks.c (default_empty_mask_is_expensive): New function.
24332 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
24333 if the target says that empty masks are expensive.
24334 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
24336 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
24338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24339 Alan Hayward <alan.hayward@arm.com>
24340 David Sherwood <david.sherwood@arm.com>
24342 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
24343 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
24344 (vect_use_loop_mask_for_alignment_p): New function.
24345 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
24346 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
24347 niters_skip argument. Make sure that the first niters_skip elements
24348 of the first iteration are inactive.
24349 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
24350 Update call to vect_set_loop_masks_directly.
24351 (get_misalign_in_elems): New function, split out from...
24352 (vect_gen_prolog_loop_niters): ...here.
24353 (vect_update_init_of_dr): Take a code argument that specifies whether
24354 the adjustment should be added or subtracted.
24355 (vect_update_init_of_drs): Likewise.
24356 (vect_prepare_for_masked_peels): New function.
24357 (vect_do_peeling): Skip prologue peeling if we're using a mask
24358 instead. Update call to vect_update_inits_of_drs.
24359 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24361 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
24362 alignment. Do not include the number of peeled iterations in
24363 the minimum threshold in that case.
24364 (vectorizable_induction): Adjust the start value down by
24365 LOOP_VINFO_MASK_SKIP_NITERS iterations.
24366 (vect_transform_loop): Call vect_prepare_for_masked_peels.
24367 Take the number of skipped iterations into account when calculating
24369 * tree-vect-stmts.c (vect_gen_while_not): New function.
24371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24372 Alan Hayward <alan.hayward@arm.com>
24373 David Sherwood <david.sherwood@arm.com>
24375 * doc/sourcebuild.texi (vect_fully_masked): Document.
24376 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
24377 default value to 0.
24378 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
24380 (vect_analyze_loop_2): ...here. Don't check the vectorization
24381 factor against the number of loop iterations if the loop is
24384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24385 Alan Hayward <alan.hayward@arm.com>
24386 David Sherwood <david.sherwood@arm.com>
24388 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
24389 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
24390 (dump_groups): Update accordingly.
24391 (iv_use::mem_type): New member variable.
24392 (address_p): New function.
24393 (record_use): Add a mem_type argument and initialize the new
24395 (record_group_use): Add a mem_type argument. Use address_p.
24396 Remove obsolete null checks of base_object. Update call to record_use.
24397 (find_interesting_uses_op): Update call to record_group_use.
24398 (find_interesting_uses_cond): Likewise.
24399 (find_interesting_uses_address): Likewise.
24400 (get_mem_type_for_internal_fn): New function.
24401 (find_address_like_use): Likewise.
24402 (find_interesting_uses_stmt): Try find_address_like_use before
24403 calling find_interesting_uses_op.
24404 (addr_offset_valid_p): Use the iv mem_type field as the type
24405 of the addressed memory.
24406 (add_autoinc_candidates): Likewise.
24407 (get_address_cost): Likewise.
24408 (split_small_address_groups_p): Use address_p.
24409 (split_address_groups): Likewise.
24410 (add_iv_candidate_for_use): Likewise.
24411 (autoinc_possible_for_pair): Likewise.
24412 (rewrite_groups): Likewise.
24413 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
24414 (determine_group_iv_cost): Update after split of USE_ADDRESS.
24415 (get_alias_ptr_type_for_ptr_address): New function.
24416 (rewrite_use_address): Rewrite address uses in calls that were
24417 identified by find_address_like_use.
24419 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24420 Alan Hayward <alan.hayward@arm.com>
24421 David Sherwood <david.sherwood@arm.com>
24423 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
24425 * gimple-expr.h (is_gimple_addressable: Likewise.
24426 * gimple-expr.c (is_gimple_address): Likewise.
24427 * internal-fn.c (expand_call_mem_ref): New function.
24428 (expand_mask_load_optab_fn): Use it.
24429 (expand_mask_store_optab_fn): Likewise.
24431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24432 Alan Hayward <alan.hayward@arm.com>
24433 David Sherwood <david.sherwood@arm.com>
24435 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
24436 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
24437 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
24438 (cond_umax@var{mode}): Document.
24439 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
24440 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
24441 (cond_umin_optab, cond_umax_optab): New optabs.
24442 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
24443 (COND_IOR, COND_XOR): New internal functions.
24444 * internal-fn.h (get_conditional_internal_fn): Declare.
24445 * internal-fn.c (cond_binary_direct): New macro.
24446 (expand_cond_binary_optab_fn): Likewise.
24447 (direct_cond_binary_optab_supported_p): Likewise.
24448 (get_conditional_internal_fn): New function.
24449 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
24450 Cope with reduction statements that are vectorized as calls rather
24452 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
24453 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
24454 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
24455 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
24456 (UNSPEC_COND_EOR): New unspecs.
24457 (optab): Add mappings for them.
24458 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
24459 (sve_int_op, sve_fp_op): New int attributes.
24461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24462 Alan Hayward <alan.hayward@arm.com>
24463 David Sherwood <david.sherwood@arm.com>
24465 * optabs.def (while_ult_optab): New optab.
24466 * doc/md.texi (while_ult@var{m}@var{n}): Document.
24467 * internal-fn.def (WHILE_ULT): New internal function.
24468 * internal-fn.h (direct_internal_fn_supported_p): New override
24469 that takes two types as argument.
24470 * internal-fn.c (while_direct): New macro.
24471 (expand_while_optab_fn): New function.
24472 (convert_optab_supported_p): Likewise.
24473 (direct_while_optab_supported_p): New macro.
24474 * wide-int.h (wi::udiv_ceil): New function.
24475 * tree-vectorizer.h (rgroup_masks): New structure.
24476 (vec_loop_masks): New typedef.
24477 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
24478 and fully_masked_p.
24479 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
24480 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
24481 (vect_max_vf): New function.
24482 (slpeel_make_loop_iterate_ntimes): Delete.
24483 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
24484 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
24485 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
24486 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
24487 internal-fn.h, stor-layout.h and optabs-query.h.
24488 (vect_set_loop_mask): New function.
24489 (add_preheader_seq): Likewise.
24490 (add_header_seq): Likewise.
24491 (interleave_supported_p): Likewise.
24492 (vect_maybe_permute_loop_masks): Likewise.
24493 (vect_set_loop_masks_directly): Likewise.
24494 (vect_set_loop_condition_masked): Likewise.
24495 (vect_set_loop_condition_unmasked): New function, split out from
24496 slpeel_make_loop_iterate_ntimes.
24497 (slpeel_make_loop_iterate_ntimes): Rename to..
24498 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
24499 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
24500 (vect_do_peeling): Update call accordingly.
24501 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
24503 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
24504 mask_compare_type, can_fully_mask_p and fully_masked_p.
24505 (release_vec_loop_masks): New function.
24506 (_loop_vec_info): Use it to free the loop masks.
24507 (can_produce_all_loop_masks_p): New function.
24508 (vect_get_max_nscalars_per_iter): Likewise.
24509 (vect_verify_full_masking): Likewise.
24510 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
24511 retries, and free the mask rgroups before retrying. Check loop-wide
24512 reasons for disallowing fully-masked loops. Make the final decision
24513 about whether use a fully-masked loop or not.
24514 (vect_estimate_min_profitable_iters): Do not assume that peeling
24515 for the number of iterations will be needed for fully-masked loops.
24516 (vectorizable_reduction): Disable fully-masked loops.
24517 (vectorizable_live_operation): Likewise.
24518 (vect_halve_mask_nunits): New function.
24519 (vect_double_mask_nunits): Likewise.
24520 (vect_record_loop_mask): Likewise.
24521 (vect_get_loop_mask): Likewise.
24522 (vect_transform_loop): Handle the case in which the final loop
24523 iteration might handle a partial vector. Call vect_set_loop_condition
24524 instead of slpeel_make_loop_iterate_ntimes.
24525 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
24526 (check_load_store_masking): New function.
24527 (prepare_load_store_mask): Likewise.
24528 (vectorizable_store): Handle fully-masked loops.
24529 (vectorizable_load): Likewise.
24530 (supportable_widening_operation): Use vect_halve_mask_nunits for
24532 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
24533 (vect_gen_while): New function.
24534 * config/aarch64/aarch64.md (umax<mode>3): New expander.
24535 (aarch64_uqdec<mode>): New insn.
24537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24538 Alan Hayward <alan.hayward@arm.com>
24539 David Sherwood <david.sherwood@arm.com>
24541 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
24542 (reduc_xor_scal_optab): New optabs.
24543 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
24544 (reduc_xor_scal_@var{m}): Document.
24545 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
24546 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
24547 internal functions.
24548 * fold-const-call.c (fold_const_call): Handle them.
24549 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
24550 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
24551 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
24552 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
24553 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
24554 (UNSPEC_XORV): New unspecs.
24555 (optab): Add entries for them.
24556 (BITWISEV): New int iterator.
24557 (bit_reduc_op): New int attributes.
24559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24560 Alan Hayward <alan.hayward@arm.com>
24561 David Sherwood <david.sherwood@arm.com>
24563 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
24564 * internal-fn.def (VEC_SHL_INSERT): New internal function.
24565 * optabs.def (vec_shl_insert_optab): New optab.
24566 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
24567 (duplicate_and_interleave): Likewise.
24568 * tree-vect-loop.c: Include internal-fn.h.
24569 (neutral_op_for_slp_reduction): New function, split out from
24570 get_initial_defs_for_reduction.
24571 (get_initial_def_for_reduction): Handle option 2 for variable-length
24572 vectors by loading the neutral value into a vector and then shifting
24573 the initial value into element 0.
24574 (get_initial_defs_for_reduction): Replace the code argument with
24575 the neutral value calculated by neutral_op_for_slp_reduction.
24576 Use gimple_build_vector for constant-length vectors.
24577 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
24578 but the first group_size elements have a neutral value.
24579 Use duplicate_and_interleave otherwise.
24580 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
24581 Update call to get_initial_defs_for_reduction. Handle SLP
24582 reductions for variable-length vectors by creating one vector
24583 result for each scalar result, with the elements associated
24584 with other scalar results stubbed out with the neutral value.
24585 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
24586 Require IFN_VEC_SHL_INSERT for double reductions on
24587 variable-length vectors, or SLP reductions that have
24588 a neutral value. Require can_duplicate_and_interleave_p
24589 support for variable-length unchained SLP reductions if there
24590 is no neutral value, such as for MIN/MAX reductions. Also require
24591 the number of vector elements to be a multiple of the number of
24592 SLP statements when doing variable-length unchained SLP reductions.
24593 Update call to vect_create_epilog_for_reduction.
24594 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
24595 and remove initial values.
24596 (duplicate_and_interleave): Make public.
24597 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
24598 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
24600 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24601 Alan Hayward <alan.hayward@arm.com>
24602 David Sherwood <david.sherwood@arm.com>
24604 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
24605 (can_duplicate_and_interleave_p): New function.
24606 (vect_get_and_check_slp_defs): Take the vector of statements
24607 rather than just the current one. Remove excess parentheses.
24608 Restriction rejectinon of vect_constant_def and vect_external_def
24609 for variable-length vectors to boolean types, or types for which
24610 can_duplicate_and_interleave_p is false.
24611 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
24612 (duplicate_and_interleave): New function.
24613 (vect_get_constant_vectors): Use gimple_build_vector for
24614 constant-length vectors and suitable variable-length constant
24615 vectors. Use duplicate_and_interleave for other variable-length
24616 vectors. Don't defer the update when inserting new statements.
24618 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24619 Alan Hayward <alan.hayward@arm.com>
24620 David Sherwood <david.sherwood@arm.com>
24622 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
24623 min_profitable_iters doesn't go negative.
24625 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24626 Alan Hayward <alan.hayward@arm.com>
24627 David Sherwood <david.sherwood@arm.com>
24629 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
24630 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
24631 * optabs.def (vec_mask_load_lanes_optab): New optab.
24632 (vec_mask_store_lanes_optab): Likewise.
24633 * internal-fn.def (MASK_LOAD_LANES): New internal function.
24634 (MASK_STORE_LANES): Likewise.
24635 * internal-fn.c (mask_load_lanes_direct): New macro.
24636 (mask_store_lanes_direct): Likewise.
24637 (expand_mask_load_optab_fn): Handle masked operations.
24638 (expand_mask_load_lanes_optab_fn): New macro.
24639 (expand_mask_store_optab_fn): Handle masked operations.
24640 (expand_mask_store_lanes_optab_fn): New macro.
24641 (direct_mask_load_lanes_optab_supported_p): Likewise.
24642 (direct_mask_store_lanes_optab_supported_p): Likewise.
24643 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
24645 (vect_load_lanes_supported): Likewise.
24646 * tree-vect-data-refs.c (strip_conversion): New function.
24647 (can_group_stmts_p): Likewise.
24648 (vect_analyze_data_ref_accesses): Use it instead of checking
24649 for a pair of assignments.
24650 (vect_store_lanes_supported): Take a masked_p parameter.
24651 (vect_load_lanes_supported): Likewise.
24652 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
24653 vect_store_lanes_supported and vect_load_lanes_supported.
24654 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
24655 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
24656 parameter. Don't allow gaps for masked accesses.
24657 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
24658 and vect_load_lanes_supported.
24659 (get_load_store_type): Take a masked_p parameter and update
24660 call to get_group_load_store_type.
24661 (vectorizable_store): Update call to get_load_store_type.
24662 Handle IFN_MASK_STORE_LANES.
24663 (vectorizable_load): Update call to get_load_store_type.
24664 Handle IFN_MASK_LOAD_LANES.
24666 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24667 Alan Hayward <alan.hayward@arm.com>
24668 David Sherwood <david.sherwood@arm.com>
24670 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
24672 * config/aarch64/aarch64-protos.h
24673 (aarch64_sve_struct_memory_operand_p): Declare.
24674 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
24675 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
24676 (VPRED, vpred): Handle SVE structure modes.
24677 * config/aarch64/constraints.md (Utx): New constraint.
24678 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
24679 (aarch64_sve_struct_nonimmediate_operand): New predicates.
24680 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
24681 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
24682 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
24683 structure modes. Split into pieces after RA.
24684 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
24685 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
24687 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
24688 SVE structure modes.
24689 (aarch64_classify_address): Likewise.
24690 (sizetochar): Move earlier in file.
24691 (aarch64_print_operand): Handle SVE register lists.
24692 (aarch64_array_mode): New function.
24693 (aarch64_sve_struct_memory_operand_p): Likewise.
24694 (TARGET_ARRAY_MODE): Redefine.
24696 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24697 Alan Hayward <alan.hayward@arm.com>
24698 David Sherwood <david.sherwood@arm.com>
24700 * target.def (array_mode): New target hook.
24701 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
24702 * doc/tm.texi: Regenerate.
24703 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
24704 * hooks.c (hook_optmode_mode_uhwi_none): New function.
24705 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
24706 targetm.array_mode.
24707 * stor-layout.c (mode_for_array): Likewise. Support polynomial
24710 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24711 Alan Hayward <alan.hayward@arm.com>
24712 David Sherwood <david.sherwood@arm.com>
24714 * fold-const.c (fold_binary_loc): Check the argument types
24715 rather than the result type when testing for a vector operation.
24717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24719 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
24720 * doc/tm.texi: Regenerate.
24722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24723 Alan Hayward <alan.hayward@arm.com>
24724 David Sherwood <david.sherwood@arm.com>
24726 * doc/invoke.texi (-msve-vector-bits=): Document new option.
24727 (sve): Document new AArch64 extension.
24728 * doc/md.texi (w): Extend the description of the AArch64
24729 constraint to include SVE vectors.
24730 (Upl, Upa): Document new AArch64 predicate constraints.
24731 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
24733 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
24734 (msve-vector-bits=): New option.
24735 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
24736 SVE when these are disabled.
24737 (sve): New extension.
24738 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
24739 modes. Adjust their number of units based on aarch64_sve_vg.
24740 (MAX_BITSIZE_MODE_ANY_MODE): Define.
24741 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
24742 aarch64_addr_query_type.
24743 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
24744 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
24745 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
24746 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
24747 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
24748 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
24749 (aarch64_simd_imm_zero_p): Delete.
24750 (aarch64_check_zero_based_sve_index_immediate): Declare.
24751 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24752 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24753 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24754 (aarch64_sve_float_mul_immediate_p): Likewise.
24755 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24756 rather than an rtx.
24757 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
24758 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
24759 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
24760 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
24761 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
24762 (aarch64_regmode_natural_size): Likewise.
24763 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
24764 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
24766 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
24767 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
24768 for VG and the SVE predicate registers.
24769 (V_ALIASES): Add a "z"-prefixed alias.
24770 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
24771 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
24772 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
24773 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
24774 (REG_CLASS_NAMES): Add entries for them.
24775 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
24776 and the predicate registers.
24777 (aarch64_sve_vg): Declare.
24778 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
24779 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
24780 (REGMODE_NATURAL_SIZE): Define.
24781 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
24783 * config/aarch64/aarch64.c: Include cfgrtl.h.
24784 (simd_immediate_info): Add a constructor for series vectors,
24785 and an associated step field.
24786 (aarch64_sve_vg): New variable.
24787 (aarch64_dbx_register_number): Handle VG and the predicate registers.
24788 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
24789 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
24790 (VEC_ANY_DATA, VEC_STRUCT): New constants.
24791 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
24792 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
24793 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
24794 (aarch64_get_mask_mode): New functions.
24795 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
24796 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
24797 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
24798 predicate modes and predicate registers. Explicitly restrict
24799 GPRs to modes of 16 bytes or smaller. Only allow FP registers
24800 to store a vector mode if it is recognized by
24801 aarch64_classify_vector_mode.
24802 (aarch64_regmode_natural_size): New function.
24803 (aarch64_hard_regno_caller_save_mode): Return the original mode
24805 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
24806 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
24807 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
24808 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
24810 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
24811 does not overlap dest if the function is frame-related. Handle
24813 (aarch64_split_add_offset): New function.
24814 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
24815 them aarch64_add_offset.
24816 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
24817 and update call to aarch64_sub_sp.
24818 (aarch64_add_cfa_expression): New function.
24819 (aarch64_expand_prologue): Pass extra temporary registers to the
24820 functions above. Handle the case in which we need to emit new
24821 DW_CFA_expressions for registers that were originally saved
24822 relative to the stack pointer, but now have to be expressed
24823 relative to the frame pointer.
24824 (aarch64_output_mi_thunk): Pass extra temporary registers to the
24826 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
24827 IP0 and IP1 values for SVE frames.
24828 (aarch64_expand_vec_series): New function.
24829 (aarch64_expand_sve_widened_duplicate): Likewise.
24830 (aarch64_expand_sve_const_vector): Likewise.
24831 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
24832 Handle SVE constants. Use emit_move_insn to move a force_const_mem
24833 into the register, rather than emitting a SET directly.
24834 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
24835 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
24836 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
24837 (offset_9bit_signed_scaled_p): New functions.
24838 (aarch64_replicate_bitmask_imm): New function.
24839 (aarch64_bitmask_imm): Use it.
24840 (aarch64_cannot_force_const_mem): Reject expressions involving
24841 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
24842 (aarch64_classify_index): Handle SVE indices, by requiring
24843 a plain register index with a scale that matches the element size.
24844 (aarch64_classify_address): Handle SVE addresses. Assert that
24845 the mode of the address is VOIDmode or an integer mode.
24846 Update call to aarch64_classify_symbol.
24847 (aarch64_classify_symbolic_expression): Update call to
24848 aarch64_classify_symbol.
24849 (aarch64_const_vec_all_in_range_p): New function.
24850 (aarch64_print_vector_float_operand): Likewise.
24851 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
24852 "vN" for FP registers with SVE modes. Handle (const ...) vectors
24853 and the FP immediates 1.0 and 0.5.
24854 (aarch64_print_address_internal): Handle SVE addresses.
24855 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
24856 (aarch64_regno_regclass): Handle predicate registers.
24857 (aarch64_secondary_reload): Handle big-endian reloads of SVE
24859 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
24860 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
24861 (aarch64_convert_sve_vector_bits): New function.
24862 (aarch64_override_options): Use it to handle -msve-vector-bits=.
24863 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
24864 rather than an rtx.
24865 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
24866 Handle SVE vector and predicate modes. Accept VL-based constants
24867 that need only one temporary register, and VL offsets that require
24868 no temporary registers.
24869 (aarch64_conditional_register_usage): Mark the predicate registers
24870 as fixed if SVE isn't available.
24871 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
24872 Return true for SVE vector and predicate modes.
24873 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
24874 rather than an unsigned int. Handle SVE modes.
24875 (aarch64_preferred_simd_mode): Update call accordingly. Handle
24877 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
24879 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
24880 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
24881 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
24882 (aarch64_sve_float_mul_immediate_p): New functions.
24883 (aarch64_sve_valid_immediate): New function.
24884 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
24885 Explicitly reject structure modes. Check for INDEX constants.
24886 Handle PTRUE and PFALSE constants.
24887 (aarch64_check_zero_based_sve_index_immediate): New function.
24888 (aarch64_simd_imm_zero_p): Delete.
24889 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
24890 vector modes. Accept constants in the range of CNT[BHWD].
24891 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
24892 ask for an Advanced SIMD mode.
24893 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
24894 (aarch64_simd_vector_alignment): Handle SVE predicates.
24895 (aarch64_vectorize_preferred_vector_alignment): New function.
24896 (aarch64_simd_vector_alignment_reachable): Use it instead of
24898 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
24899 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
24901 (MAX_VECT_LEN): Delete.
24902 (expand_vec_perm_d): Add a vec_flags field.
24903 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
24904 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
24905 (aarch64_evpc_ext): Don't apply a big-endian lane correction
24907 (aarch64_evpc_rev): Rename to...
24908 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
24909 (aarch64_evpc_rev_global): New function.
24910 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
24911 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
24913 (aarch64_evpc_sve_tbl): New function.
24914 (aarch64_expand_vec_perm_const_1): Update after rename of
24915 aarch64_evpc_rev. Handle SVE permutes too, trying
24916 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
24917 than aarch64_evpc_tbl.
24918 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
24919 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
24920 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
24921 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
24922 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
24923 (aarch64_expand_sve_vcond): New functions.
24924 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
24925 of aarch64_vector_mode_p.
24926 (aarch64_dwarf_poly_indeterminate_value): New function.
24927 (aarch64_compute_pressure_classes): Likewise.
24928 (aarch64_can_change_mode_class): Likewise.
24929 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
24930 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
24931 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
24932 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
24933 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
24934 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
24935 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
24936 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
24938 (Dn, Dl, Dr): Accept const as well as const_vector.
24939 (Dz): Likewise. Compare against CONST0_RTX.
24940 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
24941 of "vector" where appropriate.
24942 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
24943 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
24944 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
24945 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
24946 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
24947 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
24948 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
24949 (v_int_equiv): Extend to SVE modes.
24950 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
24952 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
24953 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
24954 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
24955 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
24956 (SVE_COND_FP_CMP): New int iterators.
24957 (perm_hilo): Handle the new unpack unspecs.
24958 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
24960 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
24961 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
24962 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
24963 (aarch64_equality_operator, aarch64_constant_vector_operand)
24964 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24965 (aarch64_sve_nonimmediate_operand): Likewise.
24966 (aarch64_sve_general_operand): Likewise.
24967 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24968 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24969 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24970 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24971 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24972 (aarch64_sve_float_arith_immediate): Likewise.
24973 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24974 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24975 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24976 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24977 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24978 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24979 (aarch64_sve_float_arith_operand): Likewise.
24980 (aarch64_sve_float_arith_with_sub_operand): Likewise.
24981 (aarch64_sve_float_mul_operand): Likewise.
24982 (aarch64_sve_vec_perm_operand): Likewise.
24983 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24984 (aarch64_mov_operand): Accept const_poly_int and const_vector.
24985 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24986 as well as const_vector.
24987 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24988 in file. Use CONST0_RTX and CONSTM1_RTX.
24989 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
24990 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24991 Use aarch64_simd_imm_zero.
24992 * config/aarch64/aarch64-sve.md: New file.
24993 * config/aarch64/aarch64.md: Include it.
24994 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24995 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24996 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24997 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24998 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24999 (sve): New attribute.
25000 (enabled): Disable instructions with the sve attribute unless
25002 (movqi, movhi): Pass CONST_POLY_INT operaneds through
25003 aarch64_expand_mov_immediate.
25004 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
25005 CNT[BHSD] immediates.
25006 (movti): Split CONST_POLY_INT moves into two halves.
25007 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
25008 Split additions that need a temporary here if the destination
25009 is the stack pointer.
25010 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
25011 (*add<mode>3_poly_1): New instruction.
25012 (set_clobber_cc): New expander.
25014 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25016 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
25017 parameter and use it instead of GET_MODE_SIZE (innermode). Use
25018 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
25019 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
25020 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
25021 Change innermode from fixed_mode_size to machine_mode.
25022 (simplify_subreg): Update call accordingly. Handle a constant-sized
25023 subreg of a variable-length CONST_VECTOR.
25025 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
25026 Alan Hayward <alan.hayward@arm.com>
25027 David Sherwood <david.sherwood@arm.com>
25029 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
25030 (add_offset_to_base): New function, split out from...
25031 (create_mem_ref): ...here. When handling a scale other than 1,
25032 check first whether the address is valid without the offset.
25033 Add it into the base if so, leaving the index and scale as-is.
25035 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25038 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
25039 fold_for_warn before checking if arg2 is INTEGER_CST.
25041 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
25043 * config/rs6000/predicates.md (load_multiple_operation): Delete.
25044 (store_multiple_operation): Delete.
25045 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
25046 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
25047 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
25048 guarded by TARGET_STRING.
25049 (rs6000_output_load_multiple): Delete.
25050 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
25051 OPTION_MASK_STRING / TARGET_STRING handling.
25052 (print_operand) <'N', 'O'>: Add comment that these are unused now.
25053 (const rs6000_opt_masks) <"string">: Change mask to 0.
25054 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
25055 (MASK_STRING): Delete.
25056 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
25058 (load_multiple): Delete.
25065 (store_multiple): Delete.
25072 (movmemsi_8reg): Delete.
25073 (corresponding unnamed define_insn): Delete.
25074 (movmemsi_6reg): Delete.
25075 (corresponding unnamed define_insn): Delete.
25076 (movmemsi_4reg): Delete.
25077 (corresponding unnamed define_insn): Delete.
25078 (movmemsi_2reg): Delete.
25079 (corresponding unnamed define_insn): Delete.
25080 (movmemsi_1reg): Delete.
25081 (corresponding unnamed define_insn): Delete.
25082 * config/rs6000/rs6000.opt (mno-string): New.
25083 (mstring): Replace by deprecation warning stub.
25084 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
25086 2018-01-12 Jakub Jelinek <jakub@redhat.com>
25088 * regrename.c (regrename_do_replace): If replacing the same
25089 reg multiple times, try to reuse last created gen_raw_REG.
25092 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
25093 main to workaround a bug in GDB.
25095 2018-01-12 Tom de Vries <tom@codesourcery.com>
25098 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
25100 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
25102 PR rtl-optimization/80481
25103 * ira-color.c (get_cap_member): New function.
25104 (allocnos_conflict_by_live_ranges_p): Use it.
25105 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
25106 (setup_slot_coalesced_allocno_live_ranges): Ditto.
25108 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
25111 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
25112 (*saddl_se_1): Ditto.
25113 (*ssubsi_1): Ditto.
25114 (*ssubl_se_1): Ditto.
25116 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25118 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
25119 rather than wi::to_widest for DR_INITs.
25120 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
25121 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
25122 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
25124 (vect_analyze_group_access_1): Note that here.
25126 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25128 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
25129 polynomial type sizes.
25131 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
25133 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
25134 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
25135 (gimple_add_tmp_var): Likewise.
25137 2018-01-12 Martin Liska <mliska@suse.cz>
25139 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
25140 (gimple_alloc_sizes): Likewise.
25141 (dump_gimple_statistics): Use PRIu64 in printf format.
25142 * gimple.h: Change uint64_t to int.
25144 2018-01-12 Martin Liska <mliska@suse.cz>
25146 * tree-core.h: Use uint64_t instead of int.
25147 * tree.c (tree_node_counts): Likewise.
25148 (tree_node_sizes): Likewise.
25149 (dump_tree_statistics): Use PRIu64 in printf format.
25151 2018-01-12 Martin Liska <mliska@suse.cz>
25153 * Makefile.in: As qsort_chk is implemented in vec.c, add
25154 vec.o to linkage of gencfn-macros.
25155 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
25156 passing the info to record_node_allocation_statistics.
25157 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
25159 * ggc-common.c (struct ggc_usage): Add operator== and use
25160 it in operator< and compare function.
25161 * mem-stats.h (struct mem_usage): Likewise.
25162 * vec.c (struct vec_usage): Remove operator< and compare
25163 function. Can be simply inherited.
25165 2018-01-12 Martin Jambor <mjambor@suse.cz>
25168 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
25169 * tree-ssa-math-opts.c: Include domwalk.h.
25170 (convert_mult_to_fma_1): New function.
25171 (fma_transformation_info): New type.
25172 (fma_deferring_state): Likewise.
25173 (cancel_fma_deferring): New function.
25174 (result_of_phi): Likewise.
25175 (last_fma_candidate_feeds_initial_phi): Likewise.
25176 (convert_mult_to_fma): Added deferring logic, split actual
25177 transformation to convert_mult_to_fma_1.
25178 (math_opts_dom_walker): New type.
25179 (math_opts_dom_walker::after_dom_children): New method, body moved
25180 here from pass_optimize_widening_mul::execute, added deferring logic
25182 (pass_optimize_widening_mul::execute): Moved most of code to
25183 math_opts_dom_walker::after_dom_children.
25184 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
25185 * config/i386/i386.c (ix86_option_override_internal): Added
25186 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
25188 2018-01-12 Richard Biener <rguenther@suse.de>
25191 * dwarf2out.c (gen_variable_die): Do not reset old_die for
25192 inline instance vars.
25194 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
25197 * config/rx/rx.c (rx_is_restricted_memory_address):
25198 Handle SUBREG case.
25200 2018-01-12 Richard Biener <rguenther@suse.de>
25202 PR tree-optimization/80846
25203 * target.def (split_reduction): New target hook.
25204 * targhooks.c (default_split_reduction): New function.
25205 * targhooks.h (default_split_reduction): Declare.
25206 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
25207 target requests first reduce vectors by combining low and high
25209 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
25210 (get_vectype_for_scalar_type_and_size): Export.
25211 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
25212 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
25213 * doc/tm.texi: Regenerate.
25214 * config/i386/i386.c (ix86_split_reduction): Implement
25215 TARGET_VECTORIZE_SPLIT_REDUCTION.
25217 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25220 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
25221 in PIC mode except for TARGET_VXWORKS_RTP.
25222 * config/sparc/sparc.c: Include cfgrtl.h.
25223 (TARGET_INIT_PIC_REG): Define.
25224 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
25225 (sparc_pic_register_p): New predicate.
25226 (sparc_legitimate_address_p): Use it.
25227 (sparc_legitimize_pic_address): Likewise.
25228 (sparc_delegitimize_address): Likewise.
25229 (sparc_mode_dependent_address_p): Likewise.
25230 (gen_load_pcrel_sym): Remove 4th parameter.
25231 (load_got_register): Adjust call to above. Remove obsolete stuff.
25232 (sparc_expand_prologue): Do not call load_got_register here.
25233 (sparc_flat_expand_prologue): Likewise.
25234 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
25235 (sparc_use_pseudo_pic_reg): New function.
25236 (sparc_init_pic_reg): Likewise.
25237 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
25238 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
25240 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
25242 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
25243 Add item for branch_cost.
25245 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
25247 PR rtl-optimization/83565
25248 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
25249 not extend the result to a larger mode for rotate operations.
25250 (num_sign_bit_copies1): Likewise.
25252 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25255 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
25257 Use values-Xc.o for -pedantic.
25258 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
25260 2018-01-12 Martin Liska <mliska@suse.cz>
25263 * ipa-devirt.c (final_warning_record::grow_type_warnings):
25265 (possible_polymorphic_call_targets): Use it.
25266 (ipa_devirt): Likewise.
25268 2018-01-12 Martin Liska <mliska@suse.cz>
25270 * profile-count.h (enum profile_quality): Use 0 as invalid
25271 enum value of profile_quality.
25273 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
25275 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
25276 -mext-string options.
25278 2018-01-12 Richard Biener <rguenther@suse.de>
25280 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
25281 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
25282 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
25284 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
25286 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
25288 * configure.ac (--with-long-double-format): Add support for the
25289 configuration option to change the default long double format on
25291 * config.gcc (powerpc*-linux*-*): Likewise.
25292 * configure: Regenerate.
25293 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
25294 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
25295 used without modification.
25297 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25299 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
25300 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
25301 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
25302 MISC_BUILTIN_SPEC_BARRIER.
25303 (rs6000_init_builtins): Likewise.
25304 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
25306 (speculation_barrier): New define_insn.
25307 * doc/extend.texi: Document __builtin_speculation_barrier.
25309 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25312 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
25313 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
25314 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
25316 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
25317 integral modes instead of "ss" and "sd".
25318 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
25319 vectors with 32-bit and 64-bit elements.
25320 (vecdupssescalarmodesuffix): New mode attribute.
25321 (vec_dup<mode>): Use it.
25323 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
25326 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
25327 frame if argument is passed on stack.
25329 2018-01-11 Jakub Jelinek <jakub@redhat.com>
25332 * ree.c (combine_reaching_defs): Optimize also
25333 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
25334 reg2=any_extend(exp); reg1=reg2;, formatting fix.
25336 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25338 PR middle-end/83189
25339 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
25341 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
25343 PR middle-end/83718
25344 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
25345 after they are computed.
25347 2018-01-11 Bin Cheng <bin.cheng@arm.com>
25349 PR tree-optimization/83695
25350 * gimple-loop-linterchange.cc
25351 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
25352 reset cached scev information after interchange.
25353 (pass_linterchange::execute): Remove call to scev_reset_htab.
25355 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25357 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
25358 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
25359 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
25360 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
25361 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
25362 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
25363 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
25364 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
25365 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
25366 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
25367 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
25368 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
25369 (V_lane_reg): Likewise.
25370 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
25372 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
25373 (vfmal_lane_low<mode>_intrinsic,
25374 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
25375 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
25376 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
25377 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
25378 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
25379 vfmsl_lane_high<mode>_intrinsic): New define_insns.
25381 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25383 * config/arm/arm-cpus.in (fp16fml): New feature.
25384 (ALL_SIMD): Add fp16fml.
25385 (armv8.2-a): Add fp16fml as an option.
25386 (armv8.3-a): Likewise.
25387 (armv8.4-a): Add fp16fml as part of fp16.
25388 * config/arm/arm.h (TARGET_FP16FML): Define.
25389 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
25391 * config/arm/arm-modes.def (V2HF): Define.
25392 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
25393 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
25394 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
25395 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
25396 vfmsl_low, vfmsl_high): New set of builtins.
25397 * config/arm/iterators.md (PLUSMINUS): New code iterator.
25398 (vfml_op): New code attribute.
25399 (VFMLHALVES): New int iterator.
25400 (VFML, VFMLSEL): New mode attributes.
25401 (V_reg): Define mapping for V2HF.
25402 (V_hi, V_lo): New mode attributes.
25403 (VF_constraint): Likewise.
25404 (vfml_half, vfml_half_selector): New int attributes.
25405 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
25407 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
25408 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
25410 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
25411 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
25412 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
25413 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
25415 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
25416 Document new effective target and option set.
25418 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25420 * config/arm/arm-cpus.in (armv8_4): New feature.
25421 (ARMv8_4a): New fgroup.
25422 (armv8.4-a): New arch.
25423 * config/arm/arm-tables.opt: Regenerate.
25424 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
25425 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
25426 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
25427 Add matching rules for -march=armv8.4-a and extensions.
25428 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
25430 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
25433 * config/rx/rx.md (BW): New mode attribute.
25434 (sync_lock_test_and_setsi): Add mode suffix to insn output.
25436 2018-01-11 Richard Biener <rguenther@suse.de>
25438 PR tree-optimization/83435
25439 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
25440 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
25441 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
25443 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25444 Alan Hayward <alan.hayward@arm.com>
25445 David Sherwood <david.sherwood@arm.com>
25447 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
25449 (aarch64_classify_address): Initialize it. Track polynomial offsets.
25450 (aarch64_print_address_internal): Use it to check for a zero offset.
25452 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25453 Alan Hayward <alan.hayward@arm.com>
25454 David Sherwood <david.sherwood@arm.com>
25456 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
25457 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
25458 Return a poly_int64 rather than a HOST_WIDE_INT.
25459 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
25460 rather than a HOST_WIDE_INT.
25461 * config/aarch64/aarch64.h (aarch64_frame): Protect with
25462 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
25463 hard_fp_offset, frame_size, initial_adjust, callee_offset and
25464 final_offset from HOST_WIDE_INT to poly_int64.
25465 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
25466 to_constant when getting the number of units in an Advanced SIMD
25468 (aarch64_builtin_vectorized_function): Check for a constant number
25470 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
25472 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
25473 attribute instead of GET_MODE_NUNITS.
25474 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
25475 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
25476 GET_MODE_SIZE for fixed-size registers.
25477 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
25478 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
25479 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
25480 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
25481 (aarch64_print_operand, aarch64_print_address_internal)
25482 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
25483 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
25484 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
25485 Handle polynomial GET_MODE_SIZE.
25486 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
25487 wider than SImode without modification.
25488 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
25489 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
25490 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
25491 passing and returning SVE modes.
25492 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
25493 rather than GEN_INT.
25494 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
25495 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
25496 (aarch64_allocate_and_probe_stack_space): Likewise.
25497 (aarch64_layout_frame): Cope with polynomial offsets.
25498 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
25499 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
25500 polynomial offsets.
25501 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
25502 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
25503 poly_int64 rather than a HOST_WIDE_INT.
25504 (aarch64_get_separate_components, aarch64_process_components)
25505 (aarch64_expand_prologue, aarch64_expand_epilogue)
25506 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
25507 (aarch64_anchor_offset): New function, split out from...
25508 (aarch64_legitimize_address): ...here.
25509 (aarch64_builtin_vectorization_cost): Handle polynomial
25510 TYPE_VECTOR_SUBPARTS.
25511 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
25513 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
25514 number of elements from the PARALLEL rather than the mode.
25515 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
25516 rather than GET_MODE_BITSIZE.
25517 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
25518 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
25519 (aarch64_expand_vec_perm_const_1): Handle polynomial
25520 d->perm.length () and d->perm elements.
25521 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
25522 Apply to_constant to d->perm elements.
25523 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
25524 polynomial CONST_VECTOR_NUNITS.
25525 (aarch64_move_pointer): Take amount as a poly_int64 rather
25527 (aarch64_progress_pointer): Avoid temporary variable.
25528 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
25529 the mode attribute instead of GET_MODE.
25531 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25532 Alan Hayward <alan.hayward@arm.com>
25533 David Sherwood <david.sherwood@arm.com>
25535 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
25536 x exists before using it.
25537 (aarch64_add_constant_internal): Rename to...
25538 (aarch64_add_offset_1): ...this. Replace regnum with separate
25539 src and dest rtxes. Handle the case in which they're different,
25540 including when the offset is zero. Replace scratchreg with an rtx.
25541 Use 2 additions if there is no spare register into which we can
25542 move a 16-bit constant.
25543 (aarch64_add_constant): Delete.
25544 (aarch64_add_offset): Replace reg with separate src and dest
25545 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
25546 Use aarch64_add_offset_1.
25547 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
25548 an rtx rather than an int. Take the delta as a poly_int64
25549 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
25550 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
25551 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
25552 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
25553 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
25554 and aarch64_add_sp.
25555 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
25556 aarch64_add_constant.
25558 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25560 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
25561 Use scalar_float_mode.
25563 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25565 * config/aarch64/aarch64-simd.md
25566 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
25567 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
25568 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
25569 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
25570 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
25571 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
25572 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
25573 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
25574 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
25575 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
25577 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25580 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
25581 targ_options->x_arm_arch_string is non NULL.
25583 2018-01-11 Tamar Christina <tamar.christina@arm.com>
25585 * config/aarch64/aarch64.h
25586 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
25588 2018-01-11 Sudakshina Das <sudi.das@arm.com>
25591 * expmed.c (emit_store_flag_force): Swap if const op0
25592 and change VOIDmode to mode of op0.
25594 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
25596 PR rtl-optimization/83761
25597 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
25598 than bytes to mode_for_size.
25600 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25602 PR middle-end/83189
25603 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
25604 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
25607 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25609 PR middle-end/83575
25610 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
25611 when in layout mode.
25612 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
25613 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
25616 2018-01-10 Michael Collison <michael.collison@arm.com>
25618 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
25619 * config/aarch64/aarch64-option-extension.def: Add
25620 AARCH64_OPT_EXTENSION of 'fp16fml'.
25621 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25622 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
25623 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
25624 * config/aarch64/constraints.md (Ui7): New constraint.
25625 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
25626 (VFMLA_SEL_W): Ditto.
25629 (VFMLA16_LOW): New int iterator.
25630 (VFMLA16_HIGH): Ditto.
25631 (UNSPEC_FMLAL): New unspec.
25632 (UNSPEC_FMLSL): Ditto.
25633 (UNSPEC_FMLAL2): Ditto.
25634 (UNSPEC_FMLSL2): Ditto.
25635 (f16mac): New code attribute.
25636 * config/aarch64/aarch64-simd-builtins.def
25637 (aarch64_fmlal_lowv2sf): Ditto.
25638 (aarch64_fmlsl_lowv2sf): Ditto.
25639 (aarch64_fmlalq_lowv4sf): Ditto.
25640 (aarch64_fmlslq_lowv4sf): Ditto.
25641 (aarch64_fmlal_highv2sf): Ditto.
25642 (aarch64_fmlsl_highv2sf): Ditto.
25643 (aarch64_fmlalq_highv4sf): Ditto.
25644 (aarch64_fmlslq_highv4sf): Ditto.
25645 (aarch64_fmlal_lane_lowv2sf): Ditto.
25646 (aarch64_fmlsl_lane_lowv2sf): Ditto.
25647 (aarch64_fmlal_laneq_lowv2sf): Ditto.
25648 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
25649 (aarch64_fmlalq_lane_lowv4sf): Ditto.
25650 (aarch64_fmlsl_lane_lowv4sf): Ditto.
25651 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
25652 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
25653 (aarch64_fmlal_lane_highv2sf): Ditto.
25654 (aarch64_fmlsl_lane_highv2sf): Ditto.
25655 (aarch64_fmlal_laneq_highv2sf): Ditto.
25656 (aarch64_fmlsl_laneq_highv2sf): Ditto.
25657 (aarch64_fmlalq_lane_highv4sf): Ditto.
25658 (aarch64_fmlsl_lane_highv4sf): Ditto.
25659 (aarch64_fmlalq_laneq_highv4sf): Ditto.
25660 (aarch64_fmlsl_laneq_highv4sf): Ditto.
25661 * config/aarch64/aarch64-simd.md:
25662 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
25663 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25664 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
25665 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
25666 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
25667 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
25668 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
25669 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
25670 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
25671 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
25672 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
25673 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
25674 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
25675 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
25676 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
25677 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
25678 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
25679 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
25680 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
25681 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
25682 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
25683 (vfmlsl_low_u32): Ditto.
25684 (vfmlalq_low_u32): Ditto.
25685 (vfmlslq_low_u32): Ditto.
25686 (vfmlal_high_u32): Ditto.
25687 (vfmlsl_high_u32): Ditto.
25688 (vfmlalq_high_u32): Ditto.
25689 (vfmlslq_high_u32): Ditto.
25690 (vfmlal_lane_low_u32): Ditto.
25691 (vfmlsl_lane_low_u32): Ditto.
25692 (vfmlal_laneq_low_u32): Ditto.
25693 (vfmlsl_laneq_low_u32): Ditto.
25694 (vfmlalq_lane_low_u32): Ditto.
25695 (vfmlslq_lane_low_u32): Ditto.
25696 (vfmlalq_laneq_low_u32): Ditto.
25697 (vfmlslq_laneq_low_u32): Ditto.
25698 (vfmlal_lane_high_u32): Ditto.
25699 (vfmlsl_lane_high_u32): Ditto.
25700 (vfmlal_laneq_high_u32): Ditto.
25701 (vfmlsl_laneq_high_u32): Ditto.
25702 (vfmlalq_lane_high_u32): Ditto.
25703 (vfmlslq_lane_high_u32): Ditto.
25704 (vfmlalq_laneq_high_u32): Ditto.
25705 (vfmlslq_laneq_high_u32): Ditto.
25706 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
25707 (AARCH64_FL_FOR_ARCH8_4): New.
25708 (AARCH64_ISA_F16FML): New ISA flag.
25709 (TARGET_F16FML): New feature flag for fp16fml.
25710 (doc/invoke.texi): Document new fp16fml option.
25712 2018-01-10 Michael Collison <michael.collison@arm.com>
25714 * config/aarch64/aarch64-builtins.c:
25715 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
25716 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25717 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
25718 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
25719 (AARCH64_ISA_SHA3): New ISA flag.
25720 (TARGET_SHA3): New feature flag for sha3.
25721 * config/aarch64/iterators.md (sha512_op): New int attribute.
25722 (CRYPTO_SHA512): New int iterator.
25723 (UNSPEC_SHA512H): New unspec.
25724 (UNSPEC_SHA512H2): Ditto.
25725 (UNSPEC_SHA512SU0): Ditto.
25726 (UNSPEC_SHA512SU1): Ditto.
25727 * config/aarch64/aarch64-simd-builtins.def
25728 (aarch64_crypto_sha512hqv2di): New builtin.
25729 (aarch64_crypto_sha512h2qv2di): Ditto.
25730 (aarch64_crypto_sha512su0qv2di): Ditto.
25731 (aarch64_crypto_sha512su1qv2di): Ditto.
25732 (aarch64_eor3qv8hi): Ditto.
25733 (aarch64_rax1qv2di): Ditto.
25734 (aarch64_xarqv2di): Ditto.
25735 (aarch64_bcaxqv8hi): Ditto.
25736 * config/aarch64/aarch64-simd.md:
25737 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
25738 (aarch64_crypto_sha512su0qv2di): Ditto.
25739 (aarch64_crypto_sha512su1qv2di): Ditto.
25740 (aarch64_eor3qv8hi): Ditto.
25741 (aarch64_rax1qv2di): Ditto.
25742 (aarch64_xarqv2di): Ditto.
25743 (aarch64_bcaxqv8hi): Ditto.
25744 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
25745 (vsha512h2q_u64): Ditto.
25746 (vsha512su0q_u64): Ditto.
25747 (vsha512su1q_u64): Ditto.
25748 (veor3q_u16): Ditto.
25749 (vrax1q_u64): Ditto.
25750 (vxarq_u64): Ditto.
25751 (vbcaxq_u16): Ditto.
25752 * config/arm/types.md (crypto_sha512): New type attribute.
25753 (crypto_sha3): Ditto.
25754 (doc/invoke.texi): Document new sha3 option.
25756 2018-01-10 Michael Collison <michael.collison@arm.com>
25758 * config/aarch64/aarch64-builtins.c:
25759 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
25760 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25761 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
25762 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
25763 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
25764 (AARCH64_ISA_SM4): New ISA flag.
25765 (TARGET_SM4): New feature flag for sm4.
25766 * config/aarch64/aarch64-simd-builtins.def
25767 (aarch64_sm3ss1qv4si): Ditto.
25768 (aarch64_sm3tt1aq4si): Ditto.
25769 (aarch64_sm3tt1bq4si): Ditto.
25770 (aarch64_sm3tt2aq4si): Ditto.
25771 (aarch64_sm3tt2bq4si): Ditto.
25772 (aarch64_sm3partw1qv4si): Ditto.
25773 (aarch64_sm3partw2qv4si): Ditto.
25774 (aarch64_sm4eqv4si): Ditto.
25775 (aarch64_sm4ekeyqv4si): Ditto.
25776 * config/aarch64/aarch64-simd.md:
25777 (aarch64_sm3ss1qv4si): Ditto.
25778 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
25779 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
25780 (aarch64_sm4eqv4si): Ditto.
25781 (aarch64_sm4ekeyqv4si): Ditto.
25782 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
25783 (sm3part_op): Ditto.
25784 (CRYPTO_SM3TT): Ditto.
25785 (CRYPTO_SM3PART): Ditto.
25786 (UNSPEC_SM3SS1): New unspec.
25787 (UNSPEC_SM3TT1A): Ditto.
25788 (UNSPEC_SM3TT1B): Ditto.
25789 (UNSPEC_SM3TT2A): Ditto.
25790 (UNSPEC_SM3TT2B): Ditto.
25791 (UNSPEC_SM3PARTW1): Ditto.
25792 (UNSPEC_SM3PARTW2): Ditto.
25793 (UNSPEC_SM4E): Ditto.
25794 (UNSPEC_SM4EKEY): Ditto.
25795 * config/aarch64/constraints.md (Ui2): New constraint.
25796 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
25797 * config/arm/types.md (crypto_sm3): New type attribute.
25798 (crypto_sm4): Ditto.
25799 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
25800 (vsm3tt1aq_u32): Ditto.
25801 (vsm3tt1bq_u32): Ditto.
25802 (vsm3tt2aq_u32): Ditto.
25803 (vsm3tt2bq_u32): Ditto.
25804 (vsm3partw1q_u32): Ditto.
25805 (vsm3partw2q_u32): Ditto.
25806 (vsm4eq_u32): Ditto.
25807 (vsm4ekeyq_u32): Ditto.
25808 (doc/invoke.texi): Document new sm4 option.
25810 2018-01-10 Michael Collison <michael.collison@arm.com>
25812 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
25813 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
25814 (AARCH64_FL_FOR_ARCH8_4): New.
25815 (AARCH64_FL_V8_4): New flag.
25816 (doc/invoke.texi): Document new armv8.4-a option.
25818 2018-01-10 Michael Collison <michael.collison@arm.com>
25820 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
25821 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
25822 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
25823 * config/aarch64/aarch64-option-extension.def: Add
25824 AARCH64_OPT_EXTENSION of 'sha2'.
25825 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
25826 (crypto): Disable sha2 and aes if crypto disabled.
25827 (crypto): Enable aes and sha2 if enabled.
25828 (simd): Disable sha2 and aes if simd disabled.
25829 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
25831 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
25832 (TARGET_SHA2): New feature flag for sha2.
25833 (TARGET_AES): New feature flag for aes.
25834 * config/aarch64/aarch64-simd.md:
25835 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
25836 conditional on TARGET_AES.
25837 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
25838 (aarch64_crypto_sha1hsi): Make pattern conditional
25840 (aarch64_crypto_sha1hv4si): Ditto.
25841 (aarch64_be_crypto_sha1hv4si): Ditto.
25842 (aarch64_crypto_sha1su1v4si): Ditto.
25843 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
25844 (aarch64_crypto_sha1su0v4si): Ditto.
25845 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
25846 (aarch64_crypto_sha256su0v4si): Ditto.
25847 (aarch64_crypto_sha256su1v4si): Ditto.
25848 (doc/invoke.texi): Document new aes and sha2 options.
25850 2018-01-10 Martin Sebor <msebor@redhat.com>
25852 PR tree-optimization/83781
25853 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
25856 2018-01-11 Martin Sebor <msebor@gmail.com>
25857 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
25859 PR tree-optimization/83501
25860 PR tree-optimization/81703
25862 * tree-ssa-strlen.c (get_string_cst): Rename...
25863 (get_string_len): ...to this. Handle global constants.
25864 (handle_char_store): Adjust.
25866 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
25867 Jim Wilson <jimw@sifive.com>
25869 * config/riscv/riscv-protos.h (riscv_output_return): New.
25870 * config/riscv/riscv.c (struct machine_function): New naked_p field.
25871 (riscv_attribute_table, riscv_output_return),
25872 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
25873 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
25874 (riscv_compute_frame_info): Only compute frame->mask if not a naked
25876 (riscv_expand_prologue): Add early return for naked function.
25877 (riscv_expand_epilogue): Likewise.
25878 (riscv_function_ok_for_sibcall): Return false for naked function.
25879 (riscv_set_current_function): New.
25880 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
25881 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
25882 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
25883 * doc/extend.texi (RISC-V Function Attributes): New.
25885 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
25887 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
25888 check for 128-bit long double before checking TCmode.
25889 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
25890 128-bit long doubles before checking TFmode or TCmode.
25891 (FLOAT128_IBM_P): Likewise.
25893 2018-01-10 Martin Sebor <msebor@redhat.com>
25895 PR tree-optimization/83671
25896 * builtins.c (c_strlen): Unconditionally return zero for the empty
25898 Use -Warray-bounds for warnings.
25899 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
25900 for non-constant array indices with COMPONENT_REF, arrays of
25901 arrays, and pointers to arrays.
25902 (gimple_fold_builtin_strlen): Determine and set length range for
25903 non-constant character arrays.
25905 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
25907 PR middle-end/81897
25908 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
25911 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
25913 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
25915 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25918 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
25919 VECTOR_MEM_ALTIVEC_OR_VSX_P.
25920 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
25921 indexed_or_indirect_operand predicate.
25922 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
25923 (*vsx_le_perm_load_v8hi): Likewise.
25924 (*vsx_le_perm_load_v16qi): Likewise.
25925 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
25926 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
25927 (*vsx_le_perm_store_v8hi): Likewise.
25928 (*vsx_le_perm_store_v16qi): Likewise.
25929 (eight unnamed splitters): Likewise.
25931 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
25933 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
25934 * config/rs6000/emmintrin.h: Likewise.
25935 * config/rs6000/mmintrin.h: Likewise.
25936 * config/rs6000/xmmintrin.h: Likewise.
25938 2018-01-10 David Malcolm <dmalcolm@redhat.com>
25941 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
25943 * tree.c (tree_nop_conversion): Return true for location wrapper
25945 (maybe_wrap_with_location): New function.
25946 (selftest::check_strip_nops): New function.
25947 (selftest::test_location_wrappers): New function.
25948 (selftest::tree_c_tests): Call it.
25949 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
25950 (maybe_wrap_with_location): New decl.
25951 (EXPR_LOCATION_WRAPPER_P): New macro.
25952 (location_wrapper_p): New inline function.
25953 (tree_strip_any_location_wrapper): New inline function.
25955 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
25958 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
25959 stack_realign_offset for the largest alignment of stack slot
25961 (ix86_find_max_used_stack_alignment): New function.
25962 (ix86_finalize_stack_frame_flags): Use it. Set
25963 max_used_stack_alignment if we don't realign stack.
25964 * config/i386/i386.h (machine_function): Add
25965 max_used_stack_alignment.
25967 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
25969 * config/arm/arm.opt (-mbranch-cost): New option.
25970 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25973 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
25976 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25977 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25979 2018-01-10 Richard Biener <rguenther@suse.de>
25982 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25983 early out so it also covers the case where we have a non-NULL
25986 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25988 PR tree-optimization/83753
25989 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25990 for non-strided grouped accesses if the number of elements is 1.
25992 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25995 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25996 * i386.h (TARGET_USE_GATHER): Define.
25997 * x86-tune.def (X86_TUNE_USE_GATHER): New.
25999 2018-01-10 Martin Liska <mliska@suse.cz>
26002 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
26003 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
26005 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
26006 CLEANUP_NO_PARTITIONING is not set.
26008 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
26010 * doc/rtl.texi: Remove documentation of (const ...) wrappers
26011 for vectors, as a partial revert of r254296.
26012 * rtl.h (const_vec_p): Delete.
26013 (const_vec_duplicate_p): Don't test for vector CONSTs.
26014 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
26015 * expmed.c (make_tree): Likewise.
26018 * common.md (E, F): Use CONSTANT_P instead of checking for
26020 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
26021 checking for CONST_VECTOR.
26023 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26025 PR middle-end/83575
26026 * predict.c (force_edge_cold): Handle in more sane way edges
26027 with no prediction.
26029 2018-01-09 Carl Love <cel@us.ibm.com>
26031 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
26033 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
26034 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
26035 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
26036 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
26037 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
26038 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
26039 * config/rs6000/rs6000-protos.h: Add extern defition for
26040 rs6000_generate_float2_double_code.
26041 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
26043 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
26044 (float2_v2df): Add define_expand.
26046 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
26049 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
26050 op_mode in the force_to_mode call.
26052 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26054 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
26055 instead of checking each element individually.
26056 (aarch64_evpc_uzp): Likewise.
26057 (aarch64_evpc_zip): Likewise.
26058 (aarch64_evpc_ext): Likewise.
26059 (aarch64_evpc_rev): Likewise.
26060 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
26061 instead of checking each element individually. Return true without
26063 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
26064 whether all selected elements come from the same input, instead of
26065 checking each element individually. Remove calls to gen_rtx_REG,
26066 start_sequence and end_sequence and instead assert that no rtl is
26069 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26071 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
26072 order of HIGH and CONST checks.
26074 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
26076 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
26077 if the destination isn't an SSA_NAME.
26079 2018-01-09 Richard Biener <rguenther@suse.de>
26081 PR tree-optimization/83668
26082 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
26084 (canonicalize_loop_form): ... here, renamed from ...
26085 (canonicalize_loop_closed_ssa_form): ... this and amended to
26086 swap successor edges for loop exit blocks to make us use
26087 the RPO order we need for initial schedule generation.
26089 2018-01-09 Joseph Myers <joseph@codesourcery.com>
26091 PR tree-optimization/64811
26092 * match.pd: When optimizing comparisons with Inf, avoid
26093 introducing or losing exceptions from comparisons with NaN.
26095 2018-01-09 Martin Liska <mliska@suse.cz>
26098 * asan.c (shadow_mem_size): Add gcc_assert.
26100 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
26102 Don't save registers in main().
26105 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
26106 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
26107 * config/avr/avr.c (avr_set_current_function): Don't error if
26108 naked, OS_task or OS_main are specified at the same time.
26109 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
26111 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
26113 * common/config/avr/avr-common.c (avr_option_optimization_table):
26114 Switch on -mmain-is-OS_task for optimizing compilations.
26116 2018-01-09 Richard Biener <rguenther@suse.de>
26118 PR tree-optimization/83572
26119 * graphite.c: Include cfganal.h.
26120 (graphite_transform_loops): Connect infinite loops to exit
26121 and remove fake edges at the end.
26123 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26125 * ipa-inline.c (edge_badness): Revert accidental checkin.
26127 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
26130 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
26131 symbols; not inline clones.
26133 2018-01-09 Jakub Jelinek <jakub@redhat.com>
26136 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
26137 hard registers. Formatting fixes.
26139 PR preprocessor/83722
26140 * gcc.c (try_generate_repro): Pass
26141 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
26142 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
26145 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
26146 Kito Cheng <kito.cheng@gmail.com>
26148 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
26149 (riscv_leaf_function_p): Delete.
26150 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
26152 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26154 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
26156 (do_ifelse): New function.
26157 (do_isel): New function.
26158 (do_sub3): New function.
26159 (do_add3): New function.
26160 (do_load_mask_compare): New function.
26161 (do_overlap_load_compare): New function.
26162 (expand_compare_loop): New function.
26163 (expand_block_compare): Call expand_compare_loop() when appropriate.
26164 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
26165 option description.
26166 (-mblock-compare-inline-loop-limit): New option.
26168 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26171 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
26172 Reverse order of second and third operands in first alternative.
26173 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
26174 of first and second elements in UNSPEC_VPERMR vector.
26175 (altivec_expand_vec_perm_le): Likewise.
26177 2018-01-08 Jeff Law <law@redhat.com>
26179 PR rtl-optimizatin/81308
26180 * tree-switch-conversion.c (cfg_altered): New file scoped static.
26181 (process_switch): If group_case_labels makes a change, then set
26183 (pass_convert_switch::execute): If a switch is converted, then
26184 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
26186 PR rtl-optimization/81308
26187 * recog.c (split_all_insns): Conditionally cleanup the CFG after
26190 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
26192 PR target/83663 - Revert r255946
26193 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
26194 generation for cases where splatting a value is not useful.
26195 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
26196 across a vec_duplicate and a paradoxical subreg forming a vector
26197 mode to a vec_concat.
26199 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
26201 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
26202 -march=armv8.3-a variants.
26203 * config/arm/t-multilib: Likewise.
26204 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
26206 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26208 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
26210 (cceq_ior_compare_complement): Give it a name so I can use it, and
26211 change boolean_or_operator predicate to boolean_operator so it can
26212 be used to generate a crand.
26213 (eqne): New code iterator.
26214 (bd/bd_neg): New code_attrs.
26215 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
26216 a single define_insn.
26217 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
26218 decrement (bdnzt/bdnzf/bdzt/bdzf).
26219 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
26220 with the new names of the branch decrement patterns, and added the
26221 names of the branch decrement conditional patterns.
26223 2018-01-08 Richard Biener <rguenther@suse.de>
26225 PR tree-optimization/83563
26226 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
26229 2018-01-08 Richard Biener <rguenther@suse.de>
26231 PR middle-end/83713
26232 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
26234 2018-01-08 Richard Biener <rguenther@suse.de>
26236 PR tree-optimization/83685
26237 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
26238 references to abnormals.
26240 2018-01-08 Richard Biener <rguenther@suse.de>
26243 * dwarf2out.c (output_indirect_strings): Handle empty
26244 skeleton_debug_str_hash.
26245 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
26247 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26249 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
26250 (emit_store_direct): Likewise.
26251 (arc_trampoline_adjust_address): Likewise.
26252 (arc_asm_trampoline_template): New function.
26253 (arc_initialize_trampoline): Use asm_trampoline_template.
26254 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
26255 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
26256 * config/arc/arc.md (flush_icache): Delete pattern.
26258 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
26260 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
26261 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
26264 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26267 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
26268 by not USED_FOR_TARGET.
26269 (make_pass_resolve_sw_modes): Likewise.
26271 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
26273 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
26276 2018-01-08 Richard Biener <rguenther@suse.de>
26278 PR middle-end/83580
26279 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
26281 2018-01-08 Richard Biener <rguenther@suse.de>
26283 PR middle-end/83517
26284 * match.pd ((t * 2) / 2) -> t): Add missing :c.
26286 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
26288 PR middle-end/81897
26289 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
26290 basic blocks with a small number of successors.
26291 (convert_control_dep_chain_into_preds): Improve handling of
26293 (dump_predicates): Split apart into...
26294 (dump_pred_chain): ...here...
26295 (dump_pred_info): ...and here.
26296 (can_one_predicate_be_invalidated_p): Add debugging printfs.
26297 (can_chain_union_be_invalidated_p): Improve check for invalidation
26299 (uninit_uses_cannot_happen): Avoid unnecessary if
26300 convert_control_dep_chain_into_preds yielded nothing.
26302 2018-01-06 Martin Sebor <msebor@redhat.com>
26304 PR tree-optimization/83640
26305 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
26306 subtracting negative offset from size.
26307 (builtin_access::overlap): Adjust offset bounds of the access to fall
26308 within the size of the object if possible.
26310 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
26312 PR rtl-optimization/83699
26313 * expmed.c (extract_bit_field_1): Restrict the vector usage of
26314 extract_bit_field_as_subreg to cases in which the extracted
26315 value is also a vector.
26317 * lra-constraints.c (process_alt_operands): Test for the equivalence
26318 substitutions when detecting a possible reload cycle.
26320 2018-01-06 Jakub Jelinek <jakub@redhat.com>
26323 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
26324 by default if flag_selective_schedling{,2}. Formatting fixes.
26326 PR rtl-optimization/83682
26327 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
26328 if it has non-VECTOR_MODE element mode.
26329 (vec_duplicate_p): Likewise.
26331 PR middle-end/83694
26332 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
26333 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
26335 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26338 * config/i386/i386-builtin.def
26339 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
26340 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
26341 Require also OPTION_MASK_ISA_AVX512F in addition to
26342 OPTION_MASK_ISA_GFNI.
26343 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
26344 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
26345 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
26346 to OPTION_MASK_ISA_GFNI.
26347 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
26348 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
26349 OPTION_MASK_ISA_AVX512BW.
26350 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
26351 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
26352 addition to OPTION_MASK_ISA_GFNI.
26353 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
26354 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
26355 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
26356 to OPTION_MASK_ISA_GFNI.
26357 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
26358 a requirement for all ISAs rather than any of them with a few
26360 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
26362 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
26363 bitmasks to be enabled with 3 exceptions, instead of requiring any
26364 enabled ISA with lots of exceptions.
26365 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
26366 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
26367 Change avx512bw in isa attribute to avx512f.
26368 * config/i386/sgxintrin.h: Add license boilerplate.
26369 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
26370 to __AVX512F__ and __AVX512VL to __AVX512VL__.
26371 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
26372 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
26374 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
26375 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
26376 temporarily sse2 rather than sse if not enabled already.
26379 * config/i386/sse.md (VI248_VLBW): Rename to ...
26380 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
26381 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
26382 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
26383 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
26384 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
26385 mode iterator instead of VI248_VLBW.
26387 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
26389 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
26390 (record_modified): Skip clobbers; add debug output.
26391 (param_change_prob): Use sreal frequencies.
26393 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26395 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26396 punt for user-aligned variables.
26398 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
26400 * tree-chrec.c (chrec_contains_symbols): Return true for
26403 2018-01-05 Sudakshina Das <sudi.das@arm.com>
26406 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
26407 of (x|y) == x for BICS pattern.
26409 2018-01-05 Jakub Jelinek <jakub@redhat.com>
26411 PR tree-optimization/83605
26412 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
26413 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
26416 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
26418 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
26419 * config/epiphany/rtems.h: New file.
26421 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26422 Uros Bizjak <ubizjak@gmail.com>
26425 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
26426 QIreg_operand instead of register_operand predicate.
26427 * config/i386/i386.c (ix86_rop_should_change_byte_p,
26428 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
26429 comments instead of -fmitigate[-_]rop.
26431 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26434 * cgraphunit.c (symbol_table::compile): Switch to text_section
26435 before calling assembly_start debug hook.
26436 * run-rtl-passes.c (run_rtl_passes): Likewise.
26439 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26441 * tree-vrp.c (extract_range_from_binary_expr_1): Check
26442 range_int_cst_p rather than !symbolic_range_p before calling
26443 extract_range_from_multiplicative_op_1.
26445 2018-01-04 Jeff Law <law@redhat.com>
26447 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
26448 redundant test in assertion.
26450 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26452 * doc/rtl.texi: Document machine_mode wrapper classes.
26454 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26456 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
26457 using tree_to_uhwi.
26459 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26461 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
26462 the VEC_PERM_EXPR fold to fail.
26464 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26467 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
26468 to switched_sections.
26470 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26473 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
26474 test for d.testing.
26476 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
26479 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
26480 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
26482 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26485 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
26486 is BLKmode and bitpos not zero or mode change is needed.
26488 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
26491 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
26494 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
26497 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
26498 instead of MULT rtx. Update all corresponding splitters.
26499 (*saddl_se): Ditto.
26500 (*ssub<modesuffix>): Ditto.
26501 (*ssubl_se): Ditto.
26502 (*cmp_sadd_di): Update split patterns.
26503 (*cmp_sadd_si): Ditto.
26504 (*cmp_sadd_sidi): Ditto.
26505 (*cmp_ssub_di): Ditto.
26506 (*cmp_ssub_si): Ditto.
26507 (*cmp_ssub_sidi): Ditto.
26508 * config/alpha/predicates.md (const23_operand): New predicate.
26509 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
26510 Look for ASHIFT, not MULT inner operand.
26511 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
26513 2018-01-04 Martin Liska <mliska@suse.cz>
26515 PR gcov-profile/83669
26516 * gcov.c (output_intermediate_file): Add version to intermediate
26518 * doc/gcov.texi: Document new field 'version' in intermediate
26519 file format. Fix location of '-k' option of gcov command.
26521 2018-01-04 Martin Liska <mliska@suse.cz>
26524 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
26526 2018-01-04 Jakub Jelinek <jakub@redhat.com>
26528 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
26530 2018-01-03 Martin Sebor <msebor@redhat.com>
26532 PR tree-optimization/83655
26533 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
26534 checking calls with invalid arguments.
26536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26538 * tree-vect-stmts.c (vect_get_store_rhs): New function.
26539 (vectorizable_mask_load_store): Delete.
26540 (vectorizable_call): Return false for masked loads and stores.
26541 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
26542 instead of gimple_assign_rhs1.
26543 (vectorizable_load): Handle IFN_MASK_LOAD.
26544 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
26546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26548 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
26550 (vectorizable_mask_load_store): ...here.
26551 (vectorizable_load): ...and here.
26553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26555 * tree-vect-stmts.c (vect_build_all_ones_mask)
26556 (vect_build_zero_merge_argument): New functions, split out from...
26557 (vectorizable_load): ...here.
26559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26561 * tree-vect-stmts.c (vect_check_store_rhs): New function,
26563 (vectorizable_mask_load_store): ...here.
26564 (vectorizable_store): ...and here.
26566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26568 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
26570 (vectorizable_mask_load_store): ...here.
26572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26574 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
26575 (vect_model_store_cost): Take a vec_load_store_type instead of a
26577 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
26578 (vect_model_store_cost): Take a vec_load_store_type instead of a
26580 (vectorizable_mask_load_store): Update accordingly.
26581 (vectorizable_store): Likewise.
26582 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
26584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26586 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
26587 IFN_MASK_LOAD calls here rather than...
26588 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
26590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26591 Alan Hayward <alan.hayward@arm.com>
26592 David Sherwood <david.sherwood@arm.com>
26594 * expmed.c (extract_bit_field_1): For vector extracts,
26595 fall back to extract_bit_field_as_subreg if vec_extract
26598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26599 Alan Hayward <alan.hayward@arm.com>
26600 David Sherwood <david.sherwood@arm.com>
26602 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
26603 they are variable or constant sized.
26604 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
26605 slots for constant-sized data.
26607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26608 Alan Hayward <alan.hayward@arm.com>
26609 David Sherwood <david.sherwood@arm.com>
26611 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
26612 handling COND_EXPRs with boolean comparisons, try to find a better
26613 basis for the mask type than the boolean itself.
26615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26617 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
26618 is calculated and how it can be overridden.
26619 * genmodes.c (max_bitsize_mode_any_mode): New variable.
26620 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
26622 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
26625 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26626 Alan Hayward <alan.hayward@arm.com>
26627 David Sherwood <david.sherwood@arm.com>
26629 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
26630 Remove the mode argument.
26631 (aarch64_simd_valid_immediate): Remove the mode and inverse
26633 * config/aarch64/iterators.md (bitsize): New iterator.
26634 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
26635 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
26636 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
26637 aarch64_simd_valid_immediate.
26638 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
26639 (aarch64_reg_or_bic_imm): Likewise.
26640 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
26641 with an insn_type enum and msl with a modifier_type enum.
26642 Replace element_width with a scalar_mode. Change the shift
26643 to unsigned int. Add constructors for scalar_float_mode and
26644 scalar_int_mode elements.
26645 (aarch64_vect_float_const_representable_p): Delete.
26646 (aarch64_can_const_movi_rtx_p)
26647 (aarch64_simd_scalar_immediate_valid_for_move)
26648 (aarch64_simd_make_constant): Update call to
26649 aarch64_simd_valid_immediate.
26650 (aarch64_advsimd_valid_immediate_hs): New function.
26651 (aarch64_advsimd_valid_immediate): Likewise.
26652 (aarch64_simd_valid_immediate): Remove mode and inverse
26653 arguments. Rewrite to use the above. Use const_vec_duplicate_p
26654 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
26655 and aarch64_float_const_representable_p on the result.
26656 (aarch64_output_simd_mov_immediate): Remove mode argument.
26657 Update call to aarch64_simd_valid_immediate and use of
26658 simd_immediate_info.
26659 (aarch64_output_scalar_simd_mov_immediate): Update call
26662 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26663 Alan Hayward <alan.hayward@arm.com>
26664 David Sherwood <david.sherwood@arm.com>
26666 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
26667 (mode_nunits): Likewise CONST_MODE_NUNITS.
26668 * machmode.def (ADJUST_NUNITS): Document.
26669 * genmodes.c (mode_data::need_nunits_adj): New field.
26670 (blank_mode): Update accordingly.
26671 (adj_nunits): New variable.
26672 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
26674 (emit_mode_size_inline): Set need_bytesize_adj for all modes
26675 listed in adj_nunits.
26676 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
26677 listed in adj_nunits. Don't emit case statements for such modes.
26678 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
26679 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
26680 nothing if adj_nunits is nonnull.
26681 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
26682 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
26683 (emit_mode_fbit): Update use of print_maybe_const_decl.
26684 (emit_move_size): Likewise. Treat the array as non-const
26686 (emit_mode_adjustments): Handle adj_nunits.
26688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26690 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
26691 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
26692 (VECTOR_MODES): Use it.
26693 (make_vector_modes): Take the prefix as an argument.
26695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26696 Alan Hayward <alan.hayward@arm.com>
26697 David Sherwood <david.sherwood@arm.com>
26699 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
26700 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
26701 for MODE_VECTOR_BOOL.
26702 * machmode.def (VECTOR_BOOL_MODE): Document.
26703 * genmodes.c (VECTOR_BOOL_MODE): New macro.
26704 (make_vector_bool_mode): New function.
26705 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
26707 * lto-streamer-in.c (lto_input_mode_table): Likewise.
26708 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
26710 * stor-layout.c (int_mode_for_mode): Likewise.
26711 * tree.c (build_vector_type_for_mode): Likewise.
26712 * varasm.c (output_constant_pool_2): Likewise.
26713 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
26714 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
26715 for MODE_VECTOR_BOOL.
26716 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
26717 of mode class checks.
26718 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
26719 instead of a list of mode class checks.
26720 (expand_vector_scalar_condition): Likewise.
26721 (type_for_widest_vector_mode): Handle BImode as an inner mode.
26723 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26724 Alan Hayward <alan.hayward@arm.com>
26725 David Sherwood <david.sherwood@arm.com>
26727 * machmode.h (mode_size): Change from unsigned short to
26729 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
26730 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26731 or if measurement_type is not polynomial.
26732 (fixed_size_mode::includes_p): Check for constant-sized modes.
26733 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
26734 return a poly_uint16 rather than an unsigned short.
26735 (emit_mode_size): Change the type of mode_size from unsigned short
26736 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
26737 (emit_mode_adjustments): Cope with polynomial vector sizes.
26738 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26740 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26742 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
26743 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
26744 * caller-save.c (setup_save_areas): Likewise.
26745 (replace_reg_with_saved_mem): Likewise.
26746 * calls.c (emit_library_call_value_1): Likewise.
26747 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
26748 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
26749 (gen_lowpart_for_combine): Likewise.
26750 * convert.c (convert_to_integer_1): Likewise.
26751 * cse.c (equiv_constant, cse_insn): Likewise.
26752 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
26753 (cselib_subst_to_values): Likewise.
26754 * dce.c (word_dce_process_block): Likewise.
26755 * df-problems.c (df_word_lr_mark_ref): Likewise.
26756 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
26757 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
26758 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
26759 (rtl_for_decl_location): Likewise.
26760 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
26761 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
26762 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
26763 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
26764 (expand_expr_real_1): Likewise.
26765 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
26766 (pad_below): Likewise.
26767 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26768 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
26769 * ira.c (get_subreg_tracking_sizes): Likewise.
26770 * ira-build.c (ira_create_allocno_objects): Likewise.
26771 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
26772 (ira_sort_regnos_for_alter_reg): Likewise.
26773 * ira-costs.c (record_operand_costs): Likewise.
26774 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
26775 (resolve_simple_move): Likewise.
26776 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
26777 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
26778 (lra_constraints): Likewise.
26779 (CONST_POOL_OK_P): Reject variable-sized modes.
26780 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
26781 (add_pseudo_to_slot, lra_spill): Likewise.
26782 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26783 * optabs-query.c (get_best_extraction_insn): Likewise.
26784 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26785 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
26786 (expand_mult_highpart, valid_multiword_target_p): Likewise.
26787 * recog.c (offsettable_address_addr_space_p): Likewise.
26788 * regcprop.c (maybe_mode_change): Likewise.
26789 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
26790 * regrename.c (build_def_use): Likewise.
26791 * regstat.c (dump_reg_info): Likewise.
26792 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
26793 (find_reloads, find_reloads_subreg_address): Likewise.
26794 * reload1.c (eliminate_regs_1): Likewise.
26795 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
26796 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
26797 (simplify_binary_operation_1, simplify_subreg): Likewise.
26798 * targhooks.c (default_function_arg_padding): Likewise.
26799 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
26800 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
26801 (verify_gimple_assign_ternary): Likewise.
26802 * tree-inline.c (estimate_move_cost): Likewise.
26803 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26804 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
26805 (get_address_cost_ainc): Likewise.
26806 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
26807 (vect_supportable_dr_alignment): Likewise.
26808 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26809 (vectorizable_reduction): Likewise.
26810 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
26811 (vectorizable_operation, vectorizable_load): Likewise.
26812 * tree.c (build_same_sized_truth_vector_type): Likewise.
26813 * valtrack.c (cleanup_auto_inc_dec): Likewise.
26814 * var-tracking.c (emit_note_insn_var_location): Likewise.
26815 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
26816 (ADDR_VEC_ALIGN): Likewise.
26818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26819 Alan Hayward <alan.hayward@arm.com>
26820 David Sherwood <david.sherwood@arm.com>
26822 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
26824 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
26825 or if measurement_type is polynomial.
26826 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
26827 * combine.c (make_extraction): Likewise.
26828 * dse.c (find_shift_sequence): Likewise.
26829 * dwarf2out.c (mem_loc_descriptor): Likewise.
26830 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
26831 (extract_bit_field, extract_low_bits): Likewise.
26832 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
26833 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
26834 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
26835 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
26836 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
26837 * reload.c (find_reloads): Likewise.
26838 * reload1.c (alter_reg): Likewise.
26839 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
26840 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
26841 * tree-if-conv.c (predicate_mem_writes): Likewise.
26842 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
26843 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
26844 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
26845 * valtrack.c (dead_debug_insert_temp): Likewise.
26846 * varasm.c (mergeable_constant_section): Likewise.
26847 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
26849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26850 Alan Hayward <alan.hayward@arm.com>
26851 David Sherwood <david.sherwood@arm.com>
26853 * expr.c (expand_assignment): Cope with polynomial mode sizes
26854 when assigning to a CONCAT.
26856 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26857 Alan Hayward <alan.hayward@arm.com>
26858 David Sherwood <david.sherwood@arm.com>
26860 * machmode.h (mode_precision): Change from unsigned short to
26862 (mode_to_precision): Return a poly_uint16 rather than an unsigned
26864 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
26865 or if measurement_type is not polynomial.
26866 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
26867 in which the mode is already known to be a scalar_int_mode.
26868 * genmodes.c (emit_mode_precision): Change the type of mode_precision
26869 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
26871 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26872 for GET_MODE_PRECISION.
26873 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26874 for GET_MODE_PRECISION.
26875 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
26877 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
26878 (expand_field_assignment, make_extraction): Likewise.
26879 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
26880 (get_last_value): Likewise.
26881 * convert.c (convert_to_integer_1): Likewise.
26882 * cse.c (cse_insn): Likewise.
26883 * expr.c (expand_expr_real_1): Likewise.
26884 * lra-constraints.c (simplify_operand_subreg): Likewise.
26885 * optabs-query.c (can_atomic_load_p): Likewise.
26886 * optabs.c (expand_atomic_load): Likewise.
26887 (expand_atomic_store): Likewise.
26888 * ree.c (combine_reaching_defs): Likewise.
26889 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
26890 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
26891 * tree.h (type_has_mode_precision_p): Likewise.
26892 * ubsan.c (instrument_si_overflow): Likewise.
26894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26895 Alan Hayward <alan.hayward@arm.com>
26896 David Sherwood <david.sherwood@arm.com>
26898 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
26899 polynomial numbers of units.
26900 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
26901 (valid_vector_subparts_p): New function.
26902 (build_vector_type): Remove temporary shim and take the number
26903 of units as a poly_uint64 rather than an int.
26904 (build_opaque_vector_type): Take the number of units as a
26905 poly_uint64 rather than an int.
26906 * tree.c (build_vector_from_ctor): Handle polynomial
26907 TYPE_VECTOR_SUBPARTS.
26908 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
26909 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
26910 (build_vector_from_val): If the number of units is variable,
26911 use build_vec_duplicate_cst for constant operands and
26912 VEC_DUPLICATE_EXPR otherwise.
26913 (make_vector_type): Remove temporary is_constant ().
26914 (build_vector_type, build_opaque_vector_type): Take the number of
26915 units as a poly_uint64 rather than an int.
26916 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
26918 * cfgexpand.c (expand_debug_expr): Likewise.
26919 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
26920 (store_constructor, expand_expr_real_1): Likewise.
26921 (const_scalar_mask_from_tree): Likewise.
26922 * fold-const-call.c (fold_const_reduction): Likewise.
26923 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
26924 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
26925 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
26926 (fold_relational_const): Likewise.
26927 (native_interpret_vector): Likewise. Change the size from an
26928 int to an unsigned int.
26929 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
26930 TYPE_VECTOR_SUBPARTS.
26931 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
26932 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
26933 duplicating a non-constant operand into a variable-length vector.
26934 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
26935 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
26936 * ipa-icf.c (sem_variable::equals): Likewise.
26937 * match.pd: Likewise.
26938 * omp-simd-clone.c (simd_clone_subparts): Likewise.
26939 * print-tree.c (print_node): Likewise.
26940 * stor-layout.c (layout_type): Likewise.
26941 * targhooks.c (default_builtin_vectorization_cost): Likewise.
26942 * tree-cfg.c (verify_gimple_comparison): Likewise.
26943 (verify_gimple_assign_binary): Likewise.
26944 (verify_gimple_assign_ternary): Likewise.
26945 (verify_gimple_assign_single): Likewise.
26946 * tree-pretty-print.c (dump_generic_node): Likewise.
26947 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26948 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
26949 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
26950 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
26951 (vect_shift_permute_load_chain): Likewise.
26952 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
26953 (expand_vector_condition, optimize_vector_constructor): Likewise.
26954 (lower_vec_perm, get_compute_type): Likewise.
26955 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
26956 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
26957 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
26958 (vect_recog_mask_conversion_pattern): Likewise.
26959 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
26960 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
26961 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26962 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
26963 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26964 (vectorizable_shift, vectorizable_operation, vectorizable_store)
26965 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26966 (supportable_widening_operation): Likewise.
26967 (supportable_narrowing_operation): Likewise.
26968 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26970 * varasm.c (output_constant): Likewise.
26972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26973 Alan Hayward <alan.hayward@arm.com>
26974 David Sherwood <david.sherwood@arm.com>
26976 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26977 so that both the length == 3 and length != 3 cases set up their
26978 own permute vectors. Add comments explaining why we know the
26979 number of elements is constant.
26980 (vect_permute_load_chain): Likewise.
26982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26983 Alan Hayward <alan.hayward@arm.com>
26984 David Sherwood <david.sherwood@arm.com>
26986 * machmode.h (mode_nunits): Change from unsigned char to
26988 (ONLY_FIXED_SIZE_MODES): New macro.
26989 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26990 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26991 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26993 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26994 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26995 or if measurement_type is not polynomial.
26996 * genmodes.c (ZERO_COEFFS): New macro.
26997 (emit_mode_nunits_inline): Make mode_nunits_inline return a
26999 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
27000 Use ZERO_COEFFS when emitting initializers.
27001 * data-streamer.h (bp_pack_poly_value): New function.
27002 (bp_unpack_poly_value): Likewise.
27003 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
27004 for GET_MODE_NUNITS.
27005 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
27006 for GET_MODE_NUNITS.
27007 * tree.c (make_vector_type): Remove temporary shim and make
27008 the real function take the number of units as a poly_uint64
27009 rather than an int.
27010 (build_vector_type_for_mode): Handle polynomial nunits.
27011 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
27012 * emit-rtl.c (const_vec_series_p_1): Likewise.
27013 (gen_rtx_CONST_VECTOR): Likewise.
27014 * fold-const.c (test_vec_duplicate_folding): Likewise.
27015 * genrecog.c (validate_pattern): Likewise.
27016 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
27017 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
27018 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
27019 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
27020 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
27021 * rtlanal.c (subreg_get_info): Likewise.
27022 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27023 (vect_grouped_load_supported): Likewise.
27024 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
27025 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
27026 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
27027 (simplify_const_unary_operation, simplify_binary_operation_1)
27028 (simplify_const_binary_operation, simplify_ternary_operation)
27029 (test_vector_ops_duplicate, test_vector_ops): Likewise.
27030 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
27031 instead of CONST_VECTOR_NUNITS.
27032 * varasm.c (output_constant_pool_2): Likewise.
27033 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
27034 explicit-encoded elements in the XVEC for variable-length vectors.
27036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27038 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
27040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27041 Alan Hayward <alan.hayward@arm.com>
27042 David Sherwood <david.sherwood@arm.com>
27044 * coretypes.h (fixed_size_mode): Declare.
27045 (fixed_size_mode_pod): New typedef.
27046 * builtins.h (target_builtins::x_apply_args_mode)
27047 (target_builtins::x_apply_result_mode): Change type to
27048 fixed_size_mode_pod.
27049 * builtins.c (apply_args_size, apply_result_size, result_vector)
27050 (expand_builtin_apply_args_1, expand_builtin_apply)
27051 (expand_builtin_return): Update accordingly.
27053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27055 * cse.c (hash_rtx_cb): Hash only the encoded elements.
27056 * cselib.c (cselib_hash_rtx): Likewise.
27057 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
27058 CONST_VECTOR encoding.
27060 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27061 Jeff Law <law@redhat.com>
27064 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
27065 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
27066 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
27067 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
27070 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
27071 explicitly probe *sp in a noreturn function if there were any callee
27072 register saves or frame pointer is needed.
27074 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27077 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
27078 BLKmode for ternary, binary or unary expressions.
27081 * var-tracking.c (delete_vta_debug_insn): New inline function.
27082 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
27083 insns from get_insns () to NULL instead of each bb separately.
27084 Use delete_vta_debug_insn. No longer static.
27085 (vt_debug_insns_local, variable_tracking_main_1): Adjust
27086 delete_vta_debug_insns callers.
27087 * rtl.h (delete_vta_debug_insns): Declare.
27088 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
27089 instead of variable_tracking_main.
27091 2018-01-03 Martin Sebor <msebor@redhat.com>
27093 PR tree-optimization/83603
27094 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
27095 arguments past the endof the argument list in functions declared
27096 without a prototype.
27097 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
27098 Avoid checking when arguments are null.
27100 2018-01-03 Martin Sebor <msebor@redhat.com>
27103 * doc/extend.texi (attribute const): Fix a typo.
27104 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
27105 issuing -Wsuggest-attribute for void functions.
27107 2018-01-03 Martin Sebor <msebor@redhat.com>
27109 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
27110 offset_int::from instead of wide_int::to_shwi.
27111 (maybe_diag_overlap): Remove assertion.
27112 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
27113 * gimple-ssa-sprintf.c (format_directive): Same.
27114 (parse_directive): Same.
27115 (sprintf_dom_walker::compute_format_length): Same.
27116 (try_substitute_return_value): Same.
27118 2018-01-03 Jeff Law <law@redhat.com>
27120 PR middle-end/83654
27121 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
27122 non-constant residual for zero at runtime and avoid probing in
27123 that case. Reorganize code for trailing problem to mirror handling
27126 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
27128 PR tree-optimization/83501
27129 * tree-ssa-strlen.c (get_string_cst): New.
27130 (handle_char_store): Call get_string_cst.
27132 2018-01-03 Martin Liska <mliska@suse.cz>
27134 PR tree-optimization/83593
27135 * tree-ssa-strlen.c: Include tree-cfg.h.
27136 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
27137 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
27138 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
27140 (strlen_dom_walker::before_dom_children): Call
27141 gimple_purge_dead_eh_edges. Dump tranformation with details
27143 (strlen_dom_walker::before_dom_children): Update call by adding
27144 new argument cleanup_eh.
27145 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
27147 2018-01-03 Martin Liska <mliska@suse.cz>
27150 * cif-code.def (VARIADIC_THUNK): New enum value.
27151 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
27154 2018-01-03 Jan Beulich <jbeulich@suse.com>
27156 * sse.md (mov<mode>_internal): Tighten condition for when to use
27157 vmovdqu<ssescalarsize> for TI and OI modes.
27159 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27161 Update copyright years.
27163 2018-01-03 Martin Liska <mliska@suse.cz>
27166 * ipa-visibility.c (function_and_variable_visibility): Skip
27167 functions with noipa attribure.
27169 2018-01-03 Jakub Jelinek <jakub@redhat.com>
27171 * gcc.c (process_command): Update copyright notice dates.
27172 * gcov-dump.c (print_version): Ditto.
27173 * gcov.c (print_version): Ditto.
27174 * gcov-tool.c (print_version): Ditto.
27175 * gengtype.c (create_file): Ditto.
27176 * doc/cpp.texi: Bump @copying's copyright year.
27177 * doc/cppinternals.texi: Ditto.
27178 * doc/gcc.texi: Ditto.
27179 * doc/gccint.texi: Ditto.
27180 * doc/gcov.texi: Ditto.
27181 * doc/install.texi: Ditto.
27182 * doc/invoke.texi: Ditto.
27184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27186 * vector-builder.h (vector_builder::m_full_nelts): Change from
27187 unsigned int to poly_uint64.
27188 (vector_builder::full_nelts): Update prototype accordingly.
27189 (vector_builder::new_vector): Likewise.
27190 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
27191 (vector_builder::operator ==): Likewise.
27192 (vector_builder::finalize): Likewise.
27193 * int-vector-builder.h (int_vector_builder::int_vector_builder):
27194 Take the number of elements as a poly_uint64 rather than an
27196 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
27197 from unsigned int to poly_uint64.
27198 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
27199 (vec_perm_indices::new_vector): Likewise.
27200 (vec_perm_indices::length): Likewise.
27201 (vec_perm_indices::nelts_per_input): Likewise.
27202 (vec_perm_indices::input_nelts): Likewise.
27203 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
27204 number of elements per input as a poly_uint64 rather than an
27205 unsigned int. Use the original encoding for variable-length
27206 vectors, rather than clamping each individual element.
27207 For the second and subsequent elements in each pattern,
27208 clamp the step and base before clamping their sum.
27209 (vec_perm_indices::series_p): Handle polynomial element counts.
27210 (vec_perm_indices::all_in_range_p): Likewise.
27211 (vec_perm_indices_to_tree): Likewise.
27212 (vec_perm_indices_to_rtx): Likewise.
27213 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
27214 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
27215 (tree_vector_builder::new_binary_operation): Handle polynomial
27216 element counts. Return false if we need to know the number
27217 of elements at compile time.
27218 * fold-const.c (fold_vec_perm): Punt if the number of elements
27219 isn't known at compile time.
27221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27223 * vec-perm-indices.h (vec_perm_builder): Change element type
27224 from HOST_WIDE_INT to poly_int64.
27225 (vec_perm_indices::element_type): Update accordingly.
27226 (vec_perm_indices::clamp): Handle polynomial element_types.
27227 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27228 (vec_perm_indices::all_in_range_p): Likewise.
27229 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
27231 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
27232 polynomial vec_perm_indices element types.
27233 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
27234 * fold-const.c (fold_vec_perm): Likewise.
27235 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
27236 * tree-vect-generic.c (lower_vec_perm): Likewise.
27237 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27238 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
27239 element type to HOST_WIDE_INT.
27241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27242 Alan Hayward <alan.hayward@arm.com>
27243 David Sherwood <david.sherwood@arm.com>
27245 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
27246 rather than an int. Use plus_constant.
27247 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
27248 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
27250 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27251 Alan Hayward <alan.hayward@arm.com>
27252 David Sherwood <david.sherwood@arm.com>
27254 * calls.c (emit_call_1, expand_call): Change struct_value_size from
27255 a HOST_WIDE_INT to a poly_int64.
27257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27258 Alan Hayward <alan.hayward@arm.com>
27259 David Sherwood <david.sherwood@arm.com>
27261 * calls.c (load_register_parameters): Cope with polynomial
27262 mode sizes. Require a constant size for BLKmode parameters
27263 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
27264 forces a parameter to be padded at the lsb end in order to
27265 fill a complete number of words, require the parameter size
27266 to be ordered wrt UNITS_PER_WORD.
27268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27269 Alan Hayward <alan.hayward@arm.com>
27270 David Sherwood <david.sherwood@arm.com>
27272 * reload1.c (spill_stack_slot_width): Change element type
27273 from unsigned int to poly_uint64_pod.
27274 (alter_reg): Treat mode sizes as polynomial.
27276 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27277 Alan Hayward <alan.hayward@arm.com>
27278 David Sherwood <david.sherwood@arm.com>
27280 * reload.c (complex_word_subreg_p): New function.
27281 (reload_inner_reg_of_subreg, push_reload): Use it.
27283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27284 Alan Hayward <alan.hayward@arm.com>
27285 David Sherwood <david.sherwood@arm.com>
27287 * lra-constraints.c (process_alt_operands): Reject matched
27288 operands whose sizes aren't ordered.
27289 (match_reload): Refer to this check here.
27291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27292 Alan Hayward <alan.hayward@arm.com>
27293 David Sherwood <david.sherwood@arm.com>
27295 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
27296 that the mode size is in the set {1, 2, 4, 8, 16}.
27298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27299 Alan Hayward <alan.hayward@arm.com>
27300 David Sherwood <david.sherwood@arm.com>
27302 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
27303 Use plus_constant instead of gen_rtx_PLUS.
27305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27306 Alan Hayward <alan.hayward@arm.com>
27307 David Sherwood <david.sherwood@arm.com>
27309 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
27310 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
27311 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
27312 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
27313 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
27314 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
27315 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
27316 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
27317 * config/i386/i386.c (ix86_push_rounding): ...this new function.
27318 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
27320 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
27321 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
27322 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
27323 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
27324 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
27325 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
27326 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
27327 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
27328 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
27329 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
27331 * expr.c (emit_move_resolve_push): Treat the input and result
27332 of PUSH_ROUNDING as a poly_int64.
27333 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
27334 (emit_push_insn): Likewise.
27335 * lra-eliminations.c (mark_not_eliminable): Likewise.
27336 * recog.c (push_operand): Likewise.
27337 * reload1.c (elimination_effects): Likewise.
27338 * rtlanal.c (nonzero_bits1): Likewise.
27339 * calls.c (store_one_arg): Likewise. Require the padding to be
27340 known at compile time.
27342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27343 Alan Hayward <alan.hayward@arm.com>
27344 David Sherwood <david.sherwood@arm.com>
27346 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
27347 Use plus_constant instead of gen_rtx_PLUS.
27349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27350 Alan Hayward <alan.hayward@arm.com>
27351 David Sherwood <david.sherwood@arm.com>
27353 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
27354 rather than an int.
27356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27357 Alan Hayward <alan.hayward@arm.com>
27358 David Sherwood <david.sherwood@arm.com>
27360 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
27361 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
27362 via stack temporaries. Treat the mode size as polynomial too.
27364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27365 Alan Hayward <alan.hayward@arm.com>
27366 David Sherwood <david.sherwood@arm.com>
27368 * expr.c (expand_expr_real_2): When handling conversions involving
27369 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
27370 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
27371 as a poly_uint64 too.
27373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27374 Alan Hayward <alan.hayward@arm.com>
27375 David Sherwood <david.sherwood@arm.com>
27377 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
27379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27380 Alan Hayward <alan.hayward@arm.com>
27381 David Sherwood <david.sherwood@arm.com>
27383 * combine.c (can_change_dest_mode): Handle polynomial
27384 REGMODE_NATURAL_SIZE.
27385 * expmed.c (store_bit_field_1): Likewise.
27386 * expr.c (store_constructor): Likewise.
27387 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
27388 and polynomial REGMODE_NATURAL_SIZE.
27389 (gen_lowpart_common): Likewise.
27390 * reginfo.c (record_subregs_of_mode): Likewise.
27391 * rtlanal.c (read_modify_subreg_p): Likewise.
27393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27394 Alan Hayward <alan.hayward@arm.com>
27395 David Sherwood <david.sherwood@arm.com>
27397 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
27398 numbers of elements.
27400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27401 Alan Hayward <alan.hayward@arm.com>
27402 David Sherwood <david.sherwood@arm.com>
27404 * match.pd: Cope with polynomial numbers of vector elements.
27406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27407 Alan Hayward <alan.hayward@arm.com>
27408 David Sherwood <david.sherwood@arm.com>
27410 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
27411 in a POINTER_PLUS_EXPR.
27413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27414 Alan Hayward <alan.hayward@arm.com>
27415 David Sherwood <david.sherwood@arm.com>
27417 * omp-simd-clone.c (simd_clone_subparts): New function.
27418 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27419 (ipa_simd_modify_function_body): Likewise.
27421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27422 Alan Hayward <alan.hayward@arm.com>
27423 David Sherwood <david.sherwood@arm.com>
27425 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
27426 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
27427 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
27428 (expand_vector_condition, vector_element): Likewise.
27429 (subparts_gt): New function.
27430 (get_compute_type): Use subparts_gt.
27431 (count_type_subparts): Delete.
27432 (expand_vector_operations_1): Use subparts_gt instead of
27433 count_type_subparts.
27435 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27436 Alan Hayward <alan.hayward@arm.com>
27437 David Sherwood <david.sherwood@arm.com>
27439 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
27440 (vect_compile_time_alias): ...this new function. Do the calculation
27441 on poly_ints rather than trees.
27442 (vect_prune_runtime_alias_test_list): Update call accordingly.
27444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27445 Alan Hayward <alan.hayward@arm.com>
27446 David Sherwood <david.sherwood@arm.com>
27448 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
27450 (vect_schedule_slp_instance): Likewise.
27452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27453 Alan Hayward <alan.hayward@arm.com>
27454 David Sherwood <david.sherwood@arm.com>
27456 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
27457 constant and extern definitions for variable-length vectors.
27458 (vect_get_constant_vectors): Note that the number of units
27459 is known to be constant.
27461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27462 Alan Hayward <alan.hayward@arm.com>
27463 David Sherwood <david.sherwood@arm.com>
27465 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
27466 of units as polynomial. Choose between WIDE and NARROW based
27469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27470 Alan Hayward <alan.hayward@arm.com>
27471 David Sherwood <david.sherwood@arm.com>
27473 * tree-vect-stmts.c (simd_clone_subparts): New function.
27474 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
27476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27477 Alan Hayward <alan.hayward@arm.com>
27478 David Sherwood <david.sherwood@arm.com>
27480 * tree-vect-stmts.c (vectorizable_call): Treat the number of
27481 vectors as polynomial. Use build_index_vector for
27482 IFN_GOMP_SIMD_LANE.
27484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27485 Alan Hayward <alan.hayward@arm.com>
27486 David Sherwood <david.sherwood@arm.com>
27488 * tree-vect-stmts.c (get_load_store_type): Treat the number of
27489 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
27490 for variable-length vectors.
27491 (vectorizable_mask_load_store): Treat the number of units as
27492 polynomial, asserting that it is constant if the condition has
27493 already been enforced.
27494 (vectorizable_store, vectorizable_load): Likewise.
27496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27497 Alan Hayward <alan.hayward@arm.com>
27498 David Sherwood <david.sherwood@arm.com>
27500 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
27501 of units as polynomial. Punt if we can't tell at compile time
27502 which vector contains the final result.
27504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27505 Alan Hayward <alan.hayward@arm.com>
27506 David Sherwood <david.sherwood@arm.com>
27508 * tree-vect-loop.c (vectorizable_induction): Treat the number
27509 of units as polynomial. Punt on SLP inductions. Use an integer
27510 VEC_SERIES_EXPR for variable-length integer reductions. Use a
27511 cast of such a series for variable-length floating-point
27514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27515 Alan Hayward <alan.hayward@arm.com>
27516 David Sherwood <david.sherwood@arm.com>
27518 * tree.h (build_index_vector): Declare.
27519 * tree.c (build_index_vector): New function.
27520 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
27521 of units as polynomial, forcibly converting it to a constant if
27522 vectorizable_reduction has already enforced the condition.
27523 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
27524 to create a {1,2,3,...} vector.
27525 (vectorizable_reduction): Treat the number of units as polynomial.
27526 Choose vectype_in based on the largest scalar element size rather
27527 than the smallest number of units. Enforce the restrictions
27530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27531 Alan Hayward <alan.hayward@arm.com>
27532 David Sherwood <david.sherwood@arm.com>
27534 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
27535 number of units as polynomial.
27537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27538 Alan Hayward <alan.hayward@arm.com>
27539 David Sherwood <david.sherwood@arm.com>
27541 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
27542 * target.def (autovectorize_vector_sizes): Return the vector sizes
27543 by pointer, using vector_sizes rather than a bitmask.
27544 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
27545 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
27546 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
27548 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
27549 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
27550 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
27551 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
27552 * omp-general.c (omp_max_vf): Likewise.
27553 * omp-low.c (omp_clause_aligned_alignment): Likewise.
27554 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
27555 * tree-vect-loop.c (vect_analyze_loop): Likewise.
27556 * tree-vect-slp.c (vect_slp_bb): Likewise.
27557 * doc/tm.texi: Regenerate.
27558 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
27560 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
27561 the vector size as a poly_uint64 rather than an unsigned int.
27562 (current_vector_size): Change from an unsigned int to a poly_uint64.
27563 (get_vectype_for_scalar_type): Update accordingly.
27564 * tree.h (build_truth_vector_type): Take the size and number of
27565 units as a poly_uint64 rather than an unsigned int.
27566 (build_vector_type): Add a temporary overload that takes
27567 the number of units as a poly_uint64 rather than an unsigned int.
27568 * tree.c (make_vector_type): Likewise.
27569 (build_truth_vector_type): Take the number of units as a poly_uint64
27570 rather than an unsigned int.
27572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27573 Alan Hayward <alan.hayward@arm.com>
27574 David Sherwood <david.sherwood@arm.com>
27576 * target.def (get_mask_mode): Take the number of units and length
27577 as poly_uint64s rather than unsigned ints.
27578 * targhooks.h (default_get_mask_mode): Update accordingly.
27579 * targhooks.c (default_get_mask_mode): Likewise.
27580 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
27581 * doc/tm.texi: Regenerate.
27583 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27584 Alan Hayward <alan.hayward@arm.com>
27585 David Sherwood <david.sherwood@arm.com>
27587 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
27588 * omp-general.c (omp_max_vf): Likewise.
27589 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
27590 (expand_omp_simd): Handle polynomial safelen.
27591 * omp-low.c (omplow_simd_context): Add a default constructor.
27592 (omplow_simd_context::max_vf): Change from int to poly_uint64.
27593 (lower_rec_simd_input_clauses): Update accordingly.
27594 (lower_rec_input_clauses): Likewise.
27596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27597 Alan Hayward <alan.hayward@arm.com>
27598 David Sherwood <david.sherwood@arm.com>
27600 * tree-vectorizer.h (vect_nunits_for_cost): New function.
27601 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
27602 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
27603 (vect_analyze_slp_cost): Likewise.
27604 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
27605 (vect_model_load_cost): Likewise.
27607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27608 Alan Hayward <alan.hayward@arm.com>
27609 David Sherwood <david.sherwood@arm.com>
27611 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
27612 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
27613 from an unsigned int * to a poly_uint64_pod *.
27614 (calculate_unrolling_factor): New function.
27615 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
27617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27618 Alan Hayward <alan.hayward@arm.com>
27619 David Sherwood <david.sherwood@arm.com>
27621 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
27622 from an unsigned int to a poly_uint64.
27623 (_loop_vec_info::slp_unrolling_factor): Likewise.
27624 (_loop_vec_info::vectorization_factor): Change from an int
27626 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
27627 (vect_get_num_vectors): New function.
27628 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
27629 (vect_get_num_copies): Use vect_get_num_vectors.
27630 (vect_analyze_data_ref_dependences): Change max_vf from an int *
27631 to an unsigned int *.
27632 (vect_analyze_data_refs): Change min_vf from an int * to a
27634 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27635 than an unsigned HOST_WIDE_INT.
27636 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
27637 (vect_analyze_data_ref_dependence): Change max_vf from an int *
27638 to an unsigned int *.
27639 (vect_analyze_data_ref_dependences): Likewise.
27640 (vect_compute_data_ref_alignment): Handle polynomial vf.
27641 (vect_enhance_data_refs_alignment): Likewise.
27642 (vect_prune_runtime_alias_test_list): Likewise.
27643 (vect_shift_permute_load_chain): Likewise.
27644 (vect_supportable_dr_alignment): Likewise.
27645 (dependence_distance_ge_vf): Take the vectorization factor as a
27646 poly_uint64 rather than an unsigned HOST_WIDE_INT.
27647 (vect_analyze_data_refs): Change min_vf from an int * to a
27649 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
27650 vfm1 as a poly_uint64 rather than an int. Make the same change
27651 for the returned bound_scalar.
27652 (vect_gen_vector_loop_niters): Handle polynomial vf.
27653 (vect_do_peeling): Likewise. Update call to
27654 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
27655 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
27657 * tree-vect-loop.c (vect_determine_vectorization_factor)
27658 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
27659 (vect_get_known_peeling_cost): Likewise.
27660 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
27661 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
27662 (vect_transform_loop): Likewise. Use the lowest possible VF when
27663 updating the upper bounds of the loop.
27664 (vect_min_worthwhile_factor): Make static. Return an unsigned int
27665 rather than an int.
27666 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
27667 polynomial unroll factors.
27668 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
27669 (vect_make_slp_decision): Likewise.
27670 (vect_supported_load_permutation_p): Likewise, and polynomial
27672 (vect_analyze_slp_cost): Handle polynomial vf.
27673 (vect_slp_analyze_node_operations): Likewise.
27674 (vect_slp_analyze_bb_1): Likewise.
27675 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
27676 than an unsigned HOST_WIDE_INT.
27677 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
27678 (vectorizable_load): Handle polynomial vf.
27679 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
27681 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
27683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27684 Alan Hayward <alan.hayward@arm.com>
27685 David Sherwood <david.sherwood@arm.com>
27687 * match.pd: Handle bit operations involving three constants
27688 and try to fold one pair.
27690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
27692 * tree-vect-loop-manip.c: Include gimple-fold.h.
27693 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
27694 niters_maybe_zero parameters. Handle other cases besides a step of 1.
27695 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
27696 Add a path that uses a step of VF instead of 1, but disable it
27698 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
27699 and niters_no_overflow parameters. Update calls to
27700 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
27701 Create a new SSA name if the latter choses to use a ste other
27702 than zero, and return it via niters_vector_mult_vf_var.
27703 * tree-vect-loop.c (vect_transform_loop): Update calls to
27704 vect_do_peeling, vect_gen_vector_loop_niters and
27705 slpeel_make_loop_iterate_ntimes.
27706 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
27707 (vect_gen_vector_loop_niters): Update declarations after above changes.
27709 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
27711 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
27712 128-bit round to integer instructions.
27713 (ceil<mode>2): Likewise.
27714 (btrunc<mode>2): Likewise.
27715 (round<mode>2): Likewise.
27717 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27719 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
27720 unaligned VSX load/store on P8/P9.
27721 (expand_block_clear): Allow the use of unaligned VSX
27722 load/store on P8/P9.
27724 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27726 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
27728 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
27729 swap associated with both a load and a store.
27731 2018-01-02 Andrew Waterman <andrew@sifive.com>
27733 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
27734 * config/riscv/riscv.md (clear_cache): Use it.
27736 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
27738 * web.c: Remove out-of-date comment.
27740 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27742 * expr.c (fixup_args_size_notes): Check that any existing
27743 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
27744 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
27745 (emit_single_push_insn): ...here.
27747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27749 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
27750 (const_vector_encoded_nelts): New function.
27751 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
27752 (const_vector_int_elt, const_vector_elt): Declare.
27753 * emit-rtl.c (const_vector_int_elt_1): New function.
27754 (const_vector_elt): Likewise.
27755 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
27756 of CONST_VECTOR_ELT.
27758 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27760 * expr.c: Include rtx-vector-builder.h.
27761 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
27762 directly on the tree encoding.
27763 (const_vector_from_tree): Likewise.
27764 * optabs.c: Include rtx-vector-builder.h.
27765 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
27766 sequence of "u" values.
27767 * vec-perm-indices.c: Include rtx-vector-builder.h.
27768 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
27769 directly on the vec_perm_indices encoding.
27771 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27773 * doc/rtl.texi (const_vector): Describe new encoding scheme.
27774 * Makefile.in (OBJS): Add rtx-vector-builder.o.
27775 * rtx-vector-builder.h: New file.
27776 * rtx-vector-builder.c: Likewise.
27777 * rtl.h (rtx_def::u2): Add a const_vector field.
27778 (CONST_VECTOR_NPATTERNS): New macro.
27779 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
27780 (CONST_VECTOR_DUPLICATE_P): Likewise.
27781 (CONST_VECTOR_STEPPED_P): Likewise.
27782 (CONST_VECTOR_ENCODED_ELT): Likewise.
27783 (const_vec_duplicate_p): Check for a duplicated vector encoding.
27784 (unwrap_const_vec_duplicate): Likewise.
27785 (const_vec_series_p): Check for a non-duplicated vector encoding.
27786 Say that the function only returns true for integer vectors.
27787 * emit-rtl.c: Include rtx-vector-builder.h.
27788 (gen_const_vec_duplicate_1): Delete.
27789 (gen_const_vector): Call gen_const_vec_duplicate instead of
27790 gen_const_vec_duplicate_1.
27791 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
27792 (gen_const_vec_duplicate): Use rtx_vector_builder.
27793 (gen_const_vec_series): Likewise.
27794 (gen_rtx_CONST_VECTOR): Likewise.
27795 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
27796 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27797 Build a new vector rather than modifying a CONST_VECTOR in-place.
27798 (handle_special_swappables): Update call accordingly.
27799 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
27800 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
27801 Build a new vector rather than modifying a CONST_VECTOR in-place.
27802 (handle_special_swappables): Update call accordingly.
27804 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27806 * simplify-rtx.c (simplify_const_binary_operation): Use
27807 CONST_VECTOR_ELT instead of XVECEXP.
27809 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27811 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
27812 the selector elements to be different from the data elements
27813 if the selector is a VECTOR_CST.
27814 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
27815 ssizetype for the selector.
27817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27819 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
27820 before testing each element individually.
27821 * tree-vect-generic.c (lower_vec_perm): Likewise.
27823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27825 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
27826 * selftest-run-tests.c (selftest::run_tests): Call it.
27827 * vector-builder.h (vector_builder::operator ==): New function.
27828 (vector_builder::operator !=): Likewise.
27829 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
27830 (vec_perm_indices::all_from_input_p): New function.
27831 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
27832 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
27833 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
27834 instead of reading the VECTOR_CST directly. Detect whether both
27835 vector inputs are the same before constructing the vec_perm_indices,
27836 and update the number of inputs argument accordingly. Use the
27837 utility functions added above. Only construct sel2 if we need to.
27839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27841 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
27842 the broadcast of the low byte.
27843 (expand_mult_highpart): Use an explicit encoding for the permutes.
27844 * optabs-query.c (can_mult_highpart_p): Likewise.
27845 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
27846 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27847 (vectorizable_bswap): Likewise.
27848 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
27849 explicit encoding for the power-of-2 permutes.
27850 (vect_permute_store_chain): Likewise.
27851 (vect_grouped_load_supported): Likewise.
27852 (vect_permute_load_chain): Likewise.
27854 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27856 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
27857 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
27858 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
27859 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
27860 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
27861 (vect_gen_perm_mask_any): Likewise.
27863 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27865 * int-vector-builder.h: New file.
27866 * vec-perm-indices.h: Include int-vector-builder.h.
27867 (vec_perm_indices): Redefine as an int_vector_builder.
27868 (auto_vec_perm_indices): Delete.
27869 (vec_perm_builder): Redefine as a stand-alone class.
27870 (vec_perm_indices::vec_perm_indices): New function.
27871 (vec_perm_indices::clamp): Likewise.
27872 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
27873 (vec_perm_indices::new_vector): New function.
27874 (vec_perm_indices::new_expanded_vector): Update for new
27875 vec_perm_indices class.
27876 (vec_perm_indices::rotate_inputs): New function.
27877 (vec_perm_indices::all_in_range_p): Operate directly on the
27878 encoded form, without computing elided elements.
27879 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
27880 encoding. Update for new vec_perm_indices class.
27881 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
27882 the given vec_perm_builder.
27883 (expand_vec_perm_var): Update vec_perm_builder constructor.
27884 (expand_mult_highpart): Use vec_perm_builder instead of
27885 auto_vec_perm_indices.
27886 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
27887 vec_perm_indices instead of auto_vec_perm_indices. Use a single
27888 or double series encoding as appropriate.
27889 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
27890 vec_perm_indices instead of auto_vec_perm_indices.
27891 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27892 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27893 (vect_permute_store_chain): Likewise.
27894 (vect_grouped_load_supported): Likewise.
27895 (vect_permute_load_chain): Likewise.
27896 (vect_shift_permute_load_chain): Likewise.
27897 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27898 (vect_transform_slp_perm_load): Likewise.
27899 (vect_schedule_slp_instance): Likewise.
27900 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27901 (vectorizable_mask_load_store): Likewise.
27902 (vectorizable_bswap): Likewise.
27903 (vectorizable_store): Likewise.
27904 (vectorizable_load): Likewise.
27905 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
27906 vec_perm_indices instead of auto_vec_perm_indices. Use
27907 tree_to_vec_perm_builder to read the vector from a tree.
27908 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
27909 vec_perm_builder instead of a vec_perm_indices.
27910 (have_whole_vector_shift): Use vec_perm_builder and
27911 vec_perm_indices instead of auto_vec_perm_indices. Leave the
27912 truncation to calc_vec_perm_mask_for_shift.
27913 (vect_create_epilog_for_reduction): Likewise.
27914 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
27915 from auto_vec_perm_indices to vec_perm_indices.
27916 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27917 instead of changing individual elements.
27918 (aarch64_vectorize_vec_perm_const): Use new_vector to install
27919 the vector in d.perm.
27920 * config/arm/arm.c (expand_vec_perm_d::perm): Change
27921 from auto_vec_perm_indices to vec_perm_indices.
27922 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
27923 instead of changing individual elements.
27924 (arm_vectorize_vec_perm_const): Use new_vector to install
27925 the vector in d.perm.
27926 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
27927 Update vec_perm_builder constructor.
27928 (rs6000_expand_interleave): Likewise.
27929 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
27930 (rs6000_expand_interleave): Likewise.
27932 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27934 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
27935 to qimode could truncate the indices.
27936 * optabs.c (expand_vec_perm_var): Likewise.
27938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27940 * Makefile.in (OBJS): Add vec-perm-indices.o.
27941 * vec-perm-indices.h: New file.
27942 * vec-perm-indices.c: Likewise.
27943 * target.h (vec_perm_indices): Replace with a forward class
27945 (auto_vec_perm_indices): Move to vec-perm-indices.h.
27946 * optabs.h: Include vec-perm-indices.h.
27947 (expand_vec_perm): Delete.
27948 (selector_fits_mode_p, expand_vec_perm_var): Declare.
27949 (expand_vec_perm_const): Declare.
27950 * target.def (vec_perm_const_ok): Replace with...
27951 (vec_perm_const): ...this new hook.
27952 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
27953 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
27954 * doc/tm.texi: Regenerate.
27955 * optabs.def (vec_perm_const): Delete.
27956 * doc/md.texi (vec_perm_const): Likewise.
27957 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
27958 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
27959 expand_vec_perm for constant permutation vectors. Assert that
27960 the mode of variable permutation vectors is the integer equivalent
27961 of the mode that is being permuted.
27962 * optabs-query.h (selector_fits_mode_p): Declare.
27963 * optabs-query.c: Include vec-perm-indices.h.
27964 (selector_fits_mode_p): New function.
27965 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27966 is defined, instead of checking whether the vec_perm_const_optab
27967 exists. Use targetm.vectorize.vec_perm_const instead of
27968 targetm.vectorize.vec_perm_const_ok. Check whether the indices
27969 fit in the vector mode before using a variable permute.
27970 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27971 vec_perm_indices instead of an rtx.
27972 (expand_vec_perm): Replace with...
27973 (expand_vec_perm_const): ...this new function. Take the selector
27974 as a vec_perm_indices rather than an rtx. Also take the mode of
27975 the selector. Update call to shift_amt_for_vec_perm_mask.
27976 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27977 Use vec_perm_indices::new_expanded_vector to expand the original
27978 selector into bytes. Check whether the indices fit in the vector
27979 mode before using a variable permute.
27980 (expand_vec_perm_var): Make global.
27981 (expand_mult_highpart): Use expand_vec_perm_const.
27982 * fold-const.c: Includes vec-perm-indices.h.
27983 * tree-ssa-forwprop.c: Likewise.
27984 * tree-vect-data-refs.c: Likewise.
27985 * tree-vect-generic.c: Likewise.
27986 * tree-vect-loop.c: Likewise.
27987 * tree-vect-slp.c: Likewise.
27988 * tree-vect-stmts.c: Likewise.
27989 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27991 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27992 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27993 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27994 (aarch64_vectorize_vec_perm_const): ...this new function.
27995 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27996 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27997 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27998 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27999 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28000 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28001 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
28003 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
28004 check for NEON modes.
28005 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
28006 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
28007 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
28008 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
28010 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
28011 the old VEC_PERM_CONST conditions.
28012 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
28013 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
28014 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
28015 (ia64_vectorize_vec_perm_const_ok): Merge into...
28016 (ia64_vectorize_vec_perm_const): ...this new function.
28017 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
28018 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
28019 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
28020 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
28021 * config/mips/mips.c (mips_expand_vec_perm_const)
28022 (mips_vectorize_vec_perm_const_ok): Merge into...
28023 (mips_vectorize_vec_perm_const): ...this new function.
28024 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
28025 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
28026 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
28027 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
28028 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
28029 (rs6000_expand_vec_perm_const): Delete.
28030 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
28032 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28033 (altivec_expand_vec_perm_const_le): Take each operand individually.
28034 Operate on constant selectors rather than rtxes.
28035 (altivec_expand_vec_perm_const): Likewise. Update call to
28036 altivec_expand_vec_perm_const_le.
28037 (rs6000_expand_vec_perm_const): Delete.
28038 (rs6000_vectorize_vec_perm_const_ok): Delete.
28039 (rs6000_vectorize_vec_perm_const): New function.
28040 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28041 an element count and rtx array.
28042 (rs6000_expand_extract_even): Update call accordingly.
28043 (rs6000_expand_interleave): Likewise.
28044 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
28045 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
28046 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
28047 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
28048 (rs6000_expand_vec_perm_const): Delete.
28049 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
28050 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28051 (altivec_expand_vec_perm_const_le): Take each operand individually.
28052 Operate on constant selectors rather than rtxes.
28053 (altivec_expand_vec_perm_const): Likewise. Update call to
28054 altivec_expand_vec_perm_const_le.
28055 (rs6000_expand_vec_perm_const): Delete.
28056 (rs6000_vectorize_vec_perm_const_ok): Delete.
28057 (rs6000_vectorize_vec_perm_const): New function. Remove stray
28058 reference to the SPE evmerge intructions.
28059 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
28060 an element count and rtx array.
28061 (rs6000_expand_extract_even): Update call accordingly.
28062 (rs6000_expand_interleave): Likewise.
28063 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
28064 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
28066 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
28068 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28070 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
28071 vector mode and that that mode matches the mode of the data
28073 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
28074 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
28075 directly using expand_vec_perm_1 when forcing selectors into
28077 (expand_vec_perm_var): New function, split out from expand_vec_perm.
28079 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28081 * optabs-query.h (can_vec_perm_p): Delete.
28082 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
28083 * optabs-query.c (can_vec_perm_p): Split into...
28084 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
28085 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
28086 particular selector is valid.
28087 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
28088 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
28089 (vect_grouped_load_supported): Likewise.
28090 (vect_shift_permute_load_chain): Likewise.
28091 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
28092 (vect_transform_slp_perm_load): Likewise.
28093 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
28094 (vectorizable_bswap): Likewise.
28095 (vect_gen_perm_mask_checked): Likewise.
28096 * fold-const.c (fold_ternary_loc): Likewise. Don't take
28097 implementations of variable permutation vectors into account
28098 when deciding which selector to use.
28099 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
28100 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
28101 with a false third argument.
28102 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
28103 to test whether the constant selector is valid and can_vec_perm_var_p
28104 to test whether a variable selector is valid.
28106 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28108 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
28109 * optabs-query.c (can_vec_perm_p): Likewise.
28110 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
28111 instead of vec_perm_indices.
28112 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
28113 (vect_gen_perm_mask_checked): Likewise,
28114 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
28115 (vect_gen_perm_mask_checked): Likewise,
28117 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
28119 * optabs-query.h (qimode_for_vec_perm): Declare.
28120 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
28121 (qimode_for_vec_perm): ...this new function.
28122 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
28124 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
28126 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
28127 does not have a conditional at the top.
28129 2018-01-02 Richard Biener <rguenther@suse.de>
28131 * ipa-inline.c (big_speedup_p): Fix expression.
28133 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28136 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
28139 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
28143 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
28144 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
28145 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
28146 cond_taken_branch_cost 3->4.
28148 2018-01-01 Jakub Jelinek <jakub@redhat.com>
28150 PR tree-optimization/83581
28151 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
28152 TODO_cleanup_cfg if any changes have been made.
28154 PR middle-end/83608
28155 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
28156 convert_modes if target mode has the right side, but different mode
28159 PR middle-end/83609
28160 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
28161 last argument when extracting from CONCAT. If either from_real or
28162 from_imag is NULL, use expansion through memory. If result is not
28163 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
28164 the parts directly to inner mode, if even that fails, use expansion
28167 PR middle-end/83623
28168 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
28169 check for bswap in mode rather than HImode and use that in expand_unop
28172 Copyright (C) 2018 Free Software Foundation, Inc.
28174 Copying and distribution of this file, with or without modification,
28175 are permitted in any medium without royalty provided the copyright
28176 notice and this notice are preserved.