1 2018-05-03 Tom de Vries <tom@codesourcery.com>
4 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5 dump files): Add offload-tree.
7 2018-05-03 Richard Biener <rguenther@suse.de>
9 PR tree-optimization/85615
10 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
11 to loops not nested in BBs loop father to avoid creating multi-entry
14 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16 PR tree-optimization/70291
17 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
18 arguments. Change return type to tree. Emit libcall as a new
19 statement rather than replacing existing one when inplace_p is true.
20 (expand_complex_multiplication_components): New function.
21 (expand_complex_multiplication): Expand floating-point complex
22 multiplication using the above.
23 (expand_complex_division): Rename inner_type parameter to type.
24 Update expand_complex_libcall call-site.
25 (expand_complex_operations_1): Update expand_complex_multiplication
26 and expand_complex_division call-sites.
28 2018-05-02 Jakub Jelinek <jakub@redhat.com>
31 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
32 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
33 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
34 the highest significant bit of the shift count mask is clear. In
35 check whether and[sq]i3 is needed verify that all significant bits
36 of the shift count other than the highest are set.
38 2018-05-02 Tom de Vries <tom@codesourcery.com>
41 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
42 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
43 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
44 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
45 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
46 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
47 __builtin_goacc_parlevel_size.
49 2018-05-02 Richard Biener <rguenther@suse.de>
51 PR tree-optimization/85597
52 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
53 do not use split vect_get_vec_defs call but call vect_get_slp_defs
56 2018-05-02 Tom de Vries <tom@codesourcery.com>
59 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
60 dump files): Add ltrans-tree.
62 2018-05-02 Tom de Vries <tom@codesourcery.com>
65 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
66 dump files): Add wpa-ipa.
68 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
70 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
71 powerpc*-*-linux*paired* target.
72 * config/rs6000/750cl.h: Delete.
73 * config/rs6000/paired.h: Delete.
74 * config/rs6000/paired.md: Delete.
75 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
77 * config/rs6000/rs6000-builtin.def: Remove paired float support.
78 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
79 comment. Remove paired float support.
80 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
81 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
83 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
84 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
86 * config/rs6000/rs6000.c: Remove paired float support.
87 (paired_expand_vector_init, paired_expand_vector_move,
88 paired_emit_vector_compare, paired_emit_vector_cond_expr,
89 (paired_expand_lv_builtin, paired_expand_stv_builtin,
90 paired_expand_builtin, paired_expand_predicate_builtin,
91 paired_init_builtins): Delete.
92 * config/rs6000/rs6000.h: Remove paired float support.
93 * config/rs6000/rs6000.md: Remove paired float support.
94 (move_from_CR_ov_bit): Delete.
95 * config/rs6000/rs6000.opt (mpaired): Delete.
96 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
97 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
99 2018-05-02 Richard Biener <rguenther@suse.de>
102 * gimplify.c (gimplify_save_expr): When in SSA form allow
103 SAVE_EXPRs to compute to SSA vars.
105 2018-05-02 Jakub Jelinek <jakub@redhat.com>
108 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
109 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
110 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
111 clobber operands[2], instead use a new pseudo. Formatting fixes.
113 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
115 PR tree-optimization/85586
116 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
117 exit early for statements in the same group if the accesses are
120 2018-05-02 Tom de Vries <tom@codesourcery.com>
123 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
126 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
128 PR tree-optimization/85143
129 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
131 2018-05-01 Tom de Vries <tom@codesourcery.com>
134 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
135 not found" error message.
137 2018-05-01 Tom de Vries <tom@codesourcery.com>
140 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
141 * vec.c (test_ordered_remove_if): New function.
142 (vec_c_tests): Call test_ordered_remove_if.
143 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
144 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
145 * tree-vect-patterns.c (vect_pattern_recog_1): Use
146 VEC_ORDERED_REMOVE_IF.
148 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
150 PR tree-optimization/82665
151 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
152 pointer subtraction where arguments come from a memchr call.
154 2018-05-01 Jakub Jelinek <jakub@redhat.com>
156 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
157 --push-state --as-needed and --pop-state instead of --as-needed and
158 --no-as-needed if ld supports it.
159 * configure: Regenerated.
162 * doc/install.texi2html: Replace _002d with - and _002a with * in
163 generated html files using sed.
165 2018-04-30 David Malcolm <dmalcolm@redhat.com>
168 * gcc-rich-location.c (blank_line_before_p): New function.
169 (use_new_line): New function.
170 (gcc_rich_location::add_fixit_insert_formatted): New function.
171 * gcc-rich-location.h
172 (gcc_rich_location::add_fixit_insert_formatted): New function.
174 2018-04-30 David Malcolm <dmalcolm@redhat.com>
176 * selftest.c (assert_streq): Rename "expected" and "actual" to
177 "val1" and "val2". Extend NULL-handling to cover both inputs
178 symmetrically, while still requiring both to be non-NULL for a pass.
179 * selftest.h (assert_streq): Rename "expected" and "actual" to
181 (ASSERT_EQ): Likewise.
182 (ASSERT_EQ_AT): Likewise.
183 (ASSERT_KNOWN_EQ): Likewise.
184 (ASSERT_KNOWN_EQ_AT): Likewise.
185 (ASSERT_NE): Likewise.
186 (ASSERT_MAYBE_NE): Likewise.
187 (ASSERT_MAYBE_NE_AT): Likewise.
188 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
189 the assertion to pass.
190 (ASSERT_STREQ_AT): Likewise.
192 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
194 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
195 interaction with -pie.
197 2018-04-30 David Malcolm <dmalcolm@redhat.com>
199 * selftest.h: Fix alphabetization of per-source-file selftest
202 2018-04-30 Jason Merrill <jason@redhat.com>
204 PR c++/61982 - dead stores to destroyed objects.
205 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
208 2018-04-30 Jason Merrill <jason@redhat.com>
210 * tree.c (build_clobber): New.
211 * tree.h: Declare it.
212 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
214 2018-04-30 David Malcolm <dmalcolm@redhat.com>
216 * diagnostic-show-locus.c (layout::layout): Update for
217 location_get_source_line returning a char_span.
218 (struct char_span): Move to input.h.
219 (struct correction): Update for fields in char_span becoming
221 (struct source_line): Update for location_get_source_line
222 returning a char_span.
223 (layout::print_line): Likewise.
224 * edit-context.c (edited_file::print_content): Likewise.
225 (edited_file::print_diff_hunk): Likewise.
226 (edited_file::print_run_of_changed_lines): Likewise.
227 (edited_file::get_num_lines): Likewise.
228 (edited_line::edited_line): Likewise.
229 * final.c (asm_show_source): Likewise.
230 * input.c (location_get_source_line): Convert return type
231 from const char * to char_span, losing the final "line_len"
233 (dump_location_info): Update for the above.
234 (get_substring_ranges_for_loc): Likewise. Use a char_span
235 when handling the literal within the line.
236 (test_reading_source_line): Update for location_get_source_line
237 returning a char_span.
238 * input.h (class char_span): Move here from
239 diagnostic-show-locus.c, converting from a struct to a class.
240 Make data members private.
241 (char_span::operator bool): New.
242 (char_span::length): New.
243 (char_span::get_buffer): New.
244 (char_span::operator[]): New.
245 (char_span::subspan): Make const.
246 (char_span::xstrdup): New.
247 (location_get_source_line): Convert return type from const char *
248 to char_span, losing the final "line_size" param.
250 2018-04-30 Jan Hubicka <jh@suse.cz>
252 * lto-wrapper.c (ltrans_priorities): New static var.
254 (run_gcc): Read priorities and if doing parallel build order
255 the Makefile by them.
257 2018-04-30 David Malcolm <dmalcolm@redhat.com>
259 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
261 2018-04-30 Richard Biener <rguenther@suse.de>
263 * tree-cfg.c (verify_address): Remove base argument, add
264 flag whether to check TREE_ADDRESSABLE and do that.
265 (verify_expr): Remove.
266 (verify_types_in_gimple_reference): Add pieces from verify_expr.
267 (verify_gimple_assign_single): Likewise.
268 (verify_gimple_switch): Likewise.
269 (verify_expr_location_1): Dereference tp once. Add (disabled)
270 piece from verify_expr.
271 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
273 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
275 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
277 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
279 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
280 (small_data_pattern): Likewise.
281 (arc_rewrite_small_data): Likewise.
282 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
283 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
284 (get_symbol_alignment): New function.
285 (legitimate_small_data_address_p): Likewise.
286 (legitimate_scaled_address): Update, call
287 legitimate_small_data_address_p.
288 (output_sdata): New static variable.
289 (arc_print_operand): Update how we handle small data operands.
290 (arc_print_operand_address): Likewise.
291 (arc_legitimate_address_p): Update, use
292 legitimate_small_data_address_p.
293 (arc_rewrite_small_data_p): Remove.
294 (arc_rewrite_small_data_1): Likewise.
295 (arc_rewrite_small_data): Likewise.
296 (small_data_pattern): Likewise.
297 (compact_sda_memory_operand): Update to use
298 legitimate_small_data_address_p and get_symbol_alignment.
299 (prepare_move_operands): Don't rewite sdata pattern.
300 (prepare_extend_operands): Remove.
301 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
303 (zero_extendqisi2): Likewise.
304 (zero_extendhisi2): Likewise.
305 (extendqihi2): Likewise.
306 (extendqisi2): Likewise.
307 (extendhisi2): Likewise.
311 * config/arc/constraints.md (Usd): Change it to memory constraint.
313 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
315 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
316 as source of std instructions.
317 * config/arc/arc.md (movsi_insn): Update pattern predicate to
318 allow 6-bit constants as source for store instructions.
319 (movdi_insn): Update instruction pattern to allow 6-bit constants
320 as source for store instructions.
322 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
324 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
326 2018-04-30 Nathan Sidwell <nathan@acm.org>
327 Sandra Loosemore <sandra@codesourcery.com>
329 * dumpfile.c (dump_open): Allow '-' for stdout.
330 * doc/invoke.texi (Developer Options): Document dump filename
331 determination early. Document stdin/stdout selection.
333 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
335 Microblaze Target: PIC data text relative
337 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
338 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
340 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
341 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
342 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
343 New addressing mode for data-text relative position indepenedent code.
344 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
345 'ADDRESS_SYMBOLIC_TXT_REL'.
346 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
347 (microblaze_legitimate_pic_operand): Exclude function calls from
348 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
349 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
351 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
352 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
353 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
354 for 'address + offset'.
355 (microblaze_expand_prologue): Add new function prologue call for
357 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
358 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
359 table in case of TARGET_PIC_DATA_TEXT_REL.
360 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
361 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
362 Add new macros 'UNSPEC_TEXT',
363 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
364 + exclude function calls from 'UNSPEC_PLT' in case of data text
366 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
367 new target hook for generating address diff vector tables in case of
369 * doc/tm.texi : Regenerate.
370 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
371 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
372 of addr diff vector generation.
373 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
374 target hook definition.
375 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
376 Add default function for generate_pic_addr_diff_vec -> flag_pic.
377 * doc/invoke.texi (Add new pic option): Add new microblaze pic
378 option for data text relative.
380 2018-04-30 Richard Biener <rguenther@suse.de>
382 * tree-chrec.h (evolution_function_is_constant_p): Remove
384 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
386 2018-04-30 Richard Biener <rguenther@suse.de>
389 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
391 2018-04-30 Richard Biener <rguenther@suse.de>
393 PR tree-optimization/28364
394 PR tree-optimization/85275
395 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
396 copying first exit test.
398 2018-04-28 Mark Wielaard <mark@klomp.org>
400 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
402 (dwarf_AT): Handle DW_AT_addr_base.
403 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
405 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
408 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
409 (*ashl<dwi>3_doubleword_mask_1): Ditto.
410 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
411 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
413 2018-04-28 Richard Biener <rguenther@suse.de>
415 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
416 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
417 to reflect use. Only add interesting stmts.
419 2018-04-27 Martin Jambor <mjambor@suse.cz>
422 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
423 the jump function allows for passing through aggregate values.
425 2018-04-27 David Malcolm <dmalcolm@redhat.com>
427 * input.h (in_system_header_at): Convert from macro to inline
429 (from_macro_expansion_at): Likewise.
430 (from_macro_definition_at): Likewise.
432 2018-04-27 Jeff Law <law@redhat.com>
434 * config.gcc: Mark tile* targets as deprecated/obsolete.
436 2018-04-27 Richard Biener <rguenther@suse.de>
438 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
441 2018-04-27 Richard Biener <rguenther@suse.de>
443 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
445 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
447 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
448 with Yd constraint. Set "preferred_for_speed" attribute from
449 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
451 (*movdi_internal): Ditto.
452 (movti_interunit splitters): Remove
453 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
454 (movdi_interunit splitters): Ditto.
455 * config/i386/constraints.md (Ye): Remove.
456 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
458 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
461 * config/aarch64/constraints.md (Usg): Limit to 31.
464 2018-04-27 Jakub Jelinek <jakub@redhat.com>
466 PR tree-optimization/85529
467 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
468 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
469 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
470 zero extension or masking of the MSB bit.
471 (optimize_range_tests): Add FIRST_BB argument, pass it through
472 to optimize_range_tests_var_bound.
473 (maybe_optimize_range_tests, reassociate_bb): Adjust
474 optimize_range_tests callers.
476 2018-04-26 Richard Biener <rguenther@suse.de>
477 Jakub Jelinek <jakub@redhat.com>
479 * cgraph.h (symbol_table): Just declare debug method here.
480 * symtab.c (symbol_table::debug): Define.
482 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
484 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
486 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
488 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
489 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
490 (*movdi_internal): Substitute Yi and Yj constraint with x
491 and Ym and Yn constraint with y constraint. Update "isa"
492 attribute and set "preferred_for_speed" attribute from
493 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
494 (*movsi_internal): Ditto.
495 (*movdf_internal): Ditto.
496 (*movsf_internal): Ditto.
497 (*zero_extendsidi2): Ditto.
498 * config/i386/sse.md (vec_set<mode>_0): Ditto.
499 (sse2_loadld): Ditto.
500 (*vec_extract<ssevecmodelower>_0): Ditto.
501 (*vec_extractv4si_0_zext_sse4): Ditto.
502 (vec_concatv2di): Ditto.
503 (*vec_dup<mode>): Ditto.
504 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
505 * config/i386/constraints.md (Yi): Remove.
510 2018-04-26 Nathan Sidwell <nathan@acm.org>
512 * dumpfile.c (dump_open): New.
513 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
514 (dump_finish): Detect stdio/stderr by value not name.
516 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
518 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
520 2018-04-26 Tom de Vries <tom@codesourcery.com>
523 * config/nvptx/nvptx.c (verify_neutering_jumps)
524 (verify_neutering_labels): New function
525 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
527 2018-04-26 Tom de Vries <tom@codesourcery.com>
530 * config/nvptx/nvptx.c (needs_neutering_p): New function.
531 (nvptx_single): Use needs_neutering_p to skip over insns that do not
534 2018-04-26 Richard Biener <rguenther@suse.de>
535 Tom de Vries <tom@codesourcery.com>
538 * lto-streamer-out.c (output_function): Fixup loops if required to match
539 discovery done in the reader.
541 2018-04-26 Richard Biener <rguenther@suse.de>
543 PR tree-optimization/85116
544 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
545 have a loop exit from the single latch predecessor. Remove
546 case of header with just condition.
547 (ch_base::copy_headers): Exclude infinite loops from any
549 (pass_ch::execute): Record exits.
551 2018-04-26 Richard Biener <rguenther@suse.de>
553 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
554 prologue cost vector and pass it to vect_get_load_cost.
555 (vect_get_peeling_costs_all_drs): Likewise.
556 (vect_peeling_hash_get_lowest_cost): Likewise.
557 (vect_enhance_data_refs_alignment): Likewise.
559 2018-04-26 Richard Biener <rguenther@suse.de>
562 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
563 checking of integer<->pointer conversions.
564 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
565 sign-/zero-extending pointer types.
566 (expand_omp_for_static_chunk): Likewise.
568 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
569 Jean Lee <xiaoyur347@gmail.com>
571 * config/mips/mips.c (mips_asan_shadow_offset): New function.
572 (TARGET_ASAN_SHADOW_OFFSET): Define.
573 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
574 true for -fsanitize=address.
576 2018-04-25 Mark Wielaard <mark@klomp.org>
578 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
581 2018-04-25 Jakub Jelinek <jakub@redhat.com>
583 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
584 than "alu", remove explicit "memory" and "imm_disp" attributes.
585 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
588 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
589 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
592 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
595 * config/i386/i386.c (ix86_expand_builtin): Change memory
596 operand to XI, extend p0 to Pmode.
597 * config/i386/i386.md: Change unspec volatile and operand
598 1 mode to XI, change operand 0 mode to P.
600 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
602 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
603 GET_MODE_MASK before any checking.
604 (nds32_can_use_bset_p): Likewise.
605 (nds32_can_use_btgl_p): Likewise.
607 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
609 * config/nds32/nds32-doubleword.md: New define_split pattern for
610 illegal register number.
612 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
614 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
616 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
618 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
620 2018-04-25 Richard Biener <rguenther@suse.de>
622 * lto-streamer.h (LTO_major_version): Bump to 8.
624 2018-04-25 Jakub Jelinek <jakub@redhat.com>
626 * BASE-VER: Set to 9.0.0.
628 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
630 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
631 in __abskf2 and __powikf2.
633 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
636 * config/aarch64/constraints.md (Usg, Usj): New constraints.
637 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
638 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
639 Use the above on operand 2. Reindent.
640 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
642 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
645 * common/config/i386/i386-common.c (ix86_handle_option): Don't
647 * config/i386/i386.opt (mcet): Removed.
648 * doc/install.texi: Remove -mcet documentation.
649 * doc/invoke.texi: Likewise.
651 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
654 * doc/install.texi: Remove -mcet from bootstrap-cet.
656 2018-04-24 Jakub Jelinek <jakub@redhat.com>
659 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
660 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
664 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
665 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
666 containing a CONST_VECTOR.
668 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
670 * doc/install.texi: Update newlib dependency for nvptx.
672 2018-04-24 Jakub Jelinek <jakub@redhat.com>
675 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
676 instead of INTVAL when shifting x left.
678 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
680 PR tree-optimization/85478
681 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
682 vect_grouped_store_supported for single element vectors.
684 2018-04-24 Richard Biener <rguenther@suse.de>
687 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
688 load cost increase to the case of non-constant step.
690 2018-04-24 Jakub Jelinek <jakub@redhat.com>
693 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
694 destination if any_malformed_asm.
696 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
699 * expr.c (store_field): In the bitfield case, if the value comes from
700 a function call and is returned in registers by means of a PARALLEL,
701 do not change the mode of the temporary unless BLKmode and VOIDmode.
703 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
705 PR rtl-optimization/85423
706 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
707 dependencies to debug insns when the previous insn is non-debug.
709 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
711 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
712 enums into a single definition.
713 (fls): Fix predicates and printing.
716 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
718 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
719 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
720 and short u6 immediate.
721 (check_if_valid_sleep_operand): Remove.
722 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
724 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
726 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
727 flag_always_save_lp condition.
728 * config/nds32/nds32.opt (malways-save-lp): New option.
730 2018-04-22 Shiva Chen <shiva0217@gmail.com>
732 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
733 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
734 * config/nds32/nds32.h
735 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
736 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
738 2018-04-22 Shiva Chen <shiva0217@gmail.com>
740 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
741 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
743 2018-04-22 Shiva Chen <shiva0217@gmail.com>
744 Chung-Ju Wu <jasonwucj@gmail.com>
746 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
748 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
749 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
751 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
753 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
755 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
757 * config/nds32/nds32-protos.h (nds32_data_alignment,
758 nds32_local_alignment): Declare.
759 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
760 nds32_local_alignment): New functions.
761 (TARGET_CONSTANT_ALIGNMENT): Define.
762 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
764 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
766 * config/nds32/nds32.c
767 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
768 (TARGET_MODES_TIEABLE_P): Likewise.
770 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
772 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
775 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
776 Chung-Ju Wu <jasonwucj@gmail.com>
778 * config/nds32/constants.md (unspec_volatile_element): Add enum values
779 for unaligned access.
780 * config/nds32/nds32-intrinsic.c: Implementation of expanding
782 * config/nds32/nds32-intrinsic.md: Likewise.
783 * config/nds32/nds32_intrinsic.h: Likewise.
784 * config/nds32/nds32.h (nds32_builtins): Likewise.
785 * config/nds32/nds32.opt (munaligned-access): New option.
786 * config/nds32/nds32.c (nds32_asm_file_start): Display
787 flag_unaligned_access status.
789 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
791 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
792 -mno-relax is present.
793 * config/riscv/linux.h (LINK_SPEC): Ditto.
795 2018-04-20 Martin Sebor <msebor@redhat.com>
798 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
800 (gimple_fold_builtin_stxcpy_chk): Same.
801 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
803 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
806 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
807 __powikf2 when long double is IEEE 128-bit.
809 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
811 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
812 step to make sure stack always aligned.
814 2018-04-20 Carl Love <cel@us.ibm.com>
817 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
820 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
821 Tom de Vries <tom@codesourcery.com>
824 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
825 Emit insns for calls too.
826 (nvptx_find_par): Always look for worker-level predecessor insn.
827 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
829 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
830 (nvptx_process_pars): Propagate frames for calls.
832 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
835 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
837 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
838 (ix86_handle_option): Don't handle OPT_mibt.
839 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
841 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
843 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
844 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
845 (ix86_target_macros): Define __CET__ with flag_cf_protection
847 * config/i386/i386.c (isa2_opts): Remove -mibt.
848 * config/i386/i386.h (TARGET_IBT): Removed.
849 (TARGET_IBT_P): Likewise.
850 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
851 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
852 * config/i386/i386.opt (mcet): Update help message.
855 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
856 -mcet as an alias for -mshstk.
858 2018-04-20 Richard Biener <rguenther@suse.de>
861 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
862 complexity by forcing a single use of the multiply operand.
864 2018-04-20 Martin Jambor <mjambor@suse.cz>
867 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
868 recursion dependency to only apply to non-clones.
870 2018-04-20 Martin Jambor <mjambor@suse.cz>
873 * ipa-cp.c (create_specialized_node): Check that clones of
874 self-recursive edges exist during IPA-CP.
876 2018-04-19 Toon Moene <toon@moene.org>
878 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
881 2018-04-19 Jakub Jelinek <jakub@redhat.com>
883 PR tree-optimization/85467
884 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
885 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
886 VECTOR_CST element to type.
888 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
891 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
892 * config/i386/i386.md (builtin_setjmp_setup): Removed.
893 (builtin_longjmp): Likewise.
894 (save_stack_nonlocal): New pattern.
895 (restore_stack_nonlocal): Likewise.
897 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
900 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
901 Replace ASM_OUTPUT_LABEL with fprintf.
903 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
906 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
907 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
908 * config/i386/i386-c.c (ix86_target_macros_internal): Also
909 define __IBT__ and __SHSTK__ for -fcf-protection.
910 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
912 (ix86_trampoline_init): Likewise.
913 (x86_output_mi_thunk): Likewise.
914 (ix86_notrack_prefixed_insn_p): Likewise.
915 (ix86_option_override_internal): Don't disallow -fcf-protection.
916 * config/i386/i386.md (rdssp<mode>): Also enable for
918 (incssp<mode>): Likewise.
919 (nop_endbr): Likewise.
920 * config/i386/i386.opt (mcet): Change help message to built-in
924 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
925 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
926 enable CET built-in functions.
928 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
930 * common/config/i386/i386-common.c
931 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
932 OPTION_MASK_ISA_MOVDIRI_UNSET,
933 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
934 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
935 * config.gcc (movdirintrin.h): New header.
936 * config/i386/cpuid.h (bit_MOVDIRI,
937 bit_MOVDIR64B): New bits.
938 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
940 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
941 (VOID, PVOID, PCVOID)): New function types.
942 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
943 __builtin_ia32_directstoreu_u64,
944 __builtin_ia32_movdir64b): New builtins.
945 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
946 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
948 (ix86_valid_target_attribute_inner_p): Ditto.
949 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
950 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
951 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
952 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
953 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
954 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
955 (movdiri<mode>, movdir64b_<mode>): New.
956 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
957 * config/i386/immintrin.h: Include movdirintrin.h.
958 * config/i386/movdirintrin.h: New file.
959 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
961 2018-04-19 Richard Biener <rguenther@suse.de>
964 * cfg.c (clear_bb_flags): When loop state says we have
965 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
967 2018-04-19 Richard Biener <rguenther@suse.de>
969 PR tree-optimization/84737
970 * tree-vect-data-refs.c (vect_copy_ref_info): New function
971 copying restrict info.
972 (vect_setup_realignment): Use it.
973 * tree-vectorizer.h (vect_copy_ref_info): Declare.
974 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
975 the first DR to all generated stores.
976 (vectorizable_load): Likewise for loads.
978 2018-04-19 Jakub Jelinek <jakub@redhat.com>
980 PR tree-optimization/85446
981 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
982 the integral and pointer types to have the same precision.
984 * doc/install.texi: Document --disable-cet being the default and
987 2018-04-18 Martin Liska <mliska@suse.cz>
989 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
992 2018-04-18 Martin Liska <mliska@suse.cz>
995 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
998 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
999 arguments if they are comparable.
1001 2018-04-18 Martin Liska <mliska@suse.cz>
1004 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1007 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1010 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1013 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1014 ENDBR after calling __morestack.
1016 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1019 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1020 by using gcc_base_ver to generate a gcc_driver_version, and use
1021 it when generating GCC_DRIVER_NAME.
1022 * configure: Regenerate.
1024 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1027 * config.gcc: Obsolete powerpc*-*-*spe*.
1029 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1032 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1033 (stabstr_D): Change type of unum from unsigned int to
1034 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1037 2018-04-17 Jim Wilson <jimw@sifive.com>
1040 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1041 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1042 Set arg_pointer_offset after using pretend_args_size.
1044 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1046 PR rtl-optimization/85431
1047 * dse.c (record_store): Ignore zero width stores.
1050 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1051 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1052 __builtin_stack_restore rather than after it.
1053 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1054 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1055 argument instead of virtual_dynamic_stack_rtx.
1057 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1059 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1061 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1062 Add note to error message to explain internal mapping of overloaded
1063 built-in function name to non-overloaded built-in function name.
1064 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1067 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1070 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1071 where the inputs overlap with the output.
1073 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1075 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1076 (=v, v) alternative and explicit "memory" attribute.
1077 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1078 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1080 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1081 "sselog1" type instead of "sselog".
1082 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1083 "sselog". Remove explicit "memory" attribute.
1084 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1085 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1087 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1088 "sselog1" type instead of "sselog". Remove explicit "memory"
1090 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1091 use "sselog1" type instead of "sselog". Remove explicit "memory"
1093 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1094 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1096 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1097 "sselog1" type instead of "sselog". Remove explicit "memory"
1099 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1100 use "sselog1" type instead of "sselog". Remove explicit "memory"
1104 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1107 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1110 2018-04-17 Martin Jambor <mjambor@suse.cz>
1113 * ipa-cp.c (create_specialized_node): Call
1114 expand_all_artificial_thunks if necessary.
1116 2018-04-17 Martin Liska <mliska@suse.cz>
1119 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1120 in message, remote space in between '_G' and '('.
1122 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1125 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1126 avx512f_vmcmp<mode>3<round_saeonly_name>,
1127 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1128 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1129 avx512f_rndscale<mode><round_saeonly_name>,
1130 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1131 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1132 Use %<iptr>2 instead of %2 for -masm=intel.
1133 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1134 avx512f_vcvttss2usi<round_saeonly_name>,
1135 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1137 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1138 avx512f_vcvttsd2usi<round_saeonly_name>,
1139 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1140 Use %q1 instead of %1 for -masm=intel.
1141 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1142 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1143 of %3 for -masm=intel.
1144 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1146 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1148 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1150 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1151 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1153 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1155 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1156 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1157 %g1 and one with %0 and %1.
1158 (avx512er_vmrcp28<mode><round_saeonly_name>,
1159 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1161 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1162 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1163 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1164 of %0 and %{%4%} for -masm=intel.
1165 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1166 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1167 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1168 order of %0 and %{%5%}%{z%} for -masm=intel.
1170 2018-04-17 Jan Hubicka <jh@suse.cz>
1173 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1175 2018-04-17 Martin Liska <mliska@suse.cz>
1178 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1179 for target_clone error message. Make default implementation
1180 clone to be a local declaration.
1181 (separate_attrs): Add new argument and check for an empty
1183 (expand_target_clones): Handle it.
1184 (ipa_target_clone): Make redirection just for target_clones
1187 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1188 Tom de Vries <tom@codesourcery.com>
1191 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1192 tiled basic blocks without omp continue statements.
1194 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1197 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1198 vec_extract expression as having side effects to make sure it gets
1201 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1204 * config/i386/i386.c (get_builtin_code_for_version): Check
1207 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1210 * gcc/config.gcc: Support "skylake".
1211 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1213 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1214 (processor_target_table): Add "skylake".
1215 (ix86_option_override_internal): Add "skylake".
1216 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1217 PROCESSOR_CANNONLAKE.
1218 (get_builtin_code_for_version): Fix priority for
1219 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1220 PROCESSOR_SKYLAKE-AVX512.
1221 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1222 (processor_type): Add PROCESSOR_SKYLAKE.
1224 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1225 Jason Merrill <jason@redhat.com>
1228 * convert.c (convert_to_integer_1): Use direct recursion for
1229 enumeral types and types with a precision less than the number
1230 of bits in their mode.
1232 2018-04-16 Julia Koval <julia.koval@intel.com>
1235 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1236 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1238 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1241 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1242 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1243 and -mno-direct-move.
1245 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1248 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1249 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1250 Ensure negative shifts result in {0}.
1252 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1254 PR rtl-optimization/79916
1255 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1256 regs (if any) to define how to gnerate SD moves when LRA is in
1259 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1261 PR rtl-optimization/85393
1262 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1263 * except.c (expand_dw2_landing_pad_for_region): Make static.
1264 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1265 a label and unconditional jump to old_bb, rather than
1266 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1269 PR rtl-optimization/85376
1270 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1271 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1272 instead of a specific value.
1274 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1275 Bin Cheng <bin.cheng@arm.com>
1277 PR tree-optimization/82965
1278 PR tree-optimization/83991
1279 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1280 by_profile_only parameter.
1281 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1282 information if the loop was predicted to iterate too many times.
1283 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1285 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1288 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1291 2018-04-13 Martin Liska <mliska@suse.cz>
1292 Jakub Jelinek <jakub@redhat.com>
1295 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1296 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1297 * builtins.c (expand_builtin_memory_copy_args): Use
1298 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1299 handle dest_addr == pc_rtx.
1301 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1304 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1305 asked to not generate direct moves.
1306 (fix_trunc<mode>si2_stfiwx): Similar.
1307 (fix_trunc<mode>si2_internal): Similar.
1309 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1312 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1313 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1314 lookup if dest in some wider mode is known to be const0_rtx and
1315 if so, record permanent equivalence for it to be ZERO_EXTEND of
1316 the narrower mode destination.
1318 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1320 * lto-streamer-out.c (output_function): Revert 259346.
1321 * omp-expand.c (expand_oacc_for): Likewise.
1323 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1325 PR rtl-optimization/85354
1326 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1327 * sel-sched.c (sel_global_init): ... here.
1329 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1332 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1333 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1334 mode for PE-COFF targets.
1335 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1336 (i386_pe_asm_lto_end): Likewise.
1337 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1338 (TARGET_ASM_LTO_END): Likewise.
1339 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1340 (i386_pe_asm_lto_start): New function.
1341 (i386_pe_asm_lto_end): Likewise.
1343 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1344 Richard Biener <rguenther@suse.de>
1347 * lto-streamer-out.c (output_function): Fix CFG loop state before
1349 * omp-expand.c (expand_oacc_for): Handle calls to internal
1350 functions like regular functions.
1352 2018-04-12 Richard Biener <rguenther@suse.de>
1355 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1356 for the early LTO debug to properly generate references to it
1357 during DIE emission. Do not re-use that for the skeleton for
1359 (dwarf2out_early_finish): Likewise.
1361 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1364 * config/i386/sse.md
1365 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1366 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1367 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1368 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1369 and output is a reg, avoid creating invalid lowpart subreg, but
1370 instead split into a 512-bit move. Don't split if not AVX512VL,
1371 input is xmm16+ reg and output is a mem.
1372 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1373 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1374 xmm16+ reg and output is a mem.
1376 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1378 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1379 also for flag_dwarf2_cfi_asm.
1381 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1383 PR rtl-optimization/85342
1384 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1385 a bool scalar var inside of the loop instead. Don't try to update
1386 recog_data.operand after failed apply_change_group.
1388 2018-04-12 Tom de Vries <tom@codesourcery.com>
1391 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1392 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1393 array with flexible array member as array without given dimension.
1394 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1395 argument for undefined param to true.
1397 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1400 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1401 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1402 from PowerPC section.
1403 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1404 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1405 help text that is too long.
1406 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1407 help text that is too long.
1408 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1409 help text that is too long.
1411 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1413 * config/alpha/alpha.md (stack_probe_internal): Rename
1414 from "probe_stack". Update all callers.
1416 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1418 PR rtl-optimization/84566
1419 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1420 sched_macro_fuse_insns.
1422 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1425 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1426 (compute_block_dependences): ... from here.
1428 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1430 PR tree-optimization/85331
1431 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1432 from int to HOST_WIDE_INT.
1434 2018-04-11 Martin Jambor <mjambor@suse.cz>
1437 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1438 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1439 not the same as the source val.
1440 (cgraph_edge_brings_value_p): New parameter.
1441 (gather_edges_for_value): Pass destination value to
1442 cgraph_edge_brings_value_p.
1443 (perhaps_add_new_callers): Likewise.
1444 (get_info_about_necessary_edges): Likewise and exclude values brought
1445 only by self-recursive edges.
1446 (create_specialized_node): Redirect only clones of self-calling edges.
1447 (+self_recursive_pass_through_p): New function.
1448 (find_more_scalar_values_for_callers_subset): Use it.
1449 (find_aggregate_values_for_callers_subset): Likewise.
1450 (known_aggs_to_agg_replacement_list): Removed.
1451 (decide_whether_version_node): Re-calculate known constants for all
1452 remaining context clones.
1454 2018-04-11 Richard Biener <rguenther@suse.de>
1457 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1458 from early DWARF output.
1459 (dwarf2out_early_finish): Output line info unconditionally into
1460 early DWARF and add reference to it.
1462 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1465 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1466 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1467 other than V2DFmode using iptr mode attribute.
1468 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1470 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1472 PR rtl-optimization/84659
1473 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1475 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1478 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1480 (output_loc_list): Pass address of a dummy size variable even in the
1481 locview handling loop.
1482 (index_location_lists): Add comment on why skip_loc_list_entry can't
1485 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1488 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1491 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1494 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1495 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1496 and -mstring-compare-inline-limit.
1498 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1501 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1502 for stack clash protection in a register whenever we need it to be in
1505 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1507 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1508 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1510 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1513 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1515 (mlong-double-): Ditto.
1516 * config/rs6000/sysv4.opt (msdata=): Ditto.
1517 (mtls-size=): Ditto.
1519 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1521 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1522 erroneous entries for
1523 "vector int vec_ldl (int, long int *)", and
1524 "vector unsigned int vec_ldl (int, unsigned long int *)".
1525 Add comments and entries for
1526 "vector bool char vec_ldl (int, bool char *)",
1527 "vector bool short vec_ldl (int, bool short *)",
1528 "vector bool int vec_ldl (int, bool int *)",
1529 "vector bool long long vec_ldl (int, bool long long *)",
1530 "vector pixel vec_ldl (int, pixel *)",
1531 "vector long long vec_ldl (int, long long *)",
1532 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1533 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1534 type tree bool_long_long_type_node and correct definition of
1535 bool_V2DI_type_node to make reference to this new type tree.
1536 (rs6000_mangle_type): Replace erroneous reference to
1537 bool_long_type_node with bool_long_long_type_node.
1538 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1539 comments to emphasize sign distinctions for char and int types and
1540 replace RS6000_BTI_bool_long constant with
1541 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1542 use of RS6000_BTI_pixel.
1543 (bool_long_type_node): Remove this macro definition.
1544 (bool_long_long_type_node): New macro definition
1546 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1548 PR rtl-optimization/85300
1549 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1550 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1551 simplify_unary_operation fails.
1553 2018-04-10 Martin Liska <mliska@suse.cz>
1555 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1556 cgraph_edge and ipa_ref.
1558 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1562 * config/i386/sse.md
1563 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1564 computation of the VEC_MERGE selector from mask.
1565 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1566 Fix decoding of the VEC_MERGE selector into mask.
1568 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1570 PR tree-optimization/85286
1571 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1573 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1575 * final.c (final_1): Set insn_last_address as well as
1576 insn_current_address.
1578 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1581 * explow.c (emit_stack_probe): Call validize_mem on memory location
1582 before passing it to gen_probe_stack. Create address operand and
1583 legitimize it for the probe_stack_address case.
1585 2018-04-09 Jan Hubicka <jh@suse.cz>
1588 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1589 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1590 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1592 (free_lang_data): Rebuild type inheritance graph.
1594 2018-04-09 Martin Sebor <msebor@redhat.com>
1596 * invoke.texi (-finline-small-functions): Mention other optimization
1598 (-findirect-inlining, -fpartial-inlining): Same.
1599 (-finline-functions-called-once): Same.
1600 (-freorder-blocks-and-partition): Same.
1602 2018-04-09 Jan Hubicka <jh@suse.cz>
1605 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1606 jumps; choose last target that matches the criteria (i.e.
1607 no partition changes for non-crossing jumps).
1608 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1609 support for redirecting crossing jumps to non-crossing.
1611 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1613 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1614 also for naked functions.
1616 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1618 * config/arc/arc.md (add_shift): New pattern.
1619 (add_shift2): Likewise.
1620 (sub_shift): Likewise.
1621 (sub_shift_cmp0_noout): Likewise.
1622 (compare_si_ashiftsi): Likewise.
1623 (xbfu_cmp0_noout): New combine pattern.
1624 (xbfu_cmp0"): Likewise.
1625 (movsi_set_cc_insn): Place the predicable variant first.
1626 (commutative_binary_cmp0_noout): Remove clobber.
1627 (commutative_binary_cmp0): New pattern.
1628 (noncommutative_binary_cmp0): Likewise.
1629 (noncommutative_binary_cmp0_noout): Likewise.
1630 (noncommutative_binary_comparison_result_used): Removed.
1631 (rsub_cmp0): New pattern.
1632 (rsub_cmp0_noout): Likewise.
1633 (extzvsi): Changed, keep only meaningful variants.
1634 (SQH, SEZ): New iterators.
1635 (SQH_postfix): New mode attribute.
1636 (SEZ_prefix): New code attribute.
1637 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1638 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1639 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1641 (noncommutative_operator): Check the availability of barrel
1644 2018-04-09 Richard Biener <rguenther@suse.de>
1646 PR tree-optimization/85284
1647 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1648 Only use the niter constraining form of simple_iv when the exit
1651 2018-04-09 Tom de Vries <tom@codesourcery.com>
1654 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1655 (define_expand "*memory_barrier"): New define_expand.
1656 (define_insn "memory_barrier"): New insn.
1658 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1660 PR rtl-optimization/80463
1661 PR rtl-optimization/83972
1662 PR rtl-optimization/83480
1664 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1665 correct producer for the insn.
1666 (tidy_control_flow): Fixup seqnos in case of debug insns.
1668 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1670 PR rtl-optimization/83913
1672 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1673 different sched-times when merging exprs.
1675 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1677 PR rtl-optimization/83962
1679 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1680 tidy_fallthru_edge and tidy_control_flow.
1682 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1684 PR rtl-optimization/83530
1686 * sel-sched.c (force_next_insn): New global variable.
1687 (remove_insn_for_debug): When force_next_insn is true, also leave only
1688 next insn in the ready list.
1689 (sel_sched_region): When the region wasn't scheduled, make another pass
1690 over it with force_next_insn set to 1.
1692 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1694 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1696 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1697 for interrupt control.
1698 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1699 functions for interrupt control.
1700 * config/nds32/nds32-intrinsic.md: Likewise.
1701 * config/nds32/nds32_intrinsic.h: Likewise.
1702 * config/nds32/nds32.h (nds32_builtins): Likewise.
1704 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1706 * config/nds32/nds32.c (nds32_init_machine_status,
1707 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1708 strict_aligned_p field.
1709 (nds32_expand_to_rtl_hook): New function.
1710 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1711 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1713 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1714 Chung-Ju Wu <jasonwucj@gmail.com>
1716 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1717 * config/nds32/nds32-n7.md: New file.
1718 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1719 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1721 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1722 * config/nds32/nds32.md (pipeline_model): Add n7.
1723 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1724 * config/nds32/pipelines.md: Include n7 settings.
1726 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1727 Chung-Ju Wu <jasonwucj@gmail.com>
1729 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1730 * config/nds32/nds32-e8.md: New file.
1731 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1732 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1734 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1735 * config/nds32/nds32.md (pipeline_model): Add e8.
1736 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1737 * config/nds32/pipelines.md: Include e8 settings.
1739 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1740 Chung-Ju Wu <jasonwucj@gmail.com>
1742 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1743 * config/nds32/nds32-n8.md: New file.
1744 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1745 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1747 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1748 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1749 * config/nds32/nds32.md (pipeline_model): Add n8.
1750 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1751 * config/nds32/pipelines.md: Include n8 settings.
1753 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1754 Chung-Ju Wu <jasonwucj@gmail.com>
1756 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1757 * config/nds32/nds32-n9-2r1w.md: New file.
1758 * config/nds32/nds32-n9-3r2w.md: New file.
1759 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1760 nds32_register_ports): New or modify for cpu n9.
1761 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1763 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1764 * config/nds32/nds32-utils.c: New file.
1765 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1766 TARGET_MUL_SLOW): Define.
1767 * config/nds32/nds32.md (pipeline_model): New attribute.
1768 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1769 New options that support cpu n9.
1770 * config/nds32/pipelines.md: Include n9 settings.
1771 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1773 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1775 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1776 information if necessary.
1777 (output_cond_branch_compare_zero): Likewise.
1778 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1779 (nds32_target_alignment): Refine for alignment.
1780 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1781 (FUNCTION_BOUNDARY): Modify.
1782 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1784 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1786 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1788 * config/nds32/constants.md (unspec_volatile_element): Add values for
1789 TLB operation and data prefetch.
1790 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1791 functions for TLB operation and data prefetch.
1792 * config/nds32/nds32-intrinsic.md: Likewise.
1793 * config/nds32/nds32_intrinsic.h: Likewise.
1794 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1795 (nds32_print_operand): Likewise.
1796 * config/nds32/nds32.h (nds32_builtins): Likewise.
1798 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1799 Andrew Pinski <pinsika@gcc.gnu.org>
1802 * match.pd: Use constant_boolean_node of correct type instead of
1803 boolean_true_node or boolean_false_node for simplifying
1804 pointer comparisons to zero.
1806 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1808 PR tree-optimization/80021
1809 * tree.c (verify_type_variant): Make error call in verify_variant_match
1810 translatable and remove final full stop.
1812 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1814 * config/nds32/constants.md (unspec_volatile_element): Add
1815 UNSPEC_VOLATILE_EH_RETURN.
1816 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1817 nds32_output_stack_pop): Support dwarf exception handling process.
1818 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1819 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1820 exception handling process.
1821 (nds32_compute_stack_frame): Likewise.
1822 (nds32_return_addr_rtx): Likewise.
1823 (nds32_initial_elimination_offset): Likewise.
1824 (nds32_expand_prologue): Likewise.
1825 (nds32_expand_epilogue): Likewise.
1826 (nds32_dynamic_chain_address): New function.
1827 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1829 (DYNAMIC_CHAIN_ADDRESS): Define.
1830 (EH_RETURN_DATA_REGNO): Define.
1831 (EH_RETURN_STACKADJ_RTX): Define.
1832 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1833 patterns for dwarf exception handling.
1835 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1837 * config/nds32/nds32.h: Clean up obsolete macros.
1839 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1841 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1842 Add enum values for particular instructions.
1843 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1844 particular intrinsic functions.
1845 * config/nds32/nds32-intrinsic.md: Likewise.
1846 * config/nds32/nds32_intrinsic.h: Likewise.
1847 * config/nds32/nds32.h (nds32_builtins): Likewise.
1848 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1849 (btst, ave): New patterns for particular instructions.
1851 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1853 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1854 Add enum values for atomic load/store and memory sync.
1855 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1857 * config/nds32/nds32-intrinsic.md: Likewise.
1858 * config/nds32/nds32_intrinsic.h: Likewise.
1859 * config/nds32/nds32.h (nds32_builtins): Likewise.
1861 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1863 PR tree-optimization/85257
1864 * fold-const.c (native_encode_vector): If not all elts could fit
1865 and off is -1, return 0 rather than offset.
1866 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1867 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1868 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1869 adjust buffer in native_interpret_expr call.
1871 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1873 * config/nds32/constants.md (unspec_volatile_element): Add cache
1874 control enum values.
1875 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1876 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1877 * config/nds32/nds32.c (nds32_cctl_names): New.
1878 (nds32_print_operand): Handle cache control register names.
1879 * config/nds32/nds32.h (nds32_builtins): New enum values.
1880 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1882 * config/nds32/nds32.md (type): Add mmu.
1883 * config/nds32/pipelines.md (simple_insn): Add mmu.
1885 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1887 * config/nds32/nds32.md (type): Remove call.
1888 * config/nds32/pipelines.md (simple_insn): Likewise.
1890 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1892 * config/nds32/constants.md (unspec_volatile_element): Add
1893 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1894 UNSPEC_VOLATILE_FMFCFG.
1895 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1896 description for fmfcfg and fmfcsr.
1897 (bdesc_1arg): Add fmtcsr.
1898 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1899 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1900 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1901 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1902 unspec_fmfcfg): New patterns.
1903 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1904 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1905 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1906 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1907 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1908 __nds32__fmfcfg): Define.
1910 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1912 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1913 intrinsic register names.
1914 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1915 intrinsic register enum values and macros.
1917 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1919 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1920 for load/store addressing form.
1921 (nds32_print_operand_address): Likewise.
1923 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1926 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1927 based on LABEL_REF. Remove useless assertion.
1928 (pic_address_needs_scratch): Fix formatting.
1929 (sparc_legitimize_pic_address): Minor tweaks.
1930 (sparc_delegitimize_address): Adjust assertion accordingly.
1931 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1932 into symbolic_operand.
1933 (movsi_high_pic_label_ref): Likewise.
1934 (movsi_lo_sum_pic_label_ref): Likewise.
1935 (movdi_pic_label_ref): Likewise.
1936 (movdi_high_pic_label_ref): Likewise.
1937 (movdi_lo_sum_pic_label_ref): Likewise.
1939 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1941 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1942 custom LIB_SPEC setup.
1944 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1945 Kito Cheng <kito.cheng@gmail.com>
1947 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1948 * config/riscv/freebsd.h: New.
1950 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1952 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1953 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1956 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1957 Kito Cheng <kito.cheng@gmail.com>
1959 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1960 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1961 * config/nds32/nds32-protos.h (nds32_output_call,
1962 nds32_output_return): Declare.
1963 * config/nds32/nds32.md: Refine all the call and return patterns.
1965 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1968 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1969 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1971 PR rtl-optimization/84872
1972 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1973 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1976 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1978 * expr.c (copy_blkmode_to_reg): Revert 254862.
1979 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1981 2018-04-06 Richard Biener <rguenther@suse.de>
1984 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1985 after seeing a component reference with an adjacent field. Treat
1986 refs to arrays at struct end of external decls similar to
1987 refs to unconstrained commons.
1989 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1992 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1993 look through SAVE_EXPRs with non-side-effects argument. Adjust
1995 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1998 2018-04-06 Richard Biener <rguenther@suse.de>
2001 * alias.c (find_base_term): New wrapper around find_base_term
2002 unwinding CSELIB_VAL_PTR changes.
2003 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2006 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2008 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2010 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2011 constant definitions.
2012 ("nop"): lr 0,0 -> nopr r0
2013 ("nop_lr0", "nop_lr1"): New insn definitions.
2015 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2017 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2018 NDS32_V3PUSH_AVAILABLE_P macro.
2020 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2021 Chung-Ju Wu <jasonwucj@gmail.com>
2023 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2024 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2025 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2026 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2027 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2028 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2029 * config/nds32/constraints.md: New constraints and checking for hard
2030 float configuration.
2031 * config/nds32/iterators.md: New mode iterator and attribute for hard
2032 float configuration.
2033 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2035 * config/nds32/nds32-fpu.md: New file.
2036 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2037 deal with hard float code generation.
2038 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2040 (abi_type, float_reg_number): New enum type.
2041 * config/nds32/nds32-predicates.c: New predicates for hard float.
2042 * config/nds32/nds32-protos.h: Declare functions for hard float.
2043 * config/nds32/nds32.c: Implementation for hard float configuration.
2044 * config/nds32/nds32.h: Definitions for hard float configuration.
2045 * config/nds32/nds32.md: Include hard float machine description and
2046 modify patterns for hard float configuration.
2047 * config/nds32/nds32.opt: New options for hard float configuration.
2048 * config/nds32/predicates.md: New predicates for hard float
2051 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2053 * common/config/nds32/nds32-common.c
2054 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2056 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2059 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2060 CONSTRUCTOR_ELT (ctor, ...)->value.
2062 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2065 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2067 2018-04-05 Tom de Vries <tom@codesourcery.com>
2070 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2073 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2074 Kito Cheng <kito.cheng@gmail.com>
2076 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2077 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2078 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2079 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2081 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2082 Kito Cheng <kito.cheng@gmail.com>
2084 * config/nds32/constraints.md (Ufe): New memory constraint.
2085 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2086 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2087 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2089 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2090 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2092 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2094 * config/nds32/nds32.md: Use optimize_size in the condition for
2095 alu-shift instructions.
2097 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2099 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2101 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2103 * config/nds32/nds32.md (negsi2): Refine pattern.
2105 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2106 Chung-Ju Wu <jasonwucj@gmail.com>
2108 * config/nds32/iterators.md (shift_rotate): New code iterator.
2109 (shift): New code attribute.
2110 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2111 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2112 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2113 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2114 bit-wise operations.
2115 (andsi3, *andsi3): Ditto.
2116 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2117 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2118 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2119 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2120 nds32_ior_operand, nds32_xor_operand): New predicates.
2122 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2124 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2125 (addsi3, subsi3): ... this.
2127 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2129 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2131 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2133 * config/nds32/nds32.md: Adjust indention.
2135 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2137 * config/nds32/nds32.md (feature): New attribute.
2139 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2141 * config/nds32/nds32.md (subtype): New attribute.
2143 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2146 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2147 expansion to perform a bitwise AND of the argument followed by a
2148 boolean negation of the result.
2150 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2152 PR rtl-optimization/84878
2153 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2154 the basic block. Assert the use reference is not artificial and that
2155 it has an associated insn.
2157 2018-04-04 Michael Matz <matz@suse.de>
2159 * builtins.c (compute_objsize): Pass correct operand
2160 to array_at_struct_end_p.
2162 2018-04-04 Richard Biener <rguenther@suse.de>
2165 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2166 from contexts for DINFO_LEVEL_TERSE and below.
2168 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2170 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2171 resiter_operand condition.
2172 * config/nds32/nds32.md (*move<mode>): Ditto.
2174 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2175 Monk Chiang <sh.chiang04@gmail.com>
2177 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2179 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2181 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2183 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2184 Kito Cheng <kito.cheng@gmail.com>
2186 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2187 nds32_cond_code_str, output_cond_branch,
2188 output_cond_branch_compare_zero, nds32_expand_cbranch,
2189 nds32_expand_cstore, nds32_expand_movcc,
2190 nds32_output_cbranchsi4_equality_zero,
2191 nds32_output_cbranchsi4_equality_reg,
2192 nds32_output_cbranchsi4_equality_reg_or_const_int,
2193 nds32_output_cbranchsi4_greater_less_zero: New functions.
2194 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2195 nds32_expand_cstore, nds32_expand_movcc,
2196 nds32_output_cbranchsi4_equality_zero,
2197 nds32_output_cbranchsi4_equality_reg,
2198 nds32_output_cbranchsi4_equality_reg_or_const_int,
2199 nds32_output_cbranchsi4_greater_less_zero): Declare.
2200 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2201 nds32_rimm11s_operand): New predicates.
2202 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2203 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2206 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2208 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2209 * config/nds32/nds32.md: Ditto.
2210 * config/nds32/pipelines.md: Ditto.
2212 2018-04-04 Richard Biener <rguenther@suse.de>
2214 PR tree-optimization/85168
2215 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2216 propagating abnormals.
2218 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2220 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2222 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2223 Kito Cheng <kito.cheng@gmail.com>
2225 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2226 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2227 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2228 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2229 * config/nds32/nds32.md (sibcall_internal): New.
2230 (sibcall_register): Remove.
2231 (sibcall_immediate): Remove.
2232 (sibcall_value_internal): New.
2233 (sibcall_value_register): Remove.
2234 (sibcall_value_immediate): Remove.
2235 * config/nds32/predicates.md (nds32_general_register_operand): New.
2236 (nds32_call_address_operand): New.
2238 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2240 PR rtl-optimization/85167
2241 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2242 bb_defs if *split_p, instead preinitialize it to NULL.
2244 PR tree-optimization/85156
2245 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2246 evaluating the argument multiple times.
2248 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2250 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2252 (_mm_cvtpd_ps): Likewise.
2253 (_mm_cvttpd_epi32): Likewise.
2254 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2255 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2256 vector, pixel, and bool following altivec.h include.
2258 2018-04-03 Martin Sebor <msebor@redhat.com>
2260 * doc/extend.texi (Common Function Attributes): Clarify.
2261 (const attribute): Likewise.
2262 (pure attribute): Likewise.
2264 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2267 * config/i386/i386.c (ix86_expand_vector_set): Use
2268 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2270 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2272 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2273 instructions when changing rounding bits to preserve precision bits
2274 in the x87 control word.
2276 2018-04-03 Martin Liska <mliska@suse.cz>
2278 PR tree-optimization/82491
2279 * rtl.h (strip_offset_and_add): Replace += suboffset with
2280 poly_uint64 () + suboffset.
2282 2018-03-29 Martin Liska <mliska@suse.cz>
2283 Martin Jambor <mjambor@suse.cz>
2286 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2287 param_type is not an integral or pointer type.
2289 2018-04-03 Richard Biener <rguenther@suse.de>
2291 * sese.h (recompute_all_dominators): Remove.
2293 2018-04-02 Martin Sebor <msebor@redhat.com>
2295 * doc/invoke.texi (-Wrestrict): Fix typos.
2297 2018-04-02 Jim Wilson <jimw@sifive.com>
2299 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2300 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2301 (<optab>di3, <optab>si3_extend): Likewise.
2302 (<optab>si3_mask, <optab>si3_mask_1): New.
2303 (<optab>di3_mask, <optab>di3_mask_1): New.
2304 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2305 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2306 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2308 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2310 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2313 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2315 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2316 (nds32_canonicalize_comparison): New function.
2318 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2319 Kito Cheng <kito.cheng@gmail.com>
2320 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2322 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2323 * config/nds32/constants.md (unspec_volatile_element): Add
2324 UNSPEC_VOLATILE_RELAX_GROUP.
2325 * config/nds32/nds32-relax-opt.c: New file.
2326 * config/nds32/nds32-predicates.c
2327 (nds32_symbol_load_store_p): New function.
2328 * config/nds32/nds32-protos.h
2329 (nds32_symbol_load_store_p): Declare function.
2330 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2331 * config/nds32/nds32.c
2332 (nds32_register_pass): New function to register pass.
2333 (nds32_register_passes): New function to register passes.
2334 * config/nds32/nds32.md (relax_group): New pattern.
2335 * config/nds32/nds32.opt (mrelax-hint): New option.
2336 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2338 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2340 * config/nds32/t-nds32: Modify files dependency.
2342 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2344 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2345 (PROFILE_HOOK): Define its implementation.
2347 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2349 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2350 type and 32-bit size.
2352 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2355 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2356 (V_128_256): New mode iterator.
2357 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2358 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2359 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2361 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2362 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2364 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2367 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2368 NaN inputs correctly.
2370 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2373 * config/rs6000/vsx.md (??r): New mode attribute.
2374 (*vsx_mov<mode>_64bit): Use it.
2375 (*vsx_mov<mode>_32bit): Likewise.
2377 2018-03-30 Martin Sebor <msebor@redhat.com>
2379 PR tree-optimization/84818
2380 * builtins.c (check_access): Use warning_n.
2382 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2385 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2387 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2390 2018-03-30 Julia Koval <julia.koval@intel.com>
2393 * x86-tune.def (movx, partial_reg_dependency): Enable for
2396 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2399 * lra-constraints.c (process_alt_operands): Move setting
2400 this_alternative_matches below.
2402 2018-03-29 Martin Liska <mliska@suse.cz>
2405 * doc/invoke.texi: Document how LTO works with debug info.
2406 Describe auto-load support of binutils. Mention 'x86-64'
2407 as valid option value of -march option.
2409 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2411 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2414 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2415 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2416 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2419 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2422 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2423 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2424 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2425 for RS6000_BTM_POWERPC64.
2426 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2427 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2428 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2433 2018-03-28 Carl Love <cel@us.ibm.com>
2436 2017-09-27 Carl Love <cel@us.ibm.com>
2438 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2439 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2440 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2443 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2445 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2446 instead of __vector bool.
2447 (_mm_max_pu8): Likewise.
2448 (_mm_min_pi16): Likewise.
2450 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2453 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2454 (DIVWEUO): Likewise.
2456 (DIVDEUO): Likewise.
2457 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2458 DIVWEUO and DIVDEUO.
2459 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2460 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2461 (div_extend): Likewise.
2462 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2464 (__builtin_divweuo): Likewise.
2465 (__builtin_divdeo): Likewise.
2466 (__builtin_divdeuo): Likewise.
2468 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2471 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2472 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2474 PR tree-optimization/82004
2475 * gimple-match-head.c (optimize_pow_to_exp): New function.
2476 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2477 Don't fold to exp if optimize_pow_to_exp is false.
2479 2018-03-28 Martin Liska <mliska@suse.cz>
2482 * calls.c (initialize_argument_information): Fix trailing space.
2483 * common.opt: Fix typo and provide better explanation for
2484 -fsanitize-coverage option.
2485 * config/i386/i386.opt: Fix typo.
2487 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2488 Martin Liska <mliska@suse.cz>
2491 * gimplify.c (asan_poison_variable): Don't do the check for
2492 gimplify_omp_ctxp here.
2493 (gimplify_decl_expr): Do it here.
2494 (gimplify_target_expr): Likewise.
2496 2018-03-28 Martin Liska <mliska@suse.cz>
2499 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2500 chkp_type_bounds_count if MPX is not enabled.
2502 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2504 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2506 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2509 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2510 function to create the function decl for complex long double
2511 multiply and divide for -mabi=ieeelongdouble.
2512 (init_float128_ieee): Call it.
2514 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2517 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2518 -fcf-protection=branch -mibt.
2519 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2521 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2524 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2526 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2529 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2530 extern array declarations.
2532 2018-03-27 Richard Biener <rguenther@suse.de>
2535 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2536 explicit single_use checks.
2538 2018-03-27 Richard Biener <rguenther@suse.de>
2540 PR tree-optimization/85082
2541 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2544 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2546 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2547 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2548 Turn on fasynchronous-unwind-tables and funwind-tables.
2550 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2553 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2554 (*bmi_blsr_<mode>_ccz): Ditto.
2556 2018-03-26 Tom de Vries <tom@codesourcery.com>
2558 PR tree-optimization/85063
2559 * omp-general.c (offloading_function_p): New function. Factor out
2561 * omp-offload.c (pass_omp_target_link::gate): ... here.
2562 * omp-general.h (offloading_function_p): Declare.
2563 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2564 with attribute omp declare target for offloading functions.
2566 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2568 PR tree-optimization/84005
2569 * tree-data-ref.h (get_base_for_alignment): Declare.
2570 * tree-data-ref.c (get_base_for_alignment_1): New function.
2571 (get_base_for_alignment): Likewise.
2572 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2573 get_base_for_alignment to find a suitable base object, instead
2574 of always using drb->base_address.
2576 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2579 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2580 known size by default.
2582 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2585 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2586 and non BLKmode operands.
2588 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2591 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2592 Clean up attributes.
2594 2018-03-23 Richard Biener <rguenther@suse.de>
2597 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2598 we are going to emit early debug for LTO.
2600 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2603 * function.c (match_asm_constraints_1): Don't optimize if input
2604 doesn't satisfy general_operand predicate for output's mode.
2607 * alias.c (write_dependence_p): Don't require for x_canonicalized
2608 non-VOIDmode if x has VOIDmode.
2611 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2612 just don't try to optimize it rather than assert it never happens.
2614 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2616 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2617 macro expansions for definition of ST_INTERNAL_<mode> and
2618 LD_INTERNAL_<mode> builtins.
2619 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2621 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2623 (altivec_expand_st_builtin): Likewise.
2624 (altivec_expand_builtin): Remove calls to deleted functions.
2625 (rs6000_address_for_altivec): Delete this function.
2626 * config/rs6000/vector.md: Remove expands for
2627 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2629 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2632 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2633 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2634 re-computing once computed.
2635 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2636 (arm_init_machine_status): Initialize
2637 machine->static_chain_stack_bytes.
2639 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2642 * doc/extend.texi: Add four new prototypes for vec_ld.
2643 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2644 definitions for more logical presentation.
2645 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2646 entries for V1TI variants of __builtin_altivec_ld builtin.
2647 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2648 handling of V1TI variant of LVX icode pattern.
2649 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2650 (rs6000_gimple_fold_builtin): Likewise.
2651 (altivec_init_builtins): Add code to define
2652 __builtin_altivec_lvx_v1ti function.
2654 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2657 * function.c (match_asm_constraints_1): Don't do the optimization
2658 if input isn't a REG, SUBREG, MEM or constant.
2660 2018-03-22 Tom de Vries <tom@codesourcery.com>
2662 PR tree-optimization/84956
2663 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2664 bb_has_abnormal_pred.
2666 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2669 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2670 DECL_INITIAL (decl) to decl at the end.
2671 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2674 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2676 * doc/extend.texi (__builtin_tgmath): Document when complex
2677 integer types are treated as _Complex _Float64.
2679 2018-03-21 Tom de Vries <tom@codesourcery.com>
2681 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2683 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2685 PR tree-optimization/84960
2686 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2687 if it is ENTRY block, move them into single succ of ENTRY in that case.
2689 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2691 PR tree-optimization/84811
2692 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2693 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2694 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2695 (known_subrange_p): Update accordingly. Cast each value involved
2696 in the size comparison, rather than casting the result of the
2699 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2701 PR tree-optimization/84982
2702 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2703 by flipping the least significant bit rather than all bits from
2704 bitpos to bitpos + bitsize - 1.
2706 2018-03-21 Nathan Sidwell <nathan@acm.org>
2708 * doc/extend.texi (Deprecated Features): Remove mention of
2709 long-deleted deprecations.
2711 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2714 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2715 * configure: Regenerate.
2717 2018-03-21 Tom de Vries <tom@codesourcery.com>
2719 PR tree-optimization/83126
2720 * tree-parloops.c (num_phis): New function.
2721 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2723 2018-03-21 Nathan Sidwell <nathan@acm.org>
2725 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2726 mention anon-struct/union members and trailing attributes.
2728 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2730 PR tree-optimization/84969
2731 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2732 builtin memset partitions if they set different rhs values.
2734 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2736 PR rtl-optimization/84989
2737 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2738 VEC_DUPLICATE with scalar result mode.
2740 2018-03-21 Martin Liska <mliska@suse.cz>
2743 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2744 not intended return statement.
2746 2018-03-21 Martin Liska <mliska@suse.cz>
2749 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2750 (chkp_find_bound_slots_1): Limit number of iterations.
2752 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2755 * Minor grammar fixes for x86 options.
2757 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2760 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2761 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2763 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2766 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2767 (altivec_lvx_<mode>_1op): Likewise.
2768 (altivec_stvx_<mode>_2op): Likewise.
2769 (altivec_stvx_<mode>_1op): Likewise.
2770 (altivec_lvx_<VM2:mode>): New define_expand.
2771 (altivec_stvx_<VM2:mode>): Likewise.
2772 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2773 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2774 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2775 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2776 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2777 (rs6000_gen_lvx): Likewise.
2778 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2779 (altivec_expand_stv_builtin): Likewise.
2780 (altivec_expand_builtin): Likewise.
2781 * config/rs6000/vector.md: Likewise.
2783 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2786 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2789 2018-03-20 Richard Biener <rguenther@suse.de>
2792 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2793 sign-conversions as zero, fall back to standard scalar_stmt
2796 2018-03-20 Martin Liska <mliska@suse.cz>
2799 * predict.c (rebuild_frequencies): Handle case when we have
2800 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2802 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2805 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2806 flag_section_anchors.
2807 * varasm.c (use_blocks_for_decl_p): Remove hack for
2808 dw2_force_const_mem.
2811 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2813 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2814 be created, use lowpart_subreg of operands[0] rather than operands[0]
2816 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2817 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2818 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2819 and n constraint instead of aarch64_shift_imm_di and Usd.
2820 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2821 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2823 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2826 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2827 to favor GPR over NEON registers.
2828 (<shift>di3_neon): Likewise.
2830 2018-03-20 Tom de Vries <tom@codesourcery.com>
2833 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2834 (nvptx_process_pars): Emit bar.sync asap and alap.
2836 2018-03-20 Tom de Vries <tom@codesourcery.com>
2839 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2840 seen_label if seen_label is already set.
2842 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2845 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2846 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2847 Use 1U instead of 1. Formatting fixes.
2850 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2851 instead of TREE_TYPE (s1) for the return value.
2853 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2855 PR tree-optimization/84946
2856 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2857 bitsize + bitsize in poly_uint64 rather than poly_int64.
2860 * dwarf2asm.c: Include fold-const.c.
2861 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2862 of decl rather than decl itself.
2864 PR rtl-optimization/84643
2865 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2867 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2870 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2871 calling assemble_variable.
2873 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2876 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2877 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2879 2018-03-19 Jim Wilson <jimw@sifive.com>
2882 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2883 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2884 (riscv_first_stack_step): Likewise.
2885 (riscv_option_override): Use STACK_BOUNDARY instead of
2887 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2889 (BIGGEST_ALIGNMENT): Set to 128.
2890 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2891 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2894 2018-03-19 Richard Biener <rguenther@suse.de>
2896 PR tree-optimization/84933
2897 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2898 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2900 2018-03-19 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/84859
2903 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2904 (cond_if_else_store_replacement): Perform sinking operation on
2905 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2906 Generalize what a BB with a single eligible store is.
2908 2018-03-19 Richard Biener <rguenther@suse.de>
2910 PR tree-optimization/84929
2911 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2912 chrec_is_positive against non-chrec arg.
2914 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2917 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2919 2018-03-18 Martin Liska <mliska@suse.cz>
2921 PR rtl-optimization/84635
2922 * regrename.c (build_def_use): Use matches_mode only when
2925 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2927 PR tree-optimization/84913
2928 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2929 vectorize chains of COND_EXPRs.
2931 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2933 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2935 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2937 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2939 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2941 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2943 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2944 Kito Cheng <kito.cheng@gmail.com>
2946 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2947 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2948 (nds32_adjust_reg_alloc_order): New function.
2949 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2951 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2953 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2954 nds32_print_operand, nds32_print_operand_address): Use
2955 HOST_WIDE_INT_PRINT_DEC instead.
2957 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2959 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2961 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2964 * config/i386/i386.c (initial_ix86_tune_features,
2965 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2967 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2968 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2969 rather than 1u << ix86_tune. Formatting fix.
2970 (ix86_option_override_internal): Change ix86_arch_mask from
2971 unsigned int to unsigned HOST_WIDE_INT, initialize to
2972 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2973 (ix86_function_specific_restore): Likewise.
2975 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2978 * postreload.c (reload_combine_recognize_pattern): Perform
2979 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2980 truncate_int_for_mode the result for the destination's mode.
2983 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2984 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2986 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2989 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2992 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2993 regno_allocno_class_array and sorted_pseudos.
2994 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2995 insns where regno is used.
2997 2018-03-16 Martin Liska <mliska@suse.cz>
3000 * multiple_target.c (create_dispatcher_calls): Redirect
3001 reference in the symbol table.
3003 2018-03-16 Martin Liska <mliska@suse.cz>
3006 * multiple_target.c (create_dispatcher_calls): Redirect also
3009 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3013 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3014 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3015 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3017 2018-03-16 Julia Koval <julia.koval@intel.com>
3019 * doc/invoke.texi (Skylake Server): Add CLWB.
3020 Cannonlake): Remove CLWB.
3022 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3024 PR tree-optimization/84841
3025 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3027 (FLOAT_ONE_CONST_TYPE): Define.
3028 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3029 (sort_by_operand_rank): Put entries with higher constant_type last
3030 rather than first to match comments.
3032 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3034 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3037 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3040 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3041 check and use address of target always.
3043 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3046 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3047 (indirect_thunk_bnd_needed): Likewise.
3048 (indirect_thunks_used): Likewise.
3049 (indirect_thunks_bnd_used): Likewise.
3050 (indirect_return_needed): New.
3051 (indirect_return_bnd_needed): Likewise.
3052 (output_indirect_thunk_function): Add a bool argument for
3054 (output_indirect_thunk_function): Don't generate alias for
3055 function return thunk.
3056 (ix86_code_end): Call output_indirect_thunk_function to generate
3057 function return thunks.
3058 (ix86_output_function_return): Set indirect_return_bnd_needed
3059 and indirect_return_needed instead of indirect_thunk_bnd_needed
3060 and indirect_thunk_needed.
3062 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3064 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3065 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3066 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3068 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3069 Paul Hua <paul.hua.gm@gmail.com>
3072 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3074 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3076 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3077 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3080 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3083 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3084 instead of GET_MODE_SIZE when comparing Units.
3086 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3089 * varasm.c (hash_section): Return an unchangeble hash value
3090 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3091 Return !aarch64_can_use_per_function_literal_pools_p ().
3093 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3096 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3097 rather than address of cmode as last argument to prepare_cmp_insn.
3099 2018-03-15 Julia Koval <julia.koval@intel.com>
3101 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3102 F_AVX512VNNI, F_AVX512BITALG): New.
3104 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3107 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3108 insn for floating-point loads and stores.
3110 2018-03-14 Carl Love <cel@us.ibm.com>
3112 * config/rs6000/rs6000-c.c: Add macro definitions for
3113 ALTIVEC_BUILTIN_VEC_PERMXOR.
3114 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3115 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3116 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3117 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3119 * config/doc/extend.texi: Add prototypes for vec_permxor.
3121 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3124 * diagnostic-show-locus.c (class layout_point): Convert m_line
3125 from int to linenum_type.
3126 (line_span::comparator): Use linenum "compare" function when
3127 comparing line numbers.
3128 (test_line_span): New function.
3129 (layout_range::contains_point): Convert param "row" from int to
3131 (layout_range::intersects_line_p): Likewise.
3132 (layout::will_show_line_p): Likewise.
3133 (layout::print_source_line): Likewise.
3134 (layout::should_print_annotation_line_p): Likewise.
3135 (layout::print_annotation_line): Likewise.
3136 (layout::print_leading_fixits): Likewise.
3137 (layout::annotation_line_showed_range_p): Likewise.
3138 (struct line_corrections): Likewise for field m_row.
3139 (line_corrections::line_corrections): Likewise for param "row".
3140 (layout::print_trailing_fixits): Likewise.
3141 (layout::get_state_at_point): Likewise.
3142 (layout::get_x_bound_for_row): Likewise.
3143 (layout::print_line): Likewise.
3144 (diagnostic_show_locus): Likewise for locals "last_line" and
3146 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3147 * input.c (selftest::test_linenum_comparisons): New function.
3148 (selftest::input_c_tests): Call it.
3149 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3150 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3151 * selftest.h (ASSERT_GT): New macro.
3152 (ASSERT_GT_AT): New macro.
3153 (ASSERT_LT): New macro.
3154 (ASSERT_LT_AT): New macro.
3156 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3158 PR rtl-optimization/84780
3159 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3161 2018-03-14 Martin Liska <mliska@suse.cz>
3163 * tree.c (record_node_allocation_statistics): Use
3164 get_stats_node_kind.
3165 (get_stats_node_kind): New function extracted from
3166 record_node_allocation_statistics.
3167 (free_node): Use get_stats_node_kind.
3169 2018-03-14 Richard Biener <rguenther@suse.de>
3171 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3172 that the value-set of ANTIC_IN doesn't grow.
3175 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3177 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3178 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3180 2018-03-14 Julia Koval <julia.koval@intel.com>
3182 * config.gcc (icelake-client, icelake-server): New.
3184 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3185 (initial_ix86_arch_features): Ditto.
3186 (PTA_SKYLAKE): Add SGX.
3187 (PTA_ICELAKE): Remove.
3188 (PTA_ICELAKE_CLIENT): New.
3189 (PTA_ICELAKE_SERVER): New.
3190 (ix86_option_override_internal): Split up icelake on icelake client and
3192 (get_builtin_code_for_version): Ditto.
3193 (fold_builtin_cpu): Ditto.
3194 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3195 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3196 * config/i386/i386.h (processor_type): Ditto.
3197 * doc/invoke.texi: Ditto.
3199 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3202 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3203 INTEGER_CST offset, add it together with bitpos / 8 and
3204 sign extend based on POINTER_SIZE.
3208 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3211 * config/i386/constraints.md (Yc): New register constraint.
3212 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3213 Use Yc constraint for alternative 2 of operand 0. Remove
3214 preferred_for_speed attribute.
3216 2018-03-14 Richard Biener <rguenther@suse.de>
3218 PR tree-optimization/84830
3219 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3220 with the old one to avoid oscillations.
3222 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3225 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3227 (assign_by_spills): Return a flag of reload assignment failure.
3228 Do not process the reload assignment failures. Do not spill other
3229 reload pseudos if they has the same reg class. Update n if
3231 (lra_assign): Add a return arg. Set up from the result of
3232 assign_by_spills call.
3233 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3234 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3235 usage_insns if it is not NULL.
3236 (spill_hard_reg_in_range): New function.
3237 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3238 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3239 function prototypes.
3240 (lra_assign): Change prototype.
3241 * lra.c (lra): Add code to deal with fails by splitting hard reg
3244 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3246 * config/riscv/riscv.opt (mrelax): New option.
3247 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3248 "norelax" when riscv_mrelax is disabled.
3249 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3251 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3254 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3255 reassociation for int modes.
3257 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3259 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3260 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3262 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3263 * config/aarch64/aarch64-sve.md
3264 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3265 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3266 (*extend<mode><Vwide>2): Rename to...
3267 (aarch64_sve_extend<mode><Vwide>2): ...this.
3268 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3269 renaming the old pattern to...
3270 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3272 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3273 define_expand, renaming the old pattern to...
3274 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3275 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3276 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3277 account when deciding which SVE instruction the optab should use.
3278 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3280 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3282 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3283 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3284 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3285 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3286 (tlsdesc_small_<mode>): Turn a define_expand and use
3287 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3288 (tlsdesc_small_advsimd_<mode>): ...this.
3289 (tlsdesc_small_sve_<mode>): New pattern.
3291 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3293 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3294 (UNSPEC_UMUL_HIGHPART): New constants.
3295 (MUL_HIGHPART): New int iteraor.
3296 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3297 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3299 (*<su>mul<mode>3_highpart): New define_insn.
3301 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3304 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3307 2018-03-13 Martin Liska <mliska@suse.cz>
3310 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3312 (sem_item_optimizer::~sem_item_optimizer): Release it.
3313 (sem_item_optimizer::merge_classes): Register variable aliases.
3314 (sem_item_optimizer::fixup_pt_set): New function.
3315 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3316 * ipa-icf.h: Declare new variables and functions.
3318 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3321 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3322 integer_pow2p@2 and test integer_pow2p in condition.
3323 (A < 0 ? C : 0): Similarly for @1.
3326 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3327 characters starting at p contain '\0' character, don't look beyond
3331 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3332 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3335 * reg-stack.c (change_stack): Change update_end var from int to
3336 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3337 also call set_block_for_insn on the newly added insns and rescan.
3340 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3341 on the last operand.
3344 * tree.c (stabilize_reference_1): Return save_expr (e) for
3345 STATEMENT_LIST even if it doesn't have side-effects.
3347 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3349 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3351 2018-03-12 Renlin Li <renlin.li@arm.com>
3353 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3354 aarch64_output_scalar_simd_mov_immediate.
3356 2018-03-12 Martin Sebor <msebor@redhat.com>
3358 PR tree-optimization/83456
3359 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3360 for perfectly overlapping calls to memcpy.
3361 (gimple_fold_builtin_memory_chk): Same.
3362 (gimple_fold_builtin_strcpy): Handle no-warning.
3363 (gimple_fold_builtin_stxcpy_chk): Same.
3364 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3366 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3368 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3369 parameter. Use it for SFmode.
3370 (rs6000_function_arg_advance_1): Adjust.
3371 (rs6000_function_arg): Adjust.
3372 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3374 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3376 PR rtl-optimization/84169
3377 PR rtl-optimization/84780
3378 * combine.c (can_combine_p): Check for a 2-insn combination whether
3379 the destination register is used between the two insns, too.
3381 2018-03-12 Richard Biener <rguenther@suse.de>
3383 PR tree-optimization/84803
3384 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3385 for refs DR analysis didn't process.
3387 2018-03-12 Richard Biener <rguenther@suse.de>
3389 PR tree-optimization/84777
3390 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3391 force-vectorize loops ignore whether we are optimizing for size.
3393 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3395 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3396 (TARGET_MD_ASM_ADJUST): Define.
3398 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3399 Kito Cheng <kito.cheng@gmail.com>
3400 Chung-Ju Wu <jasonwucj@gmail.com>
3402 * config/nds32/nds32.c (nds32_compute_stack_frame,
3403 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3404 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3405 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3406 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3407 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3408 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3409 * config/nds32/nds32.md (prologue, epilogue): Use macro
3410 NDS32_V3PUSH_AVAILABLE_P to do checking.
3412 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3415 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3416 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3417 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3418 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3419 addition of most attributes on !orig_type_die or the attribute not
3420 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3422 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3423 Chung-Ju Wu <jasonwucj@gmail.com>
3425 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3427 * config/nds32/nds32.opt (mvh): New option.
3429 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3430 Chung-Ju Wu <jasonwucj@gmail.com>
3432 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3434 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3435 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3438 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3439 Chung-Ju Wu <jasonwucj@gmail.com>
3441 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3443 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3444 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3446 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3447 Kito Cheng <kito.cheng@gmail.com>
3448 Chung-Ju Wu <jasonwucj@gmail.com>
3450 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3451 UNSPEC_FFMISM and UNSPEC_FLMISM.
3452 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3453 for ffb, ffmism and flmism.
3454 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3455 (unspec_ffmism): Ditto.
3456 (unspec_flmism): Ditto.
3457 (nds32_expand_builtin_impl): Check if string extension is available.
3458 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3459 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3461 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3464 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3467 * lra-assigns.c (assign_by_spills): Return a flag of reload
3468 assignment failure. Do not process the reload assignment
3469 failures. Do not spill other reload pseudos if they has the same
3471 (lra_assign): Add a return arg. Set up from the result of
3472 assign_by_spills call.
3473 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3474 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3475 usage_insns if it is not NULL.
3476 (spill_hard_reg_in_range): New function.
3477 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3478 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3479 function prototypes.
3480 (lra_assign): Change prototype.
3481 * lra.c (lra): Add code to deal with fails by splitting hard reg
3484 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3487 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3489 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3492 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3493 (dw_val_node): Add val_symbolic_view.
3494 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3495 (symview_upper_bound): New.
3496 (new_line_info_table): Initialize symviews_since_reset.
3497 (dwarf2out_source_line): Count symviews_since_reset and set
3498 symview_upper_bound.
3499 (dw_val_equal_p): Handle symview.
3500 (add_AT_symview): New.
3501 (print_dw_val): Handle symview.
3502 (attr_checksum, attr_checksum_ordered): Likewise.
3503 (same_dw_val_p, size_of_die): Likewise.
3504 (value_format, output_die): Likewise.
3505 (add_high_low_attributes): Use add_AT_symview for entry_view.
3506 (dwarf2out_finish): Reset symview_upper_bound, clear
3509 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3512 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3513 Add strict argument and use it.
3514 (rs6000_split_multireg_move): Update for new strict argument.
3515 (mem_operand_gpr): Disallow all non-offsettable addresses.
3516 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3518 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3521 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3522 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3523 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3526 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3527 decl, use remap_type if we want to use the type.
3529 2018-03-09 Martin Sebor <msebor@redhat.com>
3531 PR tree-optimization/84526
3532 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3534 (builtin_access::generic_overlap): Be prepared to handle non-array
3537 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3539 PR rtl-optimization/84682
3540 * lra-constraints.c (process_address_1): Check is_address flag
3541 for address constraints.
3542 (process_alt_operands): Likewise.
3543 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3544 preprocess_constraints.
3545 * recog.h (preprocess_constraints): Add oploc parameter.
3547 * recog.c (preprocess_constraints): Test address_operand for
3548 CT_ADDRESS constraints.
3550 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3553 * lra-assigns.c (assign_by_spills): Return a flag of reload
3554 assignment failure. Do not process the reload assignment
3555 failures. Do not spill other reload pseudos if they has the same
3557 (lra_assign): Add a return arg. Set up from the result of
3558 assign_by_spills call.
3559 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3560 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3561 usage_insns if it is not NULL.
3562 (spill_hard_reg_in_range): New function.
3563 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3564 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3565 function prototypes.
3566 (lra_assign): Change prototype.
3567 * lra.c (lra): Add code to deal with fails by splitting hard reg
3570 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3573 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3574 Accept complain bool parameter. Only emit errors if it is true.
3575 (arm_parse_cpu_option_name): Likewise.
3576 (arm_target_thumb_only): Adjust callers of the above.
3577 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3578 prototype to take a default true bool parameter.
3579 (arm_parse_arch_option_name): Likewise.
3581 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3582 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3586 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3587 * configure: Regenerate.
3588 * configure.ac ("linker --version-script option"): New.
3589 ("linker soname option"): New.
3591 2018-03-09 Richard Biener <rguenther@suse.de>
3593 PR tree-optimization/84775
3594 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3595 immediate uses of predicate stmts and mark them modified.
3598 PR tree-optimization/84178
3599 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3601 (version_loop_for_if_conversion): Delay update_ssa call.
3602 (tree_if_conversion): Delay update_ssa until after predicate
3605 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3608 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3609 when the function accesses prior frames.
3611 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3614 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3615 gen_llsym, otherwise call maybe_gen_llsym.
3618 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3619 has ',' character inside of it.
3621 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3624 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3625 as clobbering CC_REGNUM.
3627 2018-03-08 Richard Biener <rguenther@suse.de>
3630 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3631 (follow_copies_to_constant): Do not follow SSA names registered
3634 2018-03-08 Richard Biener <rguenther@suse.de>
3636 PR tree-optimization/84178
3637 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3639 (version_loop_for_if_conversion): Delay update_ssa call.
3640 (tree_if_conversion): Delay update_ssa until after predicate
3643 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3645 PR tree-optimization/84178
3646 * tree-if-conv.c (release_bb_predicate): Remove the
3647 the assertion that the stmts have NULL use_ops.
3648 Discard the statements, asserting that they haven't
3649 yet been added to a BB.
3651 2018-03-08 Richard Biener <rguenther@suse.de>
3653 PR tree-optimization/84746
3654 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3655 (phi_translate): Pass in destination ANTIC_OUT set.
3656 (phi_translate_1): Likewise. For a simplified result lookup
3657 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3658 (phi_translate_set): Adjust.
3659 (do_pre_regular_insertion): Likewise.
3660 (do_pre_partial_partial_insertion): Likewise.
3662 2018-03-08 Martin Liska <mliska@suse.cz>
3664 PR gcov-profile/84735
3665 * doc/gcov.texi: Document usage of profile files.
3666 * gcov-io.h: Document changes in the format.
3668 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3672 * dwarf2out.c (struct dw_line_info_table): Update comments for
3674 (FORCE_RESET_NEXT_VIEW): New.
3675 (FORCE_RESETTING_VIEW_P): New.
3676 (RESETTING_VIEW_P): Check for -1 too.
3677 (ZERO_VIEW_P): Likewise.
3678 (new_line_info_table): Force-reset next view.
3679 (dwarf2out_begin_function): Likewise.
3680 (dwarf2out_source_line): Simplify zero_view_p initialization.
3681 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3682 view directly. Omit view when omitting .loc at line 0.
3684 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3686 PR tree-optimization/84740
3687 * tree-switch-conversion.c (process_switch): Call build_constructors
3688 only if info.phi_count is non-zero.
3690 PR tree-optimization/84739
3691 * tree-tailcall.c (find_tail_calls): Check call arguments against
3692 DECL_ARGUMENTS (current_function_decl) rather than
3693 DECL_ARGUMENTS (func) when checking for tail recursion.
3695 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3697 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3698 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3699 Volker Reichelt's entry and add entries for people that perform
3700 GCC fuzzy testing and report numerous bugs.
3702 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3705 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3706 readonly data in sdata, if that is disabled.
3707 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3708 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3709 -mreadonly-in-sdata option.
3711 2018-03-07 Martin Sebor <msebor@redhat.com>
3713 PR tree-optimization/84468
3714 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3715 basic block when looking for nul assignment.
3717 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3720 * except.h (output_function_exception_table): Adjust prototype.
3721 * except.c (output_function_exception_table): Remove FNNAME parameter
3722 and add SECTION parameter. Ouput one part of the table at a time.
3723 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3724 the first part of the exception table and emit unwind directives.
3725 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3726 (i386_pe_seh_cold_init): Likewise.
3727 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3728 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3729 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3730 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3731 * config/i386/winnt.c: Include except.h.
3732 (struct seh_frame_state): Add reg_offset, after_prologue and
3733 in_cold_section fields.
3734 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3735 (i386_pe_seh_cold_init): New function.
3736 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3737 to seh->in_cold_section.
3738 (seh_emit_push): Record the offset of the push.
3739 (seh_emit_save): Record the offet of the save.
3740 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3741 Test seh->after_prologue to disregard the epilogue.
3742 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3743 (i386_pe_end_cold_function): New function.
3745 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3748 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3749 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3752 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3754 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3755 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3758 * multiple_target.c: Include tree-inline.h and intl.h.
3759 (expand_target_clones): Diagnose and fail if node->definition and
3760 !tree_versionable_function_p (node->decl).
3762 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3764 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3766 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3767 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3768 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3770 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3773 * combine.c (try_combine): Use reg_or_subregno instead of handling
3774 just paradoxical SUBREGs and REGs.
3776 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3778 * config/arc/arc.c (arc_finalize_pic): Remove function.
3779 (arc_must_save_register): We use single base PIC register, remove
3780 checks to save/restore the PIC register.
3781 (arc_expand_prologue): Likewise.
3782 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3784 (arc_verify_short): Likewise.
3785 (arc_attr_type): Likewise.
3786 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3787 (walk_stores): Likewise.
3788 (arc_address_cost): Make it static.
3789 (arc_verify_short): Likewise.
3790 (branch_dest): Likewise.
3791 (arc_attr_type): Likewise.
3792 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3793 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3794 (arc_final_prescan_insn): Remove inserting the nops due to
3795 hardware hazards. It is done in reorg step.
3796 (insn_length_variant_t): Remove.
3797 (insn_length_parameters_t): Likewise.
3798 (arc_insn_length_parameters): Likewise.
3799 (arc_get_insn_variants): Likewise.
3800 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3802 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3805 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3808 PR tree-optimization/84687
3809 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3811 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3813 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3815 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3816 Rename to ppc_speculation_barrier.
3817 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3818 __builtin_ppc_speculation_barrier.
3820 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3823 * combine.c (combine_simplify_rtx): Don't try to simplify if
3824 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3827 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3829 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3830 to 32 bytes when compiling for POWER9.
3832 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3835 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3836 regparm >= 3 with no arg reg available also for calls with
3837 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3840 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3842 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3844 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3847 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3849 2018-03-05 Richard Biener <rguenther@suse.de>
3851 PR tree-optimization/84486
3852 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3853 When inserting a __builtin_assume_aligned call set the LHS
3854 SSA name alignment info accordingly.
3856 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3858 PR tree-optimization/84114
3859 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3860 Avoid reassociation of FLOAT_MODE addition.
3862 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3864 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3865 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3866 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3867 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3868 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3869 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3870 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3872 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3873 __builtin_ia32_wbinvd): New builtins.
3874 (SPECIAL_ARGS2): New.
3875 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3876 (SPECIAL_ARGS2): New.
3877 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3878 (ix86_valid_target_attribute_inner_p): Ditto.
3879 (ix86_init_mmx_sse_builtins): Add special_args2.
3880 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3881 TARGET_WBNOINVD_P): New.
3882 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3883 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3884 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3885 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3886 * config/i386/pconfigintrin.h: New file.
3887 * config/i386/wbnoinvdintrin.h: Ditto.
3888 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3889 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3891 2018-03-05 Richard Biener <rguenther@suse.de>
3893 PR tree-optimization/84670
3894 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3896 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3897 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3898 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3899 doesn't grow if all successors (recursively) were visited at least
3902 2018-03-05 Richard Biener <rguenther@suse.de>
3904 PR tree-optimization/84650
3905 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3906 if executed in the loop pipeline.
3908 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3910 * doc/configfiles.texi (Configuration Files): Move info about
3911 conditionalizing $target-protos.h to...
3912 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3913 differs from $target-protos.h.
3915 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3916 Chung-Ju Wu <jasonwucj@gmail.com>
3918 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3919 * config/nds32/nds32-multiple.md (setmemsi): Define.
3920 * config/nds32/nds32-memory-manipulation.c
3921 (nds32_gen_dup_4_byte_to_word_value): New.
3922 (emit_setmem_word_loop): New.
3923 (emit_setmem_byte_loop): New.
3924 (nds32_expand_setmem_loop): New.
3925 (nds32_expand_setmem_loop_v3m): New.
3926 (nds32_expand_setmem_unroll): New.
3927 (nds32_expand_setmem): New.
3929 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3930 Chung-Ju Wu <jasonwucj@gmail.com>
3932 * config/nds32/nds32-memory-manipulation.c
3933 (nds32_emit_load_store): New.
3934 (nds32_emit_post_inc_load_store): New.
3935 (nds32_emit_mem_move): New.
3936 (nds32_emit_mem_move_block): New.
3937 (nds32_expand_movmemsi_loop_unknown_size): New.
3938 (nds32_expand_movmemsi_loop_known_size): New.
3939 (nds32_expand_movmemsi_loop): New.
3940 (nds32_expand_movmemsi_unroll): New.
3941 (nds32_expand_movmemqi): Rename ...
3942 (nds32_expand_movmemsi): ... to this.
3943 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3944 (movmemsi): ... to this.
3945 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3946 (nds32_expand_movmemsi): ... to this.
3948 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3949 Monk Chiang <sh.chiang04@gmail.com>
3950 Chung-Ju Wu <jasonwucj@gmail.com>
3952 * config/nds32/nds32-protos.h
3953 (nds32_expand_load_multiple): New arguments.
3954 (nds32_expand_store_multiple): Ditto.
3955 (nds32_valid_multiple_load_store): Rename ...
3956 (nds32_valid_multiple_load_store_p): ... to this.
3957 * config/nds32/nds32-memory-manipulation.c
3958 (nds32_expand_load_multiple): Refine implementation.
3959 (nds32_expand_store_multiple): Ditto.
3960 * config/nds32/nds32-multiple.md
3961 (load_multiple): Update nds32_expand_load_multiple interface.
3962 (store_multiple): Update nds32_expand_store_multiple interface.
3963 * config/nds32/nds32-predicates.c
3964 (nds32_valid_multiple_load_store): Rename ...
3965 (nds32_valid_multiple_load_store_p): ... to this and refine
3967 * config/nds32/predicates.md
3968 (nds32_load_multiple_and_update_address_operation): New predicate.
3969 (nds32_store_multiple_and_update_address_operation): New predicate.
3971 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3972 Chung-Ju Wu <jasonwucj@gmail.com>
3974 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3975 (combo): New attribute.
3976 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3978 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3980 * config/nds32/nds32.opt: Change -mcmodel= default value.
3982 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3983 Monk Chiang <sh.chiang04@gmail.com>
3984 Chung-Ju Wu <jasonwucj@gmail.com>
3986 * config/nds32/constants.md (unspec_element): New enum.
3987 * config/nds32/constraints.md (Umw): New constraint.
3988 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3989 * config/nds32/nds32-intrinsic.md: Likewise.
3990 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3991 (nds32_valid_smw_lwm_base_p): New.
3992 (nds32_output_smw_single_word): New.
3993 (nds32_output_lmw_single_word): New.
3994 (nds32_expand_unaligned_load): New.
3995 (nds32_expand_unaligned_store): New.
3996 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3997 (nds32_output_smw_single_word): Declare.
3998 (nds32_output_lmw_single_word): Declare.
3999 (nds32_expand_unaligned_load): Declare.
4000 (nds32_expand_unaligned_store): Declare.
4001 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4002 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4003 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4004 NDS32_BUILTIN_UASTORE_DW.
4005 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4008 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4009 Kito Cheng <kito.cheng@gmail.com>
4010 Chung-Ju Wu <jasonwucj@gmail.com>
4012 * config/nds32/nds32-intrinsic.c
4013 (nds32_expand_builtin_null_ftype_reg): Delete.
4014 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4015 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4016 (nds32_read_argument): New.
4017 (nds32_legitimize_target): Ditto.
4018 (nds32_legitimize_argument): Ditto.
4019 (nds32_check_constant_argument): Ditto.
4020 (nds32_expand_unop_builtin): Ditto.
4021 (nds32_expand_unopimm_builtin): Ditto.
4022 (nds32_expand_binop_builtin): Ditto.
4023 (nds32_builtin_decl_impl): Ditto.
4024 (builtin_description): Ditto.
4025 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4026 (nds32_init_builtins_impl): Ditto.
4027 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4028 (nds32_builtin_decl): New.
4029 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4030 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4032 2018-03-02 Jeff Law <law@redhat.com>
4034 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4035 (redundant_insn, fill_simple_delay_slots): Likewise.
4036 (fill_slots_from_thread): Likewise.
4037 * resource.c (mark_referenced_resources): Likewise.
4038 (mark_set_resources, find_dead_or_set_registers): Likewise.
4040 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4042 * substring-locations.h (format_warning_va): Formatting fix for
4044 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4046 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4047 * substring-locations.c: Include intl.h.
4048 (format_warning_va): Turned into small wrapper around
4049 format_warning_n_va, renamed to ...
4050 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4051 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4053 (format_warning_at_substring_n): New function.
4054 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4055 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4056 format_warning_at_substring with just a shorter name instead of
4057 const function pointer.
4058 (fmtwarn_n): New function.
4059 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4060 appropriate, get rid of all the fmtstr temporaries, move conditionals
4061 with G_() wrapped string literals directly into fmtwarn arguments,
4062 cast dir.len to (int), formatting fixes.
4064 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4066 * doc/invoke.texi: Remove "Cilk Plus" references.
4068 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4069 Richard Biener <rguenther@suse.de>
4072 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4073 for error or warning attributes if CALL_FROM_THUNK_P is set.
4076 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4079 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4080 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4083 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4084 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4086 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4087 instead of -1U in last predictors element's probability member.
4089 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4092 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4093 arguments if they are comparable.
4095 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4097 PR tree-optimization/84634
4098 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4099 masks and masked_loop_p with a single loop_masks, making sure it's
4100 null for bb vectorization.
4102 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4104 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4105 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4106 loop->force_vectorize to check whether there is no alias.
4108 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4111 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4113 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4115 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4116 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4117 instead of a loop around prev_real_insn.
4118 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4122 * config/i386/i386.c (ix86_print_operand): Use conditional
4123 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4126 2018-03-02 Richard Biener <rguenther@suse.de>
4128 PR tree-optimization/84427
4129 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4130 (bitmap_set_subtract_values): Rewrite to handle multiple
4133 (prune_clobbered_mems): Likewise.
4134 (phi_translate): Take edge instead of pred/phiblock.
4135 (phi_translate_1): Likewise.
4136 (phi_translate_set): Likewise. Insert all translated
4137 exprs for a value into the set, keeping possibly multiple
4138 expressions per value.
4139 (compute_antic_aux): Adjust for phi_translate changes.
4140 When intersecting union the expressions and prune those
4141 not in the final value set, keeping possibly multiple
4142 expressions per value. Do not use value-insertion
4143 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4144 all expressions. Add verification that the value-sets
4145 only shrink during iteration.
4146 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4147 (do_pre_regular_insertion): Likewise.
4148 (do_pre_partial_partial_insertion): Likewise.
4150 2018-03-02 Richard Biener <rguenther@suse.de>
4153 * config/darwin.c (saved_debug_info_level): New static global.
4154 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4155 (darwin_asm_lto_end): Restore debug info generation settings.
4157 2018-03-01 Martin Liska <mliska@suse.cz>
4160 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4163 2018-03-01 Richard Biener <rguenther@suse.de>
4166 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4167 type annotation with LTO when debug was disabled at compile-time.
4169 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4171 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4173 (mips_final_postscan_insn): Likewise.
4175 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4177 PR rtl-optimization/84528
4178 * alias.c (init_alias_target): Add commentary.
4179 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4180 a unique base value if the frame pointer is not eliminated
4181 to the stack pointer.
4183 2018-03-01 Tom de Vries <tom@codesourcery.com>
4185 PR rtl-optimization/83327
4186 * lra-int.h (hard_regs_spilled_into): Declare.
4187 * lra.c (hard_regs_spilled_into): Define.
4188 (init_reg_info): Init hard_regs_spilled_into.
4189 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4190 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4191 (process_bb_lives): Handle hard_regs_spilled_into.
4192 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4193 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4195 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4197 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4198 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4199 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4200 * config/rs6000/aix72.h: New file.
4202 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4204 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4205 instead of warning_at with conditional singular and plural messages
4209 * stor-layout.c (update_alignment_for_field): For
4210 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4211 && !DECL_PACKED (field), do the alignment update, just use
4212 only desired_align instead of MAX (type_align, desired_align)
4214 (place_field): Don't do known_align < desired_align handling
4215 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4216 is non-NULL, instead do it after rli->prev_field handling and
4217 only if not within a bitfield word. For DECL_PACKED (field)
4218 use type_align of BITS_PER_UNIT.
4220 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4222 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4223 superfluous parentheses and trailing spaces.
4225 2018-02-28 Richard Biener <rguenther@suse.de>
4227 PR tree-optimization/84584
4228 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4229 SCoPs with fake exit edge.
4231 2018-02-28 Martin Liska <mliska@suse.cz>
4234 * timevar.c (timer::print): Fix format to properly print 100%
4237 2018-02-28 Richard Biener <rguenther@suse.de>
4240 * genmatch.c (capture_info::walk_match): Do not mark
4241 captured expressions without operands as expr_p given
4242 they act more like predicates and should be subject to
4243 "lost tail" side-effect preserving.
4245 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4247 PR rtl-optimization/81611
4248 * auto-inc-dec.c (attempt_change): Move dead note from
4249 mem_insn if it's the next use of regno
4250 (find_address): Take address use of reg holding
4251 non-incremented value. Add parm to limit search to the named
4253 (merge_in_block): Attempt to use a mem insn that is the next
4254 use of the original regno.
4256 2018-02-27 Martin Sebor <msebor@redhat.com>
4259 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4260 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4262 2018-02-27 Martin Sebor <msebor@redhat.com>
4264 PR translation/84207
4265 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4266 n argument to unsigned HOST_WIDE_INT.
4267 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4268 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4269 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4270 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4272 2018-02-27 Richard Biener <rguenther@suse.de>
4274 PR tree-optimization/84512
4275 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4276 Do not use the estimate returned from record_stmt_cost for
4277 the scalar iteration cost but sum properly using add_stmt_cost.
4279 2018-02-27 Richard Biener <rguenther@suse.de>
4281 PR tree-optimization/84466
4282 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4283 Adjust last change to less strictly validate use operands.
4285 2018-02-27 Martin Liska <mliska@suse.cz>
4287 PR gcov-profile/84548
4288 * gcov.c (process_file): Allow partial overlap and consider it
4289 also as group functions.
4290 (output_lines): Properly calculate range of lines for a group.
4292 2018-02-27 Martin Liska <mliska@suse.cz>
4294 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4295 'ggc' suffixes. Change first column width.
4296 (timer::print): Fix formatting of the column.
4298 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4300 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4301 preserve inline entry blocks for the sake of debug inline
4302 entry point markers alone.
4303 (remove_unused_locals): Suggest in comments a better place to
4304 force the preservation of inline entry blocks that are
4305 otherwise unused, but do not preserve them.
4307 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4309 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4311 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4314 * config/i386/constraints.md (Bs): Replace
4315 ix86_indirect_branch_register with
4316 TARGET_INDIRECT_BRANCH_REGISTER.
4318 * config/i386/i386.md (indirect_jump): Likewise.
4319 (tablejump): Likewise.
4320 (*sibcall_memory): Likewise.
4321 (*sibcall_value_memory): Likewise.
4322 Peepholes of indirect call and jump via memory: Likewise.
4323 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4324 (*sibcall_value_GOT_32): Likewise.
4325 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4326 (GOT_memory_operand): Likewise.
4327 (call_insn_operand): Likewise.
4328 (sibcall_insn_operand): Likewise.
4329 (GOT32_symbol_operand): Likewise.
4330 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4332 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4334 PR rtl-optimization/83496
4335 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4336 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4337 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4338 redundant insn, if any.
4339 (relax_delay_slots): Likewise.
4340 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4342 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4344 PR tree-optimization/83965
4345 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4346 that grouped statements are part of a reduction chain. Return
4347 true if the statement is not marked as a reduction itself but
4349 (vect_recog_dot_prod_pattern): Don't check whether the statement
4350 is part of a group here.
4351 (vect_recog_sad_pattern): Likewise.
4352 (vect_recog_widen_sum_pattern): Likewise.
4354 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4357 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4359 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4362 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4364 (ix86_output_indirect_function_return): New prototype.
4365 (ix86_split_simple_return_pop_internal): Likewise.
4366 * config/i386/i386.c (indirect_return_via_cx): New.
4367 (indirect_return_via_cx_bnd): Likewise.
4368 (indirect_thunk_name): Handle return va CX_REG.
4369 (output_indirect_thunk_function): Create alias for
4370 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4371 (ix86_output_indirect_jmp): Remove the bool argument.
4372 (ix86_output_indirect_function_return): New function.
4373 (ix86_split_simple_return_pop_internal): Likewise.
4374 * config/i386/i386.md (*indirect_jump): Don't pass false
4375 to ix86_output_indirect_jmp.
4376 (*tablejump_1): Likewise.
4377 (simple_return_pop_internal): Change it to define_insn_and_split.
4378 Call ix86_split_simple_return_pop_internal to split it for
4380 (simple_return_indirect_internal): Call
4381 ix86_output_indirect_function_return instead of
4382 ix86_output_indirect_jmp.
4384 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4387 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4388 memset and value initialization afterwards.
4390 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4392 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4394 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4397 * common/config/aarch64/aarch64-common.c
4398 (aarch_option_optimization_table[]): Switch
4399 off fomit-frame-pointer
4401 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4402 Chung-Ju Wu <jasonwucj@gmail.com>
4404 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4406 (store_multiple): Ditto.
4408 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4410 * config.gcc: Add --with-cpu support for nds32 target.
4411 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4412 * config/nds32/nds32.opt: Add -mcpu= option.
4414 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4416 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4417 isel=yes): Warn for these deprecated options.
4419 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4421 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4422 ISA_2_5_MASKS_EMBEDDED.
4424 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4426 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4427 p->max as pointers rather than using iterative_hash_expr.
4429 2018-02-23 Carl Love <cel@us.ibm.com>
4431 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4432 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4434 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4435 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4436 P8V_BUILTIN_VEC_VUNSIGNED2.
4438 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4441 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4442 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4444 * lra-constraints.c (curr_insn_transform): Set up
4445 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4447 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4448 * lra-eliminations.c (spill_pseudos): Ditto.
4449 (process_insn_for_elimination): Ditto.
4450 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4451 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4452 LRA_NON_CLOBBERED_ALT.
4454 2018-02-22 Martin Sebor <msebor@redhat.com>
4456 PR tree-optimization/84480
4457 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4458 to maybe_diag_stxncpy_trunc. Call it.
4459 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4460 from gimple_fold_builtin_strcpy. Print inlining stack.
4461 (handle_builtin_stxncpy): Print inlining stack.
4462 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4464 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4467 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4468 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4469 and -fcheck-pointer-bounds are used together.
4470 (indirect_thunk_prefix): New enum.
4471 (indirect_thunk_need_prefix): New function.
4472 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4473 "_nt" instead of "_bnd" for NOTRACK prefix.
4474 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4475 (output_indirect_thunk_function): Likewise.
4477 (ix86_code_end): Update output_indirect_thunk_function calls.
4478 (ix86_output_indirect_branch_via_reg): Replace
4479 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4480 (ix86_output_indirect_branch_via_push): Likewise.
4481 (ix86_output_function_return): Likewise.
4482 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4483 incompatible with -fcf-protection=branch and
4484 -fcheck-pointer-bounds.
4486 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4489 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4490 Change gcc_assert call to output_operand_lossage.
4492 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4494 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4496 2018-02-22 DJ Delorie <dj@redhat.com>
4497 Sebastian Perta <sebastian.perta@renesas.com>
4498 Oleg Endo <olegendo@gcc.gnu.org>
4500 * config/rx/rx.c (rx_rtx_costs): New function.
4501 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4503 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4505 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4507 2018-02-22 Martin Liska <mliska@suse.cz>
4510 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4511 Add "native" as a possible value.
4513 2018-02-22 Martin Liska <mliska@suse.cz>
4516 * config/i386/i386.c (ix86_option_override_internal):
4517 Add "native" as a possible value for -march and -mtune.
4519 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4522 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4523 to all type variants.
4525 PR tree-optimization/84503
4526 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4527 width as info->bitpos + info->bitsize - start.
4528 (merged_store_group::merge_overlapping): Simplify width computation.
4529 (check_no_overlap): New function.
4530 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4531 start + width and last_order of the group, fail if check_no_overlap
4533 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4534 to group if check_no_overlap fails.
4536 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4538 * config/rs6000/altivec.md: Delete contraint arguments to
4539 define_expand, define_split, and define_peephole2, and in
4540 define_insn_and_split if always unused.
4541 * config/rs6000/darwin.md: Ditto.
4542 * config/rs6000/dfp.md: Ditto.
4543 * config/rs6000/rs6000.md: Ditto.
4544 * config/rs6000/sync.md: Ditto.
4545 * config/rs6000/vector.md: Ditto.
4546 * config/rs6000/vsx.md: Ditto.
4548 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4550 * config/rs6000/altivec.md: Write output control strings as braced
4551 blocks instead of double-quoted strings.
4552 * config/rs6000/darwin.md: Ditto.
4553 * config/rs6000/rs6000.md: Ditto.
4554 * config/rs6000/vector.md: Ditto.
4555 * config/rs6000/vsx.md: Ditto.
4557 2018-02-21 Jason Merrill <jason@redhat.com>
4559 PR c++/84314 - ICE with templates and fastcall attribute.
4560 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4562 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4564 * ipa-cp.c (determine_versionability): Fix comment typos.
4566 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4569 * ipa-cp.c (determine_versionability): Do not version functions caling
4572 2018-02-21 Martin Liska <mliska@suse.cz>
4575 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4576 Add "native" as a possible value.
4577 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4578 the macro when native cpu detection is available.
4580 2018-02-21 Martin Liska <mliska@suse.cz>
4583 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4584 Add "native" as a possible value.
4585 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4586 when native cpu detection is available.
4588 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4589 Martin Sebor <msebor@redhat.com>
4591 PR tree-optimization/84478
4592 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4594 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4595 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4596 support which is conservatively correct, for 2 only stay conservative
4597 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4598 argument to the 2 argument get_range_strlen, adjust 6 arg
4599 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4601 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4602 (gimple_fold_builtin_strlen): Pass true as last argument to
4605 2018-02-20 Martin Sebor <msebor@redhat.com>
4608 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4609 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4610 (builtin_memref::builtin_memref): Factor out parts into
4611 set_base_and_offset and call it.
4613 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4616 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4617 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4618 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4619 search at the associated MODE_INT.
4621 2018-02-20 Jeff Law <law@redhat.com>
4624 PR tree-optimization/81592
4626 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4627 for range data rather than using global data.
4628 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4629 range data rather than using global data.
4630 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4631 pass it to children as needed.
4632 (struct directive::fmtresult): Similarly.
4633 (struct directive::set_width): Similarly.
4634 (struct directive::set_precision): Similarly.
4635 (format_integer, format_directive, parse_directive): Similarly.
4636 (format_none): Accept unnamed vr_values parameter.
4637 (format_percent, format_floating, format_character): Similarly.
4638 (format_string, format_plain): Similarly.
4639 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4640 the EVRP range analyzer for range data rather than using global data.
4641 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4642 gimple-ssa-evrp-analyze.h
4643 (class sprintf_dom_walker): Add after_dom_children member function.
4644 Add evrp_range_analyzer member.
4645 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4646 range analyzer as needed.
4647 (sprintf_dom_walker::after_dom_children): New member function.
4648 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4650 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4651 (evrp_range_analyzer::pop_to_marker): Likewise.
4653 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4655 PR tree-optimization/84419
4656 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4657 with the required type if its current type is compatible but
4660 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4663 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4664 after vectorization.
4666 2018-02-20 Martin Liska <mliska@suse.cz>
4669 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4670 possible values if we don't have a hint.
4672 2018-02-20 Martin Liska <mliska@suse.cz>
4676 * final.c (shorten_branches): Build align_tab array with one
4678 * opts.c (finish_options): Add alignment option limit check.
4679 (MAX_CODE_ALIGN): Likewise.
4680 (MAX_CODE_ALIGN_VALUE): Likewise.
4681 * doc/invoke.texi: Document maximum allowed option value for
4682 all -falign-* options.
4684 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4687 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4688 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4689 * var-tracking.c (emit_note_insn_var_location): Remove all references
4690 to NOTE_INSN_CALL_ARG_LOCATION.
4691 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4692 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4693 Use copy_rtx_if_shared.
4694 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4695 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4696 (dwarf2out_var_location): Remove handling of
4697 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4699 * final.c (final_scan_insn): Remove all references to
4700 NOTE_INSN_CALL_ARG_LOCATION.
4701 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4702 before dumping final insns.
4703 * except.c (emit_note_eh_region_end): Remove all references to
4704 NOTE_INSN_CALL_ARG_LOCATION.
4705 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4706 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4707 * config/arc/arc.c (hwloop_optimize): Likewise.
4708 * config/arm/arm.c (create_fix_barrier): Likewise.
4709 * config/s390/s390.c (s390_chunkify_start): Likewise.
4710 * config/sh/sh.c (find_barrier): Likewise.
4711 * config/i386/i386.c (rest_of_insert_endbranch,
4712 ix86_seh_fixup_eh_fallthru): Likewise.
4713 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4714 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4715 * config/frv/frv.c (frv_function_prologue): Likewise.
4716 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4718 (note_outside_basic_block_p): Remove all references to
4719 NOTE_INSN_CALL_ARG_LOCATION.
4720 * gengtype.c (adjust_field_rtx_def): Likewise.
4721 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4723 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4724 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4727 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4730 PR tree-optimization/84452
4731 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4732 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4735 2018-02-19 Martin Liska <mliska@suse.cz>
4738 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4740 2018-02-19 Martin Liska <mliska@suse.cz>
4741 Richard Sandiford <richard.sandiford@linaro.org>
4743 PR tree-optimization/82491
4744 * gimple-fold.c (get_base_constructor): Make earlier bail out
4747 2018-02-19 Carl Love <cel@us.ibm.com>
4749 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4750 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4752 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4753 P8V_BUILTIN_VEC_NEG.
4755 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4757 * config/rl78/rl78.md (movdf): New define expand.
4759 2018-02-19 Martin Liska <mliska@suse.cz>
4762 * doc/invoke.texi: Fix typo.
4763 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4765 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4767 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4768 handle rs6000_single_float and rs6000_double_float specially for
4771 2018-02-16 Jeff Law <law@redhat.com>
4773 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4776 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4779 * ipa-inline.c (inline_small_functions): Fix a typo.
4781 2018-02-16 Nathan Sidwell <nathan@acm.org>
4783 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4785 2018-02-16 Carl Love <cel@us.ibm.com>
4787 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4788 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4789 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4790 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4791 expansion to P8V_BUILTIN_VEC_FLOAT2.
4793 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4795 PR rtl-optimization/70023
4796 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4797 src_regno into account.
4799 2018-02-16 Carl Love <cel@us.ibm.com>
4801 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4802 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4803 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4804 * config/rs6000/rs6000.c: Remove case statements for
4805 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4806 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4807 and P9V_BUILTIN_VEC_VINSERT4B.
4808 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4809 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4810 * config/rs6000/vsx.md:
4811 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4814 2018-02-16 Carl Love <cel@us.ibm.com>
4816 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4818 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4820 * config/rs6000/rs6000-c.c: Add the definitions for
4821 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4822 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4823 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4824 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4825 definition for insert4b and define insn *insert3b_internal.
4826 * doc/extend.texi: Add documentation for vec_extract4b.
4828 2018-02-16 Nathan Sidwell <nathan@acm.org>
4830 * doc/extend.texi (Backwards Compatibility): Mention friend
4831 injection. Note for-scope is deprecated.
4832 * doc/invoke.texi (-ffriend-injection): Deprecate.
4834 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4836 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4837 that moved to I2, also allow destinations that are a paradoxical
4838 subreg (instead of a normal reg).
4840 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4843 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4846 2018-02-16 Richard Biener <rguenther@suse.de>
4848 PR tree-optimization/84037
4849 PR tree-optimization/84016
4851 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4852 Adjust vec_construct for the fact we need additional higher latency
4853 128bit inserts for AVX256 and AVX512 vector builds.
4854 (ix86_add_stmt_cost): Scale vector construction cost for
4857 2018-02-16 Richard Biener <rguenther@suse.de>
4859 PR tree-optimization/84417
4860 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4861 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4862 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4864 2018-02-16 Martin Liska <mliska@suse.cz>
4867 * internal-fn.def (ASAN_CHECK): Set proper flags.
4868 (ASAN_MARK): Likewise.
4870 2018-02-16 Julia Koval <julia.koval@intel.com>
4872 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4873 from PTA_CANNONLAKE.
4875 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4878 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4879 Use ++iter rather than iter++ for std::list iterators.
4880 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4881 defer deleting them until all nodes in the forest are processed. Do
4882 free even leaf nodes. Change to_process into auto_vec.
4885 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4886 * vec.h (vec_default_construct): Use memset instead of placement new
4887 if BROKEN_VALUE_INITIALIZATION is defined.
4888 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4889 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4892 PR rtl-optimization/83723
4893 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4894 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4895 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4897 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4899 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4901 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4903 PR rtl-optimization/81443
4904 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4905 from inner REGs to paradoxical SUBREGs.
4907 2018-02-16 Richard Biener <rguenther@suse.de>
4909 PR tree-optimization/84399
4910 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4911 For operands we can analyze at their definition make sure we can
4912 analyze them at each use as well.
4914 2018-02-16 Richard Biener <rguenther@suse.de>
4916 PR tree-optimization/84190
4917 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4918 volatile accesses if the decl isn't volatile.
4920 2018-02-15 Jason Merrill <jason@redhat.com>
4922 PR c++/84314 - ICE with templates and fastcall attribute.
4923 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4924 TYPE_CANONICAL on an existing type.
4926 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4928 PR tree-optimization/84383
4929 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4930 dstoff nor call operand_equal_p if dstbase is NULL.
4932 PR tree-optimization/84334
4933 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4934 also a CONSTANT_CLASS_P, punt.
4936 2018-02-14 Jim Wilson <jimw@sifive.com>
4938 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4939 first SMALL_OPERAND check. New local min_second_step. Move assert
4940 to where locals are set. Add TARGET_RVC support.
4941 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4943 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4945 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4947 2018-02-14 Martin Sebor <msebor@redhat.com>
4949 PR tree-optimization/83698
4950 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4951 arrays constrain the offset range to their bounds.
4952 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4953 (builtin_access::overlap): Avoid setting the size of overlap if it's
4955 (maybe_diag_overlap): Also consider arrays when deciding what values
4956 of offsets to include in diagnostics.
4958 2018-02-14 Martin Sebor <msebor@redhat.com>
4961 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4962 that correspond to the kind of a declaration.
4964 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4967 * config/pa/pa.md: Load address of PIC label using the linkage table
4968 if the label is nonlocal.
4970 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4972 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4973 warning message if user requests -maltivec=be.
4974 * doc/invoke.texi: Document deprecation of -maltivec=be.
4976 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4979 * config/rs6000/rs6000-c.c: Update definitions for
4980 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4981 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4983 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4986 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4987 add _get_ssp intrinsics. Remove argument from
4988 __builtin_ia32_rdssp[d|q].
4989 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4990 * config/i386/i386-builtin.def: Remove argument from
4991 __builtin_ia32_rdssp[d|q].
4992 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4993 ix86_expand_special_args_builtin for _rdssp[d|q].
4994 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4995 Clear register before usage.
4996 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4997 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4999 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5001 PR tree-optimization/84357
5002 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5003 operand 1 of an ARRAY_REF too.
5005 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5008 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5009 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5011 (set_of_reg): New struct.
5012 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5013 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5014 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5016 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5017 Split into bitclr, bitset, bitinvert patterns if appropriate.
5018 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5019 use rx_fuse_in_memory_bitop.
5020 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5021 to named insn, correct maximum insn length.
5023 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5026 * machmode.def: Define a complex mode for PARTIAL_INT.
5027 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5029 * doc/rtl.texi: Document CSPImode.
5030 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5032 (msp430_hard_regno_nregs_with_padding): Likewise.
5034 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5037 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5039 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5041 PR rtl-optimization/84169
5042 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5043 we generated a parallel as new i3 and we split that to new i2 and i3
5044 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5045 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5046 those to i2, not i1. Partially rewrite this scan code.
5048 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5051 * stor-layout.c (place_field): For variable length fields, adjust
5052 offset_align afterwards not just based on the field's alignment,
5053 but also on the size.
5056 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5057 of exps and logs in the use_exp2 case.
5059 2018-02-13 Jeff Law <law@redhat.com>
5061 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5064 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5067 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5071 * common.opt (gas-loc-support, gas-locview-support): New.
5072 (ginline-points, ginternal-reset-location-views): New.
5073 * doc/invoke.texi: Document them. Use @itemx where intended.
5074 (gvariable-location-views): Adjust.
5075 * target.def (reset_location_view): New.
5076 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5077 (TARGET_RESET_LOCATION_VIEW): New.
5078 * doc/tm.texi: Rebuilt.
5079 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5080 (dwarf2out_default_as_locview_support): New.
5081 (output_asm_line_debug_info): Use option variables.
5082 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5083 (output_loc_list): Likewise.
5084 (add_high_low_attributes): Check option variables.
5085 Don't output entry view attribute in strict mode.
5086 (gen_inlined_subroutine_die): Check option variables.
5087 (dwarf2out_inline_entry): Likewise.
5088 (init_sections_and_labels): Likewise.
5089 (dwarf2out_early_finish): Likewise.
5090 (maybe_reset_location_view): New, from...
5091 (dwarf2out_var_location): ... here. Call it.
5092 * debug.h (dwarf2out_default_as_loc_support): Declare.
5093 (dwarf2out_default_as_locview_support): Declare.
5094 * hooks.c (hook_int_rtx_insn_0): New.
5095 * hooks.h (hook_int_rtx_insn_0): Declare.
5096 * toplev.c (process_options): Take -gas-loc-support and
5097 -gas-locview-support from dwarf2out. Enable
5098 -gvariable-location-views by default only with locview
5099 assembler support. Enable -ginternal-reset-location-views by
5100 default only if the target defines the corresponding hook.
5101 Enable -ginline-points by default if location views are
5102 enabled; force it disabled if statement frontiers are
5104 * tree-inline.c (expand_call_inline): Check option variables.
5105 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5107 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5109 PR tree-optimization/84321
5110 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5111 handling. Also check whether the anti-range contains any values
5112 that satisfy the mask; switch to a VR_RANGE if not.
5114 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5117 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5119 2018-02-13 Martin Jambor <mjambor@suse.cz>
5122 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5123 of call statements, also set location of a load to a temporary.
5125 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5127 * config/rl78/rl78.c (add_vector_labels): New function.
5128 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5129 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5130 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5131 which checks that no arguments are passed.
5132 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5133 * doc/extend.texi: Documentation for the new attribute.
5135 2018-02-13 Andreas Schwab <schwab@suse.de>
5137 * config/riscv/linux.h (CPP_SPEC): Define.
5139 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5142 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5143 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5144 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5145 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5146 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5147 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5148 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5149 temporarily for AES and PCLMUL builtins.
5151 PR tree-optimization/84339
5152 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5153 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5157 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5158 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5159 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5161 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5163 * omp-simd-clone.h: New file.
5164 * omp-simd-clone.c: Include omp-simd-clone.h.
5165 (expand_simd_clones): No longer static.
5166 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5167 cgraph.h and omp-simd-clone.h.
5168 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5169 (vect_recog_widen_shift_pattern): Formatting fix.
5170 (vect_pattern_recog_1): Don't check optab for calls.
5173 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5174 operands[2] into a REG before using gen_lowpart on it.
5176 2018-02-12 Jeff Law <law@redhat.com>
5179 * config/sh/sh.c (find_barrier): Consider a sibling call
5182 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5183 successfully back substituting a reg.
5185 2018-02-12 Richard Biener <rguenther@suse.de>
5187 PR tree-optimization/84037
5188 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5189 parameter, move visited init to caller.
5190 (vect_slp_analyze_operations): Separate cost from validity
5191 check, initialize visited once for all instances.
5192 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5194 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5196 (vect_model_promotion_demotion_cost): Likewise.
5197 (vectorizable_bswap): Guard cost modeling with !slp_node
5198 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5200 (vectorizable_call): Likewise.
5201 (vectorizable_conversion): Likewise.
5202 (vectorizable_assignment): Likewise.
5203 (vectorizable_shift): Likewise.
5204 (vectorizable_operation): Likewise.
5205 (vectorizable_store): Likewise.
5206 (vectorizable_load): Likewise.
5207 (vectorizable_condition): Likewise.
5208 (vectorizable_comparison): Likewise.
5210 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5213 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5214 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5215 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5217 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5220 * params.def (inline-min-speedup): Increase from 8 to 15.
5221 (max-inline-insns-auto): Decrease from 40 to 30.
5222 * ipa-split.c (consider_split): Add some buffer for function to
5223 be considered inlining candidate.
5224 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5227 2018-02-12 Richard Biener <rguenther@suse.de>
5229 PR tree-optimization/84037
5230 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5231 matched stmts if we cannot swap the non-matched ones.
5233 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5235 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5236 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5237 _mm_maskz_scalef_round_ss): New intrinsics.
5238 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5239 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5240 __builtin_ia32_scalefss_round): Remove.
5241 (__builtin_ia32_scalefsd_mask_round,
5242 __builtin_ia32_scalefss_mask_round): New intrinsics.
5243 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5244 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5245 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5246 "<round_constraint>")): Changed to ...
5247 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5248 "<round_scalar_constraint>")): ... this.
5249 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5250 %0, %1, %2<round_op3>}"): Changed to ...
5251 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5252 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5253 %2<round_scalar_mask_op3>}"): ... this.
5254 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5256 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5258 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5259 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5260 (_mm_maskz_sqrt_round_ss): New intrinsics.
5261 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5262 (__builtin_ia32_sqrtsd_mask_round)
5263 (__builtin_ia32_sqrtss_mask_round): New builtins.
5264 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5265 (__builtin_ia32_sqrtss_round): Remove.
5266 (__builtin_ia32_sqrtsd_mask_round)
5267 (__builtin_ia32_sqrtss_mask_round): New builtins.
5268 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5269 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5270 ((match_operand:VF_128 1 "vector_operand"
5271 "xBm,<round_constraint>")): Changed to ...
5272 ((match_operand:VF_128 1 "vector_operand"
5273 "xBm,<round_scalar_constraint>")): ... this.
5274 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5275 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5276 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5277 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5278 %<iptr>1<round_scalar_mask_op3>}): ... this.
5279 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5280 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5282 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5285 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5286 Cast vec_cmpeq result to correct type.
5287 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5288 Cast vec_cmpgt result to correct type.
5290 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5292 * final.c (final_scan_insn_1): Renamed from...
5293 (final_scan_insn): ... this. New wrapper, to recover
5294 seen from the outermost call in recursive ones.
5295 * config/sparc/sparc.c (output_return): Drop seen from call.
5296 (output_sibcall): Likewise.
5297 * config/visium/visium.c (output_branch): Likewise.
5299 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5301 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5304 2018-02-10 Alan Modra <amodra@gmail.com>
5307 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5308 Specify LR as an input.
5310 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5313 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5314 remove_member_access_dummy_vars): New functions.
5315 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5316 lower_omp_1, execute_lower_omp): Use them.
5318 PR rtl-optimization/84308
5319 * shrink-wrap.c (spread_components): Release todo vector.
5321 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5323 PR rtl-optimization/57193
5324 * ira-color.c (struct allocno_color_data): Add member
5325 conflict_allocno_hard_prefs.
5326 (update_conflict_allocno_hard_prefs): New.
5327 (bucket_allocno_compare_func): Add a preference based on
5328 conflict_allocno_hard_prefs.
5329 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5330 (color_allocnos): Remove a dead code. Initiate
5331 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5333 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5336 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5337 constraint from =wa to wa. Avoid a subreg on the output operand,
5338 instead use a pseudo and subreg it in a move.
5339 (p9_xxbrd_<mode>): Changed to ...
5340 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5341 (p9_xxbrd_v2df): New expander.
5342 (p9_xxbrw_<mode>): Changed to ...
5343 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5344 (p9_xxbrw_v4sf): New expander.
5346 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5348 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5350 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5353 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5354 multiply in 32-bit mode.
5355 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5356 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5359 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5361 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5362 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5363 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5364 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5366 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5369 * dwarf2out.c (is_trivial_indirect_ref): New function.
5370 (dwarf2out_late_global_decl): Do not generate a location
5371 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5372 and that are not defined in the current unit.
5374 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5376 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5377 instead of a libcall for UNORDERED.
5379 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5382 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5383 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5385 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5388 * config/s390/s390.c (s390_set_current_function): Invoke
5389 s390_indirect_branch_settings also if fndecl didn't change.
5391 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5393 * config/rs6000/rs6000.md (blockage): Set length to zero.
5395 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5397 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5399 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5402 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5403 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5407 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5408 PARALLEL incoming that failed vt_get_decl_and_offset check.
5411 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5412 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5414 (get_variable_section): For decls in named .bss* sections pass true as
5415 second argument to bss_initializer_p.
5417 2018-02-09 Marek Polacek <polacek@redhat.com>
5418 Jakub Jelinek <jakub@redhat.com>
5421 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5422 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5423 Sync some changes from cxx_fold_indirect_ref.
5425 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5427 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5429 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5430 (BLOCK_INLINE_ENTRY_LABEL): New.
5431 (dwarf2out_var_location): Disregard inline entry markers.
5432 (inline_entry_data): New struct.
5433 (inline_entry_data_hasher): New hashtable type.
5434 (inline_entry_data_hasher::hash): New.
5435 (inline_entry_data_hasher::equal): New.
5436 (inline_entry_data_table): New variable.
5437 (add_high_low_attributes): Add DW_AT_entry_pc and
5438 DW_AT_GNU_entry_view attributes if a pending entry is found
5439 in inline_entry_data_table. Add old entry_pc attribute only
5440 if debug nonbinding markers are disabled.
5441 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5442 markers are enabled.
5443 (block_within_block_p, dwarf2out_inline_entry): New.
5444 (dwarf2out_finish): Check that no entries remained in
5445 inline_entry_data_table.
5446 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5447 (final_scan_insn, notice_source_line): Likewise.
5448 (rest_of_clean_state): Skip inline entry markers.
5449 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5451 * gimple.c (gimple_build_debug_inline_entry): New.
5452 * gimple.h (enum gimple_debug_subcode): Add
5453 GIMPLE_DEBUG_INLINE_ENTRY.
5454 (gimple_build_debug_inline_entry): Declare.
5455 (gimple_debug_inline_entry_p): New.
5456 (gimple_debug_nonbind_marker_p): Adjust.
5457 * insn-notes.def (INLINE_ENTRY): New.
5458 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5459 inline entry marker notes.
5460 (print_insn): Likewise.
5461 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5462 (INSN_DEBUG_MARKER_KIND): Likewise.
5463 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5464 * tree-inline.c (expand_call_inline): Build and insert
5465 debug_inline_entry stmt.
5466 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5467 inline entry blocks early, if nonbind markers are enabled.
5468 (dump_scope_block): Dump fragment info.
5469 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5470 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5471 (gimple_build_debug_inline_entry): New.
5472 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5473 Enable/disable inline entry points too.
5474 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5475 (DEBUG_INSN): Describe inline entry markers.
5477 * common.opt (gvariable-location-views): New.
5478 (gvariable-location-views=incompat5): New.
5479 * config.in: Rebuilt.
5480 * configure: Rebuilt.
5481 * configure.ac: Test assembler for view support.
5482 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5483 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5484 * dwarf2out.c (var_loc_view): New typedef.
5485 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5486 (dwarf2out_locviews_in_attribute): New.
5487 (dwarf2out_locviews_in_loclist): New.
5488 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5489 (enum dw_line_info_opcode): Add LI_adv_address.
5490 (struct dw_line_info_table): Add view.
5491 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5492 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5493 (zero_view_p): New variable.
5494 (ZERO_VIEW_P): New macro.
5495 (output_asm_line_debug_info): New.
5496 (struct var_loc_node): Add view.
5497 (add_AT_view_list, AT_loc_list): New.
5498 (add_var_loc_to_decl): Add view param. Test it against last.
5499 (new_loc_list): Add view params. Record them.
5500 (AT_loc_list_ptr): Handle loc and view lists.
5501 (view_list_to_loc_list_val_node): New.
5502 (print_dw_val): Handle dw_val_class_view_list.
5503 (size_of_die): Likewise.
5504 (value_format): Likewise.
5505 (loc_list_has_views): New.
5506 (gen_llsym): Set vl_symbol too.
5507 (maybe_gen_llsym, skip_loc_list_entry): New.
5508 (dwarf2out_maybe_output_loclist_view_pair): New.
5509 (output_loc_list): Output view list or entries too.
5510 (output_view_list_offset): New.
5511 (output_die): Handle dw_val_class_view_list.
5512 (output_dwarf_version): New.
5513 (output_compilation_unit_header): Use it.
5514 (output_skeleton_debug_sections): Likewise.
5515 (output_rnglists, output_line_info): Likewise.
5516 (output_pubnames, output_aranges): Update version comments.
5517 (output_one_line_info_table): Output view numbers in asm comments.
5518 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5519 Call maybe_gen_llsym.
5520 (loc_list_from_tree_1): Adjust.
5521 (add_AT_location_description): Create view list attribute if
5522 needed, check it's absent otherwise.
5523 (convert_cfa_to_fb_loc_list): Adjust.
5524 (maybe_emit_file): Call output_asm_line_debug_info for test.
5525 (dwarf2out_var_location): Reset views as needed. Precompute
5526 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5527 attribute. Set view.
5528 (new_line_info_table): Reset next view.
5529 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5530 (dwarf2out_source_line): Likewise. Output view resets and labels to
5531 the assembler, or select appropriate line info opcodes.
5532 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5533 (optimize_string_length): Catch it. Adjust.
5534 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5535 dw_val_class_view_list, and remove it if no longer needed.
5536 (hash_loc_list): Hash view numbers.
5537 (loc_list_hasher::equal): Compare them.
5538 (optimize_location_lists): Check whether a view list symbol is
5539 needed, and whether the locview attribute is present, and
5540 whether they match. Remove the locview attribute if no longer
5542 (index_location_lists): Call skip_loc_list_entry for test.
5543 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5544 Use output_dwarf_version.
5545 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5546 (struct dw_val_node): Add val_view_list.
5547 * final.c (SEEN_NEXT_VIEW): New.
5548 (set_next_view_needed): New.
5549 (clear_next_view_needed): New.
5550 (maybe_output_next_view): New.
5551 (final_start_function): Rename to...
5552 (final_start_function_1): ... this. Take pointer to FIRST,
5553 add SEEN parameter. Emit param bindings in the initial view.
5554 (final_start_function): Reintroduce SEEN-less interface.
5555 (final): Rename to...
5556 (final_1): ... this. Take SEEN parameter. Output final pending
5557 next view at the end.
5558 (final): Reintroduce seen-less interface.
5559 (final_scan_insn): Output pending next view before switching
5560 sections or ending a block. Mark the next view as needed when
5561 outputting variable locations. Notify debug backend of section
5562 changes, and of location view changes.
5563 (rest_of_handle_final): Adjust.
5564 * toplev.c (process_options): Autodetect value for debug variable
5565 location views option. Warn on incompat5 without -gdwarf-5.
5566 * doc/invoke.texi (gvariable-location-views): New.
5567 (gvariable-location-views=incompat5): New.
5568 (gno-variable-location-views): New.
5570 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5572 PR tree-optimization/84136
5573 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5574 that the result of find_edge is non-NULL.
5576 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5579 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5580 storing integer register in SImode. Fix cost of 256 and 512
5581 byte aligned SSE register store.
5583 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5585 * config/i386/i386.c (ix86_multiplication_cost): Fix
5586 multiplication cost for TARGET_AVX512DQ.
5588 2018-02-08 Marek Polacek <polacek@redhat.com>
5590 PR tree-optimization/84238
5591 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5594 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5596 PR tree-optimization/84265
5597 * tree-vect-stmts.c (vectorizable_store): Don't treat
5598 VMAT_CONTIGUOUS accesses as grouped.
5599 (vectorizable_load): Likewise.
5601 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5603 PR tree-optimization/81635
5604 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5605 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5606 (test_round_for_mask): New functions.
5607 (wide_int_cc_tests): Call test_round_for_mask.
5608 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5609 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5610 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5611 range returned by get_range_info.
5613 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5616 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5617 * symtab.c: Include builtins.h
5618 (symtab_node::output_to_lto_symbol_table_p): Move here
5619 from lto-streamer-out.c:output_symbol_p.
5620 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5621 (output_symbol_p): Move all logic to symtab.c
5622 (produce_symtab): Update.
5624 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5626 * config/s390/s390-opts.h (enum indirect_branch): Define.
5627 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5628 (s390_indirect_branch_via_thunk)
5629 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5630 (enum s390_indirect_branch_type): Define.
5631 * config/s390/s390.c (struct s390_frame_layout, struct
5632 machine_function): Remove.
5633 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5634 (indirect_branch_table_label_no, indirect_branch_table_name):
5636 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5637 (enum s390_indirect_branch_option): Define.
5638 (s390_return_addr_from_memory): New function.
5639 (s390_handle_string_attribute): New function.
5640 (s390_attribute_table): Add new attribute handler.
5641 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5642 (s390_indirect_branch_via_thunk): New function.
5643 (s390_indirect_branch_via_inline_thunk): New function.
5644 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5645 sibling call optimization for non z10 compiles.
5646 (s390_emit_call): Force indirect branch target to be a single
5647 register. Add r1 clobber for non-z10 compiles.
5648 (s390_emit_epilogue): Emit return jump via return_use expander.
5649 (s390_reorg): Handle JUMP_INSNs as execute targets.
5650 (s390_option_override_internal): Perform validity checks for the
5651 new command line options.
5652 (s390_indirect_branch_attrvalue): New function.
5653 (s390_indirect_branch_settings): New function.
5654 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5655 (s390_output_indirect_thunk_function): New function.
5656 (s390_code_end): Implement target hook.
5657 (s390_case_values_threshold): Implement target hook.
5658 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5660 * config/s390/s390.h (struct s390_frame_layout)
5661 (struct machine_function): Move here from s390.c.
5662 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5663 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5664 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5665 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5666 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5667 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5668 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5669 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5670 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5671 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5672 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5673 (mnemonic attribute): Add values which aren't recognized
5675 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5676 pattern for branch conversion. Fix mnemonic attribute.
5677 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5678 indirect branch via thunk if requested.
5679 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5680 ("*indirect_jump"): Disable for branch conversion using out of
5682 ("indirect_jump_via_thunk<mode>_z10")
5683 ("indirect_jump_via_thunk<mode>")
5684 ("indirect_jump_via_inlinethunk<mode>_z10")
5685 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5686 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5687 ("casesi_jump_via_inlinethunk<mode>_z10")
5688 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5689 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5690 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5691 ("*indirect2_jump"): Disable for branch conversion.
5692 ("casesi_jump"): Turn into expander and expand patterns for branch
5694 ("return_use"): New expander.
5695 ("*return"): Emit return via thunk and rename it to ...
5696 ("*return<mode>"): ... this one.
5697 * config/s390/s390.opt: Add new options and and enum for the
5700 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5702 * lra-constraints.c (match_reload): Unconditionally use
5703 gen_lowpart_SUBREG, rather than selecting between that
5704 and equivalent gen_rtx_SUBREG code.
5706 2018-02-08 Richard Biener <rguenther@suse.de>
5708 PR tree-optimization/84233
5709 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5710 changed flag instead of boguously re-using phi_inserted.
5712 2018-02-08 Martin Jambor <mjambor@suse.cz>
5714 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5715 static local variables.
5717 2018-02-08 Richard Biener <rguenther@suse.de>
5719 PR tree-optimization/84278
5720 * tree-vect-stmts.c (vectorizable_store): When looking for
5721 smaller vector types to perform grouped strided loads/stores
5722 make sure the mode is supported by the target.
5723 (vectorizable_load): Likewise.
5725 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5727 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5728 Increase LDP/STP opportunities by adding adjacent callee-saves.
5730 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5732 PR rtl-optimization/84068
5733 PR rtl-optimization/83459
5734 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5736 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5738 PR tree-optimization/84224
5739 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5740 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5743 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5746 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5747 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5748 count, remove one USE.
5750 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5752 * doc/install.texi (Configuration): Document the
5753 --with-long-double-format={ibm,ieee} PowerPC configuration
5757 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5758 Convert from define_expand to be define_insn_and_split. Rework
5759 float/double/_Float128 conversions to QI/HI/SImode to work with
5760 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5761 conversions to QI/HImode types did a store and then a load to
5762 truncate the value. For conversions to VSX registers, don't split
5763 the insn, instead emit the code directly. Use the code iterator
5764 any_fix to combine signed and unsigned conversions.
5765 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5766 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5767 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5768 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5769 (fix_<mode>di2_hw): Likewise.
5770 (fixuns_<mode>di2_hw): Likewise.
5771 (fix_<mode>si2_hw): Likewise.
5772 (fixuns_<mode>si2_hw): Likewise.
5773 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5774 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5775 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5776 fix<uns>_trunc<SFDF:mode>si2_p8.
5777 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5779 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5780 (fix<uns>_<mode>_mem): Likewise.
5781 (fctiw<u>z_<mode>_mem): Likewise.
5782 (fix<uns>_<mode>_mem): Likewise.
5783 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5784 the register allocator from doing a direct move to the GPRs to do
5785 a store, and instead use the ISA 3.0 store byte/half-word from
5786 vector register instruction. For IEEE 128-bit floating point,
5787 also optimize stores of 32-bit ints.
5788 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5790 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5792 * genextract.c (push_pathstr_operand): New function to support
5794 (walk_rtx): Call push_pathstr_operand.
5795 (print_path): Support [a-zA-Z].
5797 2018-02-07 Richard Biener <rguenther@suse.de>
5799 PR tree-optimization/84037
5800 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5801 (cse_and_gimplify_to_preheader): Declare.
5802 (vect_get_place_in_interleaving_chain): Likewise.
5803 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5805 (_loop_vec_info::~_loop_vec_info): Delete it.
5806 (cse_and_gimplify_to_preheader): New function.
5807 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5808 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5809 (vectorizable_load): Likewise. For grouped stores always base
5810 the IV on the first element.
5811 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5812 condition before gimplifying.
5814 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5816 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5817 *DIV_EXPR and *MOD_EXPR.
5819 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5822 * config/i386/i386.c (ix86_option_override_internal): Mask out
5823 the CF_SET bit when checking -fcf-protection.
5825 2018-02-07 Tom de Vries <tom@codesourcery.com>
5828 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5831 2018-02-07 Richard Biener <rguenther@suse.de>
5833 PR tree-optimization/84204
5834 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5837 PR tree-optimization/84205
5838 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5839 special-case isl_ast_op_zdiv_r.
5841 PR tree-optimization/84223
5842 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5843 Only add conditions from within the region.
5844 (gather_bbs::after_dom_children): Adjust.
5846 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5849 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5850 * config/avr/avr.md: Only post-reload split REG-REG moves if
5851 either register is GENERAL_REG_P.
5853 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5855 PR tree-optimization/84235
5856 * tree-ssa-scopedtables.c
5857 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5858 if the subtraction is performed in floating point type where NaNs are
5859 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5860 build 1. Formatting fix.
5862 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5865 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5866 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5867 and skip it regardless of bb boundaries. Use CALL_P macro,
5868 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5869 unnecessarily, formatting fix.
5871 2018-02-06 Michael Collison <michael.collison@arm.com>
5873 * config/arm/thumb2.md:
5874 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5875 (*thumb_mov_notscc): Ditto.
5877 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5880 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5881 unsigned_fix, not "s".
5883 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5885 * configure.ac (gcc_fn_eh_frame_ro): New function.
5886 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5887 correct .eh_frame permissions.
5888 * configure: Regenerate.
5890 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5892 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5895 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5897 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5898 Display warning message for -mno-speculate-indirect-jumps.
5900 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5902 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5904 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5906 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5908 PR tree-optimization/84225
5909 * tree-eh.c (find_trapping_overflow): Only call
5910 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5912 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5915 * config/i386/i386.c: Reimplement the check of possible options
5916 -mibt/-mshstk conbination. Change error messages.
5917 * doc/invoke.texi: Fix a typo: remove extra '='.
5919 2018-02-06 Marek Polacek <polacek@redhat.com>
5921 PR tree-optimization/84228
5922 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5924 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5927 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5928 emitted arch directives.
5929 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5930 __ARM_FEATURE_COPROC before changing architectures.
5932 2018-02-06 Richard Biener <rguenther@suse.de>
5934 * config/i386/i386.c (print_reg): Fix typo.
5935 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5937 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5939 * configure: Regenerate.
5941 2018-02-05 Martin Sebor <msebor@redhat.com>
5943 PR tree-optimization/83369
5944 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5947 2018-02-05 Martin Liska <mliska@suse.cz>
5949 * doc/invoke.texi: Cherry-pick upstream r323995.
5951 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5953 * ira.c (ira_init_register_move_cost): Adjust comment.
5955 2018-02-05 Martin Liska <mliska@suse.cz>
5957 PR gcov-profile/84137
5958 * doc/gcov.texi: Fix typo in documentation.
5960 2018-02-05 Martin Liska <mliska@suse.cz>
5962 PR gcov-profile/83879
5963 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5964 using dlopen functionality.
5966 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5968 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5969 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5970 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5971 _mm_maskz_range_round_ss): New intrinsics.
5972 (__builtin_ia32_rangesd128_round)
5973 (__builtin_ia32_rangess128_round): Remove.
5974 (__builtin_ia32_rangesd128_mask_round,
5975 __builtin_ia32_rangess128_mask_round): New builtins.
5976 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5977 __builtin_ia32_rangess128_round): Remove.
5978 (__builtin_ia32_rangesd128_mask_round,
5979 __builtin_ia32_rangess128_mask_round): New builtins.
5980 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5981 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5982 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5983 "<round_saeonly_constraint>")): Changed to ...
5984 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5985 "<round_saeonly_scalar_constraint>")): ... this.
5986 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5987 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5988 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5989 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5990 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5992 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5994 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5996 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5997 Remove all values except native, 8540 and 8548.
5999 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6001 * config/i386/i386.c (ix86_output_function_return): Pass
6002 INVALID_REGNUM, instead of -1, as invalid register number to
6003 indirect_thunk_name and output_indirect_thunk.
6005 2018-02-02 Julia Koval <julia.koval@intel.com>
6007 * config.gcc: Add -march=icelake.
6008 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6009 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6010 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6011 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6012 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6013 (processor_target_table): Add icelake.
6014 (ix86_option_override_internal): Handle new PTAs.
6015 (get_builtin_code_for_version): Handle icelake.
6016 (M_INTEL_COREI7_ICELAKE): New.
6017 (fold_builtin_cpu): Handle icelake.
6018 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6019 * doc/invoke.texi: Add -march=icelake.
6021 2018-02-02 Julia Koval <julia.koval@intel.com>
6023 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6024 to wide_int_bitmask.
6025 * wide-int-bitmask.h: New.
6027 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6030 * config/i386/i386.md: Replace Pmode with word_mode in
6031 builtin_setjmp_setup and builtin_longjmp to support x32.
6033 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6037 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6039 (rs6000_supported_cpu_names): New static variable.
6040 (linux_cpu_translation_table): Likewise.
6041 (elf_platform) <cpu>: Define new static variable and use it.
6042 Translate kernel AT_PLATFORM name to canonical name if needed.
6043 Error if platform name is unknown.
6045 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6048 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6050 2018-02-01 Jeff Law <law@redhat.com>
6053 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6054 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6055 the scratch if RELEASE_VIA_POP is false.
6056 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6057 If we have to save a temporary register, decrement SIZE appropriately.
6058 Pass new arguments to release_scratch_register_on_entry.
6059 (ix86_adjust_stack_and_probe): Likewise.
6060 (ix86_emit_probe_stack_range): Pass new arguments to
6061 release_scratch_register_on_entry.
6063 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6065 PR rtl-optimization/84157
6066 * combine.c (change_zero_ext): Use REG_P predicate in
6067 front of HARD_REGISTER_P predicate.
6069 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6071 * config/avr/avr.c (avr_option_override): Move disabling of
6072 -fdelete-null-pointer-checks to...
6073 * common/config/avr/avr-common.c (avr_option_optimization_table):
6076 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6078 PR tree-optimization/81635
6079 * tree-data-ref.c (split_constant_offset_1): For types that
6080 wrap on overflow, try to use range info to prove that wrapping
6083 2018-02-01 Renlin Li <renlin.li@arm.com>
6086 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6088 (aarch64_register_move_cost): Likewise.
6089 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6091 (REG_CLASS_NAMES): Likewise.
6092 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6093 TAILCALL_ADDR_REGS. Remove IP registers.
6094 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6096 2018-02-01 Richard Biener <rguenther@suse.de>
6098 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6099 for specifying RPO order and allow NULL for that.
6100 * domwalk.c (dom_walker::dom_walker): Likewise.
6101 (dom_walker::walk): Handle NULL RPO order.
6102 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6104 (rewrite_update_dom_walker): Likewise.
6105 (mark_def_dom_walker): Likewise.
6107 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6109 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6110 (aarch64_maybe_expand_sve_subreg_move): Declare.
6111 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6112 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6114 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6115 that are semantically a reverse operation.
6116 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6117 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6118 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6120 (aarch64_can_change_mode_class): For big-endian, forbid changes
6121 between two SVE modes if they have different element sizes.
6123 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6125 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6126 the TImode handling for big-endian targets.
6128 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6130 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6131 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6133 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6134 Remove BSWAP handing for big-endian targets and use the form of
6135 LD1RQ appropariate for the mode.
6137 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6139 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6140 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6143 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6146 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6147 check for operands that need to go through aarch64_sve_reload_be.
6149 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6151 PR tree-optimization/81661
6152 PR tree-optimization/84117
6153 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6154 * tree-eh.c: Include gimplify.h.
6155 (find_trapping_overflow, replace_trapping_overflow,
6156 rewrite_to_non_trapping_overflow): New functions.
6157 * tree-vect-loop.c: Include tree-eh.h.
6158 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6159 * tree-data-ref.c: Include tree-eh.h.
6160 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6162 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6164 PR rtl-optimization/84123
6165 * combine.c (change_zero_ext): Check if hard register satisfies
6166 can_change_dest_mode before calling gen_lowpart_SUBREG.
6168 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6171 * ira.c (ira_init_register_move_cost): Remove assert.
6173 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6175 PR rtl-optimization/84071
6176 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6177 * doc/tm.texi: Regenerate.
6179 2018-01-31 Richard Biener <rguenther@suse.de>
6181 PR tree-optimization/84132
6182 * tree-data-ref.c (analyze_miv_subscript): Properly
6183 check whether evolution_function_is_affine_multivariate_p
6184 before calling gcd_of_steps_may_divide_p.
6186 2018-01-31 Julia Koval <julia.koval@intel.com>
6189 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6190 * config/i386/i386.md (rdpid_rex64) New.
6191 (rdpid): Make 32bit only.
6193 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6196 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6197 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6199 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6202 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6204 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6206 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6208 PR rtl-optimization/84071
6209 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6210 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6212 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6214 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6215 (arc_attribute_table): Add 'aux' attribute.
6216 (arc_in_small_data_p): Consider aux like variables.
6217 (arc_is_aux_reg_p): New function.
6218 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6219 (arc_get_aux_arg): New function.
6220 (prepare_move_operands): Handle aux-register access.
6221 (arc_handle_aux_attribute): New function.
6222 * doc/extend.texi (ARC Variable attributes): Add subsection.
6224 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6226 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6227 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6228 (arc_attribute_table): Add 'uncached' attribute.
6229 (arc_print_operand): Print '.di' flag for uncached memory
6231 (arc_in_small_data_p): Do not consider for small data the uncached
6233 (arc_is_uncached_mem_p): New function.
6234 * config/arc/predicates.md (compact_store_memory_operand): Check
6235 for uncached memory accesses.
6236 (nonvol_nonimm_operand): Likewise.
6237 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6239 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6242 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6243 falign-loops=): Add Optimization flag.
6245 2018-01-30 Jeff Law <law@redhat.com>
6248 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6249 INT_REGISTERS_SAVED. Check it prior to calling
6250 get_scratch_register_on_entry.
6251 (ix86_adjust_stack_and_probe): Similarly.
6252 (ix86_emit_probe_stack_range): Similarly.
6253 (ix86_expand_prologue): Corresponding changes.
6255 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6258 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6259 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6261 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6264 * lra-constraints.c (curr_insn_transform): Process AND in the
6267 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6269 PR rtl-optimization/83986
6270 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6271 dependence against last_pending_memory_flush in addition to
6274 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6276 PR tree-optimization/81611
6277 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6280 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6283 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6286 2018-01-30 Richard Biener <rguenther@suse.de>
6287 Jakub Jelinek <jakub@redhat.com>
6289 PR tree-optimization/84111
6290 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6291 inner loops added during recursion, as they don't have up-to-date
6294 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6297 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6298 (can_inline_edge_by_limits_p): ... here.
6299 (can_early_inline_edge_p, check_callers,
6300 update_caller_keys, update_callee_keys, recursive_inlining,
6301 add_new_edges_to_heap, speculation_useful_p,
6302 inline_small_functions,
6303 inline_small_functions, flatten_function,
6304 inline_to_all_callers_1): Update.
6306 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6308 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6311 2018-01-30 Richard Biener <rguenther@suse.de>
6313 PR tree-optimization/83008
6314 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6315 invariant and constant vector uses in stmts when they need
6318 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6321 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6322 * configure: Regenerate.
6324 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6326 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6328 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6329 Use gen_rtx_REG rather than gen_lowpart.
6331 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6333 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6334 rather than 0 when creating partial subregs.
6336 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6338 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6341 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6344 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6345 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6346 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6347 flags. This restores the settings used before the 2017-07-24.
6348 Turning off pre increment/decrement/modify allows IVOPTS to
6349 optimize DF/SF loops where the index is an int.
6351 2018-01-29 Richard Biener <rguenther@suse.de>
6352 Kelvin Nilsen <kelvin@gcc.gnu.org>
6355 * tree-vect-stmts.c (vectorizable_call): Don't call
6356 targetm.vectorize_builtin_md_vectorized_function if callee is
6359 2018-01-22 Carl Love <cel@us.ibm.com>
6361 * doc/extend.tex: Fix typo in second arg in
6362 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6364 2018-01-29 Richard Biener <rguenther@suse.de>
6366 PR tree-optimization/84086
6367 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6368 (flush_ssaname_freelist): When SSA names were released reset
6369 the SCEV hash table.
6371 2018-01-29 Richard Biener <rguenther@suse.de>
6373 PR tree-optimization/84057
6374 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6375 removed paths when removing edges.
6377 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6379 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6380 -mfunction-return=@var{choice}.
6382 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6385 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6387 (layout::print_source_line): Likewise.
6388 (test_get_line_width_without_trailing_whitespace): Add test cases.
6390 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6393 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6396 2018-01-26 Jim Wilson <jimw@sifive.com>
6398 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6400 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6403 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6405 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6406 and CMP + SUB-immediate -> SUBS.
6408 2018-01-26 Martin Sebor <msebor@redhat.com>
6410 PR tree-optimization/83896
6411 * tree-ssa-strlen.c (get_string_len): Rename...
6412 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6413 Avoid assuming length is constant.
6414 (handle_char_store): Use HOST_WIDE_INT for string length.
6416 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6419 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6420 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6422 2018-01-26 Richard Biener <rguenther@suse.de>
6424 PR rtl-optimization/84003
6425 * dse.c (record_store): Only record redundant stores when
6426 the earlier store aliases at least all accesses the later one does.
6428 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6430 PR rtl-optimization/83985
6431 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6432 REG_CFA_RESTORE insns.
6433 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6434 REG_CFA_RESTORE insns here.
6437 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6438 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6440 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6442 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6443 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6444 (arc_init): Likewise.
6445 (arc_override_options): Likewise.
6446 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6448 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6450 * config/arc/arc.h (TARGET_DBNZ): Define.
6451 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6452 properly set the tune attribute.
6453 (dbnz): Use TARGET_DBNZ guard.
6454 * config/arc/arc.opt (mtune): Add core3 option.
6456 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6458 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6459 recognize new pic like addresses.
6460 (arc_delegitimize_address): Clean up.
6462 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6464 * config/arc/arc-arches.def: Option mrf16 valid for all
6466 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6467 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6468 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6469 * config/arc/arc-tables.opt: Regenerate.
6470 * config/arc/arc.c (arc_conditional_register_usage): Handle
6471 reduced register file case.
6472 (arc_file_start): Set must have build attributes.
6473 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6475 * config/arc/arc.opt (mrf16): Add new option.
6476 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6477 * config/arc/genmultilib.awk: Handle new mrf16 option.
6478 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6479 * config/arc/t-multilib: Regenerate.
6480 * doc/invoke.texi (ARC Options): Document mrf16 option.
6482 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6484 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6485 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6486 (arc_attribute_table): Add 'secure_call' attribute.
6487 (arc_print_operand): Print secure call operand.
6488 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6490 (arc_is_secure_call_p): New function. * config/arc/arc.md
6491 (call_i): Add support for sjli instruction.
6492 (call_value_i): Likewise.
6493 * config/arc/constraints.md (Csc): New constraint.
6495 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6496 John Eric Martin <John.Martin@emmicro-us.com>
6498 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6499 * config/arc/arc.c (_arc_jli_section): New struct.
6500 (arc_jli_section): New type.
6501 (rc_jli_sections): New static variable.
6502 (arc_handle_jli_attribute): New function.
6503 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6504 (arc_file_end): New function.
6505 (TARGET_ASM_FILE_END): Define.
6506 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6507 (arc_add_jli_section): New function.
6508 (jli_call_scan): Likewise.
6509 (arc_reorg): Call jli_call_scan.
6510 (arc_output_addsi): Remove 'S' from printing asm operand.
6511 (arc_is_jli_call_p): New function.
6512 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6514 (movhi_insn): Likewise.
6515 (movsi_insn): Likewise.
6516 (movsi_set_cc_insn): Likewise.
6517 (loadqi_update): Likewise.
6518 (load_zeroextendqisi_update): Likewise.
6519 (load_signextendqisi_update): Likewise.
6520 (loadhi_update): Likewise.
6521 (load_zeroextendhisi_update): Likewise.
6522 (load_signextendhisi_update): Likewise.
6523 (loadsi_update): Likewise.
6524 (loadsf_update): Likewise.
6525 (movsicc_insn): Likewise.
6526 (bset_insn): Likewise.
6527 (bxor_insn): Likewise.
6528 (bclr_insn): Likewise.
6529 (bmsk_insn): Likewise.
6530 (bicsi3_insn): Likewise.
6531 (cmpsi_cc_c_insn): Likewise.
6532 (movsi_ne): Likewise.
6533 (movsi_cond_exec): Likewise.
6534 (clrsbsi2): Likewise.
6545 (call_i): Remove 'S' asm letter, add jli instruction.
6546 (call_value_i): Likewise.
6547 * config/arc/arc.op (mjli-always): New option.
6548 * config/arc/constraints.md (Cji): New constraint.
6549 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6551 (subsf3_fpx): Likewise.
6552 (mulsf3_fpx): Likewise.
6553 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6555 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6557 * doc/invoke.texi (ARC): Document mjli-always option.
6559 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6561 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6562 avoid addition with 0 and use incw and decw where possible.
6564 2018-01-26 Richard Biener <rguenther@suse.de>
6566 PR tree-optimization/81082
6567 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6568 association if it requires casting to unsigned.
6569 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6570 from fold_plusminus_mult_expr to catch important cases late when
6571 range info is available.
6573 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6575 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6576 * configure.ac (hidden_linkonce): New test.
6577 * configure: Regenerate.
6578 * config.in: Regenerate.
6580 2018-01-26 Julia Koval <julia.koval@intel.com>
6582 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6583 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6584 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6585 _mm_mask_bitshuffle_epi64_mask): Fix type.
6586 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6587 USI_FTYPE_V4DI_V4DI_USI): Remove.
6588 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6589 __builtin_ia32_vpshufbitqmb256_mask,
6590 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6591 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6592 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6594 2018-01-26 Alan Modra <amodra@gmail.com>
6597 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6598 UNSPEC_VBPERMQ. Sort other unspecs.
6600 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6602 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6604 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6607 * predict.c (drop_profile): Do not push/pop cfun; update also
6609 (handle_missing_profiles): Fix logic looking for zero profiles.
6611 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6614 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6615 on functions with #pragma omp declare simd or functions with simd
6617 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6618 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6619 Remove trailing \n from warning_at calls.
6621 2018-01-25 Tom de Vries <tom@codesourcery.com>
6624 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6625 for neutered workers.
6627 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6630 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6631 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6633 2018-01-24 Jeff Law <law@redhat.com>
6636 * i386.c (get_probe_interval): Move to earlier point.
6637 (ix86_compute_frame_layout): If -fstack-clash-protection and
6638 the frame is larger than the probe interval, then use pushes
6639 to save registers rather than reg->mem moves.
6640 (ix86_expand_prologue): Remove conditional for int_registers_saved
6643 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6646 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6647 min/max for never referenced object.
6649 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6652 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6654 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6655 attributes from DECL_ATTRIBUTES (decl) without affecting
6656 DECL_ATTRIBUTES (current_function_decl).
6657 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6658 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6660 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6662 PR tree-optimization/83979
6663 * fold-const.c (fold_comparison): Use constant_boolean_node
6664 instead of boolean_{true,false}_node.
6666 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6668 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6671 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6673 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6674 Simplify the clause that sets the length attribute.
6675 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6676 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6677 clause that sets the length attribute.
6678 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6680 2018-01-24 Tom de Vries <tom@codesourcery.com>
6683 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6684 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6685 Add strict parameter.
6686 (prevent_branch_around_nothing): Insert dummy insn between branch to
6687 label and label with no ptx insn inbetween.
6688 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6690 2018-01-24 Tom de Vries <tom@codesourcery.com>
6693 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6694 for neutered threads in warp.
6695 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6697 2018-01-24 Richard Biener <rguenther@suse.de>
6699 PR tree-optimization/83176
6700 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6703 2018-01-24 Richard Biener <rguenther@suse.de>
6705 PR tree-optimization/82819
6706 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6707 code generating pluses that are no-ops in the target precision.
6709 2018-01-24 Richard Biener <rguenther@suse.de>
6712 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6714 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6716 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6717 to merge probabilities.
6718 * predict.c (probably_never_executed): Also mark as cold functions
6719 with global 0 profile and guessed local profile.
6720 * profile-count.c (profile_probability::combine_with_count): New
6722 * profile-count.h (profile_probability::operator*,
6723 profile_probability::operator*=, profile_probability::operator/,
6724 profile_probability::operator/=): Reduce precision to adjusted
6725 and set value to guessed on contradictory divisions.
6726 (profile_probability::combine_with_freq): Remove.
6727 (profile_probability::combine_wiht_count): Declare.
6728 (profile_count::force_nonzero):: Set to adjusted.
6729 (profile_count::probability_in):: Set quality to adjusted.
6730 * tree-ssa-tail-merge.c (replace_block_by): Use
6733 2018-01-23 Andrew Waterman <andrew@sifive.com>
6734 Jim Wilson <jimw@sifive.com>
6736 * config/riscv/riscv.c (riscv_stack_boundary): New.
6737 (riscv_option_override): Set riscv_stack_boundary. Handle
6738 riscv_preferred_stack_boundary_arg.
6739 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6740 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6741 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6742 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6743 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6744 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6746 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6749 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6750 of struct ix86_frame.
6751 (ix86_expand_epilogue): Likewise. Add a local variable for
6752 the reg_save_offset field in struct ix86_frame.
6754 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6756 PR tree-optimization/82604
6757 * tree-loop-distribution.c (enum partition_kind): New enum item
6758 PKIND_PARTIAL_MEMSET.
6759 (partition_builtin_p): Support above new enum item.
6760 (generate_code_for_partition): Ditto.
6761 (compute_access_range): Differentiate cases that equality can be
6762 proven at all loops, the innermost loops or no loops.
6763 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6764 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6765 (finalize_partitions, distribute_loop): Don't fuse partition of
6766 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6767 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6770 2018-01-23 Martin Liska <mliska@suse.cz>
6772 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6773 order to ignore the predictor.
6774 (PRED_POLYMORPHIC_CALL): Likewise.
6775 (PRED_RECURSIVE_CALL): Likewise.
6777 2018-01-23 Martin Liska <mliska@suse.cz>
6779 * tree-profile.c (tree_profiling): Print function header to
6780 aware reader which function we are working on.
6781 * value-prof.c (gimple_find_values_to_profile): Do not print
6782 not interesting value histograms.
6784 2018-01-23 Martin Liska <mliska@suse.cz>
6786 * profile-count.h (enum profile_quality): Add
6787 profile_uninitialized as the first value. Do not number values
6788 as they are zero based.
6789 (profile_count::verify): Update sanity check.
6790 (profile_probability::verify): Likewise.
6792 2018-01-23 Nathan Sidwell <nathan@acm.org>
6794 * doc/invoke.texi (ffor-scope): Deprecate.
6796 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6798 PR tree-optimization/83510
6799 * domwalk.c (set_all_edges_as_executable): New function.
6800 (dom_walker::dom_walker): Convert bool param
6801 "skip_unreachable_blocks" to enum reachability. Move setup of
6802 edge flags to set_all_edges_as_executable and only do it when
6803 reachability is REACHABLE_BLOCKS.
6804 * domwalk.h (enum dom_walker::reachability): New enum.
6805 (dom_walker::dom_walker): Convert bool param
6806 "skip_unreachable_blocks" to enum reachability.
6807 (set_all_edges_as_executable): New decl.
6808 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6809 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6811 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6812 but converting true to REACHABLE_BLOCKS.
6813 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6815 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6816 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6817 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6819 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6820 if check_all_array_refs will be called.
6822 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6824 * tree.c (selftest::test_location_wrappers): Add more test
6827 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6829 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6830 (selftest::test_bit_in_range): Likewise.
6832 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6835 * doc/sourcebuild.texi (vect_float): Say that the selector
6836 only describes the situation when -funsafe-math-optimizations is on.
6837 (vect_float_strict): Document.
6839 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6841 PR tree-optimization/83965
6842 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6843 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6844 instead of checking only for a reduction.
6845 (vect_recog_widen_sum_pattern): Likewise.
6847 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6849 * predict.c (probably_never_executed): Only use precise profile info.
6850 (compute_function_frequency): Skip after inlining hack since we now
6851 have quality checking.
6853 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6855 * profile-count.h (profile_probability::very_unlikely,
6856 profile_probability::unlikely, profile_probability::even): Set
6857 precision to guessed.
6859 2018-01-23 Richard Biener <rguenther@suse.de>
6861 PR tree-optimization/83963
6862 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6863 Properly terminate dominator walk when crossing the exit edge not
6864 when visiting its source block.
6866 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6869 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6870 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6872 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6874 PR tree-optimization/83957
6875 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6876 semicolon after for body surrounded by braces.
6878 PR tree-optimization/83081
6879 * profile-count.h (profile_probability::split): New method.
6880 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6881 Use profile_probability::split.
6882 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6883 when splitting a single conditional jump into 2.
6885 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6887 PR tree-optimization/69452
6888 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6891 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6893 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6894 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6895 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6899 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6900 * config/rl78/rl78.md (movdi): New define_expand.
6901 * config/rl78/rl78.c (rl78_split_movdi): New function.
6903 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6906 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6908 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6909 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6910 128-bit to produce an UNSPEC move to get the double word with the
6911 signbit and then a shift directly to do signbit.
6912 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6913 implementation with a new version that just does either a direct
6914 move or a regular move. Move memory interface to separate insns.
6915 Move insns so they are next to the expander.
6916 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6917 with signbit move. Split big and little endian case.
6918 (signbit<mode>2_dm_mem_le): Likewise.
6919 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6920 (signbit<mode>2_dm2): Likewise.
6922 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6924 * config/rl78/rl78.md (anddi3): New define_expand.
6926 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6928 * config/rl78/rl78.md (umindi3): New define_expand.
6930 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6932 * config/rl78/rl78.md (smindi3): New define_expand.
6934 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6936 * config/rl78/rl78.md (smaxdi3): New define_expand.
6938 2018-01-22 Carl Love <cel@us.ibm.com>
6940 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6941 LVX_V1TI): Add macro expansion.
6942 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6943 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6944 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6945 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6946 Change check to determine if the instruction is a byte reversing
6947 entry. Fix typo in comment.
6948 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6949 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6950 Add def_builtin calls for new builtins.
6951 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6952 Add define_insn expansion.
6954 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6956 * config/rl78/rl78.md (umaxdi3): New define_expand.
6958 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6960 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6961 for non-QImode registers.
6963 2018-01-22 Richard Biener <rguenther@suse.de>
6965 PR tree-optimization/83963
6966 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6967 including the loop exit block.
6968 (scop_detection::merge_sese): Likewise.
6969 (scop_detection::add_scop): Do it here instead.
6971 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6973 * doc/sourcebuild.texi (arm_softfloat): Document.
6975 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6978 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6979 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6980 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6982 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6983 David Edelsohn <dje.gcc@gmail.com>
6986 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6987 Change "crset eq" to "crset 2".
6988 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6989 (*call_indirect_aix<mode>_nospec): Likewise.
6990 (*call_value_indirect_aix<mode>_nospec): Likewise.
6991 (*call_indirect_elfv2<mode>_nospec): Likewise.
6992 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6993 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6994 change assembly output from . to $.
6995 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6996 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6997 (*tablejump<mode>_internal1_nospec): Likewise.
6999 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7002 * config/sh/sh_optimize_sett_clrt.cc:
7003 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7005 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7007 PR tree-optimization/83940
7008 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7009 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7010 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7011 use it to pass back the definition type.
7012 (vect_check_store_rhs): Likewise rhs_dt_out.
7013 (vect_build_gather_load_calls): Add a mask_dt argument and use
7014 it instead of a call to vect_is_simple_use.
7015 (vectorizable_store): Update calls to vect_check_load_store_mask
7016 and vect_check_store_rhs. Use the dt returned by the latter instead
7017 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7018 instead of calls to vect_is_simple_use. Pass the scalar rather
7019 than the vector operand to vect_is_simple_use when handling
7020 second and subsequent copies of an rhs value.
7021 (vectorizable_load): Update calls to vect_check_load_store_mask
7022 and vect_build_gather_load_calls. Use the cached mask_dt and
7023 gs_info.offset_dt instead of calls to vect_is_simple_use.
7025 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7028 * tree-emutls.c: Include gimplify.h.
7029 (lower_emutls_2): New function.
7030 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7031 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7032 it before further processing.
7035 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7036 UINTVAL (trueop1) instead of INTVAL (op1).
7038 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7042 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7043 INCOMING_FRAME_SP_OFFSET if not defined.
7044 (scan_trace): Add ENTRY argument. If true and
7045 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7046 emit a note to adjust the CFA offset.
7047 (create_cfi_notes): Adjust scan_trace callers.
7048 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7049 INCOMING_FRAME_SP_OFFSET in the CIE.
7050 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7051 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7053 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7054 * doc/tm.texi: Regenerated.
7056 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7058 PR rtl-optimization/83147
7059 * lra-constraints.c (remove_inheritance_pseudos): Use
7060 lra_substitute_pseudo_within_insn.
7062 2018-01-19 Tom de Vries <tom@codesourcery.com>
7063 Cesar Philippidis <cesar@codesourcery.com>
7066 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7068 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7071 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7072 spaces for function labels.
7074 2018-01-19 Martin Liska <mliska@suse.cz>
7076 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7077 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7078 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7079 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7080 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7081 (PRED_CONST_RETURN): Change from 69 to 65.
7082 (PRED_NULL_RETURN): Change from 91 to 71.
7083 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7084 (PRED_LOOP_GUARD): Change from 66 to 73.
7086 2018-01-19 Martin Liska <mliska@suse.cz>
7088 * predict.c (predict_insn_def): Add new assert.
7089 (struct branch_predictor): Change type to signed integer.
7090 (test_prediction_value_range): Amend test to cover
7092 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7093 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7094 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7095 (PRED_LOOP_IV_COMPARE): Likewise.
7096 * predict.h (PROB_UNINITIALIZED): Define new constant.
7098 2018-01-19 Martin Liska <mliska@suse.cz>
7100 * predict.c (dump_prediction): Add new format for
7101 analyze_brprob.py script which is enabled with -details
7103 * profile-count.h (precise_p): New function.
7105 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7107 PR tree-optimization/83922
7108 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7109 there are no statements that need masking.
7110 (vect_active_double_reduction_p): New function.
7111 (vect_analyze_loop_operations): Use it when handling phis that
7112 are not in the loop header.
7114 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7116 PR tree-optimization/83914
7117 * tree-vect-loop.c (vectorizable_induction): Don't convert
7118 init_expr or apply the peeling adjustment for inductions
7119 that are nested within the vectorized loop.
7121 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7123 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7126 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7130 * function.h (gimplify_parameters): Add gimple_seq * argument.
7131 * function.c: Include gimple.h and options.h.
7132 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7133 for the added local temporaries if needed.
7134 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7135 if there are any parameter cleanups, wrap whole body into a
7136 try/finally with the cleanups.
7138 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7141 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7142 Use GET_MODE_CLASS for scalar floating point.
7144 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7148 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7149 Fix call of call_cgraph_insertion_hooks.
7151 2018-01-18 Martin Sebor <msebor@redhat.com>
7153 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7155 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7158 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7161 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7164 * common.opt: (-ffile-prefix-map): New option.
7165 * opts.c (common_handle_option): Defer it.
7166 * opts-global.c (handle_common_deferred_options): Handle it.
7167 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7168 * file-prefix-map.h: New file.
7169 (remap_debug_filename, add_debug_prefix_map): ...here.
7170 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7171 * final.c (debug_prefix_map, add_debug_prefix_map
7172 remap_debug_filename): Move to...
7173 * file-prefix-map.c: New file.
7174 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7175 generalize, get rid of alloca(), use strrchr() instead of strchr().
7176 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7177 Implement in terms of add_prefix_map().
7178 (remap_macro_filename, remap_debug_filename): Implement in term of
7180 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7181 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7182 * dbxout.c: Include file-prefix-map.h.
7183 * varasm.c: Likewise.
7184 * vmsdbgout.c: Likewise.
7185 * xcoffout.c: Likewise.
7186 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7187 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7188 * doc/invoke.texi (-ffile-prefix-map): Document.
7189 (-fdebug-prefix-map): Update description.
7191 2018-01-18 Martin Liska <mliska@suse.cz>
7193 * config/i386/i386.c (indirect_thunk_name): Document that also
7195 (output_indirect_thunk): Document why both instructions
7196 (pause and lfence) are generated.
7198 2018-01-18 Richard Biener <rguenther@suse.de>
7200 PR tree-optimization/83887
7201 * graphite-scop-detection.c
7202 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7203 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7204 (scop_detection::merge_sese): Re-implement with a flood-fill
7205 algorithm that properly finds a SESE region if it exists.
7207 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7210 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7211 pointer_diff optimizations use view_convert instead of convert.
7213 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7215 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7216 Generate different code for -mno-speculate-indirect-jumps.
7217 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7218 (*call_indirect_aix<mode>): Disable for
7219 -mno-speculate-indirect-jumps.
7220 (*call_indirect_aix<mode>_nospec): New define_insn.
7221 (*call_value_indirect_aix<mode>): Disable for
7222 -mno-speculate-indirect-jumps.
7223 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7224 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7225 -mno-speculate-indirect-jumps.
7226 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7228 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7230 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7231 long double type, set the flags for noting the default long double
7232 type, even if we don't pass or return a long double type.
7234 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7237 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7240 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7242 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7243 support for merge[hl].
7244 (fold_mergehl_helper): New helper function.
7245 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7246 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7247 (altivec_vmrglw_direct): Add xxmrglw insn.
7249 2018-01-17 Andrew Waterman <andrew@sifive.com>
7251 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7252 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7254 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7257 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7258 call the lto_location_cache before reading the
7259 DECL_SOURCE_LOCATION of the types.
7261 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7262 Richard Sandiford <richard.sandiford@linaro.org>
7264 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7265 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7266 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7267 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7268 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7270 * config/aarch64/constraints.md (aarch64_movti_operand):
7272 * config/aarch64/predicates.md (Uti): Add new constraint.
7274 2018-01-17 Carl Love <cel@us.ibm.com>
7276 * config/rs6000/vsx.md (define_expand xl_len_r,
7277 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7278 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7280 (define_expand, define_insn): Move the shift left from the
7281 define_insn to the define_expand for lxvl and stxvl instructions.
7282 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7283 and XL_LEN_R definitions to PURE.
7285 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7287 * config/i386/i386.c (indirect_thunk_name): Declare regno
7288 as unsigned int. Compare regno with INVALID_REGNUM.
7289 (output_indirect_thunk): Ditto.
7290 (output_indirect_thunk_function): Ditto.
7291 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7292 in the call to output_indirect_thunk_function.
7294 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7297 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7298 rather than the size of inner_type to determine the stack slot size
7299 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7301 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7304 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7307 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7309 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7310 endian Linux systems to optionally enable multilibs for selecting
7311 the long double type if the user configured an explicit type.
7312 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7313 have no long double multilibs if not defined.
7314 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7315 warn if the user used -mabi={ieee,ibm}longdouble and we built
7316 multilibs for long double.
7317 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7318 appropriate multilib option.
7319 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7321 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7322 for building long double multilibs.
7323 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7325 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7327 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7330 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7332 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7335 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7338 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7341 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7343 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7344 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7346 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7348 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7349 different rtl trees depending on TARGET_64BIT.
7350 (rs6000_gen_lvx): Likewise.
7352 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7354 * config/visium/visium.md (nop): Tweak comment.
7355 (hazard_nop): Likewise.
7357 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7359 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7360 -mspeculate-indirect-jumps.
7361 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7362 for -mno-speculate-indirect-jumps.
7363 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7364 (*call_value_indirect_elfv2<mode>): Disable for
7365 -mno-speculate-indirect-jumps.
7366 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7367 (indirect_jump): Emit different RTL for
7368 -mno-speculate-indirect-jumps.
7369 (*indirect_jump<mode>): Disable for
7370 -mno-speculate-indirect-jumps.
7371 (*indirect_jump<mode>_nospec): New define_insn.
7372 (tablejump): Emit different RTL for
7373 -mno-speculate-indirect-jumps.
7374 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7375 (tablejumpsi_nospec): New define_expand.
7376 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7377 (tablejumpdi_nospec): New define_expand.
7378 (*tablejump<mode>_internal1): Disable for
7379 -mno-speculate-indirect-jumps.
7380 (*tablejump<mode>_internal1_nospec): New define_insn.
7381 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7384 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7386 * caller-save.c (insert_save): Drop unnecessary parameter. All
7389 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7390 Richard Biener <rguenth@suse.de>
7393 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7394 return early, inline manually is_gimple_sizepos. Make sure if we
7395 call gimplify_expr we don't end up with a gimple constant.
7396 * tree.c (variably_modified_type_p): Don't return true for
7397 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7398 * gimplify.h (is_gimple_sizepos): Remove.
7400 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7402 PR tree-optimization/83857
7403 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7404 vectorizable_live_operation for pure SLP statements.
7405 (vectorizable_live_operation): Handle PHIs.
7407 2018-01-16 Richard Biener <rguenther@suse.de>
7409 PR tree-optimization/83867
7410 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7411 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7413 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7416 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7417 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7418 If off is not INTEGER_CST, issue a may not be aligned warning
7419 rather than isn't aligned. Use isn%'t rather than isn't.
7420 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7422 <case MULT_EXPR>: Improve the case when bottom and one of the
7423 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7424 operand, in that case check if the other operand is multiple of
7425 bottom divided by the INTEGER_CST operand.
7427 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7430 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7431 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7432 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7433 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7434 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7435 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7436 (pa_function_arg_size): New function.
7438 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7440 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7441 in a separate statement.
7443 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7445 PR tree-optimization/83847
7446 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7447 group gathers and scatters.
7449 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7451 PR rtl-optimization/86620
7452 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7454 PR rtl-optimization/83213
7455 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7456 to last if both are JUMP_INSNs.
7458 PR tree-optimization/83843
7459 * gimple-ssa-store-merging.c
7460 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7461 store_immediate_info for bswap/nop orig_stores.
7463 2018-01-15 Andrew Waterman <andrew@sifive.com>
7465 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7467 <UDIV>: Increase cost if !TARGET_DIV.
7469 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7471 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7472 (define_attr "cr_logical_3op"): New.
7473 (cceq_ior_compare): Adjust.
7474 (cceq_ior_compare_complement): Adjust.
7475 (*cceq_rev_compare): Adjust.
7476 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7477 (is_cracked_insn): Adjust.
7478 (insn_must_be_first_in_group): Adjust.
7479 * config/rs6000/40x.md: Adjust.
7480 * config/rs6000/440.md: Adjust.
7481 * config/rs6000/476.md: Adjust.
7482 * config/rs6000/601.md: Adjust.
7483 * config/rs6000/603.md: Adjust.
7484 * config/rs6000/6xx.md: Adjust.
7485 * config/rs6000/7450.md: Adjust.
7486 * config/rs6000/7xx.md: Adjust.
7487 * config/rs6000/8540.md: Adjust.
7488 * config/rs6000/cell.md: Adjust.
7489 * config/rs6000/e300c2c3.md: Adjust.
7490 * config/rs6000/e500mc.md: Adjust.
7491 * config/rs6000/e500mc64.md: Adjust.
7492 * config/rs6000/e5500.md: Adjust.
7493 * config/rs6000/e6500.md: Adjust.
7494 * config/rs6000/mpc.md: Adjust.
7495 * config/rs6000/power4.md: Adjust.
7496 * config/rs6000/power5.md: Adjust.
7497 * config/rs6000/power6.md: Adjust.
7498 * config/rs6000/power7.md: Adjust.
7499 * config/rs6000/power8.md: Adjust.
7500 * config/rs6000/power9.md: Adjust.
7501 * config/rs6000/rs64.md: Adjust.
7502 * config/rs6000/titan.md: Adjust.
7504 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7506 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7507 ix86_indirect_branch_register logic.
7509 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7511 * config/i386/constraints.md (Bs): Update
7512 ix86_indirect_branch_register check. Don't check
7513 ix86_indirect_branch_register with GOT_memory_operand.
7515 * config/i386/predicates.md (GOT_memory_operand): Don't check
7516 ix86_indirect_branch_register here.
7517 (GOT32_symbol_operand): Likewise.
7519 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7521 * config/i386/predicates.md (constant_call_address_operand):
7522 Rewrite ix86_indirect_branch_register logic.
7523 (sibcall_insn_operand): Likewise.
7525 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7527 * config/i386/constraints.md (Bs): Replace
7528 ix86_indirect_branch_thunk_register with
7529 ix86_indirect_branch_register.
7531 * config/i386/i386.md (indirect_jump): Likewise.
7532 (tablejump): Likewise.
7533 (*sibcall_memory): Likewise.
7534 (*sibcall_value_memory): Likewise.
7535 Peepholes of indirect call and jump via memory: Likewise.
7536 * config/i386/i386.opt: Likewise.
7537 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7538 (GOT_memory_operand): Likewise.
7539 (call_insn_operand): Likewise.
7540 (sibcall_insn_operand): Likewise.
7541 (GOT32_symbol_operand): Likewise.
7543 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7546 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7547 type rather than type addr's type points to.
7548 (expand_omp_atomic_mutex): Likewise.
7549 (expand_omp_atomic): Likewise.
7551 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7554 * config/i386/i386.c (output_indirect_thunk_function): Use
7555 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7556 for __x86_return_thunk.
7558 2018-01-15 Richard Biener <rguenther@suse.de>
7561 * expmed.c (extract_bit_field_1): Fix typo.
7563 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7566 * config/arm/iterators.md (VF): New mode iterator.
7567 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7568 Remove integer-related logic from pattern.
7569 (neon_vabd<mode>_3): Likewise.
7571 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7574 * common.opt (fstrict-overflow): No longer an alias.
7575 (fwrapv-pointer): New option.
7576 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7577 also for pointer types based on flag_wrapv_pointer.
7578 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7579 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7580 opts->x_flag_wrapv got set.
7581 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7582 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7583 POINTER_TYPE_OVERFLOW_UNDEFINED.
7584 * match.pd: Likewise in address comparison pattern.
7585 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7587 2018-01-15 Richard Biener <rguenther@suse.de>
7590 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7591 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7592 Reset type names to their identifier if their TYPE_DECL doesn't
7593 have linkage (and thus is used for ODR and devirt).
7594 (save_debug_info_for_decl): Remove.
7595 (save_debug_info_for_type): Likewise.
7596 (add_tree_to_fld_list): Adjust.
7597 * tree-pretty-print.c (dump_generic_node): Make dumping of
7598 type names more robust.
7600 2018-01-15 Richard Biener <rguenther@suse.de>
7602 * BASE-VER: Bump to 8.0.1.
7604 2018-01-14 Martin Sebor <msebor@redhat.com>
7607 * builtins.c (check_access): Avoid warning when the no-warning bit
7610 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7612 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7613 * ira-color (allocno_hard_regs_compare): Likewise.
7615 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7618 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7619 Use .pushsection/.popsection.
7621 2018-01-14 Martin Sebor <msebor@redhat.com>
7624 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7626 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7628 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7629 entry from extra_headers.
7630 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7631 extra_headers, make the list bitwise identical to the i?86-*-* one.
7633 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7635 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7636 -mcmodel=large with -mindirect-branch=thunk,
7637 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7638 -mfunction-return=thunk-extern.
7639 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7640 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7641 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7643 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7645 * config/i386/i386.c (print_reg): Print the name of the full
7646 integer register without '%'.
7647 (ix86_print_operand): Handle 'V'.
7648 * doc/extend.texi: Document 'V' modifier.
7650 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7652 * config/i386/constraints.md (Bs): Disallow memory operand for
7653 -mindirect-branch-register.
7655 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7656 (GOT_memory_operand): Likewise.
7657 (call_insn_operand): Likewise.
7658 (sibcall_insn_operand): Likewise.
7659 (GOT32_symbol_operand): Likewise.
7660 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7661 for -mindirect-branch-register.
7662 (tablejump): Likewise.
7663 (*sibcall_memory): Likewise.
7664 (*sibcall_value_memory): Likewise.
7665 Disallow peepholes of indirect call and jump via memory for
7666 -mindirect-branch-register.
7667 (*call_pop): Replace m with Bw.
7668 (*call_value_pop): Likewise.
7669 (*sibcall_pop_memory): Replace m with Bs.
7670 * config/i386/i386.opt (mindirect-branch-register): New option.
7671 * doc/invoke.texi: Document -mindirect-branch-register option.
7673 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7675 * config/i386/i386-protos.h (ix86_output_function_return): New.
7676 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7677 set function_return_type.
7678 (indirect_thunk_name): Add ret_p to indicate thunk for function
7680 (output_indirect_thunk_function): Pass false to
7681 indirect_thunk_name.
7682 (ix86_output_indirect_branch_via_reg): Likewise.
7683 (ix86_output_indirect_branch_via_push): Likewise.
7684 (output_indirect_thunk_function): Create alias for function
7685 return thunk if regno < 0.
7686 (ix86_output_function_return): New function.
7687 (ix86_handle_fndecl_attribute): Handle function_return.
7688 (ix86_attribute_table): Add function_return.
7689 * config/i386/i386.h (machine_function): Add
7690 function_return_type.
7691 * config/i386/i386.md (simple_return_internal): Use
7692 ix86_output_function_return.
7693 (simple_return_internal_long): Likewise.
7694 * config/i386/i386.opt (mfunction-return=): New option.
7695 (indirect_branch): Mention -mfunction-return=.
7696 * doc/extend.texi: Document function_return function attribute.
7697 * doc/invoke.texi: Document -mfunction-return= option.
7699 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7701 * config/i386/i386-opts.h (indirect_branch): New.
7702 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7703 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7704 with local indirect jump when converting indirect call and jump.
7705 (ix86_set_indirect_branch_type): New.
7706 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7707 (indirectlabelno): New.
7708 (indirect_thunk_needed): Likewise.
7709 (indirect_thunk_bnd_needed): Likewise.
7710 (indirect_thunks_used): Likewise.
7711 (indirect_thunks_bnd_used): Likewise.
7712 (INDIRECT_LABEL): Likewise.
7713 (indirect_thunk_name): Likewise.
7714 (output_indirect_thunk): Likewise.
7715 (output_indirect_thunk_function): Likewise.
7716 (ix86_output_indirect_branch_via_reg): Likewise.
7717 (ix86_output_indirect_branch_via_push): Likewise.
7718 (ix86_output_indirect_branch): Likewise.
7719 (ix86_output_indirect_jmp): Likewise.
7720 (ix86_code_end): Call output_indirect_thunk_function if needed.
7721 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7723 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7724 (ix86_attribute_table): Add indirect_branch.
7725 * config/i386/i386.h (machine_function): Add indirect_branch_type
7726 and has_local_indirect_jump.
7727 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7729 (tablejump): Likewise.
7730 (*indirect_jump): Use ix86_output_indirect_jmp.
7731 (*tablejump_1): Likewise.
7732 (simple_return_indirect_internal): Likewise.
7733 * config/i386/i386.opt (mindirect-branch=): New option.
7734 (indirect_branch): New.
7737 (thunk-inline): Likewise.
7738 (thunk-extern): Likewise.
7739 * doc/extend.texi: Document indirect_branch function attribute.
7740 * doc/invoke.texi: Document -mindirect-branch= option.
7742 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7745 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7747 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7749 * ipa-inline.c (want_inline_small_function_p): Return false if
7750 inlining has already failed with CIF_FINAL_ERROR.
7751 (update_caller_keys): Call want_inline_small_function_p before
7753 (update_callee_keys): Likewise.
7755 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7757 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7759 (rs6000_quadword_masked_address_p): Likewise.
7760 (quad_aligned_load_p): Likewise.
7761 (quad_aligned_store_p): Likewise.
7762 (const_load_sequence_p): Add comment to describe the outer-most loop.
7763 (mimic_memory_attributes_and_flags): New function.
7764 (rs6000_gen_stvx): Likewise.
7765 (replace_swapped_aligned_store): Likewise.
7766 (rs6000_gen_lvx): Likewise.
7767 (replace_swapped_aligned_load): Likewise.
7768 (replace_swapped_load_constant): Capitalize argument name in
7769 comment describing this function.
7770 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7771 and stores that access quad-word aligned addresses and replace
7772 with stvx or lvx instructions when appropriate.
7773 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7774 New function prototype.
7775 (rs6000_quadword_masked_address_p): Likewise.
7776 (rs6000_gen_lvx): Likewise.
7777 (rs6000_gen_stvx): Likewise.
7778 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7779 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7780 when memory address is aligned.
7781 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7782 this split to select lvx instruction when memory address is aligned.
7783 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7784 instruction when memory address is aligned.
7785 (*vsx_le_perm_load_v16qi): Likewise.
7786 (four unnamed splitters): Modify to select the stvx instruction
7787 when memory is aligned.
7789 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7791 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7792 which appears in the queue multiple times.
7794 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7795 Alan Hayward <alan.hayward@arm.com>
7796 David Sherwood <david.sherwood@arm.com>
7798 * tree-vectorizer.h (vec_lower_bound): New structure.
7799 (_loop_vec_info): Add check_nonzero and lower_bounds.
7800 (LOOP_VINFO_CHECK_NONZERO): New macro.
7801 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7802 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7803 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7804 fields. Make seg_len the distance travelled, not including the
7806 (dr_direction_indicator): Declare.
7807 (dr_zero_step_indicator): Likewise.
7808 (dr_known_forward_stride_p): Likewise.
7809 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7811 (runtime_alias_check_p): Allow runtime alias checks with
7813 (operator ==): Compare access_size and align.
7814 (prune_runtime_alias_test_list): Rework for new distinction between
7815 the access_size and seg_len.
7816 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7818 (get_segment_min_max): New function.
7819 (create_intersect_range_checks): Use it.
7820 (dr_step_indicator): New function.
7821 (dr_direction_indicator): Likewise.
7822 (dr_zero_step_indicator): Likewise.
7823 (dr_known_forward_stride_p): Likewise.
7824 * tree-loop-distribution.c (data_ref_segment_size): Return
7825 DR_STEP * (niters - 1).
7826 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7828 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7829 (vect_preserves_scalar_order_p): New function, split out from...
7830 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7831 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7832 (vect_vfa_access_size): New function.
7833 (vect_vfa_align): Likewise.
7834 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7835 (dump_lower_bound): New function.
7836 (vect_check_lower_bound): Likewise.
7837 (vect_small_gap_p): Likewise.
7838 (vectorizable_with_step_bound_p): Likewise.
7839 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7840 depencies if the vectorization factor is 1. Convert the checks
7841 for nonzero steps into checks on the bounds of DR_STEP. Try using
7842 a bunds check for variable steps if the minimum required step is
7843 relatively small. Update calls to the dr_with_seg_len
7844 constructor and to vect_compile_time_alias.
7845 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7847 (vect_loop_versioning): Call it.
7848 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7850 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7853 Alan Hayward <alan.hayward@arm.com>
7854 David Sherwood <david.sherwood@arm.com>
7856 * doc/sourcebuild.texi (vect_scatter_store): Document.
7857 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7859 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7861 * genopinit.c (main): Add supports_vec_scatter_store and
7862 supports_vec_scatter_store_cached to target_optabs.
7863 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7864 IFN_MASK_SCATTER_STORE.
7865 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7867 * internal-fn.h (internal_store_fn_p): Declare.
7868 (internal_fn_stored_value_index): Likewise.
7869 * internal-fn.c (scatter_store_direct): New macro.
7870 (expand_scatter_store_optab_fn): New function.
7871 (direct_scatter_store_optab_supported_p): New macro.
7872 (internal_store_fn_p): New function.
7873 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7874 IFN_MASK_SCATTER_STORE.
7875 (internal_fn_mask_index): Likewise.
7876 (internal_fn_stored_value_index): New function.
7877 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7879 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7880 * optabs-query.c (supports_vec_scatter_store_p): New function.
7881 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7882 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7883 true for scatter stores.
7884 (vect_gather_scatter_fn_p): Handle scatter stores too.
7885 (vect_check_gather_scatter): Consider using scatter stores if
7886 supports_vec_scatter_store_p.
7887 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7889 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7890 internal_fn_stored_value_index.
7891 (check_load_store_masking): Handle scatter stores too.
7892 (vect_get_store_rhs): Make public.
7893 (vectorizable_call): Use internal_store_fn_p.
7894 (vectorizable_store): Handle scatter store internal functions.
7895 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7896 when deciding whether the end of the group has been reached.
7897 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7898 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7899 (mask_scatter_store<mode>): New insns.
7901 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7902 Alan Hayward <alan.hayward@arm.com>
7903 David Sherwood <david.sherwood@arm.com>
7905 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7906 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7907 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7909 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7910 Use vect_truncate_gather_scatter_offset if we can't treat the
7911 operation as a normal gather load or scatter store.
7912 (get_group_load_store_type): Take the gather_scatter_info
7913 as argument. Try using a gather load or scatter store for
7914 single-element groups.
7915 (get_load_store_type): Update calls to get_group_load_store_type
7916 and vect_use_strided_gather_scatters_p.
7918 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7919 Alan Hayward <alan.hayward@arm.com>
7920 David Sherwood <david.sherwood@arm.com>
7922 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7923 optional tree argument.
7924 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7926 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7927 but continue to use the current value as a fallback.
7928 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7929 to compare the updates.
7930 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7931 (get_load_store_type): Use it when handling a strided access.
7932 (vect_get_strided_load_store_ops): New function.
7933 (vect_get_data_ptr_increment): Likewise.
7934 (vectorizable_load): Handle strided gather loads. Always pass
7935 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7937 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7938 Alan Hayward <alan.hayward@arm.com>
7939 David Sherwood <david.sherwood@arm.com>
7941 * doc/md.texi (gather_load@var{m}): Document.
7942 (mask_gather_load@var{m}): Likewise.
7943 * genopinit.c (main): Add supports_vec_gather_load and
7944 supports_vec_gather_load_cached to target_optabs.
7945 * optabs-tree.c (init_tree_optimization_optabs): Use
7946 ggc_cleared_alloc to allocate target_optabs.
7947 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7948 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7950 * internal-fn.h (internal_load_fn_p): Declare.
7951 (internal_gather_scatter_fn_p): Likewise.
7952 (internal_fn_mask_index): Likewise.
7953 (internal_gather_scatter_fn_supported_p): Likewise.
7954 * internal-fn.c (gather_load_direct): New macro.
7955 (expand_gather_load_optab_fn): New function.
7956 (direct_gather_load_optab_supported_p): New macro.
7957 (direct_internal_fn_optab): New function.
7958 (internal_load_fn_p): Likewise.
7959 (internal_gather_scatter_fn_p): Likewise.
7960 (internal_fn_mask_index): Likewise.
7961 (internal_gather_scatter_fn_supported_p): Likewise.
7962 * optabs-query.c (supports_at_least_one_mode_p): New function.
7963 (supports_vec_gather_load_p): Likewise.
7964 * optabs-query.h (supports_vec_gather_load_p): Declare.
7965 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7966 and memory_type field.
7967 (NUM_PATTERNS): Bump to 15.
7968 * tree-vect-data-refs.c: Include internal-fn.h.
7969 (vect_gather_scatter_fn_p): New function.
7970 (vect_describe_gather_scatter_call): Likewise.
7971 (vect_check_gather_scatter): Try using internal functions for
7972 gather loads. Recognize existing calls to a gather load function.
7973 (vect_analyze_data_refs): Consider using gather loads if
7974 supports_vec_gather_load_p.
7975 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7976 (vect_get_gather_scatter_offset_type): Likewise.
7977 (vect_convert_mask_for_vectype): Likewise.
7978 (vect_add_conversion_to_patterm): Likewise.
7979 (vect_try_gather_scatter_pattern): Likewise.
7980 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7981 (vect_vect_recog_func_ptrs): Add it.
7982 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7983 internal_fn_mask_index and internal_gather_scatter_fn_p.
7984 (check_load_store_masking): Take the gather_scatter_info as an
7985 argument and handle gather loads.
7986 (vect_get_gather_scatter_ops): New function.
7987 (vectorizable_call): Check internal_load_fn_p.
7988 (vectorizable_load): Likewise. Handle gather load internal
7990 (vectorizable_store): Update call to check_load_store_masking.
7991 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7992 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7993 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7994 (aarch64_gather_scale_operand_d): New predicates.
7995 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7996 (mask_gather_load<mode>): New insns.
7998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7999 Alan Hayward <alan.hayward@arm.com>
8000 David Sherwood <david.sherwood@arm.com>
8002 * optabs.def (fold_left_plus_optab): New optab.
8003 * doc/md.texi (fold_left_plus_@var{m}): Document.
8004 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8005 * internal-fn.c (fold_left_direct): Define.
8006 (expand_fold_left_optab_fn): Likewise.
8007 (direct_fold_left_optab_supported_p): Likewise.
8008 * fold-const-call.c (fold_const_fold_left): New function.
8009 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8010 * tree-parloops.c (valid_reduction_p): New function.
8011 (gather_scalar_reductions): Use it.
8012 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8013 (vect_finish_replace_stmt): Declare.
8014 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8015 (needs_fold_left_reduction_p): New function, split out from...
8016 (vect_is_simple_reduction): ...here. Accept reductions that
8017 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8018 (vect_force_simple_reduction): Also store the reduction type in
8019 the assignment's STMT_VINFO_REDUC_TYPE.
8020 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8021 (merge_with_identity): New function.
8022 (vect_expand_fold_left): Likewise.
8023 (vectorize_fold_left_reduction): Likewise.
8024 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8025 scalar phi in place for it. Check for target support and reject
8026 cases that would reassociate the operation. Defer the transform
8027 phase to vectorize_fold_left_reduction.
8028 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8029 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8030 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8032 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8034 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8035 call to ifc_temp_var.
8037 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8038 Alan Hayward <alan.hayward@arm.com>
8039 David Sherwood <david.sherwood@arm.com>
8041 * target.def (legitimize_address_displacement): Take the original
8042 offset as a poly_int.
8043 * targhooks.h (default_legitimize_address_displacement): Update
8045 * targhooks.c (default_legitimize_address_displacement): Likewise.
8046 * doc/tm.texi: Regenerate.
8047 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8048 as an argument, moving assert of ad->disp == ad->disp_term to...
8049 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8050 Try calling targetm.legitimize_address_displacement before expanding
8051 the address rather than afterwards, and adjust for the new interface.
8052 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8053 Match the new hook interface. Handle SVE addresses.
8054 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8057 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8059 * Makefile.in (OBJS): Add early-remat.o.
8060 * target.def (select_early_remat_modes): New hook.
8061 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8062 * doc/tm.texi: Regenerate.
8063 * targhooks.h (default_select_early_remat_modes): Declare.
8064 * targhooks.c (default_select_early_remat_modes): New function.
8065 * timevar.def (TV_EARLY_REMAT): New timevar.
8066 * passes.def (pass_early_remat): New pass.
8067 * tree-pass.h (make_pass_early_remat): Declare.
8068 * early-remat.c: New file.
8069 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8071 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8073 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8074 Alan Hayward <alan.hayward@arm.com>
8075 David Sherwood <david.sherwood@arm.com>
8077 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8078 vfm1 with a bound_epilog parameter.
8079 (vect_do_peeling): Update calls accordingly, and move the prologue
8080 call earlier in the function. Treat the base bound_epilog as 0 for
8081 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8082 this base when peeling for gaps.
8083 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8084 with fully-masked loops.
8085 (vect_estimate_min_profitable_iters): Handle the single peeled
8086 iteration in that case.
8088 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8089 Alan Hayward <alan.hayward@arm.com>
8090 David Sherwood <david.sherwood@arm.com>
8092 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8093 single-element interleaving even if the size is not a power of 2.
8094 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8095 accesses for single-element interleaving if the group size is
8098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8099 Alan Hayward <alan.hayward@arm.com>
8100 David Sherwood <david.sherwood@arm.com>
8102 * doc/md.texi (fold_extract_last_@var{m}): Document.
8103 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8104 * optabs.def (fold_extract_last_optab): New optab.
8105 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8106 * internal-fn.c (fold_extract_direct): New macro.
8107 (expand_fold_extract_optab_fn): Likewise.
8108 (direct_fold_extract_optab_supported_p): Likewise.
8109 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8110 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8111 EXTRACT_LAST_REDUCTION.
8112 (get_initial_def_for_reduction): Do not create an initial vector
8113 for EXTRACT_LAST_REDUCTION reductions.
8114 (vectorizable_reduction): Leave the scalar phi in place for
8115 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8116 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8117 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8118 transform phase to vectorizable_condition.
8119 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8121 (vect_finish_stmt_generation): ...here.
8122 (vect_finish_replace_stmt): New function.
8123 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8124 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8126 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8128 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8129 Alan Hayward <alan.hayward@arm.com>
8130 David Sherwood <david.sherwood@arm.com>
8132 * doc/md.texi (extract_last_@var{m}): Document.
8133 * optabs.def (extract_last_optab): New optab.
8134 * internal-fn.def (EXTRACT_LAST): New internal function.
8135 * internal-fn.c (cond_unary_direct): New macro.
8136 (expand_cond_unary_optab_fn): Likewise.
8137 (direct_cond_unary_optab_supported_p): Likewise.
8138 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8139 loops using EXTRACT_LAST.
8140 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8141 (extract_last_<mode>): ...this optab.
8142 (vec_extract<mode><Vel>): Update accordingly.
8144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8145 Alan Hayward <alan.hayward@arm.com>
8146 David Sherwood <david.sherwood@arm.com>
8148 * target.def (empty_mask_is_expensive): New hook.
8149 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8150 * doc/tm.texi: Regenerate.
8151 * targhooks.h (default_empty_mask_is_expensive): Declare.
8152 * targhooks.c (default_empty_mask_is_expensive): New function.
8153 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8154 if the target says that empty masks are expensive.
8155 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8157 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8159 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8160 Alan Hayward <alan.hayward@arm.com>
8161 David Sherwood <david.sherwood@arm.com>
8163 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8164 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8165 (vect_use_loop_mask_for_alignment_p): New function.
8166 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8167 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8168 niters_skip argument. Make sure that the first niters_skip elements
8169 of the first iteration are inactive.
8170 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8171 Update call to vect_set_loop_masks_directly.
8172 (get_misalign_in_elems): New function, split out from...
8173 (vect_gen_prolog_loop_niters): ...here.
8174 (vect_update_init_of_dr): Take a code argument that specifies whether
8175 the adjustment should be added or subtracted.
8176 (vect_update_init_of_drs): Likewise.
8177 (vect_prepare_for_masked_peels): New function.
8178 (vect_do_peeling): Skip prologue peeling if we're using a mask
8179 instead. Update call to vect_update_inits_of_drs.
8180 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8182 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8183 alignment. Do not include the number of peeled iterations in
8184 the minimum threshold in that case.
8185 (vectorizable_induction): Adjust the start value down by
8186 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8187 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8188 Take the number of skipped iterations into account when calculating
8190 * tree-vect-stmts.c (vect_gen_while_not): New function.
8192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8193 Alan Hayward <alan.hayward@arm.com>
8194 David Sherwood <david.sherwood@arm.com>
8196 * doc/sourcebuild.texi (vect_fully_masked): Document.
8197 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8199 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8201 (vect_analyze_loop_2): ...here. Don't check the vectorization
8202 factor against the number of loop iterations if the loop is
8205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8206 Alan Hayward <alan.hayward@arm.com>
8207 David Sherwood <david.sherwood@arm.com>
8209 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8210 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8211 (dump_groups): Update accordingly.
8212 (iv_use::mem_type): New member variable.
8213 (address_p): New function.
8214 (record_use): Add a mem_type argument and initialize the new
8216 (record_group_use): Add a mem_type argument. Use address_p.
8217 Remove obsolete null checks of base_object. Update call to record_use.
8218 (find_interesting_uses_op): Update call to record_group_use.
8219 (find_interesting_uses_cond): Likewise.
8220 (find_interesting_uses_address): Likewise.
8221 (get_mem_type_for_internal_fn): New function.
8222 (find_address_like_use): Likewise.
8223 (find_interesting_uses_stmt): Try find_address_like_use before
8224 calling find_interesting_uses_op.
8225 (addr_offset_valid_p): Use the iv mem_type field as the type
8226 of the addressed memory.
8227 (add_autoinc_candidates): Likewise.
8228 (get_address_cost): Likewise.
8229 (split_small_address_groups_p): Use address_p.
8230 (split_address_groups): Likewise.
8231 (add_iv_candidate_for_use): Likewise.
8232 (autoinc_possible_for_pair): Likewise.
8233 (rewrite_groups): Likewise.
8234 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8235 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8236 (get_alias_ptr_type_for_ptr_address): New function.
8237 (rewrite_use_address): Rewrite address uses in calls that were
8238 identified by find_address_like_use.
8240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8241 Alan Hayward <alan.hayward@arm.com>
8242 David Sherwood <david.sherwood@arm.com>
8244 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8246 * gimple-expr.h (is_gimple_addressable: Likewise.
8247 * gimple-expr.c (is_gimple_address): Likewise.
8248 * internal-fn.c (expand_call_mem_ref): New function.
8249 (expand_mask_load_optab_fn): Use it.
8250 (expand_mask_store_optab_fn): Likewise.
8252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8253 Alan Hayward <alan.hayward@arm.com>
8254 David Sherwood <david.sherwood@arm.com>
8256 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8257 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8258 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8259 (cond_umax@var{mode}): Document.
8260 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8261 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8262 (cond_umin_optab, cond_umax_optab): New optabs.
8263 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8264 (COND_IOR, COND_XOR): New internal functions.
8265 * internal-fn.h (get_conditional_internal_fn): Declare.
8266 * internal-fn.c (cond_binary_direct): New macro.
8267 (expand_cond_binary_optab_fn): Likewise.
8268 (direct_cond_binary_optab_supported_p): Likewise.
8269 (get_conditional_internal_fn): New function.
8270 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8271 Cope with reduction statements that are vectorized as calls rather
8273 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8274 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8275 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8276 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8277 (UNSPEC_COND_EOR): New unspecs.
8278 (optab): Add mappings for them.
8279 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8280 (sve_int_op, sve_fp_op): New int attributes.
8282 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8283 Alan Hayward <alan.hayward@arm.com>
8284 David Sherwood <david.sherwood@arm.com>
8286 * optabs.def (while_ult_optab): New optab.
8287 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8288 * internal-fn.def (WHILE_ULT): New internal function.
8289 * internal-fn.h (direct_internal_fn_supported_p): New override
8290 that takes two types as argument.
8291 * internal-fn.c (while_direct): New macro.
8292 (expand_while_optab_fn): New function.
8293 (convert_optab_supported_p): Likewise.
8294 (direct_while_optab_supported_p): New macro.
8295 * wide-int.h (wi::udiv_ceil): New function.
8296 * tree-vectorizer.h (rgroup_masks): New structure.
8297 (vec_loop_masks): New typedef.
8298 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8300 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8301 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8302 (vect_max_vf): New function.
8303 (slpeel_make_loop_iterate_ntimes): Delete.
8304 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8305 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8306 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8307 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8308 internal-fn.h, stor-layout.h and optabs-query.h.
8309 (vect_set_loop_mask): New function.
8310 (add_preheader_seq): Likewise.
8311 (add_header_seq): Likewise.
8312 (interleave_supported_p): Likewise.
8313 (vect_maybe_permute_loop_masks): Likewise.
8314 (vect_set_loop_masks_directly): Likewise.
8315 (vect_set_loop_condition_masked): Likewise.
8316 (vect_set_loop_condition_unmasked): New function, split out from
8317 slpeel_make_loop_iterate_ntimes.
8318 (slpeel_make_loop_iterate_ntimes): Rename to..
8319 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8320 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8321 (vect_do_peeling): Update call accordingly.
8322 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8324 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8325 mask_compare_type, can_fully_mask_p and fully_masked_p.
8326 (release_vec_loop_masks): New function.
8327 (_loop_vec_info): Use it to free the loop masks.
8328 (can_produce_all_loop_masks_p): New function.
8329 (vect_get_max_nscalars_per_iter): Likewise.
8330 (vect_verify_full_masking): Likewise.
8331 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8332 retries, and free the mask rgroups before retrying. Check loop-wide
8333 reasons for disallowing fully-masked loops. Make the final decision
8334 about whether use a fully-masked loop or not.
8335 (vect_estimate_min_profitable_iters): Do not assume that peeling
8336 for the number of iterations will be needed for fully-masked loops.
8337 (vectorizable_reduction): Disable fully-masked loops.
8338 (vectorizable_live_operation): Likewise.
8339 (vect_halve_mask_nunits): New function.
8340 (vect_double_mask_nunits): Likewise.
8341 (vect_record_loop_mask): Likewise.
8342 (vect_get_loop_mask): Likewise.
8343 (vect_transform_loop): Handle the case in which the final loop
8344 iteration might handle a partial vector. Call vect_set_loop_condition
8345 instead of slpeel_make_loop_iterate_ntimes.
8346 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8347 (check_load_store_masking): New function.
8348 (prepare_load_store_mask): Likewise.
8349 (vectorizable_store): Handle fully-masked loops.
8350 (vectorizable_load): Likewise.
8351 (supportable_widening_operation): Use vect_halve_mask_nunits for
8353 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8354 (vect_gen_while): New function.
8355 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8356 (aarch64_uqdec<mode>): New insn.
8358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8359 Alan Hayward <alan.hayward@arm.com>
8360 David Sherwood <david.sherwood@arm.com>
8362 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8363 (reduc_xor_scal_optab): New optabs.
8364 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8365 (reduc_xor_scal_@var{m}): Document.
8366 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8367 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8369 * fold-const-call.c (fold_const_call): Handle them.
8370 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8371 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8372 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8373 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8374 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8375 (UNSPEC_XORV): New unspecs.
8376 (optab): Add entries for them.
8377 (BITWISEV): New int iterator.
8378 (bit_reduc_op): New int attributes.
8380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8381 Alan Hayward <alan.hayward@arm.com>
8382 David Sherwood <david.sherwood@arm.com>
8384 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8385 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8386 * optabs.def (vec_shl_insert_optab): New optab.
8387 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8388 (duplicate_and_interleave): Likewise.
8389 * tree-vect-loop.c: Include internal-fn.h.
8390 (neutral_op_for_slp_reduction): New function, split out from
8391 get_initial_defs_for_reduction.
8392 (get_initial_def_for_reduction): Handle option 2 for variable-length
8393 vectors by loading the neutral value into a vector and then shifting
8394 the initial value into element 0.
8395 (get_initial_defs_for_reduction): Replace the code argument with
8396 the neutral value calculated by neutral_op_for_slp_reduction.
8397 Use gimple_build_vector for constant-length vectors.
8398 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8399 but the first group_size elements have a neutral value.
8400 Use duplicate_and_interleave otherwise.
8401 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8402 Update call to get_initial_defs_for_reduction. Handle SLP
8403 reductions for variable-length vectors by creating one vector
8404 result for each scalar result, with the elements associated
8405 with other scalar results stubbed out with the neutral value.
8406 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8407 Require IFN_VEC_SHL_INSERT for double reductions on
8408 variable-length vectors, or SLP reductions that have
8409 a neutral value. Require can_duplicate_and_interleave_p
8410 support for variable-length unchained SLP reductions if there
8411 is no neutral value, such as for MIN/MAX reductions. Also require
8412 the number of vector elements to be a multiple of the number of
8413 SLP statements when doing variable-length unchained SLP reductions.
8414 Update call to vect_create_epilog_for_reduction.
8415 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8416 and remove initial values.
8417 (duplicate_and_interleave): Make public.
8418 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8419 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8422 Alan Hayward <alan.hayward@arm.com>
8423 David Sherwood <david.sherwood@arm.com>
8425 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8426 (can_duplicate_and_interleave_p): New function.
8427 (vect_get_and_check_slp_defs): Take the vector of statements
8428 rather than just the current one. Remove excess parentheses.
8429 Restriction rejectinon of vect_constant_def and vect_external_def
8430 for variable-length vectors to boolean types, or types for which
8431 can_duplicate_and_interleave_p is false.
8432 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8433 (duplicate_and_interleave): New function.
8434 (vect_get_constant_vectors): Use gimple_build_vector for
8435 constant-length vectors and suitable variable-length constant
8436 vectors. Use duplicate_and_interleave for other variable-length
8437 vectors. Don't defer the update when inserting new statements.
8439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8440 Alan Hayward <alan.hayward@arm.com>
8441 David Sherwood <david.sherwood@arm.com>
8443 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8444 min_profitable_iters doesn't go negative.
8446 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8447 Alan Hayward <alan.hayward@arm.com>
8448 David Sherwood <david.sherwood@arm.com>
8450 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8451 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8452 * optabs.def (vec_mask_load_lanes_optab): New optab.
8453 (vec_mask_store_lanes_optab): Likewise.
8454 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8455 (MASK_STORE_LANES): Likewise.
8456 * internal-fn.c (mask_load_lanes_direct): New macro.
8457 (mask_store_lanes_direct): Likewise.
8458 (expand_mask_load_optab_fn): Handle masked operations.
8459 (expand_mask_load_lanes_optab_fn): New macro.
8460 (expand_mask_store_optab_fn): Handle masked operations.
8461 (expand_mask_store_lanes_optab_fn): New macro.
8462 (direct_mask_load_lanes_optab_supported_p): Likewise.
8463 (direct_mask_store_lanes_optab_supported_p): Likewise.
8464 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8466 (vect_load_lanes_supported): Likewise.
8467 * tree-vect-data-refs.c (strip_conversion): New function.
8468 (can_group_stmts_p): Likewise.
8469 (vect_analyze_data_ref_accesses): Use it instead of checking
8470 for a pair of assignments.
8471 (vect_store_lanes_supported): Take a masked_p parameter.
8472 (vect_load_lanes_supported): Likewise.
8473 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8474 vect_store_lanes_supported and vect_load_lanes_supported.
8475 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8476 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8477 parameter. Don't allow gaps for masked accesses.
8478 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8479 and vect_load_lanes_supported.
8480 (get_load_store_type): Take a masked_p parameter and update
8481 call to get_group_load_store_type.
8482 (vectorizable_store): Update call to get_load_store_type.
8483 Handle IFN_MASK_STORE_LANES.
8484 (vectorizable_load): Update call to get_load_store_type.
8485 Handle IFN_MASK_LOAD_LANES.
8487 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8488 Alan Hayward <alan.hayward@arm.com>
8489 David Sherwood <david.sherwood@arm.com>
8491 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8493 * config/aarch64/aarch64-protos.h
8494 (aarch64_sve_struct_memory_operand_p): Declare.
8495 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8496 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8497 (VPRED, vpred): Handle SVE structure modes.
8498 * config/aarch64/constraints.md (Utx): New constraint.
8499 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8500 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8501 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8502 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8503 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8504 structure modes. Split into pieces after RA.
8505 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8506 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8508 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8509 SVE structure modes.
8510 (aarch64_classify_address): Likewise.
8511 (sizetochar): Move earlier in file.
8512 (aarch64_print_operand): Handle SVE register lists.
8513 (aarch64_array_mode): New function.
8514 (aarch64_sve_struct_memory_operand_p): Likewise.
8515 (TARGET_ARRAY_MODE): Redefine.
8517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8518 Alan Hayward <alan.hayward@arm.com>
8519 David Sherwood <david.sherwood@arm.com>
8521 * target.def (array_mode): New target hook.
8522 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8523 * doc/tm.texi: Regenerate.
8524 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8525 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8526 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8528 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8531 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8532 Alan Hayward <alan.hayward@arm.com>
8533 David Sherwood <david.sherwood@arm.com>
8535 * fold-const.c (fold_binary_loc): Check the argument types
8536 rather than the result type when testing for a vector operation.
8538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8540 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8541 * doc/tm.texi: Regenerate.
8543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8544 Alan Hayward <alan.hayward@arm.com>
8545 David Sherwood <david.sherwood@arm.com>
8547 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8548 (sve): Document new AArch64 extension.
8549 * doc/md.texi (w): Extend the description of the AArch64
8550 constraint to include SVE vectors.
8551 (Upl, Upa): Document new AArch64 predicate constraints.
8552 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8554 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8555 (msve-vector-bits=): New option.
8556 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8557 SVE when these are disabled.
8558 (sve): New extension.
8559 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8560 modes. Adjust their number of units based on aarch64_sve_vg.
8561 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8562 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8563 aarch64_addr_query_type.
8564 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8565 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8566 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8567 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8568 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8569 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8570 (aarch64_simd_imm_zero_p): Delete.
8571 (aarch64_check_zero_based_sve_index_immediate): Declare.
8572 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8573 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8574 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8575 (aarch64_sve_float_mul_immediate_p): Likewise.
8576 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8578 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8579 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8580 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8581 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8582 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8583 (aarch64_regmode_natural_size): Likewise.
8584 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8585 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8587 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8588 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8589 for VG and the SVE predicate registers.
8590 (V_ALIASES): Add a "z"-prefixed alias.
8591 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8592 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8593 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8594 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8595 (REG_CLASS_NAMES): Add entries for them.
8596 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8597 and the predicate registers.
8598 (aarch64_sve_vg): Declare.
8599 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8600 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8601 (REGMODE_NATURAL_SIZE): Define.
8602 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8604 * config/aarch64/aarch64.c: Include cfgrtl.h.
8605 (simd_immediate_info): Add a constructor for series vectors,
8606 and an associated step field.
8607 (aarch64_sve_vg): New variable.
8608 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8609 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8610 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8611 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8612 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8613 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8614 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8615 (aarch64_get_mask_mode): New functions.
8616 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8617 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8618 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8619 predicate modes and predicate registers. Explicitly restrict
8620 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8621 to store a vector mode if it is recognized by
8622 aarch64_classify_vector_mode.
8623 (aarch64_regmode_natural_size): New function.
8624 (aarch64_hard_regno_caller_save_mode): Return the original mode
8626 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8627 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8628 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8629 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8631 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8632 does not overlap dest if the function is frame-related. Handle
8634 (aarch64_split_add_offset): New function.
8635 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8636 them aarch64_add_offset.
8637 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8638 and update call to aarch64_sub_sp.
8639 (aarch64_add_cfa_expression): New function.
8640 (aarch64_expand_prologue): Pass extra temporary registers to the
8641 functions above. Handle the case in which we need to emit new
8642 DW_CFA_expressions for registers that were originally saved
8643 relative to the stack pointer, but now have to be expressed
8644 relative to the frame pointer.
8645 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8647 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8648 IP0 and IP1 values for SVE frames.
8649 (aarch64_expand_vec_series): New function.
8650 (aarch64_expand_sve_widened_duplicate): Likewise.
8651 (aarch64_expand_sve_const_vector): Likewise.
8652 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8653 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8654 into the register, rather than emitting a SET directly.
8655 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8656 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8657 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8658 (offset_9bit_signed_scaled_p): New functions.
8659 (aarch64_replicate_bitmask_imm): New function.
8660 (aarch64_bitmask_imm): Use it.
8661 (aarch64_cannot_force_const_mem): Reject expressions involving
8662 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8663 (aarch64_classify_index): Handle SVE indices, by requiring
8664 a plain register index with a scale that matches the element size.
8665 (aarch64_classify_address): Handle SVE addresses. Assert that
8666 the mode of the address is VOIDmode or an integer mode.
8667 Update call to aarch64_classify_symbol.
8668 (aarch64_classify_symbolic_expression): Update call to
8669 aarch64_classify_symbol.
8670 (aarch64_const_vec_all_in_range_p): New function.
8671 (aarch64_print_vector_float_operand): Likewise.
8672 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8673 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8674 and the FP immediates 1.0 and 0.5.
8675 (aarch64_print_address_internal): Handle SVE addresses.
8676 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8677 (aarch64_regno_regclass): Handle predicate registers.
8678 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8680 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8681 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8682 (aarch64_convert_sve_vector_bits): New function.
8683 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8684 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8686 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8687 Handle SVE vector and predicate modes. Accept VL-based constants
8688 that need only one temporary register, and VL offsets that require
8689 no temporary registers.
8690 (aarch64_conditional_register_usage): Mark the predicate registers
8691 as fixed if SVE isn't available.
8692 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8693 Return true for SVE vector and predicate modes.
8694 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8695 rather than an unsigned int. Handle SVE modes.
8696 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8698 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8700 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8701 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8702 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8703 (aarch64_sve_float_mul_immediate_p): New functions.
8704 (aarch64_sve_valid_immediate): New function.
8705 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8706 Explicitly reject structure modes. Check for INDEX constants.
8707 Handle PTRUE and PFALSE constants.
8708 (aarch64_check_zero_based_sve_index_immediate): New function.
8709 (aarch64_simd_imm_zero_p): Delete.
8710 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8711 vector modes. Accept constants in the range of CNT[BHWD].
8712 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8713 ask for an Advanced SIMD mode.
8714 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8715 (aarch64_simd_vector_alignment): Handle SVE predicates.
8716 (aarch64_vectorize_preferred_vector_alignment): New function.
8717 (aarch64_simd_vector_alignment_reachable): Use it instead of
8719 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8720 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8722 (MAX_VECT_LEN): Delete.
8723 (expand_vec_perm_d): Add a vec_flags field.
8724 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8725 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8726 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8728 (aarch64_evpc_rev): Rename to...
8729 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8730 (aarch64_evpc_rev_global): New function.
8731 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8732 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8734 (aarch64_evpc_sve_tbl): New function.
8735 (aarch64_expand_vec_perm_const_1): Update after rename of
8736 aarch64_evpc_rev. Handle SVE permutes too, trying
8737 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8738 than aarch64_evpc_tbl.
8739 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8740 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8741 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8742 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8743 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8744 (aarch64_expand_sve_vcond): New functions.
8745 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8746 of aarch64_vector_mode_p.
8747 (aarch64_dwarf_poly_indeterminate_value): New function.
8748 (aarch64_compute_pressure_classes): Likewise.
8749 (aarch64_can_change_mode_class): Likewise.
8750 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8751 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8752 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8753 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8754 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8755 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8756 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8757 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8759 (Dn, Dl, Dr): Accept const as well as const_vector.
8760 (Dz): Likewise. Compare against CONST0_RTX.
8761 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8762 of "vector" where appropriate.
8763 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8764 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8765 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8766 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8767 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8768 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8769 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8770 (v_int_equiv): Extend to SVE modes.
8771 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8773 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8774 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8775 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8776 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8777 (SVE_COND_FP_CMP): New int iterators.
8778 (perm_hilo): Handle the new unpack unspecs.
8779 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8781 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8782 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8783 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8784 (aarch64_equality_operator, aarch64_constant_vector_operand)
8785 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8786 (aarch64_sve_nonimmediate_operand): Likewise.
8787 (aarch64_sve_general_operand): Likewise.
8788 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8789 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8790 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8791 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8792 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8793 (aarch64_sve_float_arith_immediate): Likewise.
8794 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8795 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8796 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8797 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8798 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8799 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8800 (aarch64_sve_float_arith_operand): Likewise.
8801 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8802 (aarch64_sve_float_mul_operand): Likewise.
8803 (aarch64_sve_vec_perm_operand): Likewise.
8804 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8805 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8806 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8807 as well as const_vector.
8808 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8809 in file. Use CONST0_RTX and CONSTM1_RTX.
8810 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8811 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8812 Use aarch64_simd_imm_zero.
8813 * config/aarch64/aarch64-sve.md: New file.
8814 * config/aarch64/aarch64.md: Include it.
8815 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8816 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8817 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8818 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8819 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8820 (sve): New attribute.
8821 (enabled): Disable instructions with the sve attribute unless
8823 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8824 aarch64_expand_mov_immediate.
8825 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8826 CNT[BHSD] immediates.
8827 (movti): Split CONST_POLY_INT moves into two halves.
8828 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8829 Split additions that need a temporary here if the destination
8830 is the stack pointer.
8831 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8832 (*add<mode>3_poly_1): New instruction.
8833 (set_clobber_cc): New expander.
8835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8837 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8838 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8839 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8840 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8841 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8842 Change innermode from fixed_mode_size to machine_mode.
8843 (simplify_subreg): Update call accordingly. Handle a constant-sized
8844 subreg of a variable-length CONST_VECTOR.
8846 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8847 Alan Hayward <alan.hayward@arm.com>
8848 David Sherwood <david.sherwood@arm.com>
8850 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8851 (add_offset_to_base): New function, split out from...
8852 (create_mem_ref): ...here. When handling a scale other than 1,
8853 check first whether the address is valid without the offset.
8854 Add it into the base if so, leaving the index and scale as-is.
8856 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8859 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8860 fold_for_warn before checking if arg2 is INTEGER_CST.
8862 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8864 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8865 (store_multiple_operation): Delete.
8866 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8867 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8868 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8869 guarded by TARGET_STRING.
8870 (rs6000_output_load_multiple): Delete.
8871 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8872 OPTION_MASK_STRING / TARGET_STRING handling.
8873 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8874 (const rs6000_opt_masks) <"string">: Change mask to 0.
8875 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8876 (MASK_STRING): Delete.
8877 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8879 (load_multiple): Delete.
8886 (store_multiple): Delete.
8893 (movmemsi_8reg): Delete.
8894 (corresponding unnamed define_insn): Delete.
8895 (movmemsi_6reg): Delete.
8896 (corresponding unnamed define_insn): Delete.
8897 (movmemsi_4reg): Delete.
8898 (corresponding unnamed define_insn): Delete.
8899 (movmemsi_2reg): Delete.
8900 (corresponding unnamed define_insn): Delete.
8901 (movmemsi_1reg): Delete.
8902 (corresponding unnamed define_insn): Delete.
8903 * config/rs6000/rs6000.opt (mno-string): New.
8904 (mstring): Replace by deprecation warning stub.
8905 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8907 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8909 * regrename.c (regrename_do_replace): If replacing the same
8910 reg multiple times, try to reuse last created gen_raw_REG.
8913 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8914 main to workaround a bug in GDB.
8916 2018-01-12 Tom de Vries <tom@codesourcery.com>
8919 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8921 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8923 PR rtl-optimization/80481
8924 * ira-color.c (get_cap_member): New function.
8925 (allocnos_conflict_by_live_ranges_p): Use it.
8926 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8927 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8929 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8932 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8933 (*saddl_se_1): Ditto.
8935 (*ssubl_se_1): Ditto.
8937 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8939 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8940 rather than wi::to_widest for DR_INITs.
8941 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8942 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8943 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8945 (vect_analyze_group_access_1): Note that here.
8947 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8949 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8950 polynomial type sizes.
8952 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8954 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8955 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8956 (gimple_add_tmp_var): Likewise.
8958 2018-01-12 Martin Liska <mliska@suse.cz>
8960 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8961 (gimple_alloc_sizes): Likewise.
8962 (dump_gimple_statistics): Use PRIu64 in printf format.
8963 * gimple.h: Change uint64_t to int.
8965 2018-01-12 Martin Liska <mliska@suse.cz>
8967 * tree-core.h: Use uint64_t instead of int.
8968 * tree.c (tree_node_counts): Likewise.
8969 (tree_node_sizes): Likewise.
8970 (dump_tree_statistics): Use PRIu64 in printf format.
8972 2018-01-12 Martin Liska <mliska@suse.cz>
8974 * Makefile.in: As qsort_chk is implemented in vec.c, add
8975 vec.o to linkage of gencfn-macros.
8976 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8977 passing the info to record_node_allocation_statistics.
8978 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8980 * ggc-common.c (struct ggc_usage): Add operator== and use
8981 it in operator< and compare function.
8982 * mem-stats.h (struct mem_usage): Likewise.
8983 * vec.c (struct vec_usage): Remove operator< and compare
8984 function. Can be simply inherited.
8986 2018-01-12 Martin Jambor <mjambor@suse.cz>
8989 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8990 * tree-ssa-math-opts.c: Include domwalk.h.
8991 (convert_mult_to_fma_1): New function.
8992 (fma_transformation_info): New type.
8993 (fma_deferring_state): Likewise.
8994 (cancel_fma_deferring): New function.
8995 (result_of_phi): Likewise.
8996 (last_fma_candidate_feeds_initial_phi): Likewise.
8997 (convert_mult_to_fma): Added deferring logic, split actual
8998 transformation to convert_mult_to_fma_1.
8999 (math_opts_dom_walker): New type.
9000 (math_opts_dom_walker::after_dom_children): New method, body moved
9001 here from pass_optimize_widening_mul::execute, added deferring logic
9003 (pass_optimize_widening_mul::execute): Moved most of code to
9004 math_opts_dom_walker::after_dom_children.
9005 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9006 * config/i386/i386.c (ix86_option_override_internal): Added
9007 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9009 2018-01-12 Richard Biener <rguenther@suse.de>
9012 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9013 inline instance vars.
9015 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9018 * config/rx/rx.c (rx_is_restricted_memory_address):
9021 2018-01-12 Richard Biener <rguenther@suse.de>
9023 PR tree-optimization/80846
9024 * target.def (split_reduction): New target hook.
9025 * targhooks.c (default_split_reduction): New function.
9026 * targhooks.h (default_split_reduction): Declare.
9027 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9028 target requests first reduce vectors by combining low and high
9030 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9031 (get_vectype_for_scalar_type_and_size): Export.
9032 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9033 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9034 * doc/tm.texi: Regenerate.
9035 * config/i386/i386.c (ix86_split_reduction): Implement
9036 TARGET_VECTORIZE_SPLIT_REDUCTION.
9038 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9041 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9042 in PIC mode except for TARGET_VXWORKS_RTP.
9043 * config/sparc/sparc.c: Include cfgrtl.h.
9044 (TARGET_INIT_PIC_REG): Define.
9045 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9046 (sparc_pic_register_p): New predicate.
9047 (sparc_legitimate_address_p): Use it.
9048 (sparc_legitimize_pic_address): Likewise.
9049 (sparc_delegitimize_address): Likewise.
9050 (sparc_mode_dependent_address_p): Likewise.
9051 (gen_load_pcrel_sym): Remove 4th parameter.
9052 (load_got_register): Adjust call to above. Remove obsolete stuff.
9053 (sparc_expand_prologue): Do not call load_got_register here.
9054 (sparc_flat_expand_prologue): Likewise.
9055 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9056 (sparc_use_pseudo_pic_reg): New function.
9057 (sparc_init_pic_reg): Likewise.
9058 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9059 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9061 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9063 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9064 Add item for branch_cost.
9066 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9068 PR rtl-optimization/83565
9069 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9070 not extend the result to a larger mode for rotate operations.
9071 (num_sign_bit_copies1): Likewise.
9073 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9076 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9078 Use values-Xc.o for -pedantic.
9079 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9081 2018-01-12 Martin Liska <mliska@suse.cz>
9084 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9086 (possible_polymorphic_call_targets): Use it.
9087 (ipa_devirt): Likewise.
9089 2018-01-12 Martin Liska <mliska@suse.cz>
9091 * profile-count.h (enum profile_quality): Use 0 as invalid
9092 enum value of profile_quality.
9094 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9096 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9097 -mext-string options.
9099 2018-01-12 Richard Biener <rguenther@suse.de>
9101 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9102 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9103 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9105 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9107 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9109 * configure.ac (--with-long-double-format): Add support for the
9110 configuration option to change the default long double format on
9112 * config.gcc (powerpc*-linux*-*): Likewise.
9113 * configure: Regenerate.
9114 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9115 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9116 used without modification.
9118 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9120 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9121 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9122 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9123 MISC_BUILTIN_SPEC_BARRIER.
9124 (rs6000_init_builtins): Likewise.
9125 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9127 (speculation_barrier): New define_insn.
9128 * doc/extend.texi: Document __builtin_speculation_barrier.
9130 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9133 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9134 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9135 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9137 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9138 integral modes instead of "ss" and "sd".
9139 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9140 vectors with 32-bit and 64-bit elements.
9141 (vecdupssescalarmodesuffix): New mode attribute.
9142 (vec_dup<mode>): Use it.
9144 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9147 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9148 frame if argument is passed on stack.
9150 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9153 * ree.c (combine_reaching_defs): Optimize also
9154 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9155 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9157 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9160 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9162 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9165 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9166 after they are computed.
9168 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9170 PR tree-optimization/83695
9171 * gimple-loop-linterchange.cc
9172 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9173 reset cached scev information after interchange.
9174 (pass_linterchange::execute): Remove call to scev_reset_htab.
9176 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9178 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9179 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9180 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9181 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9182 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9183 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9184 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9185 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9186 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9187 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9188 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9189 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9190 (V_lane_reg): Likewise.
9191 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9193 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9194 (vfmal_lane_low<mode>_intrinsic,
9195 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9196 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9197 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9198 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9199 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9200 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9202 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9204 * config/arm/arm-cpus.in (fp16fml): New feature.
9205 (ALL_SIMD): Add fp16fml.
9206 (armv8.2-a): Add fp16fml as an option.
9207 (armv8.3-a): Likewise.
9208 (armv8.4-a): Add fp16fml as part of fp16.
9209 * config/arm/arm.h (TARGET_FP16FML): Define.
9210 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9212 * config/arm/arm-modes.def (V2HF): Define.
9213 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9214 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9215 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9216 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9217 vfmsl_low, vfmsl_high): New set of builtins.
9218 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9219 (vfml_op): New code attribute.
9220 (VFMLHALVES): New int iterator.
9221 (VFML, VFMLSEL): New mode attributes.
9222 (V_reg): Define mapping for V2HF.
9223 (V_hi, V_lo): New mode attributes.
9224 (VF_constraint): Likewise.
9225 (vfml_half, vfml_half_selector): New int attributes.
9226 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9228 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9229 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9231 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9232 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9233 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9234 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9236 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9237 Document new effective target and option set.
9239 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9241 * config/arm/arm-cpus.in (armv8_4): New feature.
9242 (ARMv8_4a): New fgroup.
9243 (armv8.4-a): New arch.
9244 * config/arm/arm-tables.opt: Regenerate.
9245 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9246 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9247 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9248 Add matching rules for -march=armv8.4-a and extensions.
9249 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9251 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9254 * config/rx/rx.md (BW): New mode attribute.
9255 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9257 2018-01-11 Richard Biener <rguenther@suse.de>
9259 PR tree-optimization/83435
9260 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9261 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9262 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9264 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9265 Alan Hayward <alan.hayward@arm.com>
9266 David Sherwood <david.sherwood@arm.com>
9268 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9270 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9271 (aarch64_print_address_internal): Use it to check for a zero offset.
9273 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9274 Alan Hayward <alan.hayward@arm.com>
9275 David Sherwood <david.sherwood@arm.com>
9277 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9278 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9279 Return a poly_int64 rather than a HOST_WIDE_INT.
9280 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9281 rather than a HOST_WIDE_INT.
9282 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9283 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9284 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9285 final_offset from HOST_WIDE_INT to poly_int64.
9286 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9287 to_constant when getting the number of units in an Advanced SIMD
9289 (aarch64_builtin_vectorized_function): Check for a constant number
9291 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9293 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9294 attribute instead of GET_MODE_NUNITS.
9295 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9296 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9297 GET_MODE_SIZE for fixed-size registers.
9298 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9299 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9300 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9301 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9302 (aarch64_print_operand, aarch64_print_address_internal)
9303 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9304 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9305 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9306 Handle polynomial GET_MODE_SIZE.
9307 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9308 wider than SImode without modification.
9309 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9310 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9311 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9312 passing and returning SVE modes.
9313 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9314 rather than GEN_INT.
9315 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9316 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9317 (aarch64_allocate_and_probe_stack_space): Likewise.
9318 (aarch64_layout_frame): Cope with polynomial offsets.
9319 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9320 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9322 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9323 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9324 poly_int64 rather than a HOST_WIDE_INT.
9325 (aarch64_get_separate_components, aarch64_process_components)
9326 (aarch64_expand_prologue, aarch64_expand_epilogue)
9327 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9328 (aarch64_anchor_offset): New function, split out from...
9329 (aarch64_legitimize_address): ...here.
9330 (aarch64_builtin_vectorization_cost): Handle polynomial
9331 TYPE_VECTOR_SUBPARTS.
9332 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9334 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9335 number of elements from the PARALLEL rather than the mode.
9336 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9337 rather than GET_MODE_BITSIZE.
9338 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9339 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9340 (aarch64_expand_vec_perm_const_1): Handle polynomial
9341 d->perm.length () and d->perm elements.
9342 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9343 Apply to_constant to d->perm elements.
9344 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9345 polynomial CONST_VECTOR_NUNITS.
9346 (aarch64_move_pointer): Take amount as a poly_int64 rather
9348 (aarch64_progress_pointer): Avoid temporary variable.
9349 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9350 the mode attribute instead of GET_MODE.
9352 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9353 Alan Hayward <alan.hayward@arm.com>
9354 David Sherwood <david.sherwood@arm.com>
9356 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9357 x exists before using it.
9358 (aarch64_add_constant_internal): Rename to...
9359 (aarch64_add_offset_1): ...this. Replace regnum with separate
9360 src and dest rtxes. Handle the case in which they're different,
9361 including when the offset is zero. Replace scratchreg with an rtx.
9362 Use 2 additions if there is no spare register into which we can
9363 move a 16-bit constant.
9364 (aarch64_add_constant): Delete.
9365 (aarch64_add_offset): Replace reg with separate src and dest
9366 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9367 Use aarch64_add_offset_1.
9368 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9369 an rtx rather than an int. Take the delta as a poly_int64
9370 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9371 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9372 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9373 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9374 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9376 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9377 aarch64_add_constant.
9379 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9381 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9382 Use scalar_float_mode.
9384 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9386 * config/aarch64/aarch64-simd.md
9387 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9388 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9389 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9390 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9391 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9392 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9393 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9394 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9395 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9396 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9398 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9401 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9402 targ_options->x_arm_arch_string is non NULL.
9404 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9406 * config/aarch64/aarch64.h
9407 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9409 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9412 * expmed.c (emit_store_flag_force): Swap if const op0
9413 and change VOIDmode to mode of op0.
9415 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9417 PR rtl-optimization/83761
9418 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9419 than bytes to mode_for_size.
9421 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9424 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9425 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9428 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9431 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9432 when in layout mode.
9433 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9434 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9437 2018-01-10 Michael Collison <michael.collison@arm.com>
9439 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9440 * config/aarch64/aarch64-option-extension.def: Add
9441 AARCH64_OPT_EXTENSION of 'fp16fml'.
9442 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9443 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9444 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9445 * config/aarch64/constraints.md (Ui7): New constraint.
9446 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9447 (VFMLA_SEL_W): Ditto.
9450 (VFMLA16_LOW): New int iterator.
9451 (VFMLA16_HIGH): Ditto.
9452 (UNSPEC_FMLAL): New unspec.
9453 (UNSPEC_FMLSL): Ditto.
9454 (UNSPEC_FMLAL2): Ditto.
9455 (UNSPEC_FMLSL2): Ditto.
9456 (f16mac): New code attribute.
9457 * config/aarch64/aarch64-simd-builtins.def
9458 (aarch64_fmlal_lowv2sf): Ditto.
9459 (aarch64_fmlsl_lowv2sf): Ditto.
9460 (aarch64_fmlalq_lowv4sf): Ditto.
9461 (aarch64_fmlslq_lowv4sf): Ditto.
9462 (aarch64_fmlal_highv2sf): Ditto.
9463 (aarch64_fmlsl_highv2sf): Ditto.
9464 (aarch64_fmlalq_highv4sf): Ditto.
9465 (aarch64_fmlslq_highv4sf): Ditto.
9466 (aarch64_fmlal_lane_lowv2sf): Ditto.
9467 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9468 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9469 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9470 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9471 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9472 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9473 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9474 (aarch64_fmlal_lane_highv2sf): Ditto.
9475 (aarch64_fmlsl_lane_highv2sf): Ditto.
9476 (aarch64_fmlal_laneq_highv2sf): Ditto.
9477 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9478 (aarch64_fmlalq_lane_highv4sf): Ditto.
9479 (aarch64_fmlsl_lane_highv4sf): Ditto.
9480 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9481 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9482 * config/aarch64/aarch64-simd.md:
9483 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9484 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9485 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9486 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9487 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9488 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9489 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9490 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9491 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9492 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9493 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9494 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9495 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9496 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9497 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9498 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9499 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9500 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9501 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9502 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9503 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9504 (vfmlsl_low_u32): Ditto.
9505 (vfmlalq_low_u32): Ditto.
9506 (vfmlslq_low_u32): Ditto.
9507 (vfmlal_high_u32): Ditto.
9508 (vfmlsl_high_u32): Ditto.
9509 (vfmlalq_high_u32): Ditto.
9510 (vfmlslq_high_u32): Ditto.
9511 (vfmlal_lane_low_u32): Ditto.
9512 (vfmlsl_lane_low_u32): Ditto.
9513 (vfmlal_laneq_low_u32): Ditto.
9514 (vfmlsl_laneq_low_u32): Ditto.
9515 (vfmlalq_lane_low_u32): Ditto.
9516 (vfmlslq_lane_low_u32): Ditto.
9517 (vfmlalq_laneq_low_u32): Ditto.
9518 (vfmlslq_laneq_low_u32): Ditto.
9519 (vfmlal_lane_high_u32): Ditto.
9520 (vfmlsl_lane_high_u32): Ditto.
9521 (vfmlal_laneq_high_u32): Ditto.
9522 (vfmlsl_laneq_high_u32): Ditto.
9523 (vfmlalq_lane_high_u32): Ditto.
9524 (vfmlslq_lane_high_u32): Ditto.
9525 (vfmlalq_laneq_high_u32): Ditto.
9526 (vfmlslq_laneq_high_u32): Ditto.
9527 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9528 (AARCH64_FL_FOR_ARCH8_4): New.
9529 (AARCH64_ISA_F16FML): New ISA flag.
9530 (TARGET_F16FML): New feature flag for fp16fml.
9531 (doc/invoke.texi): Document new fp16fml option.
9533 2018-01-10 Michael Collison <michael.collison@arm.com>
9535 * config/aarch64/aarch64-builtins.c:
9536 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9537 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9538 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9539 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9540 (AARCH64_ISA_SHA3): New ISA flag.
9541 (TARGET_SHA3): New feature flag for sha3.
9542 * config/aarch64/iterators.md (sha512_op): New int attribute.
9543 (CRYPTO_SHA512): New int iterator.
9544 (UNSPEC_SHA512H): New unspec.
9545 (UNSPEC_SHA512H2): Ditto.
9546 (UNSPEC_SHA512SU0): Ditto.
9547 (UNSPEC_SHA512SU1): Ditto.
9548 * config/aarch64/aarch64-simd-builtins.def
9549 (aarch64_crypto_sha512hqv2di): New builtin.
9550 (aarch64_crypto_sha512h2qv2di): Ditto.
9551 (aarch64_crypto_sha512su0qv2di): Ditto.
9552 (aarch64_crypto_sha512su1qv2di): Ditto.
9553 (aarch64_eor3qv8hi): Ditto.
9554 (aarch64_rax1qv2di): Ditto.
9555 (aarch64_xarqv2di): Ditto.
9556 (aarch64_bcaxqv8hi): Ditto.
9557 * config/aarch64/aarch64-simd.md:
9558 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9559 (aarch64_crypto_sha512su0qv2di): Ditto.
9560 (aarch64_crypto_sha512su1qv2di): Ditto.
9561 (aarch64_eor3qv8hi): Ditto.
9562 (aarch64_rax1qv2di): Ditto.
9563 (aarch64_xarqv2di): Ditto.
9564 (aarch64_bcaxqv8hi): Ditto.
9565 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9566 (vsha512h2q_u64): Ditto.
9567 (vsha512su0q_u64): Ditto.
9568 (vsha512su1q_u64): Ditto.
9569 (veor3q_u16): Ditto.
9570 (vrax1q_u64): Ditto.
9572 (vbcaxq_u16): Ditto.
9573 * config/arm/types.md (crypto_sha512): New type attribute.
9574 (crypto_sha3): Ditto.
9575 (doc/invoke.texi): Document new sha3 option.
9577 2018-01-10 Michael Collison <michael.collison@arm.com>
9579 * config/aarch64/aarch64-builtins.c:
9580 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9581 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9582 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9583 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9584 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9585 (AARCH64_ISA_SM4): New ISA flag.
9586 (TARGET_SM4): New feature flag for sm4.
9587 * config/aarch64/aarch64-simd-builtins.def
9588 (aarch64_sm3ss1qv4si): Ditto.
9589 (aarch64_sm3tt1aq4si): Ditto.
9590 (aarch64_sm3tt1bq4si): Ditto.
9591 (aarch64_sm3tt2aq4si): Ditto.
9592 (aarch64_sm3tt2bq4si): Ditto.
9593 (aarch64_sm3partw1qv4si): Ditto.
9594 (aarch64_sm3partw2qv4si): Ditto.
9595 (aarch64_sm4eqv4si): Ditto.
9596 (aarch64_sm4ekeyqv4si): Ditto.
9597 * config/aarch64/aarch64-simd.md:
9598 (aarch64_sm3ss1qv4si): Ditto.
9599 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9600 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9601 (aarch64_sm4eqv4si): Ditto.
9602 (aarch64_sm4ekeyqv4si): Ditto.
9603 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9604 (sm3part_op): Ditto.
9605 (CRYPTO_SM3TT): Ditto.
9606 (CRYPTO_SM3PART): Ditto.
9607 (UNSPEC_SM3SS1): New unspec.
9608 (UNSPEC_SM3TT1A): Ditto.
9609 (UNSPEC_SM3TT1B): Ditto.
9610 (UNSPEC_SM3TT2A): Ditto.
9611 (UNSPEC_SM3TT2B): Ditto.
9612 (UNSPEC_SM3PARTW1): Ditto.
9613 (UNSPEC_SM3PARTW2): Ditto.
9614 (UNSPEC_SM4E): Ditto.
9615 (UNSPEC_SM4EKEY): Ditto.
9616 * config/aarch64/constraints.md (Ui2): New constraint.
9617 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9618 * config/arm/types.md (crypto_sm3): New type attribute.
9619 (crypto_sm4): Ditto.
9620 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9621 (vsm3tt1aq_u32): Ditto.
9622 (vsm3tt1bq_u32): Ditto.
9623 (vsm3tt2aq_u32): Ditto.
9624 (vsm3tt2bq_u32): Ditto.
9625 (vsm3partw1q_u32): Ditto.
9626 (vsm3partw2q_u32): Ditto.
9627 (vsm4eq_u32): Ditto.
9628 (vsm4ekeyq_u32): Ditto.
9629 (doc/invoke.texi): Document new sm4 option.
9631 2018-01-10 Michael Collison <michael.collison@arm.com>
9633 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9634 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9635 (AARCH64_FL_FOR_ARCH8_4): New.
9636 (AARCH64_FL_V8_4): New flag.
9637 (doc/invoke.texi): Document new armv8.4-a option.
9639 2018-01-10 Michael Collison <michael.collison@arm.com>
9641 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9642 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9643 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9644 * config/aarch64/aarch64-option-extension.def: Add
9645 AARCH64_OPT_EXTENSION of 'sha2'.
9646 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9647 (crypto): Disable sha2 and aes if crypto disabled.
9648 (crypto): Enable aes and sha2 if enabled.
9649 (simd): Disable sha2 and aes if simd disabled.
9650 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9652 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9653 (TARGET_SHA2): New feature flag for sha2.
9654 (TARGET_AES): New feature flag for aes.
9655 * config/aarch64/aarch64-simd.md:
9656 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9657 conditional on TARGET_AES.
9658 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9659 (aarch64_crypto_sha1hsi): Make pattern conditional
9661 (aarch64_crypto_sha1hv4si): Ditto.
9662 (aarch64_be_crypto_sha1hv4si): Ditto.
9663 (aarch64_crypto_sha1su1v4si): Ditto.
9664 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9665 (aarch64_crypto_sha1su0v4si): Ditto.
9666 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9667 (aarch64_crypto_sha256su0v4si): Ditto.
9668 (aarch64_crypto_sha256su1v4si): Ditto.
9669 (doc/invoke.texi): Document new aes and sha2 options.
9671 2018-01-10 Martin Sebor <msebor@redhat.com>
9673 PR tree-optimization/83781
9674 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9677 2018-01-11 Martin Sebor <msebor@gmail.com>
9678 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9680 PR tree-optimization/83501
9681 PR tree-optimization/81703
9683 * tree-ssa-strlen.c (get_string_cst): Rename...
9684 (get_string_len): ...to this. Handle global constants.
9685 (handle_char_store): Adjust.
9687 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9688 Jim Wilson <jimw@sifive.com>
9690 * config/riscv/riscv-protos.h (riscv_output_return): New.
9691 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9692 (riscv_attribute_table, riscv_output_return),
9693 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9694 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9695 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9697 (riscv_expand_prologue): Add early return for naked function.
9698 (riscv_expand_epilogue): Likewise.
9699 (riscv_function_ok_for_sibcall): Return false for naked function.
9700 (riscv_set_current_function): New.
9701 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9702 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9703 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9704 * doc/extend.texi (RISC-V Function Attributes): New.
9706 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9708 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9709 check for 128-bit long double before checking TCmode.
9710 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9711 128-bit long doubles before checking TFmode or TCmode.
9712 (FLOAT128_IBM_P): Likewise.
9714 2018-01-10 Martin Sebor <msebor@redhat.com>
9716 PR tree-optimization/83671
9717 * builtins.c (c_strlen): Unconditionally return zero for the empty
9719 Use -Warray-bounds for warnings.
9720 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9721 for non-constant array indices with COMPONENT_REF, arrays of
9722 arrays, and pointers to arrays.
9723 (gimple_fold_builtin_strlen): Determine and set length range for
9724 non-constant character arrays.
9726 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9729 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9732 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9734 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9736 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9739 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9740 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9741 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9742 indexed_or_indirect_operand predicate.
9743 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9744 (*vsx_le_perm_load_v8hi): Likewise.
9745 (*vsx_le_perm_load_v16qi): Likewise.
9746 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9747 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9748 (*vsx_le_perm_store_v8hi): Likewise.
9749 (*vsx_le_perm_store_v16qi): Likewise.
9750 (eight unnamed splitters): Likewise.
9752 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9754 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9755 * config/rs6000/emmintrin.h: Likewise.
9756 * config/rs6000/mmintrin.h: Likewise.
9757 * config/rs6000/xmmintrin.h: Likewise.
9759 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9762 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9764 * tree.c (tree_nop_conversion): Return true for location wrapper
9766 (maybe_wrap_with_location): New function.
9767 (selftest::check_strip_nops): New function.
9768 (selftest::test_location_wrappers): New function.
9769 (selftest::tree_c_tests): Call it.
9770 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9771 (maybe_wrap_with_location): New decl.
9772 (EXPR_LOCATION_WRAPPER_P): New macro.
9773 (location_wrapper_p): New inline function.
9774 (tree_strip_any_location_wrapper): New inline function.
9776 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9779 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9780 stack_realign_offset for the largest alignment of stack slot
9782 (ix86_find_max_used_stack_alignment): New function.
9783 (ix86_finalize_stack_frame_flags): Use it. Set
9784 max_used_stack_alignment if we don't realign stack.
9785 * config/i386/i386.h (machine_function): Add
9786 max_used_stack_alignment.
9788 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9790 * config/arm/arm.opt (-mbranch-cost): New option.
9791 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9794 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9797 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9798 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9800 2018-01-10 Richard Biener <rguenther@suse.de>
9803 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9804 early out so it also covers the case where we have a non-NULL
9807 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9809 PR tree-optimization/83753
9810 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9811 for non-strided grouped accesses if the number of elements is 1.
9813 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9816 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9817 * i386.h (TARGET_USE_GATHER): Define.
9818 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9820 2018-01-10 Martin Liska <mliska@suse.cz>
9823 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9824 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9826 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9827 CLEANUP_NO_PARTITIONING is not set.
9829 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9831 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9832 for vectors, as a partial revert of r254296.
9833 * rtl.h (const_vec_p): Delete.
9834 (const_vec_duplicate_p): Don't test for vector CONSTs.
9835 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9836 * expmed.c (make_tree): Likewise.
9839 * common.md (E, F): Use CONSTANT_P instead of checking for
9841 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9842 checking for CONST_VECTOR.
9844 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9847 * predict.c (force_edge_cold): Handle in more sane way edges
9850 2018-01-09 Carl Love <cel@us.ibm.com>
9852 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9854 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9855 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9856 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9857 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9858 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9859 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9860 * config/rs6000/rs6000-protos.h: Add extern defition for
9861 rs6000_generate_float2_double_code.
9862 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9864 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9865 (float2_v2df): Add define_expand.
9867 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9870 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9871 op_mode in the force_to_mode call.
9873 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9875 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9876 instead of checking each element individually.
9877 (aarch64_evpc_uzp): Likewise.
9878 (aarch64_evpc_zip): Likewise.
9879 (aarch64_evpc_ext): Likewise.
9880 (aarch64_evpc_rev): Likewise.
9881 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9882 instead of checking each element individually. Return true without
9884 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9885 whether all selected elements come from the same input, instead of
9886 checking each element individually. Remove calls to gen_rtx_REG,
9887 start_sequence and end_sequence and instead assert that no rtl is
9890 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9892 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9893 order of HIGH and CONST checks.
9895 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9897 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9898 if the destination isn't an SSA_NAME.
9900 2018-01-09 Richard Biener <rguenther@suse.de>
9902 PR tree-optimization/83668
9903 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9905 (canonicalize_loop_form): ... here, renamed from ...
9906 (canonicalize_loop_closed_ssa_form): ... this and amended to
9907 swap successor edges for loop exit blocks to make us use
9908 the RPO order we need for initial schedule generation.
9910 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9912 PR tree-optimization/64811
9913 * match.pd: When optimizing comparisons with Inf, avoid
9914 introducing or losing exceptions from comparisons with NaN.
9916 2018-01-09 Martin Liska <mliska@suse.cz>
9919 * asan.c (shadow_mem_size): Add gcc_assert.
9921 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9923 Don't save registers in main().
9926 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9927 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9928 * config/avr/avr.c (avr_set_current_function): Don't error if
9929 naked, OS_task or OS_main are specified at the same time.
9930 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9932 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9934 * common/config/avr/avr-common.c (avr_option_optimization_table):
9935 Switch on -mmain-is-OS_task for optimizing compilations.
9937 2018-01-09 Richard Biener <rguenther@suse.de>
9939 PR tree-optimization/83572
9940 * graphite.c: Include cfganal.h.
9941 (graphite_transform_loops): Connect infinite loops to exit
9942 and remove fake edges at the end.
9944 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9946 * ipa-inline.c (edge_badness): Revert accidental checkin.
9948 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9951 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9952 symbols; not inline clones.
9954 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9957 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9958 hard registers. Formatting fixes.
9960 PR preprocessor/83722
9961 * gcc.c (try_generate_repro): Pass
9962 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9963 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9966 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9967 Kito Cheng <kito.cheng@gmail.com>
9969 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9970 (riscv_leaf_function_p): Delete.
9971 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9973 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9975 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9977 (do_ifelse): New function.
9978 (do_isel): New function.
9979 (do_sub3): New function.
9980 (do_add3): New function.
9981 (do_load_mask_compare): New function.
9982 (do_overlap_load_compare): New function.
9983 (expand_compare_loop): New function.
9984 (expand_block_compare): Call expand_compare_loop() when appropriate.
9985 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9987 (-mblock-compare-inline-loop-limit): New option.
9989 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9992 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9993 Reverse order of second and third operands in first alternative.
9994 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9995 of first and second elements in UNSPEC_VPERMR vector.
9996 (altivec_expand_vec_perm_le): Likewise.
9998 2018-01-08 Jeff Law <law@redhat.com>
10000 PR rtl-optimizatin/81308
10001 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10002 (process_switch): If group_case_labels makes a change, then set
10004 (pass_convert_switch::execute): If a switch is converted, then
10005 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10007 PR rtl-optimization/81308
10008 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10011 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10013 PR target/83663 - Revert r255946
10014 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10015 generation for cases where splatting a value is not useful.
10016 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10017 across a vec_duplicate and a paradoxical subreg forming a vector
10018 mode to a vec_concat.
10020 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10022 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10023 -march=armv8.3-a variants.
10024 * config/arm/t-multilib: Likewise.
10025 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10027 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10029 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10031 (cceq_ior_compare_complement): Give it a name so I can use it, and
10032 change boolean_or_operator predicate to boolean_operator so it can
10033 be used to generate a crand.
10034 (eqne): New code iterator.
10035 (bd/bd_neg): New code_attrs.
10036 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10037 a single define_insn.
10038 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10039 decrement (bdnzt/bdnzf/bdzt/bdzf).
10040 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10041 with the new names of the branch decrement patterns, and added the
10042 names of the branch decrement conditional patterns.
10044 2018-01-08 Richard Biener <rguenther@suse.de>
10046 PR tree-optimization/83563
10047 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10050 2018-01-08 Richard Biener <rguenther@suse.de>
10052 PR middle-end/83713
10053 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10055 2018-01-08 Richard Biener <rguenther@suse.de>
10057 PR tree-optimization/83685
10058 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10059 references to abnormals.
10061 2018-01-08 Richard Biener <rguenther@suse.de>
10064 * dwarf2out.c (output_indirect_strings): Handle empty
10065 skeleton_debug_str_hash.
10066 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10068 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10070 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10071 (emit_store_direct): Likewise.
10072 (arc_trampoline_adjust_address): Likewise.
10073 (arc_asm_trampoline_template): New function.
10074 (arc_initialize_trampoline): Use asm_trampoline_template.
10075 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10076 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10077 * config/arc/arc.md (flush_icache): Delete pattern.
10079 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10081 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10082 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10085 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10088 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10089 by not USED_FOR_TARGET.
10090 (make_pass_resolve_sw_modes): Likewise.
10092 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10094 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10097 2018-01-08 Richard Biener <rguenther@suse.de>
10099 PR middle-end/83580
10100 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10102 2018-01-08 Richard Biener <rguenther@suse.de>
10104 PR middle-end/83517
10105 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10107 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10109 PR middle-end/81897
10110 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10111 basic blocks with a small number of successors.
10112 (convert_control_dep_chain_into_preds): Improve handling of
10114 (dump_predicates): Split apart into...
10115 (dump_pred_chain): ...here...
10116 (dump_pred_info): ...and here.
10117 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10118 (can_chain_union_be_invalidated_p): Improve check for invalidation
10120 (uninit_uses_cannot_happen): Avoid unnecessary if
10121 convert_control_dep_chain_into_preds yielded nothing.
10123 2018-01-06 Martin Sebor <msebor@redhat.com>
10125 PR tree-optimization/83640
10126 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10127 subtracting negative offset from size.
10128 (builtin_access::overlap): Adjust offset bounds of the access to fall
10129 within the size of the object if possible.
10131 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10133 PR rtl-optimization/83699
10134 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10135 extract_bit_field_as_subreg to cases in which the extracted
10136 value is also a vector.
10138 * lra-constraints.c (process_alt_operands): Test for the equivalence
10139 substitutions when detecting a possible reload cycle.
10141 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10144 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10145 by default if flag_selective_schedling{,2}. Formatting fixes.
10147 PR rtl-optimization/83682
10148 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10149 if it has non-VECTOR_MODE element mode.
10150 (vec_duplicate_p): Likewise.
10152 PR middle-end/83694
10153 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10154 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10156 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10159 * config/i386/i386-builtin.def
10160 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10161 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10162 Require also OPTION_MASK_ISA_AVX512F in addition to
10163 OPTION_MASK_ISA_GFNI.
10164 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10165 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10166 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10167 to OPTION_MASK_ISA_GFNI.
10168 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10169 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10170 OPTION_MASK_ISA_AVX512BW.
10171 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10172 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10173 addition to OPTION_MASK_ISA_GFNI.
10174 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10175 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10176 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10177 to OPTION_MASK_ISA_GFNI.
10178 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10179 a requirement for all ISAs rather than any of them with a few
10181 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10183 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10184 bitmasks to be enabled with 3 exceptions, instead of requiring any
10185 enabled ISA with lots of exceptions.
10186 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10187 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10188 Change avx512bw in isa attribute to avx512f.
10189 * config/i386/sgxintrin.h: Add license boilerplate.
10190 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10191 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10192 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10193 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10195 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10196 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10197 temporarily sse2 rather than sse if not enabled already.
10200 * config/i386/sse.md (VI248_VLBW): Rename to ...
10201 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10202 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10203 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10204 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10205 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10206 mode iterator instead of VI248_VLBW.
10208 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10210 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10211 (record_modified): Skip clobbers; add debug output.
10212 (param_change_prob): Use sreal frequencies.
10214 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10216 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10217 punt for user-aligned variables.
10219 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10221 * tree-chrec.c (chrec_contains_symbols): Return true for
10224 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10227 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10228 of (x|y) == x for BICS pattern.
10230 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10232 PR tree-optimization/83605
10233 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10234 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10237 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10239 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10240 * config/epiphany/rtems.h: New file.
10242 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10243 Uros Bizjak <ubizjak@gmail.com>
10246 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10247 QIreg_operand instead of register_operand predicate.
10248 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10249 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10250 comments instead of -fmitigate[-_]rop.
10252 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10255 * cgraphunit.c (symbol_table::compile): Switch to text_section
10256 before calling assembly_start debug hook.
10257 * run-rtl-passes.c (run_rtl_passes): Likewise.
10260 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10262 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10263 range_int_cst_p rather than !symbolic_range_p before calling
10264 extract_range_from_multiplicative_op_1.
10266 2018-01-04 Jeff Law <law@redhat.com>
10268 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10269 redundant test in assertion.
10271 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10273 * doc/rtl.texi: Document machine_mode wrapper classes.
10275 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10277 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10278 using tree_to_uhwi.
10280 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10282 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10283 the VEC_PERM_EXPR fold to fail.
10285 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10288 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10289 to switched_sections.
10291 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10294 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10295 test for d.testing.
10297 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10300 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10301 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10303 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10306 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10307 is BLKmode and bitpos not zero or mode change is needed.
10309 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10312 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10315 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10318 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10319 instead of MULT rtx. Update all corresponding splitters.
10320 (*saddl_se): Ditto.
10321 (*ssub<modesuffix>): Ditto.
10322 (*ssubl_se): Ditto.
10323 (*cmp_sadd_di): Update split patterns.
10324 (*cmp_sadd_si): Ditto.
10325 (*cmp_sadd_sidi): Ditto.
10326 (*cmp_ssub_di): Ditto.
10327 (*cmp_ssub_si): Ditto.
10328 (*cmp_ssub_sidi): Ditto.
10329 * config/alpha/predicates.md (const23_operand): New predicate.
10330 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10331 Look for ASHIFT, not MULT inner operand.
10332 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10334 2018-01-04 Martin Liska <mliska@suse.cz>
10336 PR gcov-profile/83669
10337 * gcov.c (output_intermediate_file): Add version to intermediate
10339 * doc/gcov.texi: Document new field 'version' in intermediate
10340 file format. Fix location of '-k' option of gcov command.
10342 2018-01-04 Martin Liska <mliska@suse.cz>
10345 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10347 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10349 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10351 2018-01-03 Martin Sebor <msebor@redhat.com>
10353 PR tree-optimization/83655
10354 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10355 checking calls with invalid arguments.
10357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10359 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10360 (vectorizable_mask_load_store): Delete.
10361 (vectorizable_call): Return false for masked loads and stores.
10362 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10363 instead of gimple_assign_rhs1.
10364 (vectorizable_load): Handle IFN_MASK_LOAD.
10365 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10367 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10369 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10371 (vectorizable_mask_load_store): ...here.
10372 (vectorizable_load): ...and here.
10374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10376 * tree-vect-stmts.c (vect_build_all_ones_mask)
10377 (vect_build_zero_merge_argument): New functions, split out from...
10378 (vectorizable_load): ...here.
10380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10382 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10384 (vectorizable_mask_load_store): ...here.
10385 (vectorizable_store): ...and here.
10387 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10389 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10391 (vectorizable_mask_load_store): ...here.
10393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10395 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10396 (vect_model_store_cost): Take a vec_load_store_type instead of a
10398 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10399 (vect_model_store_cost): Take a vec_load_store_type instead of a
10401 (vectorizable_mask_load_store): Update accordingly.
10402 (vectorizable_store): Likewise.
10403 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10405 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10407 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10408 IFN_MASK_LOAD calls here rather than...
10409 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10412 Alan Hayward <alan.hayward@arm.com>
10413 David Sherwood <david.sherwood@arm.com>
10415 * expmed.c (extract_bit_field_1): For vector extracts,
10416 fall back to extract_bit_field_as_subreg if vec_extract
10419 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10420 Alan Hayward <alan.hayward@arm.com>
10421 David Sherwood <david.sherwood@arm.com>
10423 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10424 they are variable or constant sized.
10425 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10426 slots for constant-sized data.
10428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10429 Alan Hayward <alan.hayward@arm.com>
10430 David Sherwood <david.sherwood@arm.com>
10432 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10433 handling COND_EXPRs with boolean comparisons, try to find a better
10434 basis for the mask type than the boolean itself.
10436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10438 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10439 is calculated and how it can be overridden.
10440 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10441 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10443 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10447 Alan Hayward <alan.hayward@arm.com>
10448 David Sherwood <david.sherwood@arm.com>
10450 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10451 Remove the mode argument.
10452 (aarch64_simd_valid_immediate): Remove the mode and inverse
10454 * config/aarch64/iterators.md (bitsize): New iterator.
10455 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10456 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10457 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10458 aarch64_simd_valid_immediate.
10459 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10460 (aarch64_reg_or_bic_imm): Likewise.
10461 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10462 with an insn_type enum and msl with a modifier_type enum.
10463 Replace element_width with a scalar_mode. Change the shift
10464 to unsigned int. Add constructors for scalar_float_mode and
10465 scalar_int_mode elements.
10466 (aarch64_vect_float_const_representable_p): Delete.
10467 (aarch64_can_const_movi_rtx_p)
10468 (aarch64_simd_scalar_immediate_valid_for_move)
10469 (aarch64_simd_make_constant): Update call to
10470 aarch64_simd_valid_immediate.
10471 (aarch64_advsimd_valid_immediate_hs): New function.
10472 (aarch64_advsimd_valid_immediate): Likewise.
10473 (aarch64_simd_valid_immediate): Remove mode and inverse
10474 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10475 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10476 and aarch64_float_const_representable_p on the result.
10477 (aarch64_output_simd_mov_immediate): Remove mode argument.
10478 Update call to aarch64_simd_valid_immediate and use of
10479 simd_immediate_info.
10480 (aarch64_output_scalar_simd_mov_immediate): Update call
10483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10484 Alan Hayward <alan.hayward@arm.com>
10485 David Sherwood <david.sherwood@arm.com>
10487 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10488 (mode_nunits): Likewise CONST_MODE_NUNITS.
10489 * machmode.def (ADJUST_NUNITS): Document.
10490 * genmodes.c (mode_data::need_nunits_adj): New field.
10491 (blank_mode): Update accordingly.
10492 (adj_nunits): New variable.
10493 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10495 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10496 listed in adj_nunits.
10497 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10498 listed in adj_nunits. Don't emit case statements for such modes.
10499 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10500 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10501 nothing if adj_nunits is nonnull.
10502 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10503 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10504 (emit_mode_fbit): Update use of print_maybe_const_decl.
10505 (emit_move_size): Likewise. Treat the array as non-const
10507 (emit_mode_adjustments): Handle adj_nunits.
10509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10511 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10512 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10513 (VECTOR_MODES): Use it.
10514 (make_vector_modes): Take the prefix as an argument.
10516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10517 Alan Hayward <alan.hayward@arm.com>
10518 David Sherwood <david.sherwood@arm.com>
10520 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10521 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10522 for MODE_VECTOR_BOOL.
10523 * machmode.def (VECTOR_BOOL_MODE): Document.
10524 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10525 (make_vector_bool_mode): New function.
10526 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10528 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10529 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10531 * stor-layout.c (int_mode_for_mode): Likewise.
10532 * tree.c (build_vector_type_for_mode): Likewise.
10533 * varasm.c (output_constant_pool_2): Likewise.
10534 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10535 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10536 for MODE_VECTOR_BOOL.
10537 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10538 of mode class checks.
10539 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10540 instead of a list of mode class checks.
10541 (expand_vector_scalar_condition): Likewise.
10542 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10545 Alan Hayward <alan.hayward@arm.com>
10546 David Sherwood <david.sherwood@arm.com>
10548 * machmode.h (mode_size): Change from unsigned short to
10550 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10551 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10552 or if measurement_type is not polynomial.
10553 (fixed_size_mode::includes_p): Check for constant-sized modes.
10554 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10555 return a poly_uint16 rather than an unsigned short.
10556 (emit_mode_size): Change the type of mode_size from unsigned short
10557 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10558 (emit_mode_adjustments): Cope with polynomial vector sizes.
10559 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10561 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10563 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10564 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10565 * caller-save.c (setup_save_areas): Likewise.
10566 (replace_reg_with_saved_mem): Likewise.
10567 * calls.c (emit_library_call_value_1): Likewise.
10568 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10569 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10570 (gen_lowpart_for_combine): Likewise.
10571 * convert.c (convert_to_integer_1): Likewise.
10572 * cse.c (equiv_constant, cse_insn): Likewise.
10573 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10574 (cselib_subst_to_values): Likewise.
10575 * dce.c (word_dce_process_block): Likewise.
10576 * df-problems.c (df_word_lr_mark_ref): Likewise.
10577 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10578 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10579 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10580 (rtl_for_decl_location): Likewise.
10581 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10582 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10583 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10584 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10585 (expand_expr_real_1): Likewise.
10586 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10587 (pad_below): Likewise.
10588 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10589 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10590 * ira.c (get_subreg_tracking_sizes): Likewise.
10591 * ira-build.c (ira_create_allocno_objects): Likewise.
10592 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10593 (ira_sort_regnos_for_alter_reg): Likewise.
10594 * ira-costs.c (record_operand_costs): Likewise.
10595 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10596 (resolve_simple_move): Likewise.
10597 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10598 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10599 (lra_constraints): Likewise.
10600 (CONST_POOL_OK_P): Reject variable-sized modes.
10601 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10602 (add_pseudo_to_slot, lra_spill): Likewise.
10603 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10604 * optabs-query.c (get_best_extraction_insn): Likewise.
10605 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10606 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10607 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10608 * recog.c (offsettable_address_addr_space_p): Likewise.
10609 * regcprop.c (maybe_mode_change): Likewise.
10610 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10611 * regrename.c (build_def_use): Likewise.
10612 * regstat.c (dump_reg_info): Likewise.
10613 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10614 (find_reloads, find_reloads_subreg_address): Likewise.
10615 * reload1.c (eliminate_regs_1): Likewise.
10616 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10617 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10618 (simplify_binary_operation_1, simplify_subreg): Likewise.
10619 * targhooks.c (default_function_arg_padding): Likewise.
10620 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10621 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10622 (verify_gimple_assign_ternary): Likewise.
10623 * tree-inline.c (estimate_move_cost): Likewise.
10624 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10625 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10626 (get_address_cost_ainc): Likewise.
10627 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10628 (vect_supportable_dr_alignment): Likewise.
10629 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10630 (vectorizable_reduction): Likewise.
10631 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10632 (vectorizable_operation, vectorizable_load): Likewise.
10633 * tree.c (build_same_sized_truth_vector_type): Likewise.
10634 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10635 * var-tracking.c (emit_note_insn_var_location): Likewise.
10636 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10637 (ADDR_VEC_ALIGN): Likewise.
10639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10640 Alan Hayward <alan.hayward@arm.com>
10641 David Sherwood <david.sherwood@arm.com>
10643 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10645 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10646 or if measurement_type is polynomial.
10647 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10648 * combine.c (make_extraction): Likewise.
10649 * dse.c (find_shift_sequence): Likewise.
10650 * dwarf2out.c (mem_loc_descriptor): Likewise.
10651 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10652 (extract_bit_field, extract_low_bits): Likewise.
10653 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10654 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10655 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10656 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10657 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10658 * reload.c (find_reloads): Likewise.
10659 * reload1.c (alter_reg): Likewise.
10660 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10661 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10662 * tree-if-conv.c (predicate_mem_writes): Likewise.
10663 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10664 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10665 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10666 * valtrack.c (dead_debug_insert_temp): Likewise.
10667 * varasm.c (mergeable_constant_section): Likewise.
10668 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10671 Alan Hayward <alan.hayward@arm.com>
10672 David Sherwood <david.sherwood@arm.com>
10674 * expr.c (expand_assignment): Cope with polynomial mode sizes
10675 when assigning to a CONCAT.
10677 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10678 Alan Hayward <alan.hayward@arm.com>
10679 David Sherwood <david.sherwood@arm.com>
10681 * machmode.h (mode_precision): Change from unsigned short to
10683 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10685 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10686 or if measurement_type is not polynomial.
10687 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10688 in which the mode is already known to be a scalar_int_mode.
10689 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10690 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10692 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10693 for GET_MODE_PRECISION.
10694 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10695 for GET_MODE_PRECISION.
10696 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10698 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10699 (expand_field_assignment, make_extraction): Likewise.
10700 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10701 (get_last_value): Likewise.
10702 * convert.c (convert_to_integer_1): Likewise.
10703 * cse.c (cse_insn): Likewise.
10704 * expr.c (expand_expr_real_1): Likewise.
10705 * lra-constraints.c (simplify_operand_subreg): Likewise.
10706 * optabs-query.c (can_atomic_load_p): Likewise.
10707 * optabs.c (expand_atomic_load): Likewise.
10708 (expand_atomic_store): Likewise.
10709 * ree.c (combine_reaching_defs): Likewise.
10710 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10711 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10712 * tree.h (type_has_mode_precision_p): Likewise.
10713 * ubsan.c (instrument_si_overflow): Likewise.
10715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10716 Alan Hayward <alan.hayward@arm.com>
10717 David Sherwood <david.sherwood@arm.com>
10719 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10720 polynomial numbers of units.
10721 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10722 (valid_vector_subparts_p): New function.
10723 (build_vector_type): Remove temporary shim and take the number
10724 of units as a poly_uint64 rather than an int.
10725 (build_opaque_vector_type): Take the number of units as a
10726 poly_uint64 rather than an int.
10727 * tree.c (build_vector_from_ctor): Handle polynomial
10728 TYPE_VECTOR_SUBPARTS.
10729 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10730 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10731 (build_vector_from_val): If the number of units is variable,
10732 use build_vec_duplicate_cst for constant operands and
10733 VEC_DUPLICATE_EXPR otherwise.
10734 (make_vector_type): Remove temporary is_constant ().
10735 (build_vector_type, build_opaque_vector_type): Take the number of
10736 units as a poly_uint64 rather than an int.
10737 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10739 * cfgexpand.c (expand_debug_expr): Likewise.
10740 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10741 (store_constructor, expand_expr_real_1): Likewise.
10742 (const_scalar_mask_from_tree): Likewise.
10743 * fold-const-call.c (fold_const_reduction): Likewise.
10744 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10745 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10746 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10747 (fold_relational_const): Likewise.
10748 (native_interpret_vector): Likewise. Change the size from an
10749 int to an unsigned int.
10750 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10751 TYPE_VECTOR_SUBPARTS.
10752 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10753 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10754 duplicating a non-constant operand into a variable-length vector.
10755 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10756 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10757 * ipa-icf.c (sem_variable::equals): Likewise.
10758 * match.pd: Likewise.
10759 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10760 * print-tree.c (print_node): Likewise.
10761 * stor-layout.c (layout_type): Likewise.
10762 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10763 * tree-cfg.c (verify_gimple_comparison): Likewise.
10764 (verify_gimple_assign_binary): Likewise.
10765 (verify_gimple_assign_ternary): Likewise.
10766 (verify_gimple_assign_single): Likewise.
10767 * tree-pretty-print.c (dump_generic_node): Likewise.
10768 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10769 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10770 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10771 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10772 (vect_shift_permute_load_chain): Likewise.
10773 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10774 (expand_vector_condition, optimize_vector_constructor): Likewise.
10775 (lower_vec_perm, get_compute_type): Likewise.
10776 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10777 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10778 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10779 (vect_recog_mask_conversion_pattern): Likewise.
10780 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10781 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10782 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10783 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10784 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10785 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10786 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10787 (supportable_widening_operation): Likewise.
10788 (supportable_narrowing_operation): Likewise.
10789 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10791 * varasm.c (output_constant): Likewise.
10793 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10794 Alan Hayward <alan.hayward@arm.com>
10795 David Sherwood <david.sherwood@arm.com>
10797 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10798 so that both the length == 3 and length != 3 cases set up their
10799 own permute vectors. Add comments explaining why we know the
10800 number of elements is constant.
10801 (vect_permute_load_chain): Likewise.
10803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10804 Alan Hayward <alan.hayward@arm.com>
10805 David Sherwood <david.sherwood@arm.com>
10807 * machmode.h (mode_nunits): Change from unsigned char to
10809 (ONLY_FIXED_SIZE_MODES): New macro.
10810 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10811 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10812 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10814 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10815 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10816 or if measurement_type is not polynomial.
10817 * genmodes.c (ZERO_COEFFS): New macro.
10818 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10820 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10821 Use ZERO_COEFFS when emitting initializers.
10822 * data-streamer.h (bp_pack_poly_value): New function.
10823 (bp_unpack_poly_value): Likewise.
10824 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10825 for GET_MODE_NUNITS.
10826 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10827 for GET_MODE_NUNITS.
10828 * tree.c (make_vector_type): Remove temporary shim and make
10829 the real function take the number of units as a poly_uint64
10830 rather than an int.
10831 (build_vector_type_for_mode): Handle polynomial nunits.
10832 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10833 * emit-rtl.c (const_vec_series_p_1): Likewise.
10834 (gen_rtx_CONST_VECTOR): Likewise.
10835 * fold-const.c (test_vec_duplicate_folding): Likewise.
10836 * genrecog.c (validate_pattern): Likewise.
10837 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10838 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10839 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10840 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10841 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10842 * rtlanal.c (subreg_get_info): Likewise.
10843 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10844 (vect_grouped_load_supported): Likewise.
10845 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10846 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10847 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10848 (simplify_const_unary_operation, simplify_binary_operation_1)
10849 (simplify_const_binary_operation, simplify_ternary_operation)
10850 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10851 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10852 instead of CONST_VECTOR_NUNITS.
10853 * varasm.c (output_constant_pool_2): Likewise.
10854 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10855 explicit-encoded elements in the XVEC for variable-length vectors.
10857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10859 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10862 Alan Hayward <alan.hayward@arm.com>
10863 David Sherwood <david.sherwood@arm.com>
10865 * coretypes.h (fixed_size_mode): Declare.
10866 (fixed_size_mode_pod): New typedef.
10867 * builtins.h (target_builtins::x_apply_args_mode)
10868 (target_builtins::x_apply_result_mode): Change type to
10869 fixed_size_mode_pod.
10870 * builtins.c (apply_args_size, apply_result_size, result_vector)
10871 (expand_builtin_apply_args_1, expand_builtin_apply)
10872 (expand_builtin_return): Update accordingly.
10874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10876 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10877 * cselib.c (cselib_hash_rtx): Likewise.
10878 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10879 CONST_VECTOR encoding.
10881 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10882 Jeff Law <law@redhat.com>
10885 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10886 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10887 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10888 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10891 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10892 explicitly probe *sp in a noreturn function if there were any callee
10893 register saves or frame pointer is needed.
10895 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10898 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10899 BLKmode for ternary, binary or unary expressions.
10902 * var-tracking.c (delete_vta_debug_insn): New inline function.
10903 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10904 insns from get_insns () to NULL instead of each bb separately.
10905 Use delete_vta_debug_insn. No longer static.
10906 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10907 delete_vta_debug_insns callers.
10908 * rtl.h (delete_vta_debug_insns): Declare.
10909 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10910 instead of variable_tracking_main.
10912 2018-01-03 Martin Sebor <msebor@redhat.com>
10914 PR tree-optimization/83603
10915 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10916 arguments past the endof the argument list in functions declared
10917 without a prototype.
10918 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10919 Avoid checking when arguments are null.
10921 2018-01-03 Martin Sebor <msebor@redhat.com>
10924 * doc/extend.texi (attribute const): Fix a typo.
10925 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10926 issuing -Wsuggest-attribute for void functions.
10928 2018-01-03 Martin Sebor <msebor@redhat.com>
10930 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10931 offset_int::from instead of wide_int::to_shwi.
10932 (maybe_diag_overlap): Remove assertion.
10933 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10934 * gimple-ssa-sprintf.c (format_directive): Same.
10935 (parse_directive): Same.
10936 (sprintf_dom_walker::compute_format_length): Same.
10937 (try_substitute_return_value): Same.
10939 2018-01-03 Jeff Law <law@redhat.com>
10941 PR middle-end/83654
10942 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10943 non-constant residual for zero at runtime and avoid probing in
10944 that case. Reorganize code for trailing problem to mirror handling
10947 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10949 PR tree-optimization/83501
10950 * tree-ssa-strlen.c (get_string_cst): New.
10951 (handle_char_store): Call get_string_cst.
10953 2018-01-03 Martin Liska <mliska@suse.cz>
10955 PR tree-optimization/83593
10956 * tree-ssa-strlen.c: Include tree-cfg.h.
10957 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10958 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10959 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10961 (strlen_dom_walker::before_dom_children): Call
10962 gimple_purge_dead_eh_edges. Dump tranformation with details
10964 (strlen_dom_walker::before_dom_children): Update call by adding
10965 new argument cleanup_eh.
10966 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10968 2018-01-03 Martin Liska <mliska@suse.cz>
10971 * cif-code.def (VARIADIC_THUNK): New enum value.
10972 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10975 2018-01-03 Jan Beulich <jbeulich@suse.com>
10977 * sse.md (mov<mode>_internal): Tighten condition for when to use
10978 vmovdqu<ssescalarsize> for TI and OI modes.
10980 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10982 Update copyright years.
10984 2018-01-03 Martin Liska <mliska@suse.cz>
10987 * ipa-visibility.c (function_and_variable_visibility): Skip
10988 functions with noipa attribure.
10990 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10992 * gcc.c (process_command): Update copyright notice dates.
10993 * gcov-dump.c (print_version): Ditto.
10994 * gcov.c (print_version): Ditto.
10995 * gcov-tool.c (print_version): Ditto.
10996 * gengtype.c (create_file): Ditto.
10997 * doc/cpp.texi: Bump @copying's copyright year.
10998 * doc/cppinternals.texi: Ditto.
10999 * doc/gcc.texi: Ditto.
11000 * doc/gccint.texi: Ditto.
11001 * doc/gcov.texi: Ditto.
11002 * doc/install.texi: Ditto.
11003 * doc/invoke.texi: Ditto.
11005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11007 * vector-builder.h (vector_builder::m_full_nelts): Change from
11008 unsigned int to poly_uint64.
11009 (vector_builder::full_nelts): Update prototype accordingly.
11010 (vector_builder::new_vector): Likewise.
11011 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11012 (vector_builder::operator ==): Likewise.
11013 (vector_builder::finalize): Likewise.
11014 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11015 Take the number of elements as a poly_uint64 rather than an
11017 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11018 from unsigned int to poly_uint64.
11019 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11020 (vec_perm_indices::new_vector): Likewise.
11021 (vec_perm_indices::length): Likewise.
11022 (vec_perm_indices::nelts_per_input): Likewise.
11023 (vec_perm_indices::input_nelts): Likewise.
11024 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11025 number of elements per input as a poly_uint64 rather than an
11026 unsigned int. Use the original encoding for variable-length
11027 vectors, rather than clamping each individual element.
11028 For the second and subsequent elements in each pattern,
11029 clamp the step and base before clamping their sum.
11030 (vec_perm_indices::series_p): Handle polynomial element counts.
11031 (vec_perm_indices::all_in_range_p): Likewise.
11032 (vec_perm_indices_to_tree): Likewise.
11033 (vec_perm_indices_to_rtx): Likewise.
11034 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11035 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11036 (tree_vector_builder::new_binary_operation): Handle polynomial
11037 element counts. Return false if we need to know the number
11038 of elements at compile time.
11039 * fold-const.c (fold_vec_perm): Punt if the number of elements
11040 isn't known at compile time.
11042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11044 * vec-perm-indices.h (vec_perm_builder): Change element type
11045 from HOST_WIDE_INT to poly_int64.
11046 (vec_perm_indices::element_type): Update accordingly.
11047 (vec_perm_indices::clamp): Handle polynomial element_types.
11048 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11049 (vec_perm_indices::all_in_range_p): Likewise.
11050 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11052 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11053 polynomial vec_perm_indices element types.
11054 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11055 * fold-const.c (fold_vec_perm): Likewise.
11056 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11057 * tree-vect-generic.c (lower_vec_perm): Likewise.
11058 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11059 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11060 element type to HOST_WIDE_INT.
11062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11063 Alan Hayward <alan.hayward@arm.com>
11064 David Sherwood <david.sherwood@arm.com>
11066 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11067 rather than an int. Use plus_constant.
11068 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11069 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11072 Alan Hayward <alan.hayward@arm.com>
11073 David Sherwood <david.sherwood@arm.com>
11075 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11076 a HOST_WIDE_INT to a poly_int64.
11078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11079 Alan Hayward <alan.hayward@arm.com>
11080 David Sherwood <david.sherwood@arm.com>
11082 * calls.c (load_register_parameters): Cope with polynomial
11083 mode sizes. Require a constant size for BLKmode parameters
11084 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11085 forces a parameter to be padded at the lsb end in order to
11086 fill a complete number of words, require the parameter size
11087 to be ordered wrt UNITS_PER_WORD.
11089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11090 Alan Hayward <alan.hayward@arm.com>
11091 David Sherwood <david.sherwood@arm.com>
11093 * reload1.c (spill_stack_slot_width): Change element type
11094 from unsigned int to poly_uint64_pod.
11095 (alter_reg): Treat mode sizes as polynomial.
11097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11098 Alan Hayward <alan.hayward@arm.com>
11099 David Sherwood <david.sherwood@arm.com>
11101 * reload.c (complex_word_subreg_p): New function.
11102 (reload_inner_reg_of_subreg, push_reload): Use it.
11104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11105 Alan Hayward <alan.hayward@arm.com>
11106 David Sherwood <david.sherwood@arm.com>
11108 * lra-constraints.c (process_alt_operands): Reject matched
11109 operands whose sizes aren't ordered.
11110 (match_reload): Refer to this check here.
11112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11113 Alan Hayward <alan.hayward@arm.com>
11114 David Sherwood <david.sherwood@arm.com>
11116 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11117 that the mode size is in the set {1, 2, 4, 8, 16}.
11119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11120 Alan Hayward <alan.hayward@arm.com>
11121 David Sherwood <david.sherwood@arm.com>
11123 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11124 Use plus_constant instead of gen_rtx_PLUS.
11126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11127 Alan Hayward <alan.hayward@arm.com>
11128 David Sherwood <david.sherwood@arm.com>
11130 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11131 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11132 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11133 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11134 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11135 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11136 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11137 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11138 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11139 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11141 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11142 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11143 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11144 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11145 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11146 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11147 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11148 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11149 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11150 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11152 * expr.c (emit_move_resolve_push): Treat the input and result
11153 of PUSH_ROUNDING as a poly_int64.
11154 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11155 (emit_push_insn): Likewise.
11156 * lra-eliminations.c (mark_not_eliminable): Likewise.
11157 * recog.c (push_operand): Likewise.
11158 * reload1.c (elimination_effects): Likewise.
11159 * rtlanal.c (nonzero_bits1): Likewise.
11160 * calls.c (store_one_arg): Likewise. Require the padding to be
11161 known at compile time.
11163 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11164 Alan Hayward <alan.hayward@arm.com>
11165 David Sherwood <david.sherwood@arm.com>
11167 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11168 Use plus_constant instead of gen_rtx_PLUS.
11170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11171 Alan Hayward <alan.hayward@arm.com>
11172 David Sherwood <david.sherwood@arm.com>
11174 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11175 rather than an int.
11177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11178 Alan Hayward <alan.hayward@arm.com>
11179 David Sherwood <david.sherwood@arm.com>
11181 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11182 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11183 via stack temporaries. Treat the mode size as polynomial too.
11185 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11186 Alan Hayward <alan.hayward@arm.com>
11187 David Sherwood <david.sherwood@arm.com>
11189 * expr.c (expand_expr_real_2): When handling conversions involving
11190 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11191 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11192 as a poly_uint64 too.
11194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11195 Alan Hayward <alan.hayward@arm.com>
11196 David Sherwood <david.sherwood@arm.com>
11198 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11201 Alan Hayward <alan.hayward@arm.com>
11202 David Sherwood <david.sherwood@arm.com>
11204 * combine.c (can_change_dest_mode): Handle polynomial
11205 REGMODE_NATURAL_SIZE.
11206 * expmed.c (store_bit_field_1): Likewise.
11207 * expr.c (store_constructor): Likewise.
11208 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11209 and polynomial REGMODE_NATURAL_SIZE.
11210 (gen_lowpart_common): Likewise.
11211 * reginfo.c (record_subregs_of_mode): Likewise.
11212 * rtlanal.c (read_modify_subreg_p): Likewise.
11214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11215 Alan Hayward <alan.hayward@arm.com>
11216 David Sherwood <david.sherwood@arm.com>
11218 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11219 numbers of elements.
11221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11222 Alan Hayward <alan.hayward@arm.com>
11223 David Sherwood <david.sherwood@arm.com>
11225 * match.pd: Cope with polynomial numbers of vector elements.
11227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11228 Alan Hayward <alan.hayward@arm.com>
11229 David Sherwood <david.sherwood@arm.com>
11231 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11232 in a POINTER_PLUS_EXPR.
11234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11235 Alan Hayward <alan.hayward@arm.com>
11236 David Sherwood <david.sherwood@arm.com>
11238 * omp-simd-clone.c (simd_clone_subparts): New function.
11239 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11240 (ipa_simd_modify_function_body): Likewise.
11242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11243 Alan Hayward <alan.hayward@arm.com>
11244 David Sherwood <david.sherwood@arm.com>
11246 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11247 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11248 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11249 (expand_vector_condition, vector_element): Likewise.
11250 (subparts_gt): New function.
11251 (get_compute_type): Use subparts_gt.
11252 (count_type_subparts): Delete.
11253 (expand_vector_operations_1): Use subparts_gt instead of
11254 count_type_subparts.
11256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11257 Alan Hayward <alan.hayward@arm.com>
11258 David Sherwood <david.sherwood@arm.com>
11260 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11261 (vect_compile_time_alias): ...this new function. Do the calculation
11262 on poly_ints rather than trees.
11263 (vect_prune_runtime_alias_test_list): Update call accordingly.
11265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11266 Alan Hayward <alan.hayward@arm.com>
11267 David Sherwood <david.sherwood@arm.com>
11269 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11271 (vect_schedule_slp_instance): Likewise.
11273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11274 Alan Hayward <alan.hayward@arm.com>
11275 David Sherwood <david.sherwood@arm.com>
11277 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11278 constant and extern definitions for variable-length vectors.
11279 (vect_get_constant_vectors): Note that the number of units
11280 is known to be constant.
11282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11283 Alan Hayward <alan.hayward@arm.com>
11284 David Sherwood <david.sherwood@arm.com>
11286 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11287 of units as polynomial. Choose between WIDE and NARROW based
11290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11291 Alan Hayward <alan.hayward@arm.com>
11292 David Sherwood <david.sherwood@arm.com>
11294 * tree-vect-stmts.c (simd_clone_subparts): New function.
11295 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11298 Alan Hayward <alan.hayward@arm.com>
11299 David Sherwood <david.sherwood@arm.com>
11301 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11302 vectors as polynomial. Use build_index_vector for
11303 IFN_GOMP_SIMD_LANE.
11305 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11306 Alan Hayward <alan.hayward@arm.com>
11307 David Sherwood <david.sherwood@arm.com>
11309 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11310 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11311 for variable-length vectors.
11312 (vectorizable_mask_load_store): Treat the number of units as
11313 polynomial, asserting that it is constant if the condition has
11314 already been enforced.
11315 (vectorizable_store, vectorizable_load): Likewise.
11317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11318 Alan Hayward <alan.hayward@arm.com>
11319 David Sherwood <david.sherwood@arm.com>
11321 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11322 of units as polynomial. Punt if we can't tell at compile time
11323 which vector contains the final result.
11325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11326 Alan Hayward <alan.hayward@arm.com>
11327 David Sherwood <david.sherwood@arm.com>
11329 * tree-vect-loop.c (vectorizable_induction): Treat the number
11330 of units as polynomial. Punt on SLP inductions. Use an integer
11331 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11332 cast of such a series for variable-length floating-point
11335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11336 Alan Hayward <alan.hayward@arm.com>
11337 David Sherwood <david.sherwood@arm.com>
11339 * tree.h (build_index_vector): Declare.
11340 * tree.c (build_index_vector): New function.
11341 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11342 of units as polynomial, forcibly converting it to a constant if
11343 vectorizable_reduction has already enforced the condition.
11344 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11345 to create a {1,2,3,...} vector.
11346 (vectorizable_reduction): Treat the number of units as polynomial.
11347 Choose vectype_in based on the largest scalar element size rather
11348 than the smallest number of units. Enforce the restrictions
11351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11352 Alan Hayward <alan.hayward@arm.com>
11353 David Sherwood <david.sherwood@arm.com>
11355 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11356 number of units as polynomial.
11358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11359 Alan Hayward <alan.hayward@arm.com>
11360 David Sherwood <david.sherwood@arm.com>
11362 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11363 * target.def (autovectorize_vector_sizes): Return the vector sizes
11364 by pointer, using vector_sizes rather than a bitmask.
11365 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11366 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11367 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11369 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11370 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11371 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11372 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11373 * omp-general.c (omp_max_vf): Likewise.
11374 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11375 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11376 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11377 * tree-vect-slp.c (vect_slp_bb): Likewise.
11378 * doc/tm.texi: Regenerate.
11379 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11381 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11382 the vector size as a poly_uint64 rather than an unsigned int.
11383 (current_vector_size): Change from an unsigned int to a poly_uint64.
11384 (get_vectype_for_scalar_type): Update accordingly.
11385 * tree.h (build_truth_vector_type): Take the size and number of
11386 units as a poly_uint64 rather than an unsigned int.
11387 (build_vector_type): Add a temporary overload that takes
11388 the number of units as a poly_uint64 rather than an unsigned int.
11389 * tree.c (make_vector_type): Likewise.
11390 (build_truth_vector_type): Take the number of units as a poly_uint64
11391 rather than an unsigned int.
11393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11394 Alan Hayward <alan.hayward@arm.com>
11395 David Sherwood <david.sherwood@arm.com>
11397 * target.def (get_mask_mode): Take the number of units and length
11398 as poly_uint64s rather than unsigned ints.
11399 * targhooks.h (default_get_mask_mode): Update accordingly.
11400 * targhooks.c (default_get_mask_mode): Likewise.
11401 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11402 * doc/tm.texi: Regenerate.
11404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11405 Alan Hayward <alan.hayward@arm.com>
11406 David Sherwood <david.sherwood@arm.com>
11408 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11409 * omp-general.c (omp_max_vf): Likewise.
11410 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11411 (expand_omp_simd): Handle polynomial safelen.
11412 * omp-low.c (omplow_simd_context): Add a default constructor.
11413 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11414 (lower_rec_simd_input_clauses): Update accordingly.
11415 (lower_rec_input_clauses): Likewise.
11417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11418 Alan Hayward <alan.hayward@arm.com>
11419 David Sherwood <david.sherwood@arm.com>
11421 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11422 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11423 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11424 (vect_analyze_slp_cost): Likewise.
11425 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11426 (vect_model_load_cost): Likewise.
11428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11429 Alan Hayward <alan.hayward@arm.com>
11430 David Sherwood <david.sherwood@arm.com>
11432 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11433 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11434 from an unsigned int * to a poly_uint64_pod *.
11435 (calculate_unrolling_factor): New function.
11436 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11439 Alan Hayward <alan.hayward@arm.com>
11440 David Sherwood <david.sherwood@arm.com>
11442 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11443 from an unsigned int to a poly_uint64.
11444 (_loop_vec_info::slp_unrolling_factor): Likewise.
11445 (_loop_vec_info::vectorization_factor): Change from an int
11447 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11448 (vect_get_num_vectors): New function.
11449 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11450 (vect_get_num_copies): Use vect_get_num_vectors.
11451 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11452 to an unsigned int *.
11453 (vect_analyze_data_refs): Change min_vf from an int * to a
11455 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11456 than an unsigned HOST_WIDE_INT.
11457 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11458 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11459 to an unsigned int *.
11460 (vect_analyze_data_ref_dependences): Likewise.
11461 (vect_compute_data_ref_alignment): Handle polynomial vf.
11462 (vect_enhance_data_refs_alignment): Likewise.
11463 (vect_prune_runtime_alias_test_list): Likewise.
11464 (vect_shift_permute_load_chain): Likewise.
11465 (vect_supportable_dr_alignment): Likewise.
11466 (dependence_distance_ge_vf): Take the vectorization factor as a
11467 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11468 (vect_analyze_data_refs): Change min_vf from an int * to a
11470 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11471 vfm1 as a poly_uint64 rather than an int. Make the same change
11472 for the returned bound_scalar.
11473 (vect_gen_vector_loop_niters): Handle polynomial vf.
11474 (vect_do_peeling): Likewise. Update call to
11475 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11476 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11478 * tree-vect-loop.c (vect_determine_vectorization_factor)
11479 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11480 (vect_get_known_peeling_cost): Likewise.
11481 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11482 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11483 (vect_transform_loop): Likewise. Use the lowest possible VF when
11484 updating the upper bounds of the loop.
11485 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11486 rather than an int.
11487 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11488 polynomial unroll factors.
11489 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11490 (vect_make_slp_decision): Likewise.
11491 (vect_supported_load_permutation_p): Likewise, and polynomial
11493 (vect_analyze_slp_cost): Handle polynomial vf.
11494 (vect_slp_analyze_node_operations): Likewise.
11495 (vect_slp_analyze_bb_1): Likewise.
11496 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11497 than an unsigned HOST_WIDE_INT.
11498 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11499 (vectorizable_load): Handle polynomial vf.
11500 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11502 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11505 Alan Hayward <alan.hayward@arm.com>
11506 David Sherwood <david.sherwood@arm.com>
11508 * match.pd: Handle bit operations involving three constants
11509 and try to fold one pair.
11511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11513 * tree-vect-loop-manip.c: Include gimple-fold.h.
11514 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11515 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11516 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11517 Add a path that uses a step of VF instead of 1, but disable it
11519 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11520 and niters_no_overflow parameters. Update calls to
11521 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11522 Create a new SSA name if the latter choses to use a ste other
11523 than zero, and return it via niters_vector_mult_vf_var.
11524 * tree-vect-loop.c (vect_transform_loop): Update calls to
11525 vect_do_peeling, vect_gen_vector_loop_niters and
11526 slpeel_make_loop_iterate_ntimes.
11527 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11528 (vect_gen_vector_loop_niters): Update declarations after above changes.
11530 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11532 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11533 128-bit round to integer instructions.
11534 (ceil<mode>2): Likewise.
11535 (btrunc<mode>2): Likewise.
11536 (round<mode>2): Likewise.
11538 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11540 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11541 unaligned VSX load/store on P8/P9.
11542 (expand_block_clear): Allow the use of unaligned VSX
11543 load/store on P8/P9.
11545 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11547 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11549 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11550 swap associated with both a load and a store.
11552 2018-01-02 Andrew Waterman <andrew@sifive.com>
11554 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11555 * config/riscv/riscv.md (clear_cache): Use it.
11557 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11559 * web.c: Remove out-of-date comment.
11561 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11563 * expr.c (fixup_args_size_notes): Check that any existing
11564 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11565 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11566 (emit_single_push_insn): ...here.
11568 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11570 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11571 (const_vector_encoded_nelts): New function.
11572 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11573 (const_vector_int_elt, const_vector_elt): Declare.
11574 * emit-rtl.c (const_vector_int_elt_1): New function.
11575 (const_vector_elt): Likewise.
11576 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11577 of CONST_VECTOR_ELT.
11579 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11581 * expr.c: Include rtx-vector-builder.h.
11582 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11583 directly on the tree encoding.
11584 (const_vector_from_tree): Likewise.
11585 * optabs.c: Include rtx-vector-builder.h.
11586 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11587 sequence of "u" values.
11588 * vec-perm-indices.c: Include rtx-vector-builder.h.
11589 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11590 directly on the vec_perm_indices encoding.
11592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11594 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11595 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11596 * rtx-vector-builder.h: New file.
11597 * rtx-vector-builder.c: Likewise.
11598 * rtl.h (rtx_def::u2): Add a const_vector field.
11599 (CONST_VECTOR_NPATTERNS): New macro.
11600 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11601 (CONST_VECTOR_DUPLICATE_P): Likewise.
11602 (CONST_VECTOR_STEPPED_P): Likewise.
11603 (CONST_VECTOR_ENCODED_ELT): Likewise.
11604 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11605 (unwrap_const_vec_duplicate): Likewise.
11606 (const_vec_series_p): Check for a non-duplicated vector encoding.
11607 Say that the function only returns true for integer vectors.
11608 * emit-rtl.c: Include rtx-vector-builder.h.
11609 (gen_const_vec_duplicate_1): Delete.
11610 (gen_const_vector): Call gen_const_vec_duplicate instead of
11611 gen_const_vec_duplicate_1.
11612 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11613 (gen_const_vec_duplicate): Use rtx_vector_builder.
11614 (gen_const_vec_series): Likewise.
11615 (gen_rtx_CONST_VECTOR): Likewise.
11616 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11617 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11618 Build a new vector rather than modifying a CONST_VECTOR in-place.
11619 (handle_special_swappables): Update call accordingly.
11620 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11621 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11622 Build a new vector rather than modifying a CONST_VECTOR in-place.
11623 (handle_special_swappables): Update call accordingly.
11625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11627 * simplify-rtx.c (simplify_const_binary_operation): Use
11628 CONST_VECTOR_ELT instead of XVECEXP.
11630 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11632 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11633 the selector elements to be different from the data elements
11634 if the selector is a VECTOR_CST.
11635 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11636 ssizetype for the selector.
11638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11640 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11641 before testing each element individually.
11642 * tree-vect-generic.c (lower_vec_perm): Likewise.
11644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11646 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11647 * selftest-run-tests.c (selftest::run_tests): Call it.
11648 * vector-builder.h (vector_builder::operator ==): New function.
11649 (vector_builder::operator !=): Likewise.
11650 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11651 (vec_perm_indices::all_from_input_p): New function.
11652 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11653 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11654 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11655 instead of reading the VECTOR_CST directly. Detect whether both
11656 vector inputs are the same before constructing the vec_perm_indices,
11657 and update the number of inputs argument accordingly. Use the
11658 utility functions added above. Only construct sel2 if we need to.
11660 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11662 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11663 the broadcast of the low byte.
11664 (expand_mult_highpart): Use an explicit encoding for the permutes.
11665 * optabs-query.c (can_mult_highpart_p): Likewise.
11666 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11667 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11668 (vectorizable_bswap): Likewise.
11669 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11670 explicit encoding for the power-of-2 permutes.
11671 (vect_permute_store_chain): Likewise.
11672 (vect_grouped_load_supported): Likewise.
11673 (vect_permute_load_chain): Likewise.
11675 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11677 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11678 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11679 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11680 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11681 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11682 (vect_gen_perm_mask_any): Likewise.
11684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11686 * int-vector-builder.h: New file.
11687 * vec-perm-indices.h: Include int-vector-builder.h.
11688 (vec_perm_indices): Redefine as an int_vector_builder.
11689 (auto_vec_perm_indices): Delete.
11690 (vec_perm_builder): Redefine as a stand-alone class.
11691 (vec_perm_indices::vec_perm_indices): New function.
11692 (vec_perm_indices::clamp): Likewise.
11693 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11694 (vec_perm_indices::new_vector): New function.
11695 (vec_perm_indices::new_expanded_vector): Update for new
11696 vec_perm_indices class.
11697 (vec_perm_indices::rotate_inputs): New function.
11698 (vec_perm_indices::all_in_range_p): Operate directly on the
11699 encoded form, without computing elided elements.
11700 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11701 encoding. Update for new vec_perm_indices class.
11702 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11703 the given vec_perm_builder.
11704 (expand_vec_perm_var): Update vec_perm_builder constructor.
11705 (expand_mult_highpart): Use vec_perm_builder instead of
11706 auto_vec_perm_indices.
11707 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11708 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11709 or double series encoding as appropriate.
11710 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11711 vec_perm_indices instead of auto_vec_perm_indices.
11712 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11713 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11714 (vect_permute_store_chain): Likewise.
11715 (vect_grouped_load_supported): Likewise.
11716 (vect_permute_load_chain): Likewise.
11717 (vect_shift_permute_load_chain): Likewise.
11718 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11719 (vect_transform_slp_perm_load): Likewise.
11720 (vect_schedule_slp_instance): Likewise.
11721 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11722 (vectorizable_mask_load_store): Likewise.
11723 (vectorizable_bswap): Likewise.
11724 (vectorizable_store): Likewise.
11725 (vectorizable_load): Likewise.
11726 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11727 vec_perm_indices instead of auto_vec_perm_indices. Use
11728 tree_to_vec_perm_builder to read the vector from a tree.
11729 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11730 vec_perm_builder instead of a vec_perm_indices.
11731 (have_whole_vector_shift): Use vec_perm_builder and
11732 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11733 truncation to calc_vec_perm_mask_for_shift.
11734 (vect_create_epilog_for_reduction): Likewise.
11735 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11736 from auto_vec_perm_indices to vec_perm_indices.
11737 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11738 instead of changing individual elements.
11739 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11740 the vector in d.perm.
11741 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11742 from auto_vec_perm_indices to vec_perm_indices.
11743 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11744 instead of changing individual elements.
11745 (arm_vectorize_vec_perm_const): Use new_vector to install
11746 the vector in d.perm.
11747 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11748 Update vec_perm_builder constructor.
11749 (rs6000_expand_interleave): Likewise.
11750 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11751 (rs6000_expand_interleave): Likewise.
11753 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11755 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11756 to qimode could truncate the indices.
11757 * optabs.c (expand_vec_perm_var): Likewise.
11759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11761 * Makefile.in (OBJS): Add vec-perm-indices.o.
11762 * vec-perm-indices.h: New file.
11763 * vec-perm-indices.c: Likewise.
11764 * target.h (vec_perm_indices): Replace with a forward class
11766 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11767 * optabs.h: Include vec-perm-indices.h.
11768 (expand_vec_perm): Delete.
11769 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11770 (expand_vec_perm_const): Declare.
11771 * target.def (vec_perm_const_ok): Replace with...
11772 (vec_perm_const): ...this new hook.
11773 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11774 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11775 * doc/tm.texi: Regenerate.
11776 * optabs.def (vec_perm_const): Delete.
11777 * doc/md.texi (vec_perm_const): Likewise.
11778 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11779 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11780 expand_vec_perm for constant permutation vectors. Assert that
11781 the mode of variable permutation vectors is the integer equivalent
11782 of the mode that is being permuted.
11783 * optabs-query.h (selector_fits_mode_p): Declare.
11784 * optabs-query.c: Include vec-perm-indices.h.
11785 (selector_fits_mode_p): New function.
11786 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11787 is defined, instead of checking whether the vec_perm_const_optab
11788 exists. Use targetm.vectorize.vec_perm_const instead of
11789 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11790 fit in the vector mode before using a variable permute.
11791 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11792 vec_perm_indices instead of an rtx.
11793 (expand_vec_perm): Replace with...
11794 (expand_vec_perm_const): ...this new function. Take the selector
11795 as a vec_perm_indices rather than an rtx. Also take the mode of
11796 the selector. Update call to shift_amt_for_vec_perm_mask.
11797 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11798 Use vec_perm_indices::new_expanded_vector to expand the original
11799 selector into bytes. Check whether the indices fit in the vector
11800 mode before using a variable permute.
11801 (expand_vec_perm_var): Make global.
11802 (expand_mult_highpart): Use expand_vec_perm_const.
11803 * fold-const.c: Includes vec-perm-indices.h.
11804 * tree-ssa-forwprop.c: Likewise.
11805 * tree-vect-data-refs.c: Likewise.
11806 * tree-vect-generic.c: Likewise.
11807 * tree-vect-loop.c: Likewise.
11808 * tree-vect-slp.c: Likewise.
11809 * tree-vect-stmts.c: Likewise.
11810 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11812 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11813 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11814 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11815 (aarch64_vectorize_vec_perm_const): ...this new function.
11816 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11817 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11818 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11819 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11820 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11821 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11822 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11824 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11825 check for NEON modes.
11826 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11827 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11828 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11829 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11831 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11832 the old VEC_PERM_CONST conditions.
11833 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11834 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11835 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11836 (ia64_vectorize_vec_perm_const_ok): Merge into...
11837 (ia64_vectorize_vec_perm_const): ...this new function.
11838 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11839 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11840 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11841 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11842 * config/mips/mips.c (mips_expand_vec_perm_const)
11843 (mips_vectorize_vec_perm_const_ok): Merge into...
11844 (mips_vectorize_vec_perm_const): ...this new function.
11845 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11846 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11847 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11848 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11849 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11850 (rs6000_expand_vec_perm_const): Delete.
11851 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11853 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11854 (altivec_expand_vec_perm_const_le): Take each operand individually.
11855 Operate on constant selectors rather than rtxes.
11856 (altivec_expand_vec_perm_const): Likewise. Update call to
11857 altivec_expand_vec_perm_const_le.
11858 (rs6000_expand_vec_perm_const): Delete.
11859 (rs6000_vectorize_vec_perm_const_ok): Delete.
11860 (rs6000_vectorize_vec_perm_const): New function.
11861 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11862 an element count and rtx array.
11863 (rs6000_expand_extract_even): Update call accordingly.
11864 (rs6000_expand_interleave): Likewise.
11865 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11866 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11867 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11868 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11869 (rs6000_expand_vec_perm_const): Delete.
11870 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11871 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11872 (altivec_expand_vec_perm_const_le): Take each operand individually.
11873 Operate on constant selectors rather than rtxes.
11874 (altivec_expand_vec_perm_const): Likewise. Update call to
11875 altivec_expand_vec_perm_const_le.
11876 (rs6000_expand_vec_perm_const): Delete.
11877 (rs6000_vectorize_vec_perm_const_ok): Delete.
11878 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11879 reference to the SPE evmerge intructions.
11880 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11881 an element count and rtx array.
11882 (rs6000_expand_extract_even): Update call accordingly.
11883 (rs6000_expand_interleave): Likewise.
11884 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11885 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11887 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11889 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11891 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11892 vector mode and that that mode matches the mode of the data
11894 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11895 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11896 directly using expand_vec_perm_1 when forcing selectors into
11898 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11900 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11902 * optabs-query.h (can_vec_perm_p): Delete.
11903 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11904 * optabs-query.c (can_vec_perm_p): Split into...
11905 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11906 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11907 particular selector is valid.
11908 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11909 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11910 (vect_grouped_load_supported): Likewise.
11911 (vect_shift_permute_load_chain): Likewise.
11912 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11913 (vect_transform_slp_perm_load): Likewise.
11914 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11915 (vectorizable_bswap): Likewise.
11916 (vect_gen_perm_mask_checked): Likewise.
11917 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11918 implementations of variable permutation vectors into account
11919 when deciding which selector to use.
11920 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11921 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11922 with a false third argument.
11923 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11924 to test whether the constant selector is valid and can_vec_perm_var_p
11925 to test whether a variable selector is valid.
11927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11929 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11930 * optabs-query.c (can_vec_perm_p): Likewise.
11931 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11932 instead of vec_perm_indices.
11933 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11934 (vect_gen_perm_mask_checked): Likewise,
11935 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11936 (vect_gen_perm_mask_checked): Likewise,
11938 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11940 * optabs-query.h (qimode_for_vec_perm): Declare.
11941 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11942 (qimode_for_vec_perm): ...this new function.
11943 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11945 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11947 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11948 does not have a conditional at the top.
11950 2018-01-02 Richard Biener <rguenther@suse.de>
11952 * ipa-inline.c (big_speedup_p): Fix expression.
11954 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11957 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11960 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11964 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11965 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11966 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11967 cond_taken_branch_cost 3->4.
11969 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11971 PR tree-optimization/83581
11972 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11973 TODO_cleanup_cfg if any changes have been made.
11975 PR middle-end/83608
11976 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11977 convert_modes if target mode has the right side, but different mode
11980 PR middle-end/83609
11981 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11982 last argument when extracting from CONCAT. If either from_real or
11983 from_imag is NULL, use expansion through memory. If result is not
11984 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11985 the parts directly to inner mode, if even that fails, use expansion
11988 PR middle-end/83623
11989 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11990 check for bswap in mode rather than HImode and use that in expand_unop
11993 Copyright (C) 2018 Free Software Foundation, Inc.
11995 Copying and distribution of this file, with or without modification,
11996 are permitted in any medium without royalty provided the copyright
11997 notice and this notice are preserved.