1 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
5 _mm512_mask_mullox_epi64): New intrinsics.
7 2018-05-03 Tom de Vries <tom@codesourcery.com>
10 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
11 dump files): Add offload-tree.
13 2018-05-03 Richard Biener <rguenther@suse.de>
15 PR tree-optimization/85615
16 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
17 to loops not nested in BBs loop father to avoid creating multi-entry
20 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22 PR tree-optimization/70291
23 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
24 arguments. Change return type to tree. Emit libcall as a new
25 statement rather than replacing existing one when inplace_p is true.
26 (expand_complex_multiplication_components): New function.
27 (expand_complex_multiplication): Expand floating-point complex
28 multiplication using the above.
29 (expand_complex_division): Rename inner_type parameter to type.
30 Update expand_complex_libcall call-site.
31 (expand_complex_operations_1): Update expand_complex_multiplication
32 and expand_complex_division call-sites.
34 2018-05-02 Jakub Jelinek <jakub@redhat.com>
37 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
38 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
39 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
40 the highest significant bit of the shift count mask is clear. In
41 check whether and[sq]i3 is needed verify that all significant bits
42 of the shift count other than the highest are set.
44 2018-05-02 Tom de Vries <tom@codesourcery.com>
47 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
48 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
49 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
50 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
51 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
52 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
53 __builtin_goacc_parlevel_size.
55 2018-05-02 Richard Biener <rguenther@suse.de>
57 PR tree-optimization/85597
58 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
59 do not use split vect_get_vec_defs call but call vect_get_slp_defs
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 ltrans-tree.
68 2018-05-02 Tom de Vries <tom@codesourcery.com>
71 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
72 dump files): Add wpa-ipa.
74 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
76 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
77 powerpc*-*-linux*paired* target.
78 * config/rs6000/750cl.h: Delete.
79 * config/rs6000/paired.h: Delete.
80 * config/rs6000/paired.md: Delete.
81 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
83 * config/rs6000/rs6000-builtin.def: Remove paired float support.
84 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
85 comment. Remove paired float support.
86 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
87 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
89 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
90 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
92 * config/rs6000/rs6000.c: Remove paired float support.
93 (paired_expand_vector_init, paired_expand_vector_move,
94 paired_emit_vector_compare, paired_emit_vector_cond_expr,
95 (paired_expand_lv_builtin, paired_expand_stv_builtin,
96 paired_expand_builtin, paired_expand_predicate_builtin,
97 paired_init_builtins): Delete.
98 * config/rs6000/rs6000.h: Remove paired float support.
99 * config/rs6000/rs6000.md: Remove paired float support.
100 (move_from_CR_ov_bit): Delete.
101 * config/rs6000/rs6000.opt (mpaired): Delete.
102 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
103 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
105 2018-05-02 Richard Biener <rguenther@suse.de>
108 * gimplify.c (gimplify_save_expr): When in SSA form allow
109 SAVE_EXPRs to compute to SSA vars.
111 2018-05-02 Jakub Jelinek <jakub@redhat.com>
114 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
115 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
116 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
117 clobber operands[2], instead use a new pseudo. Formatting fixes.
119 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
121 PR tree-optimization/85586
122 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
123 exit early for statements in the same group if the accesses are
126 2018-05-02 Tom de Vries <tom@codesourcery.com>
129 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
132 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
134 PR tree-optimization/85143
135 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
137 2018-05-01 Tom de Vries <tom@codesourcery.com>
140 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
141 not found" error message.
143 2018-05-01 Tom de Vries <tom@codesourcery.com>
146 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
147 * vec.c (test_ordered_remove_if): New function.
148 (vec_c_tests): Call test_ordered_remove_if.
149 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
150 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
151 * tree-vect-patterns.c (vect_pattern_recog_1): Use
152 VEC_ORDERED_REMOVE_IF.
154 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
156 PR tree-optimization/82665
157 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
158 pointer subtraction where arguments come from a memchr call.
160 2018-05-01 Jakub Jelinek <jakub@redhat.com>
162 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
163 --push-state --as-needed and --pop-state instead of --as-needed and
164 --no-as-needed if ld supports it.
165 * configure: Regenerated.
168 * doc/install.texi2html: Replace _002d with - and _002a with * in
169 generated html files using sed.
171 2018-04-30 David Malcolm <dmalcolm@redhat.com>
174 * gcc-rich-location.c (blank_line_before_p): New function.
175 (use_new_line): New function.
176 (gcc_rich_location::add_fixit_insert_formatted): New function.
177 * gcc-rich-location.h
178 (gcc_rich_location::add_fixit_insert_formatted): New function.
180 2018-04-30 David Malcolm <dmalcolm@redhat.com>
182 * selftest.c (assert_streq): Rename "expected" and "actual" to
183 "val1" and "val2". Extend NULL-handling to cover both inputs
184 symmetrically, while still requiring both to be non-NULL for a pass.
185 * selftest.h (assert_streq): Rename "expected" and "actual" to
187 (ASSERT_EQ): Likewise.
188 (ASSERT_EQ_AT): Likewise.
189 (ASSERT_KNOWN_EQ): Likewise.
190 (ASSERT_KNOWN_EQ_AT): Likewise.
191 (ASSERT_NE): Likewise.
192 (ASSERT_MAYBE_NE): Likewise.
193 (ASSERT_MAYBE_NE_AT): Likewise.
194 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
195 the assertion to pass.
196 (ASSERT_STREQ_AT): Likewise.
198 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
200 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
201 interaction with -pie.
203 2018-04-30 David Malcolm <dmalcolm@redhat.com>
205 * selftest.h: Fix alphabetization of per-source-file selftest
208 2018-04-30 Jason Merrill <jason@redhat.com>
210 PR c++/61982 - dead stores to destroyed objects.
211 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
214 2018-04-30 Jason Merrill <jason@redhat.com>
216 * tree.c (build_clobber): New.
217 * tree.h: Declare it.
218 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
220 2018-04-30 David Malcolm <dmalcolm@redhat.com>
222 * diagnostic-show-locus.c (layout::layout): Update for
223 location_get_source_line returning a char_span.
224 (struct char_span): Move to input.h.
225 (struct correction): Update for fields in char_span becoming
227 (struct source_line): Update for location_get_source_line
228 returning a char_span.
229 (layout::print_line): Likewise.
230 * edit-context.c (edited_file::print_content): Likewise.
231 (edited_file::print_diff_hunk): Likewise.
232 (edited_file::print_run_of_changed_lines): Likewise.
233 (edited_file::get_num_lines): Likewise.
234 (edited_line::edited_line): Likewise.
235 * final.c (asm_show_source): Likewise.
236 * input.c (location_get_source_line): Convert return type
237 from const char * to char_span, losing the final "line_len"
239 (dump_location_info): Update for the above.
240 (get_substring_ranges_for_loc): Likewise. Use a char_span
241 when handling the literal within the line.
242 (test_reading_source_line): Update for location_get_source_line
243 returning a char_span.
244 * input.h (class char_span): Move here from
245 diagnostic-show-locus.c, converting from a struct to a class.
246 Make data members private.
247 (char_span::operator bool): New.
248 (char_span::length): New.
249 (char_span::get_buffer): New.
250 (char_span::operator[]): New.
251 (char_span::subspan): Make const.
252 (char_span::xstrdup): New.
253 (location_get_source_line): Convert return type from const char *
254 to char_span, losing the final "line_size" param.
256 2018-04-30 Jan Hubicka <jh@suse.cz>
258 * lto-wrapper.c (ltrans_priorities): New static var.
260 (run_gcc): Read priorities and if doing parallel build order
261 the Makefile by them.
263 2018-04-30 David Malcolm <dmalcolm@redhat.com>
265 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
267 2018-04-30 Richard Biener <rguenther@suse.de>
269 * tree-cfg.c (verify_address): Remove base argument, add
270 flag whether to check TREE_ADDRESSABLE and do that.
271 (verify_expr): Remove.
272 (verify_types_in_gimple_reference): Add pieces from verify_expr.
273 (verify_gimple_assign_single): Likewise.
274 (verify_gimple_switch): Likewise.
275 (verify_expr_location_1): Dereference tp once. Add (disabled)
276 piece from verify_expr.
277 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
279 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
281 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
283 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
285 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
286 (small_data_pattern): Likewise.
287 (arc_rewrite_small_data): Likewise.
288 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
289 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
290 (get_symbol_alignment): New function.
291 (legitimate_small_data_address_p): Likewise.
292 (legitimate_scaled_address): Update, call
293 legitimate_small_data_address_p.
294 (output_sdata): New static variable.
295 (arc_print_operand): Update how we handle small data operands.
296 (arc_print_operand_address): Likewise.
297 (arc_legitimate_address_p): Update, use
298 legitimate_small_data_address_p.
299 (arc_rewrite_small_data_p): Remove.
300 (arc_rewrite_small_data_1): Likewise.
301 (arc_rewrite_small_data): Likewise.
302 (small_data_pattern): Likewise.
303 (compact_sda_memory_operand): Update to use
304 legitimate_small_data_address_p and get_symbol_alignment.
305 (prepare_move_operands): Don't rewite sdata pattern.
306 (prepare_extend_operands): Remove.
307 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
309 (zero_extendqisi2): Likewise.
310 (zero_extendhisi2): Likewise.
311 (extendqihi2): Likewise.
312 (extendqisi2): Likewise.
313 (extendhisi2): Likewise.
317 * config/arc/constraints.md (Usd): Change it to memory constraint.
319 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
321 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
322 as source of std instructions.
323 * config/arc/arc.md (movsi_insn): Update pattern predicate to
324 allow 6-bit constants as source for store instructions.
325 (movdi_insn): Update instruction pattern to allow 6-bit constants
326 as source for store instructions.
328 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
330 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
332 2018-04-30 Nathan Sidwell <nathan@acm.org>
333 Sandra Loosemore <sandra@codesourcery.com>
335 * dumpfile.c (dump_open): Allow '-' for stdout.
336 * doc/invoke.texi (Developer Options): Document dump filename
337 determination early. Document stdin/stdout selection.
339 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
341 Microblaze Target: PIC data text relative
343 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
344 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
346 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
347 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
348 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
349 New addressing mode for data-text relative position indepenedent code.
350 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
351 'ADDRESS_SYMBOLIC_TXT_REL'.
352 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
353 (microblaze_legitimate_pic_operand): Exclude function calls from
354 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
355 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
357 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
358 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
359 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
360 for 'address + offset'.
361 (microblaze_expand_prologue): Add new function prologue call for
363 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
364 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
365 table in case of TARGET_PIC_DATA_TEXT_REL.
366 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
367 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
368 Add new macros 'UNSPEC_TEXT',
369 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
370 + exclude function calls from 'UNSPEC_PLT' in case of data text
372 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
373 new target hook for generating address diff vector tables in case of
375 * doc/tm.texi : Regenerate.
376 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
377 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
378 of addr diff vector generation.
379 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
380 target hook definition.
381 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
382 Add default function for generate_pic_addr_diff_vec -> flag_pic.
383 * doc/invoke.texi (Add new pic option): Add new microblaze pic
384 option for data text relative.
386 2018-04-30 Richard Biener <rguenther@suse.de>
388 * tree-chrec.h (evolution_function_is_constant_p): Remove
390 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
392 2018-04-30 Richard Biener <rguenther@suse.de>
395 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
397 2018-04-30 Richard Biener <rguenther@suse.de>
399 PR tree-optimization/28364
400 PR tree-optimization/85275
401 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
402 copying first exit test.
404 2018-04-28 Mark Wielaard <mark@klomp.org>
406 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
408 (dwarf_AT): Handle DW_AT_addr_base.
409 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
411 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
414 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
415 (*ashl<dwi>3_doubleword_mask_1): Ditto.
416 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
417 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
419 2018-04-28 Richard Biener <rguenther@suse.de>
421 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
422 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
423 to reflect use. Only add interesting stmts.
425 2018-04-27 Martin Jambor <mjambor@suse.cz>
428 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
429 the jump function allows for passing through aggregate values.
431 2018-04-27 David Malcolm <dmalcolm@redhat.com>
433 * input.h (in_system_header_at): Convert from macro to inline
435 (from_macro_expansion_at): Likewise.
436 (from_macro_definition_at): Likewise.
438 2018-04-27 Jeff Law <law@redhat.com>
440 * config.gcc: Mark tile* targets as deprecated/obsolete.
442 2018-04-27 Richard Biener <rguenther@suse.de>
444 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
447 2018-04-27 Richard Biener <rguenther@suse.de>
449 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
451 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
453 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
454 with Yd constraint. Set "preferred_for_speed" attribute from
455 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
457 (*movdi_internal): Ditto.
458 (movti_interunit splitters): Remove
459 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
460 (movdi_interunit splitters): Ditto.
461 * config/i386/constraints.md (Ye): Remove.
462 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
464 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
467 * config/aarch64/constraints.md (Usg): Limit to 31.
470 2018-04-27 Jakub Jelinek <jakub@redhat.com>
472 PR tree-optimization/85529
473 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
474 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
475 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
476 zero extension or masking of the MSB bit.
477 (optimize_range_tests): Add FIRST_BB argument, pass it through
478 to optimize_range_tests_var_bound.
479 (maybe_optimize_range_tests, reassociate_bb): Adjust
480 optimize_range_tests callers.
482 2018-04-26 Richard Biener <rguenther@suse.de>
483 Jakub Jelinek <jakub@redhat.com>
485 * cgraph.h (symbol_table): Just declare debug method here.
486 * symtab.c (symbol_table::debug): Define.
488 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
490 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
492 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
494 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
495 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
496 (*movdi_internal): Substitute Yi and Yj constraint with x
497 and Ym and Yn constraint with y constraint. Update "isa"
498 attribute and set "preferred_for_speed" attribute from
499 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
500 (*movsi_internal): Ditto.
501 (*movdf_internal): Ditto.
502 (*movsf_internal): Ditto.
503 (*zero_extendsidi2): Ditto.
504 * config/i386/sse.md (vec_set<mode>_0): Ditto.
505 (sse2_loadld): Ditto.
506 (*vec_extract<ssevecmodelower>_0): Ditto.
507 (*vec_extractv4si_0_zext_sse4): Ditto.
508 (vec_concatv2di): Ditto.
509 (*vec_dup<mode>): Ditto.
510 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
511 * config/i386/constraints.md (Yi): Remove.
516 2018-04-26 Nathan Sidwell <nathan@acm.org>
518 * dumpfile.c (dump_open): New.
519 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
520 (dump_finish): Detect stdio/stderr by value not name.
522 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
524 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
526 2018-04-26 Tom de Vries <tom@codesourcery.com>
529 * config/nvptx/nvptx.c (verify_neutering_jumps)
530 (verify_neutering_labels): New function
531 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
533 2018-04-26 Tom de Vries <tom@codesourcery.com>
536 * config/nvptx/nvptx.c (needs_neutering_p): New function.
537 (nvptx_single): Use needs_neutering_p to skip over insns that do not
540 2018-04-26 Richard Biener <rguenther@suse.de>
541 Tom de Vries <tom@codesourcery.com>
544 * lto-streamer-out.c (output_function): Fixup loops if required to match
545 discovery done in the reader.
547 2018-04-26 Richard Biener <rguenther@suse.de>
549 PR tree-optimization/85116
550 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
551 have a loop exit from the single latch predecessor. Remove
552 case of header with just condition.
553 (ch_base::copy_headers): Exclude infinite loops from any
555 (pass_ch::execute): Record exits.
557 2018-04-26 Richard Biener <rguenther@suse.de>
559 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
560 prologue cost vector and pass it to vect_get_load_cost.
561 (vect_get_peeling_costs_all_drs): Likewise.
562 (vect_peeling_hash_get_lowest_cost): Likewise.
563 (vect_enhance_data_refs_alignment): Likewise.
565 2018-04-26 Richard Biener <rguenther@suse.de>
568 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
569 checking of integer<->pointer conversions.
570 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
571 sign-/zero-extending pointer types.
572 (expand_omp_for_static_chunk): Likewise.
574 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
575 Jean Lee <xiaoyur347@gmail.com>
577 * config/mips/mips.c (mips_asan_shadow_offset): New function.
578 (TARGET_ASAN_SHADOW_OFFSET): Define.
579 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
580 true for -fsanitize=address.
582 2018-04-25 Mark Wielaard <mark@klomp.org>
584 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
587 2018-04-25 Jakub Jelinek <jakub@redhat.com>
589 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
590 than "alu", remove explicit "memory" and "imm_disp" attributes.
591 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
594 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
595 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
598 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
601 * config/i386/i386.c (ix86_expand_builtin): Change memory
602 operand to XI, extend p0 to Pmode.
603 * config/i386/i386.md: Change unspec volatile and operand
604 1 mode to XI, change operand 0 mode to P.
606 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
608 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
609 GET_MODE_MASK before any checking.
610 (nds32_can_use_bset_p): Likewise.
611 (nds32_can_use_btgl_p): Likewise.
613 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
615 * config/nds32/nds32-doubleword.md: New define_split pattern for
616 illegal register number.
618 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
620 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
622 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
624 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
626 2018-04-25 Richard Biener <rguenther@suse.de>
628 * lto-streamer.h (LTO_major_version): Bump to 8.
630 2018-04-25 Jakub Jelinek <jakub@redhat.com>
632 * BASE-VER: Set to 9.0.0.
634 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
636 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
637 in __abskf2 and __powikf2.
639 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
642 * config/aarch64/constraints.md (Usg, Usj): New constraints.
643 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
644 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
645 Use the above on operand 2. Reindent.
646 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
648 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
651 * common/config/i386/i386-common.c (ix86_handle_option): Don't
653 * config/i386/i386.opt (mcet): Removed.
654 * doc/install.texi: Remove -mcet documentation.
655 * doc/invoke.texi: Likewise.
657 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
660 * doc/install.texi: Remove -mcet from bootstrap-cet.
662 2018-04-24 Jakub Jelinek <jakub@redhat.com>
665 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
666 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
670 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
671 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
672 containing a CONST_VECTOR.
674 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
676 * doc/install.texi: Update newlib dependency for nvptx.
678 2018-04-24 Jakub Jelinek <jakub@redhat.com>
681 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
682 instead of INTVAL when shifting x left.
684 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
686 PR tree-optimization/85478
687 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
688 vect_grouped_store_supported for single element vectors.
690 2018-04-24 Richard Biener <rguenther@suse.de>
693 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
694 load cost increase to the case of non-constant step.
696 2018-04-24 Jakub Jelinek <jakub@redhat.com>
699 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
700 destination if any_malformed_asm.
702 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
705 * expr.c (store_field): In the bitfield case, if the value comes from
706 a function call and is returned in registers by means of a PARALLEL,
707 do not change the mode of the temporary unless BLKmode and VOIDmode.
709 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
711 PR rtl-optimization/85423
712 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
713 dependencies to debug insns when the previous insn is non-debug.
715 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
717 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
718 enums into a single definition.
719 (fls): Fix predicates and printing.
722 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
724 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
725 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
726 and short u6 immediate.
727 (check_if_valid_sleep_operand): Remove.
728 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
730 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
732 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
733 flag_always_save_lp condition.
734 * config/nds32/nds32.opt (malways-save-lp): New option.
736 2018-04-22 Shiva Chen <shiva0217@gmail.com>
738 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
739 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
740 * config/nds32/nds32.h
741 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
742 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
744 2018-04-22 Shiva Chen <shiva0217@gmail.com>
746 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
747 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
749 2018-04-22 Shiva Chen <shiva0217@gmail.com>
750 Chung-Ju Wu <jasonwucj@gmail.com>
752 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
754 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
755 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
757 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
759 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
761 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
763 * config/nds32/nds32-protos.h (nds32_data_alignment,
764 nds32_local_alignment): Declare.
765 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
766 nds32_local_alignment): New functions.
767 (TARGET_CONSTANT_ALIGNMENT): Define.
768 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
770 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
772 * config/nds32/nds32.c
773 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
774 (TARGET_MODES_TIEABLE_P): Likewise.
776 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
778 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
781 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
782 Chung-Ju Wu <jasonwucj@gmail.com>
784 * config/nds32/constants.md (unspec_volatile_element): Add enum values
785 for unaligned access.
786 * config/nds32/nds32-intrinsic.c: Implementation of expanding
788 * config/nds32/nds32-intrinsic.md: Likewise.
789 * config/nds32/nds32_intrinsic.h: Likewise.
790 * config/nds32/nds32.h (nds32_builtins): Likewise.
791 * config/nds32/nds32.opt (munaligned-access): New option.
792 * config/nds32/nds32.c (nds32_asm_file_start): Display
793 flag_unaligned_access status.
795 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
797 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
798 -mno-relax is present.
799 * config/riscv/linux.h (LINK_SPEC): Ditto.
801 2018-04-20 Martin Sebor <msebor@redhat.com>
804 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
806 (gimple_fold_builtin_stxcpy_chk): Same.
807 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
809 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
812 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
813 __powikf2 when long double is IEEE 128-bit.
815 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
817 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
818 step to make sure stack always aligned.
820 2018-04-20 Carl Love <cel@us.ibm.com>
823 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
826 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
827 Tom de Vries <tom@codesourcery.com>
830 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
831 Emit insns for calls too.
832 (nvptx_find_par): Always look for worker-level predecessor insn.
833 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
835 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
836 (nvptx_process_pars): Propagate frames for calls.
838 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
841 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
843 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
844 (ix86_handle_option): Don't handle OPT_mibt.
845 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
847 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
849 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
850 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
851 (ix86_target_macros): Define __CET__ with flag_cf_protection
853 * config/i386/i386.c (isa2_opts): Remove -mibt.
854 * config/i386/i386.h (TARGET_IBT): Removed.
855 (TARGET_IBT_P): Likewise.
856 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
857 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
858 * config/i386/i386.opt (mcet): Update help message.
861 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
862 -mcet as an alias for -mshstk.
864 2018-04-20 Richard Biener <rguenther@suse.de>
867 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
868 complexity by forcing a single use of the multiply operand.
870 2018-04-20 Martin Jambor <mjambor@suse.cz>
873 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
874 recursion dependency to only apply to non-clones.
876 2018-04-20 Martin Jambor <mjambor@suse.cz>
879 * ipa-cp.c (create_specialized_node): Check that clones of
880 self-recursive edges exist during IPA-CP.
882 2018-04-19 Toon Moene <toon@moene.org>
884 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
887 2018-04-19 Jakub Jelinek <jakub@redhat.com>
889 PR tree-optimization/85467
890 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
891 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
892 VECTOR_CST element to type.
894 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
897 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
898 * config/i386/i386.md (builtin_setjmp_setup): Removed.
899 (builtin_longjmp): Likewise.
900 (save_stack_nonlocal): New pattern.
901 (restore_stack_nonlocal): Likewise.
903 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
906 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
907 Replace ASM_OUTPUT_LABEL with fprintf.
909 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
912 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
913 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
914 * config/i386/i386-c.c (ix86_target_macros_internal): Also
915 define __IBT__ and __SHSTK__ for -fcf-protection.
916 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
918 (ix86_trampoline_init): Likewise.
919 (x86_output_mi_thunk): Likewise.
920 (ix86_notrack_prefixed_insn_p): Likewise.
921 (ix86_option_override_internal): Don't disallow -fcf-protection.
922 * config/i386/i386.md (rdssp<mode>): Also enable for
924 (incssp<mode>): Likewise.
925 (nop_endbr): Likewise.
926 * config/i386/i386.opt (mcet): Change help message to built-in
930 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
931 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
932 enable CET built-in functions.
934 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
936 * common/config/i386/i386-common.c
937 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
938 OPTION_MASK_ISA_MOVDIRI_UNSET,
939 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
940 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
941 * config.gcc (movdirintrin.h): New header.
942 * config/i386/cpuid.h (bit_MOVDIRI,
943 bit_MOVDIR64B): New bits.
944 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
946 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
947 (VOID, PVOID, PCVOID)): New function types.
948 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
949 __builtin_ia32_directstoreu_u64,
950 __builtin_ia32_movdir64b): New builtins.
951 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
952 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
954 (ix86_valid_target_attribute_inner_p): Ditto.
955 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
956 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
957 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
958 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
959 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
960 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
961 (movdiri<mode>, movdir64b_<mode>): New.
962 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
963 * config/i386/immintrin.h: Include movdirintrin.h.
964 * config/i386/movdirintrin.h: New file.
965 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
967 2018-04-19 Richard Biener <rguenther@suse.de>
970 * cfg.c (clear_bb_flags): When loop state says we have
971 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
973 2018-04-19 Richard Biener <rguenther@suse.de>
975 PR tree-optimization/84737
976 * tree-vect-data-refs.c (vect_copy_ref_info): New function
977 copying restrict info.
978 (vect_setup_realignment): Use it.
979 * tree-vectorizer.h (vect_copy_ref_info): Declare.
980 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
981 the first DR to all generated stores.
982 (vectorizable_load): Likewise for loads.
984 2018-04-19 Jakub Jelinek <jakub@redhat.com>
986 PR tree-optimization/85446
987 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
988 the integral and pointer types to have the same precision.
990 * doc/install.texi: Document --disable-cet being the default and
993 2018-04-18 Martin Liska <mliska@suse.cz>
995 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
998 2018-04-18 Martin Liska <mliska@suse.cz>
1001 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
1004 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
1005 arguments if they are comparable.
1007 2018-04-18 Martin Liska <mliska@suse.cz>
1010 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1013 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1016 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
1019 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
1020 ENDBR after calling __morestack.
1022 2018-04-18 David Malcolm <dmalcolm@redhat.com>
1025 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
1026 by using gcc_base_ver to generate a gcc_driver_version, and use
1027 it when generating GCC_DRIVER_NAME.
1028 * configure: Regenerate.
1030 2018-04-18 Jakub Jelinek <jakub@redhat.com>
1033 * config.gcc: Obsolete powerpc*-*-*spe*.
1035 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1038 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
1039 (stabstr_D): Change type of unum from unsigned int to
1040 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
1043 2018-04-17 Jim Wilson <jimw@sifive.com>
1046 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
1047 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
1048 Set arg_pointer_offset after using pretend_args_size.
1050 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1052 PR rtl-optimization/85431
1053 * dse.c (record_store): Ignore zero width stores.
1056 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
1057 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
1058 __builtin_stack_restore rather than after it.
1059 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
1060 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
1061 argument instead of virtual_dynamic_stack_rtx.
1063 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
1065 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
1067 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
1068 Add note to error message to explain internal mapping of overloaded
1069 built-in function name to non-overloaded built-in function name.
1070 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
1073 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
1076 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
1077 where the inputs overlap with the output.
1079 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1081 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
1082 (=v, v) alternative and explicit "memory" attribute.
1083 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
1084 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1086 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
1087 "sselog1" type instead of "sselog".
1088 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
1089 "sselog". Remove explicit "memory" attribute.
1090 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
1091 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1093 (vec_extract_hi_v32hi): Merge all alternatives into one, use
1094 "sselog1" type instead of "sselog". Remove explicit "memory"
1096 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
1097 use "sselog1" type instead of "sselog". Remove explicit "memory"
1099 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
1100 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
1102 (vec_extract_hi_v64qi): Merge all alternatives into one, use
1103 "sselog1" type instead of "sselog". Remove explicit "memory"
1105 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
1106 use "sselog1" type instead of "sselog". Remove explicit "memory"
1110 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
1113 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
1116 2018-04-17 Martin Jambor <mjambor@suse.cz>
1119 * ipa-cp.c (create_specialized_node): Call
1120 expand_all_artificial_thunks if necessary.
1122 2018-04-17 Martin Liska <mliska@suse.cz>
1125 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
1126 in message, remote space in between '_G' and '('.
1128 2018-04-17 Jakub Jelinek <jakub@redhat.com>
1131 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
1132 avx512f_vmcmp<mode>3<round_saeonly_name>,
1133 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
1134 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
1135 avx512f_rndscale<mode><round_saeonly_name>,
1136 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
1137 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
1138 Use %<iptr>2 instead of %2 for -masm=intel.
1139 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
1140 avx512f_vcvttss2usi<round_saeonly_name>,
1141 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
1143 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
1144 avx512f_vcvttsd2usi<round_saeonly_name>,
1145 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
1146 Use %q1 instead of %1 for -masm=intel.
1147 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
1148 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
1149 of %3 for -masm=intel.
1150 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
1152 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
1154 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
1156 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
1157 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
1159 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
1161 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
1162 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
1163 %g1 and one with %0 and %1.
1164 (avx512er_vmrcp28<mode><round_saeonly_name>,
1165 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
1167 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
1168 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
1169 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
1170 of %0 and %{%4%} for -masm=intel.
1171 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
1172 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
1173 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
1174 order of %0 and %{%5%}%{z%} for -masm=intel.
1176 2018-04-17 Jan Hubicka <jh@suse.cz>
1179 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
1181 2018-04-17 Martin Liska <mliska@suse.cz>
1184 * multiple_target.c (create_dispatcher_calls): Set apostrophes
1185 for target_clone error message. Make default implementation
1186 clone to be a local declaration.
1187 (separate_attrs): Add new argument and check for an empty
1189 (expand_target_clones): Handle it.
1190 (ipa_target_clone): Make redirection just for target_clones
1193 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
1194 Tom de Vries <tom@codesourcery.com>
1197 * omp-expand.c (expand_oacc_for): Add dummy false branch for
1198 tiled basic blocks without omp continue statements.
1200 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
1203 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
1204 vec_extract expression as having side effects to make sure it gets
1207 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1210 * config/i386/i386.c (get_builtin_code_for_version): Check
1213 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1216 * gcc/config.gcc: Support "skylake".
1217 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1219 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1220 (processor_target_table): Add "skylake".
1221 (ix86_option_override_internal): Add "skylake".
1222 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1223 PROCESSOR_CANNONLAKE.
1224 (get_builtin_code_for_version): Fix priority for
1225 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1226 PROCESSOR_SKYLAKE-AVX512.
1227 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1228 (processor_type): Add PROCESSOR_SKYLAKE.
1230 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1231 Jason Merrill <jason@redhat.com>
1234 * convert.c (convert_to_integer_1): Use direct recursion for
1235 enumeral types and types with a precision less than the number
1236 of bits in their mode.
1238 2018-04-16 Julia Koval <julia.koval@intel.com>
1241 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1242 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1244 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1247 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1248 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1249 and -mno-direct-move.
1251 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1254 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1255 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1256 Ensure negative shifts result in {0}.
1258 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1260 PR rtl-optimization/79916
1261 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1262 regs (if any) to define how to gnerate SD moves when LRA is in
1265 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1267 PR rtl-optimization/85393
1268 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1269 * except.c (expand_dw2_landing_pad_for_region): Make static.
1270 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1271 a label and unconditional jump to old_bb, rather than
1272 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1275 PR rtl-optimization/85376
1276 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1277 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1278 instead of a specific value.
1280 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1281 Bin Cheng <bin.cheng@arm.com>
1283 PR tree-optimization/82965
1284 PR tree-optimization/83991
1285 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1286 by_profile_only parameter.
1287 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1288 information if the loop was predicted to iterate too many times.
1289 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1291 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1294 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1297 2018-04-13 Martin Liska <mliska@suse.cz>
1298 Jakub Jelinek <jakub@redhat.com>
1301 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1302 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1303 * builtins.c (expand_builtin_memory_copy_args): Use
1304 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1305 handle dest_addr == pc_rtx.
1307 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1310 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1311 asked to not generate direct moves.
1312 (fix_trunc<mode>si2_stfiwx): Similar.
1313 (fix_trunc<mode>si2_internal): Similar.
1315 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1318 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1319 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1320 lookup if dest in some wider mode is known to be const0_rtx and
1321 if so, record permanent equivalence for it to be ZERO_EXTEND of
1322 the narrower mode destination.
1324 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1326 * lto-streamer-out.c (output_function): Revert 259346.
1327 * omp-expand.c (expand_oacc_for): Likewise.
1329 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1331 PR rtl-optimization/85354
1332 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1333 * sel-sched.c (sel_global_init): ... here.
1335 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1338 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1339 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1340 mode for PE-COFF targets.
1341 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1342 (i386_pe_asm_lto_end): Likewise.
1343 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1344 (TARGET_ASM_LTO_END): Likewise.
1345 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1346 (i386_pe_asm_lto_start): New function.
1347 (i386_pe_asm_lto_end): Likewise.
1349 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1350 Richard Biener <rguenther@suse.de>
1353 * lto-streamer-out.c (output_function): Fix CFG loop state before
1355 * omp-expand.c (expand_oacc_for): Handle calls to internal
1356 functions like regular functions.
1358 2018-04-12 Richard Biener <rguenther@suse.de>
1361 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1362 for the early LTO debug to properly generate references to it
1363 during DIE emission. Do not re-use that for the skeleton for
1365 (dwarf2out_early_finish): Likewise.
1367 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1370 * config/i386/sse.md
1371 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1372 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1373 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1374 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1375 and output is a reg, avoid creating invalid lowpart subreg, but
1376 instead split into a 512-bit move. Don't split if not AVX512VL,
1377 input is xmm16+ reg and output is a mem.
1378 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1379 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1380 xmm16+ reg and output is a mem.
1382 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1384 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1385 also for flag_dwarf2_cfi_asm.
1387 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1389 PR rtl-optimization/85342
1390 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1391 a bool scalar var inside of the loop instead. Don't try to update
1392 recog_data.operand after failed apply_change_group.
1394 2018-04-12 Tom de Vries <tom@codesourcery.com>
1397 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1398 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1399 array with flexible array member as array without given dimension.
1400 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1401 argument for undefined param to true.
1403 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1406 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1407 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1408 from PowerPC section.
1409 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1410 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1411 help text that is too long.
1412 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1413 help text that is too long.
1414 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1415 help text that is too long.
1417 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1419 * config/alpha/alpha.md (stack_probe_internal): Rename
1420 from "probe_stack". Update all callers.
1422 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1424 PR rtl-optimization/84566
1425 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1426 sched_macro_fuse_insns.
1428 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1431 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1432 (compute_block_dependences): ... from here.
1434 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1436 PR tree-optimization/85331
1437 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1438 from int to HOST_WIDE_INT.
1440 2018-04-11 Martin Jambor <mjambor@suse.cz>
1443 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1444 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1445 not the same as the source val.
1446 (cgraph_edge_brings_value_p): New parameter.
1447 (gather_edges_for_value): Pass destination value to
1448 cgraph_edge_brings_value_p.
1449 (perhaps_add_new_callers): Likewise.
1450 (get_info_about_necessary_edges): Likewise and exclude values brought
1451 only by self-recursive edges.
1452 (create_specialized_node): Redirect only clones of self-calling edges.
1453 (+self_recursive_pass_through_p): New function.
1454 (find_more_scalar_values_for_callers_subset): Use it.
1455 (find_aggregate_values_for_callers_subset): Likewise.
1456 (known_aggs_to_agg_replacement_list): Removed.
1457 (decide_whether_version_node): Re-calculate known constants for all
1458 remaining context clones.
1460 2018-04-11 Richard Biener <rguenther@suse.de>
1463 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1464 from early DWARF output.
1465 (dwarf2out_early_finish): Output line info unconditionally into
1466 early DWARF and add reference to it.
1468 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1471 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1472 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1473 other than V2DFmode using iptr mode attribute.
1474 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1476 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1478 PR rtl-optimization/84659
1479 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1481 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1484 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1486 (output_loc_list): Pass address of a dummy size variable even in the
1487 locview handling loop.
1488 (index_location_lists): Add comment on why skip_loc_list_entry can't
1491 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1494 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1497 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1500 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1501 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1502 and -mstring-compare-inline-limit.
1504 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1507 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1508 for stack clash protection in a register whenever we need it to be in
1511 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1513 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1514 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1516 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1519 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1521 (mlong-double-): Ditto.
1522 * config/rs6000/sysv4.opt (msdata=): Ditto.
1523 (mtls-size=): Ditto.
1525 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1527 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1528 erroneous entries for
1529 "vector int vec_ldl (int, long int *)", and
1530 "vector unsigned int vec_ldl (int, unsigned long int *)".
1531 Add comments and entries for
1532 "vector bool char vec_ldl (int, bool char *)",
1533 "vector bool short vec_ldl (int, bool short *)",
1534 "vector bool int vec_ldl (int, bool int *)",
1535 "vector bool long long vec_ldl (int, bool long long *)",
1536 "vector pixel vec_ldl (int, pixel *)",
1537 "vector long long vec_ldl (int, long long *)",
1538 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1539 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1540 type tree bool_long_long_type_node and correct definition of
1541 bool_V2DI_type_node to make reference to this new type tree.
1542 (rs6000_mangle_type): Replace erroneous reference to
1543 bool_long_type_node with bool_long_long_type_node.
1544 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1545 comments to emphasize sign distinctions for char and int types and
1546 replace RS6000_BTI_bool_long constant with
1547 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1548 use of RS6000_BTI_pixel.
1549 (bool_long_type_node): Remove this macro definition.
1550 (bool_long_long_type_node): New macro definition
1552 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1554 PR rtl-optimization/85300
1555 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1556 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1557 simplify_unary_operation fails.
1559 2018-04-10 Martin Liska <mliska@suse.cz>
1561 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1562 cgraph_edge and ipa_ref.
1564 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1568 * config/i386/sse.md
1569 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1570 computation of the VEC_MERGE selector from mask.
1571 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1572 Fix decoding of the VEC_MERGE selector into mask.
1574 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1576 PR tree-optimization/85286
1577 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1579 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1581 * final.c (final_1): Set insn_last_address as well as
1582 insn_current_address.
1584 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1587 * explow.c (emit_stack_probe): Call validize_mem on memory location
1588 before passing it to gen_probe_stack. Create address operand and
1589 legitimize it for the probe_stack_address case.
1591 2018-04-09 Jan Hubicka <jh@suse.cz>
1594 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1595 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1596 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1598 (free_lang_data): Rebuild type inheritance graph.
1600 2018-04-09 Martin Sebor <msebor@redhat.com>
1602 * invoke.texi (-finline-small-functions): Mention other optimization
1604 (-findirect-inlining, -fpartial-inlining): Same.
1605 (-finline-functions-called-once): Same.
1606 (-freorder-blocks-and-partition): Same.
1608 2018-04-09 Jan Hubicka <jh@suse.cz>
1611 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1612 jumps; choose last target that matches the criteria (i.e.
1613 no partition changes for non-crossing jumps).
1614 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1615 support for redirecting crossing jumps to non-crossing.
1617 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1619 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1620 also for naked functions.
1622 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1624 * config/arc/arc.md (add_shift): New pattern.
1625 (add_shift2): Likewise.
1626 (sub_shift): Likewise.
1627 (sub_shift_cmp0_noout): Likewise.
1628 (compare_si_ashiftsi): Likewise.
1629 (xbfu_cmp0_noout): New combine pattern.
1630 (xbfu_cmp0"): Likewise.
1631 (movsi_set_cc_insn): Place the predicable variant first.
1632 (commutative_binary_cmp0_noout): Remove clobber.
1633 (commutative_binary_cmp0): New pattern.
1634 (noncommutative_binary_cmp0): Likewise.
1635 (noncommutative_binary_cmp0_noout): Likewise.
1636 (noncommutative_binary_comparison_result_used): Removed.
1637 (rsub_cmp0): New pattern.
1638 (rsub_cmp0_noout): Likewise.
1639 (extzvsi): Changed, keep only meaningful variants.
1640 (SQH, SEZ): New iterators.
1641 (SQH_postfix): New mode attribute.
1642 (SEZ_prefix): New code attribute.
1643 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1644 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1645 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1647 (noncommutative_operator): Check the availability of barrel
1650 2018-04-09 Richard Biener <rguenther@suse.de>
1652 PR tree-optimization/85284
1653 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1654 Only use the niter constraining form of simple_iv when the exit
1657 2018-04-09 Tom de Vries <tom@codesourcery.com>
1660 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1661 (define_expand "*memory_barrier"): New define_expand.
1662 (define_insn "memory_barrier"): New insn.
1664 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1666 PR rtl-optimization/80463
1667 PR rtl-optimization/83972
1668 PR rtl-optimization/83480
1670 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1671 correct producer for the insn.
1672 (tidy_control_flow): Fixup seqnos in case of debug insns.
1674 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1676 PR rtl-optimization/83913
1678 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1679 different sched-times when merging exprs.
1681 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1683 PR rtl-optimization/83962
1685 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1686 tidy_fallthru_edge and tidy_control_flow.
1688 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1690 PR rtl-optimization/83530
1692 * sel-sched.c (force_next_insn): New global variable.
1693 (remove_insn_for_debug): When force_next_insn is true, also leave only
1694 next insn in the ready list.
1695 (sel_sched_region): When the region wasn't scheduled, make another pass
1696 over it with force_next_insn set to 1.
1698 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1700 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1702 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1703 for interrupt control.
1704 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1705 functions for interrupt control.
1706 * config/nds32/nds32-intrinsic.md: Likewise.
1707 * config/nds32/nds32_intrinsic.h: Likewise.
1708 * config/nds32/nds32.h (nds32_builtins): Likewise.
1710 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1712 * config/nds32/nds32.c (nds32_init_machine_status,
1713 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1714 strict_aligned_p field.
1715 (nds32_expand_to_rtl_hook): New function.
1716 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1717 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1719 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1720 Chung-Ju Wu <jasonwucj@gmail.com>
1722 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1723 * config/nds32/nds32-n7.md: New file.
1724 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1725 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1727 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1728 * config/nds32/nds32.md (pipeline_model): Add n7.
1729 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1730 * config/nds32/pipelines.md: Include n7 settings.
1732 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1733 Chung-Ju Wu <jasonwucj@gmail.com>
1735 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1736 * config/nds32/nds32-e8.md: New file.
1737 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1738 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1740 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1741 * config/nds32/nds32.md (pipeline_model): Add e8.
1742 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1743 * config/nds32/pipelines.md: Include e8 settings.
1745 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1746 Chung-Ju Wu <jasonwucj@gmail.com>
1748 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1749 * config/nds32/nds32-n8.md: New file.
1750 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1751 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1753 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1754 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1755 * config/nds32/nds32.md (pipeline_model): Add n8.
1756 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1757 * config/nds32/pipelines.md: Include n8 settings.
1759 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1760 Chung-Ju Wu <jasonwucj@gmail.com>
1762 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1763 * config/nds32/nds32-n9-2r1w.md: New file.
1764 * config/nds32/nds32-n9-3r2w.md: New file.
1765 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1766 nds32_register_ports): New or modify for cpu n9.
1767 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1769 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1770 * config/nds32/nds32-utils.c: New file.
1771 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1772 TARGET_MUL_SLOW): Define.
1773 * config/nds32/nds32.md (pipeline_model): New attribute.
1774 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1775 New options that support cpu n9.
1776 * config/nds32/pipelines.md: Include n9 settings.
1777 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1779 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1781 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1782 information if necessary.
1783 (output_cond_branch_compare_zero): Likewise.
1784 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1785 (nds32_target_alignment): Refine for alignment.
1786 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1787 (FUNCTION_BOUNDARY): Modify.
1788 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1790 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1792 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1794 * config/nds32/constants.md (unspec_volatile_element): Add values for
1795 TLB operation and data prefetch.
1796 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1797 functions for TLB operation and data prefetch.
1798 * config/nds32/nds32-intrinsic.md: Likewise.
1799 * config/nds32/nds32_intrinsic.h: Likewise.
1800 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1801 (nds32_print_operand): Likewise.
1802 * config/nds32/nds32.h (nds32_builtins): Likewise.
1804 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1805 Andrew Pinski <pinsika@gcc.gnu.org>
1808 * match.pd: Use constant_boolean_node of correct type instead of
1809 boolean_true_node or boolean_false_node for simplifying
1810 pointer comparisons to zero.
1812 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1814 PR tree-optimization/80021
1815 * tree.c (verify_type_variant): Make error call in verify_variant_match
1816 translatable and remove final full stop.
1818 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1820 * config/nds32/constants.md (unspec_volatile_element): Add
1821 UNSPEC_VOLATILE_EH_RETURN.
1822 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1823 nds32_output_stack_pop): Support dwarf exception handling process.
1824 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1825 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1826 exception handling process.
1827 (nds32_compute_stack_frame): Likewise.
1828 (nds32_return_addr_rtx): Likewise.
1829 (nds32_initial_elimination_offset): Likewise.
1830 (nds32_expand_prologue): Likewise.
1831 (nds32_expand_epilogue): Likewise.
1832 (nds32_dynamic_chain_address): New function.
1833 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1835 (DYNAMIC_CHAIN_ADDRESS): Define.
1836 (EH_RETURN_DATA_REGNO): Define.
1837 (EH_RETURN_STACKADJ_RTX): Define.
1838 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1839 patterns for dwarf exception handling.
1841 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.h: Clean up obsolete macros.
1845 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1847 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1848 Add enum values for particular instructions.
1849 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1850 particular intrinsic functions.
1851 * config/nds32/nds32-intrinsic.md: Likewise.
1852 * config/nds32/nds32_intrinsic.h: Likewise.
1853 * config/nds32/nds32.h (nds32_builtins): Likewise.
1854 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1855 (btst, ave): New patterns for particular instructions.
1857 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1859 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1860 Add enum values for atomic load/store and memory sync.
1861 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1863 * config/nds32/nds32-intrinsic.md: Likewise.
1864 * config/nds32/nds32_intrinsic.h: Likewise.
1865 * config/nds32/nds32.h (nds32_builtins): Likewise.
1867 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1869 PR tree-optimization/85257
1870 * fold-const.c (native_encode_vector): If not all elts could fit
1871 and off is -1, return 0 rather than offset.
1872 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1873 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1874 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1875 adjust buffer in native_interpret_expr call.
1877 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1879 * config/nds32/constants.md (unspec_volatile_element): Add cache
1880 control enum values.
1881 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1882 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1883 * config/nds32/nds32.c (nds32_cctl_names): New.
1884 (nds32_print_operand): Handle cache control register names.
1885 * config/nds32/nds32.h (nds32_builtins): New enum values.
1886 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1888 * config/nds32/nds32.md (type): Add mmu.
1889 * config/nds32/pipelines.md (simple_insn): Add mmu.
1891 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1893 * config/nds32/nds32.md (type): Remove call.
1894 * config/nds32/pipelines.md (simple_insn): Likewise.
1896 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1898 * config/nds32/constants.md (unspec_volatile_element): Add
1899 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1900 UNSPEC_VOLATILE_FMFCFG.
1901 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1902 description for fmfcfg and fmfcsr.
1903 (bdesc_1arg): Add fmtcsr.
1904 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1905 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1906 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1907 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1908 unspec_fmfcfg): New patterns.
1909 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1910 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1911 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1912 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1913 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1914 __nds32__fmfcfg): Define.
1916 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1918 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1919 intrinsic register names.
1920 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1921 intrinsic register enum values and macros.
1923 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1925 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1926 for load/store addressing form.
1927 (nds32_print_operand_address): Likewise.
1929 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1932 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1933 based on LABEL_REF. Remove useless assertion.
1934 (pic_address_needs_scratch): Fix formatting.
1935 (sparc_legitimize_pic_address): Minor tweaks.
1936 (sparc_delegitimize_address): Adjust assertion accordingly.
1937 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1938 into symbolic_operand.
1939 (movsi_high_pic_label_ref): Likewise.
1940 (movsi_lo_sum_pic_label_ref): Likewise.
1941 (movdi_pic_label_ref): Likewise.
1942 (movdi_high_pic_label_ref): Likewise.
1943 (movdi_lo_sum_pic_label_ref): Likewise.
1945 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1947 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1948 custom LIB_SPEC setup.
1950 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1951 Kito Cheng <kito.cheng@gmail.com>
1953 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1954 * config/riscv/freebsd.h: New.
1956 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1958 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1959 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1962 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1963 Kito Cheng <kito.cheng@gmail.com>
1965 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1966 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1967 * config/nds32/nds32-protos.h (nds32_output_call,
1968 nds32_output_return): Declare.
1969 * config/nds32/nds32.md: Refine all the call and return patterns.
1971 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1974 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1975 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1977 PR rtl-optimization/84872
1978 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1979 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1982 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1984 * expr.c (copy_blkmode_to_reg): Revert 254862.
1985 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1987 2018-04-06 Richard Biener <rguenther@suse.de>
1990 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1991 after seeing a component reference with an adjacent field. Treat
1992 refs to arrays at struct end of external decls similar to
1993 refs to unconstrained commons.
1995 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1998 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1999 look through SAVE_EXPRs with non-side-effects argument. Adjust
2001 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
2004 2018-04-06 Richard Biener <rguenther@suse.de>
2007 * alias.c (find_base_term): New wrapper around find_base_term
2008 unwinding CSELIB_VAL_PTR changes.
2009 (find_base_term): Do not restore CSELIB_VAL_PTR during the
2012 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2014 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
2016 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
2017 constant definitions.
2018 ("nop"): lr 0,0 -> nopr r0
2019 ("nop_lr0", "nop_lr1"): New insn definitions.
2021 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
2023 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
2024 NDS32_V3PUSH_AVAILABLE_P macro.
2026 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
2027 Chung-Ju Wu <jasonwucj@gmail.com>
2029 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
2030 (nds32*-*-*): Add float and fpu_config into supported_defaults.
2031 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
2032 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
2033 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
2034 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
2035 * config/nds32/constraints.md: New constraints and checking for hard
2036 float configuration.
2037 * config/nds32/iterators.md: New mode iterator and attribute for hard
2038 float configuration.
2039 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
2041 * config/nds32/nds32-fpu.md: New file.
2042 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
2043 deal with hard float code generation.
2044 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
2046 (abi_type, float_reg_number): New enum type.
2047 * config/nds32/nds32-predicates.c: New predicates for hard float.
2048 * config/nds32/nds32-protos.h: Declare functions for hard float.
2049 * config/nds32/nds32.c: Implementation for hard float configuration.
2050 * config/nds32/nds32.h: Definitions for hard float configuration.
2051 * config/nds32/nds32.md: Include hard float machine description and
2052 modify patterns for hard float configuration.
2053 * config/nds32/nds32.opt: New options for hard float configuration.
2054 * config/nds32/predicates.md: New predicates for hard float
2057 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2059 * common/config/nds32/nds32-common.c
2060 (nds32_option_optimization_table): Enable -mreleax-hint by default.
2062 2018-04-05 Jakub Jelinek <jakub@redhat.com>
2065 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
2066 CONSTRUCTOR_ELT (ctor, ...)->value.
2068 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
2071 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
2073 2018-04-05 Tom de Vries <tom@codesourcery.com>
2076 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
2079 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2080 Kito Cheng <kito.cheng@gmail.com>
2082 * config/nds32/constraints.md (U33): Fine-tune checking condition.
2083 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
2084 * config/nds32/nds32.h (nds32_16bit_address_type): Add
2085 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
2087 2018-04-05 Shiva Chen <shiva0217@gmail.com>
2088 Kito Cheng <kito.cheng@gmail.com>
2090 * config/nds32/constraints.md (Ufe): New memory constraint.
2091 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
2092 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
2093 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
2095 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
2096 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
2098 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2100 * config/nds32/nds32.md: Use optimize_size in the condition for
2101 alu-shift instructions.
2103 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2105 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
2107 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2109 * config/nds32/nds32.md (negsi2): Refine pattern.
2111 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2112 Chung-Ju Wu <jasonwucj@gmail.com>
2114 * config/nds32/iterators.md (shift_rotate): New code iterator.
2115 (shift): New code attribute.
2116 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
2117 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
2118 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
2119 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
2120 bit-wise operations.
2121 (andsi3, *andsi3): Ditto.
2122 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
2123 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
2124 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
2125 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
2126 nds32_ior_operand, nds32_xor_operand): New predicates.
2128 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2130 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
2131 (addsi3, subsi3): ... this.
2133 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2135 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
2137 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2139 * config/nds32/nds32.md: Adjust indention.
2141 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
2143 * config/nds32/nds32.md (feature): New attribute.
2145 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
2147 * config/nds32/nds32.md (subtype): New attribute.
2149 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
2152 * config/arm/arm-builtins.c (arm_expand_builtin): Change
2153 expansion to perform a bitwise AND of the argument followed by a
2154 boolean negation of the result.
2156 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
2158 PR rtl-optimization/84878
2159 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
2160 the basic block. Assert the use reference is not artificial and that
2161 it has an associated insn.
2163 2018-04-04 Michael Matz <matz@suse.de>
2165 * builtins.c (compute_objsize): Pass correct operand
2166 to array_at_struct_end_p.
2168 2018-04-04 Richard Biener <rguenther@suse.de>
2171 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
2172 from contexts for DINFO_LEVEL_TERSE and below.
2174 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2176 * config/nds32/nds32-doubleword.md (move_<mode>): Require
2177 resiter_operand condition.
2178 * config/nds32/nds32.md (*move<mode>): Ditto.
2180 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2181 Monk Chiang <sh.chiang04@gmail.com>
2183 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
2185 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2187 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
2189 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2190 Kito Cheng <kito.cheng@gmail.com>
2192 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
2193 nds32_cond_code_str, output_cond_branch,
2194 output_cond_branch_compare_zero, 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: New functions.
2200 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
2201 nds32_expand_cstore, nds32_expand_movcc,
2202 nds32_output_cbranchsi4_equality_zero,
2203 nds32_output_cbranchsi4_equality_reg,
2204 nds32_output_cbranchsi4_equality_reg_or_const_int,
2205 nds32_output_cbranchsi4_greater_less_zero): Declare.
2206 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2207 nds32_rimm11s_operand): New predicates.
2208 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2209 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2212 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2214 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2215 * config/nds32/nds32.md: Ditto.
2216 * config/nds32/pipelines.md: Ditto.
2218 2018-04-04 Richard Biener <rguenther@suse.de>
2220 PR tree-optimization/85168
2221 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2222 propagating abnormals.
2224 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2226 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2228 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2229 Kito Cheng <kito.cheng@gmail.com>
2231 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2232 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2233 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2234 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2235 * config/nds32/nds32.md (sibcall_internal): New.
2236 (sibcall_register): Remove.
2237 (sibcall_immediate): Remove.
2238 (sibcall_value_internal): New.
2239 (sibcall_value_register): Remove.
2240 (sibcall_value_immediate): Remove.
2241 * config/nds32/predicates.md (nds32_general_register_operand): New.
2242 (nds32_call_address_operand): New.
2244 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2246 PR rtl-optimization/85167
2247 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2248 bb_defs if *split_p, instead preinitialize it to NULL.
2250 PR tree-optimization/85156
2251 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2252 evaluating the argument multiple times.
2254 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2256 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2258 (_mm_cvtpd_ps): Likewise.
2259 (_mm_cvttpd_epi32): Likewise.
2260 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2261 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2262 vector, pixel, and bool following altivec.h include.
2264 2018-04-03 Martin Sebor <msebor@redhat.com>
2266 * doc/extend.texi (Common Function Attributes): Clarify.
2267 (const attribute): Likewise.
2268 (pure attribute): Likewise.
2270 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2273 * config/i386/i386.c (ix86_expand_vector_set): Use
2274 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2276 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2278 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2279 instructions when changing rounding bits to preserve precision bits
2280 in the x87 control word.
2282 2018-04-03 Martin Liska <mliska@suse.cz>
2284 PR tree-optimization/82491
2285 * rtl.h (strip_offset_and_add): Replace += suboffset with
2286 poly_uint64 () + suboffset.
2288 2018-03-29 Martin Liska <mliska@suse.cz>
2289 Martin Jambor <mjambor@suse.cz>
2292 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2293 param_type is not an integral or pointer type.
2295 2018-04-03 Richard Biener <rguenther@suse.de>
2297 * sese.h (recompute_all_dominators): Remove.
2299 2018-04-02 Martin Sebor <msebor@redhat.com>
2301 * doc/invoke.texi (-Wrestrict): Fix typos.
2303 2018-04-02 Jim Wilson <jimw@sifive.com>
2305 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2306 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2307 (<optab>di3, <optab>si3_extend): Likewise.
2308 (<optab>si3_mask, <optab>si3_mask_1): New.
2309 (<optab>di3_mask, <optab>di3_mask_1): New.
2310 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2311 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2312 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2314 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2316 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2319 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2321 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2322 (nds32_canonicalize_comparison): New function.
2324 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2325 Kito Cheng <kito.cheng@gmail.com>
2326 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2328 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2329 * config/nds32/constants.md (unspec_volatile_element): Add
2330 UNSPEC_VOLATILE_RELAX_GROUP.
2331 * config/nds32/nds32-relax-opt.c: New file.
2332 * config/nds32/nds32-predicates.c
2333 (nds32_symbol_load_store_p): New function.
2334 * config/nds32/nds32-protos.h
2335 (nds32_symbol_load_store_p): Declare function.
2336 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2337 * config/nds32/nds32.c
2338 (nds32_register_pass): New function to register pass.
2339 (nds32_register_passes): New function to register passes.
2340 * config/nds32/nds32.md (relax_group): New pattern.
2341 * config/nds32/nds32.opt (mrelax-hint): New option.
2342 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2344 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2346 * config/nds32/t-nds32: Modify files dependency.
2348 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2350 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2351 (PROFILE_HOOK): Define its implementation.
2353 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2355 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2356 type and 32-bit size.
2358 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2361 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2362 (V_128_256): New mode iterator.
2363 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2364 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2365 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2367 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2368 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2370 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2373 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2374 NaN inputs correctly.
2376 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2379 * config/rs6000/vsx.md (??r): New mode attribute.
2380 (*vsx_mov<mode>_64bit): Use it.
2381 (*vsx_mov<mode>_32bit): Likewise.
2383 2018-03-30 Martin Sebor <msebor@redhat.com>
2385 PR tree-optimization/84818
2386 * builtins.c (check_access): Use warning_n.
2388 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2391 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2393 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2396 2018-03-30 Julia Koval <julia.koval@intel.com>
2399 * x86-tune.def (movx, partial_reg_dependency): Enable for
2402 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2405 * lra-constraints.c (process_alt_operands): Move setting
2406 this_alternative_matches below.
2408 2018-03-29 Martin Liska <mliska@suse.cz>
2411 * doc/invoke.texi: Document how LTO works with debug info.
2412 Describe auto-load support of binutils. Mention 'x86-64'
2413 as valid option value of -march option.
2415 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2417 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2420 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2421 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2422 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2425 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2428 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2429 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2430 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2431 for RS6000_BTM_POWERPC64.
2432 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2433 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2434 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2439 2018-03-28 Carl Love <cel@us.ibm.com>
2442 2017-09-27 Carl Love <cel@us.ibm.com>
2444 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2445 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2446 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2449 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2451 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2452 instead of __vector bool.
2453 (_mm_max_pu8): Likewise.
2454 (_mm_min_pi16): Likewise.
2456 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2459 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2460 (DIVWEUO): Likewise.
2462 (DIVDEUO): Likewise.
2463 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2464 DIVWEUO and DIVDEUO.
2465 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2466 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2467 (div_extend): Likewise.
2468 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2470 (__builtin_divweuo): Likewise.
2471 (__builtin_divdeo): Likewise.
2472 (__builtin_divdeuo): Likewise.
2474 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2477 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2478 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2480 PR tree-optimization/82004
2481 * gimple-match-head.c (optimize_pow_to_exp): New function.
2482 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2483 Don't fold to exp if optimize_pow_to_exp is false.
2485 2018-03-28 Martin Liska <mliska@suse.cz>
2488 * calls.c (initialize_argument_information): Fix trailing space.
2489 * common.opt: Fix typo and provide better explanation for
2490 -fsanitize-coverage option.
2491 * config/i386/i386.opt: Fix typo.
2493 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2494 Martin Liska <mliska@suse.cz>
2497 * gimplify.c (asan_poison_variable): Don't do the check for
2498 gimplify_omp_ctxp here.
2499 (gimplify_decl_expr): Do it here.
2500 (gimplify_target_expr): Likewise.
2502 2018-03-28 Martin Liska <mliska@suse.cz>
2505 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2506 chkp_type_bounds_count if MPX is not enabled.
2508 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2510 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2512 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2515 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2516 function to create the function decl for complex long double
2517 multiply and divide for -mabi=ieeelongdouble.
2518 (init_float128_ieee): Call it.
2520 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2523 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2524 -fcf-protection=branch -mibt.
2525 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2527 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2530 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2532 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2535 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2536 extern array declarations.
2538 2018-03-27 Richard Biener <rguenther@suse.de>
2541 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2542 explicit single_use checks.
2544 2018-03-27 Richard Biener <rguenther@suse.de>
2546 PR tree-optimization/85082
2547 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2550 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2552 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2553 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2554 Turn on fasynchronous-unwind-tables and funwind-tables.
2556 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2559 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2560 (*bmi_blsr_<mode>_ccz): Ditto.
2562 2018-03-26 Tom de Vries <tom@codesourcery.com>
2564 PR tree-optimization/85063
2565 * omp-general.c (offloading_function_p): New function. Factor out
2567 * omp-offload.c (pass_omp_target_link::gate): ... here.
2568 * omp-general.h (offloading_function_p): Declare.
2569 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2570 with attribute omp declare target for offloading functions.
2572 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2574 PR tree-optimization/84005
2575 * tree-data-ref.h (get_base_for_alignment): Declare.
2576 * tree-data-ref.c (get_base_for_alignment_1): New function.
2577 (get_base_for_alignment): Likewise.
2578 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2579 get_base_for_alignment to find a suitable base object, instead
2580 of always using drb->base_address.
2582 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2585 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2586 known size by default.
2588 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2591 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2592 and non BLKmode operands.
2594 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2597 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2598 Clean up attributes.
2600 2018-03-23 Richard Biener <rguenther@suse.de>
2603 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2604 we are going to emit early debug for LTO.
2606 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2609 * function.c (match_asm_constraints_1): Don't optimize if input
2610 doesn't satisfy general_operand predicate for output's mode.
2613 * alias.c (write_dependence_p): Don't require for x_canonicalized
2614 non-VOIDmode if x has VOIDmode.
2617 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2618 just don't try to optimize it rather than assert it never happens.
2620 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2622 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2623 macro expansions for definition of ST_INTERNAL_<mode> and
2624 LD_INTERNAL_<mode> builtins.
2625 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2627 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2629 (altivec_expand_st_builtin): Likewise.
2630 (altivec_expand_builtin): Remove calls to deleted functions.
2631 (rs6000_address_for_altivec): Delete this function.
2632 * config/rs6000/vector.md: Remove expands for
2633 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2635 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2638 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2639 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2640 re-computing once computed.
2641 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2642 (arm_init_machine_status): Initialize
2643 machine->static_chain_stack_bytes.
2645 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2648 * doc/extend.texi: Add four new prototypes for vec_ld.
2649 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2650 definitions for more logical presentation.
2651 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2652 entries for V1TI variants of __builtin_altivec_ld builtin.
2653 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2654 handling of V1TI variant of LVX icode pattern.
2655 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2656 (rs6000_gimple_fold_builtin): Likewise.
2657 (altivec_init_builtins): Add code to define
2658 __builtin_altivec_lvx_v1ti function.
2660 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2663 * function.c (match_asm_constraints_1): Don't do the optimization
2664 if input isn't a REG, SUBREG, MEM or constant.
2666 2018-03-22 Tom de Vries <tom@codesourcery.com>
2668 PR tree-optimization/84956
2669 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2670 bb_has_abnormal_pred.
2672 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2675 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2676 DECL_INITIAL (decl) to decl at the end.
2677 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2680 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2682 * doc/extend.texi (__builtin_tgmath): Document when complex
2683 integer types are treated as _Complex _Float64.
2685 2018-03-21 Tom de Vries <tom@codesourcery.com>
2687 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2689 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2691 PR tree-optimization/84960
2692 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2693 if it is ENTRY block, move them into single succ of ENTRY in that case.
2695 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2697 PR tree-optimization/84811
2698 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2699 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2700 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2701 (known_subrange_p): Update accordingly. Cast each value involved
2702 in the size comparison, rather than casting the result of the
2705 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2707 PR tree-optimization/84982
2708 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2709 by flipping the least significant bit rather than all bits from
2710 bitpos to bitpos + bitsize - 1.
2712 2018-03-21 Nathan Sidwell <nathan@acm.org>
2714 * doc/extend.texi (Deprecated Features): Remove mention of
2715 long-deleted deprecations.
2717 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2720 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2721 * configure: Regenerate.
2723 2018-03-21 Tom de Vries <tom@codesourcery.com>
2725 PR tree-optimization/83126
2726 * tree-parloops.c (num_phis): New function.
2727 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2729 2018-03-21 Nathan Sidwell <nathan@acm.org>
2731 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2732 mention anon-struct/union members and trailing attributes.
2734 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2736 PR tree-optimization/84969
2737 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2738 builtin memset partitions if they set different rhs values.
2740 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2742 PR rtl-optimization/84989
2743 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2744 VEC_DUPLICATE with scalar result mode.
2746 2018-03-21 Martin Liska <mliska@suse.cz>
2749 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2750 not intended return statement.
2752 2018-03-21 Martin Liska <mliska@suse.cz>
2755 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2756 (chkp_find_bound_slots_1): Limit number of iterations.
2758 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2761 * Minor grammar fixes for x86 options.
2763 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2766 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2767 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2769 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2772 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2773 (altivec_lvx_<mode>_1op): Likewise.
2774 (altivec_stvx_<mode>_2op): Likewise.
2775 (altivec_stvx_<mode>_1op): Likewise.
2776 (altivec_lvx_<VM2:mode>): New define_expand.
2777 (altivec_stvx_<VM2:mode>): Likewise.
2778 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2779 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2780 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2781 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2782 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2783 (rs6000_gen_lvx): Likewise.
2784 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2785 (altivec_expand_stv_builtin): Likewise.
2786 (altivec_expand_builtin): Likewise.
2787 * config/rs6000/vector.md: Likewise.
2789 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2792 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2795 2018-03-20 Richard Biener <rguenther@suse.de>
2798 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2799 sign-conversions as zero, fall back to standard scalar_stmt
2802 2018-03-20 Martin Liska <mliska@suse.cz>
2805 * predict.c (rebuild_frequencies): Handle case when we have
2806 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2808 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2811 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2812 flag_section_anchors.
2813 * varasm.c (use_blocks_for_decl_p): Remove hack for
2814 dw2_force_const_mem.
2817 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2819 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2820 be created, use lowpart_subreg of operands[0] rather than operands[0]
2822 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2823 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2824 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2825 and n constraint instead of aarch64_shift_imm_di and Usd.
2826 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2827 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2829 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2832 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2833 to favor GPR over NEON registers.
2834 (<shift>di3_neon): Likewise.
2836 2018-03-20 Tom de Vries <tom@codesourcery.com>
2839 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2840 (nvptx_process_pars): Emit bar.sync asap and alap.
2842 2018-03-20 Tom de Vries <tom@codesourcery.com>
2845 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2846 seen_label if seen_label is already set.
2848 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2851 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2852 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2853 Use 1U instead of 1. Formatting fixes.
2856 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2857 instead of TREE_TYPE (s1) for the return value.
2859 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2861 PR tree-optimization/84946
2862 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2863 bitsize + bitsize in poly_uint64 rather than poly_int64.
2866 * dwarf2asm.c: Include fold-const.c.
2867 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2868 of decl rather than decl itself.
2870 PR rtl-optimization/84643
2871 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2873 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2876 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2877 calling assemble_variable.
2879 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2882 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2883 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2885 2018-03-19 Jim Wilson <jimw@sifive.com>
2888 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2889 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2890 (riscv_first_stack_step): Likewise.
2891 (riscv_option_override): Use STACK_BOUNDARY instead of
2893 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2895 (BIGGEST_ALIGNMENT): Set to 128.
2896 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2897 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2900 2018-03-19 Richard Biener <rguenther@suse.de>
2902 PR tree-optimization/84933
2903 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2904 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2906 2018-03-19 Richard Biener <rguenther@suse.de>
2908 PR tree-optimization/84859
2909 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2910 (cond_if_else_store_replacement): Perform sinking operation on
2911 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2912 Generalize what a BB with a single eligible store is.
2914 2018-03-19 Richard Biener <rguenther@suse.de>
2916 PR tree-optimization/84929
2917 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2918 chrec_is_positive against non-chrec arg.
2920 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2923 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2925 2018-03-18 Martin Liska <mliska@suse.cz>
2927 PR rtl-optimization/84635
2928 * regrename.c (build_def_use): Use matches_mode only when
2931 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2933 PR tree-optimization/84913
2934 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2935 vectorize chains of COND_EXPRs.
2937 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2939 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2941 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2943 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2945 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2947 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2949 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2950 Kito Cheng <kito.cheng@gmail.com>
2952 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2953 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2954 (nds32_adjust_reg_alloc_order): New function.
2955 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2957 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2959 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2960 nds32_print_operand, nds32_print_operand_address): Use
2961 HOST_WIDE_INT_PRINT_DEC instead.
2963 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2965 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2967 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2970 * config/i386/i386.c (initial_ix86_tune_features,
2971 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2973 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2974 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2975 rather than 1u << ix86_tune. Formatting fix.
2976 (ix86_option_override_internal): Change ix86_arch_mask from
2977 unsigned int to unsigned HOST_WIDE_INT, initialize to
2978 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2979 (ix86_function_specific_restore): Likewise.
2981 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2984 * postreload.c (reload_combine_recognize_pattern): Perform
2985 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2986 truncate_int_for_mode the result for the destination's mode.
2989 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2990 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2992 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2995 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2998 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2999 regno_allocno_class_array and sorted_pseudos.
3000 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
3001 insns where regno is used.
3003 2018-03-16 Martin Liska <mliska@suse.cz>
3006 * multiple_target.c (create_dispatcher_calls): Redirect
3007 reference in the symbol table.
3009 2018-03-16 Martin Liska <mliska@suse.cz>
3012 * multiple_target.c (create_dispatcher_calls): Redirect also
3015 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3019 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
3020 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
3021 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
3023 2018-03-16 Julia Koval <julia.koval@intel.com>
3025 * doc/invoke.texi (Skylake Server): Add CLWB.
3026 Cannonlake): Remove CLWB.
3028 2018-03-16 Jakub Jelinek <jakub@redhat.com>
3030 PR tree-optimization/84841
3031 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
3033 (FLOAT_ONE_CONST_TYPE): Define.
3034 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
3035 (sort_by_operand_rank): Put entries with higher constant_type last
3036 rather than first to match comments.
3038 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
3040 * config/nios2/nios2.md (movsi_internal): Fix thinko in
3043 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3046 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
3047 check and use address of target always.
3049 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
3052 * config/i386/i386.c (indirect_thunk_needed): Update comments.
3053 (indirect_thunk_bnd_needed): Likewise.
3054 (indirect_thunks_used): Likewise.
3055 (indirect_thunks_bnd_used): Likewise.
3056 (indirect_return_needed): New.
3057 (indirect_return_bnd_needed): Likewise.
3058 (output_indirect_thunk_function): Add a bool argument for
3060 (output_indirect_thunk_function): Don't generate alias for
3061 function return thunk.
3062 (ix86_code_end): Call output_indirect_thunk_function to generate
3063 function return thunks.
3064 (ix86_output_function_return): Set indirect_return_bnd_needed
3065 and indirect_return_needed instead of indirect_thunk_bnd_needed
3066 and indirect_thunk_needed.
3068 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
3070 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
3071 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
3072 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
3074 2018-03-15 David Malcolm <dmalcolm@redhat.com>
3075 Paul Hua <paul.hua.gm@gmail.com>
3078 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
3080 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
3082 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
3083 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
3086 2018-03-15 Tamar Christina <tamar.christina@arm.com>
3089 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
3090 instead of GET_MODE_SIZE when comparing Units.
3092 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3095 * varasm.c (hash_section): Return an unchangeble hash value
3096 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
3097 Return !aarch64_can_use_per_function_literal_pools_p ().
3099 2018-03-15 Jakub Jelinek <jakub@redhat.com>
3102 * optabs.c (emit_conditional_move): Pass address of cmode's copy
3103 rather than address of cmode as last argument to prepare_cmp_insn.
3105 2018-03-15 Julia Koval <julia.koval@intel.com>
3107 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
3108 F_AVX512VNNI, F_AVX512BITALG): New.
3110 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
3113 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
3114 insn for floating-point loads and stores.
3116 2018-03-14 Carl Love <cel@us.ibm.com>
3118 * config/rs6000/rs6000-c.c: Add macro definitions for
3119 ALTIVEC_BUILTIN_VEC_PERMXOR.
3120 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
3121 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
3122 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
3123 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
3125 * config/doc/extend.texi: Add prototypes for vec_permxor.
3127 2018-03-14 David Malcolm <dmalcolm@redhat.com>
3130 * diagnostic-show-locus.c (class layout_point): Convert m_line
3131 from int to linenum_type.
3132 (line_span::comparator): Use linenum "compare" function when
3133 comparing line numbers.
3134 (test_line_span): New function.
3135 (layout_range::contains_point): Convert param "row" from int to
3137 (layout_range::intersects_line_p): Likewise.
3138 (layout::will_show_line_p): Likewise.
3139 (layout::print_source_line): Likewise.
3140 (layout::should_print_annotation_line_p): Likewise.
3141 (layout::print_annotation_line): Likewise.
3142 (layout::print_leading_fixits): Likewise.
3143 (layout::annotation_line_showed_range_p): Likewise.
3144 (struct line_corrections): Likewise for field m_row.
3145 (line_corrections::line_corrections): Likewise for param "row".
3146 (layout::print_trailing_fixits): Likewise.
3147 (layout::get_state_at_point): Likewise.
3148 (layout::get_x_bound_for_row): Likewise.
3149 (layout::print_line): Likewise.
3150 (diagnostic_show_locus): Likewise for locals "last_line" and
3152 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
3153 * input.c (selftest::test_linenum_comparisons): New function.
3154 (selftest::input_c_tests): Call it.
3155 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
3156 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
3157 * selftest.h (ASSERT_GT): New macro.
3158 (ASSERT_GT_AT): New macro.
3159 (ASSERT_LT): New macro.
3160 (ASSERT_LT_AT): New macro.
3162 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
3164 PR rtl-optimization/84780
3165 * combine.c (distribute_links): Don't make a link based on pc_rtx.
3167 2018-03-14 Martin Liska <mliska@suse.cz>
3169 * tree.c (record_node_allocation_statistics): Use
3170 get_stats_node_kind.
3171 (get_stats_node_kind): New function extracted from
3172 record_node_allocation_statistics.
3173 (free_node): Use get_stats_node_kind.
3175 2018-03-14 Richard Biener <rguenther@suse.de>
3177 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
3178 that the value-set of ANTIC_IN doesn't grow.
3181 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3183 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3184 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3186 2018-03-14 Julia Koval <julia.koval@intel.com>
3188 * config.gcc (icelake-client, icelake-server): New.
3190 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
3191 (initial_ix86_arch_features): Ditto.
3192 (PTA_SKYLAKE): Add SGX.
3193 (PTA_ICELAKE): Remove.
3194 (PTA_ICELAKE_CLIENT): New.
3195 (PTA_ICELAKE_SERVER): New.
3196 (ix86_option_override_internal): Split up icelake on icelake client and
3198 (get_builtin_code_for_version): Ditto.
3199 (fold_builtin_cpu): Ditto.
3200 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
3201 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
3202 * config/i386/i386.h (processor_type): Ditto.
3203 * doc/invoke.texi: Ditto.
3205 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3208 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3209 INTEGER_CST offset, add it together with bitpos / 8 and
3210 sign extend based on POINTER_SIZE.
3214 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3217 * config/i386/constraints.md (Yc): New register constraint.
3218 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3219 Use Yc constraint for alternative 2 of operand 0. Remove
3220 preferred_for_speed attribute.
3222 2018-03-14 Richard Biener <rguenther@suse.de>
3224 PR tree-optimization/84830
3225 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3226 with the old one to avoid oscillations.
3228 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3231 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3233 (assign_by_spills): Return a flag of reload assignment failure.
3234 Do not process the reload assignment failures. Do not spill other
3235 reload pseudos if they has the same reg class. Update n if
3237 (lra_assign): Add a return arg. Set up from the result of
3238 assign_by_spills call.
3239 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3240 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3241 usage_insns if it is not NULL.
3242 (spill_hard_reg_in_range): New function.
3243 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3244 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3245 function prototypes.
3246 (lra_assign): Change prototype.
3247 * lra.c (lra): Add code to deal with fails by splitting hard reg
3250 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3252 * config/riscv/riscv.opt (mrelax): New option.
3253 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3254 "norelax" when riscv_mrelax is disabled.
3255 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3257 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3260 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3261 reassociation for int modes.
3263 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3265 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3266 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3268 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3269 * config/aarch64/aarch64-sve.md
3270 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3271 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3272 (*extend<mode><Vwide>2): Rename to...
3273 (aarch64_sve_extend<mode><Vwide>2): ...this.
3274 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3275 renaming the old pattern to...
3276 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3278 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3279 define_expand, renaming the old pattern to...
3280 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3281 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3282 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3283 account when deciding which SVE instruction the optab should use.
3284 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3286 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3288 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3289 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3290 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3291 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3292 (tlsdesc_small_<mode>): Turn a define_expand and use
3293 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3294 (tlsdesc_small_advsimd_<mode>): ...this.
3295 (tlsdesc_small_sve_<mode>): New pattern.
3297 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3299 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3300 (UNSPEC_UMUL_HIGHPART): New constants.
3301 (MUL_HIGHPART): New int iteraor.
3302 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3303 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3305 (*<su>mul<mode>3_highpart): New define_insn.
3307 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3310 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3313 2018-03-13 Martin Liska <mliska@suse.cz>
3316 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3318 (sem_item_optimizer::~sem_item_optimizer): Release it.
3319 (sem_item_optimizer::merge_classes): Register variable aliases.
3320 (sem_item_optimizer::fixup_pt_set): New function.
3321 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3322 * ipa-icf.h: Declare new variables and functions.
3324 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3327 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3328 integer_pow2p@2 and test integer_pow2p in condition.
3329 (A < 0 ? C : 0): Similarly for @1.
3332 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3333 characters starting at p contain '\0' character, don't look beyond
3337 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3338 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3341 * reg-stack.c (change_stack): Change update_end var from int to
3342 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3343 also call set_block_for_insn on the newly added insns and rescan.
3346 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3347 on the last operand.
3350 * tree.c (stabilize_reference_1): Return save_expr (e) for
3351 STATEMENT_LIST even if it doesn't have side-effects.
3353 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3355 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3357 2018-03-12 Renlin Li <renlin.li@arm.com>
3359 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3360 aarch64_output_scalar_simd_mov_immediate.
3362 2018-03-12 Martin Sebor <msebor@redhat.com>
3364 PR tree-optimization/83456
3365 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3366 for perfectly overlapping calls to memcpy.
3367 (gimple_fold_builtin_memory_chk): Same.
3368 (gimple_fold_builtin_strcpy): Handle no-warning.
3369 (gimple_fold_builtin_stxcpy_chk): Same.
3370 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3372 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3374 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3375 parameter. Use it for SFmode.
3376 (rs6000_function_arg_advance_1): Adjust.
3377 (rs6000_function_arg): Adjust.
3378 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3380 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3382 PR rtl-optimization/84169
3383 PR rtl-optimization/84780
3384 * combine.c (can_combine_p): Check for a 2-insn combination whether
3385 the destination register is used between the two insns, too.
3387 2018-03-12 Richard Biener <rguenther@suse.de>
3389 PR tree-optimization/84803
3390 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3391 for refs DR analysis didn't process.
3393 2018-03-12 Richard Biener <rguenther@suse.de>
3395 PR tree-optimization/84777
3396 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3397 force-vectorize loops ignore whether we are optimizing for size.
3399 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3401 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3402 (TARGET_MD_ASM_ADJUST): Define.
3404 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3405 Kito Cheng <kito.cheng@gmail.com>
3406 Chung-Ju Wu <jasonwucj@gmail.com>
3408 * config/nds32/nds32.c (nds32_compute_stack_frame,
3409 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3410 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3411 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3412 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3413 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3414 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3415 * config/nds32/nds32.md (prologue, epilogue): Use macro
3416 NDS32_V3PUSH_AVAILABLE_P to do checking.
3418 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3421 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3422 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3423 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3424 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3425 addition of most attributes on !orig_type_die or the attribute not
3426 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3428 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3429 Chung-Ju Wu <jasonwucj@gmail.com>
3431 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3433 * config/nds32/nds32.opt (mvh): New option.
3435 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3436 Chung-Ju Wu <jasonwucj@gmail.com>
3438 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3440 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3441 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3444 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3445 Chung-Ju Wu <jasonwucj@gmail.com>
3447 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3449 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3450 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3452 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3453 Kito Cheng <kito.cheng@gmail.com>
3454 Chung-Ju Wu <jasonwucj@gmail.com>
3456 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3457 UNSPEC_FFMISM and UNSPEC_FLMISM.
3458 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3459 for ffb, ffmism and flmism.
3460 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3461 (unspec_ffmism): Ditto.
3462 (unspec_flmism): Ditto.
3463 (nds32_expand_builtin_impl): Check if string extension is available.
3464 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3465 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3467 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3470 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3473 * lra-assigns.c (assign_by_spills): Return a flag of reload
3474 assignment failure. Do not process the reload assignment
3475 failures. Do not spill other reload pseudos if they has the same
3477 (lra_assign): Add a return arg. Set up from the result of
3478 assign_by_spills call.
3479 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3480 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3481 usage_insns if it is not NULL.
3482 (spill_hard_reg_in_range): New function.
3483 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3484 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3485 function prototypes.
3486 (lra_assign): Change prototype.
3487 * lra.c (lra): Add code to deal with fails by splitting hard reg
3490 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3493 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3495 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3498 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3499 (dw_val_node): Add val_symbolic_view.
3500 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3501 (symview_upper_bound): New.
3502 (new_line_info_table): Initialize symviews_since_reset.
3503 (dwarf2out_source_line): Count symviews_since_reset and set
3504 symview_upper_bound.
3505 (dw_val_equal_p): Handle symview.
3506 (add_AT_symview): New.
3507 (print_dw_val): Handle symview.
3508 (attr_checksum, attr_checksum_ordered): Likewise.
3509 (same_dw_val_p, size_of_die): Likewise.
3510 (value_format, output_die): Likewise.
3511 (add_high_low_attributes): Use add_AT_symview for entry_view.
3512 (dwarf2out_finish): Reset symview_upper_bound, clear
3515 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3518 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3519 Add strict argument and use it.
3520 (rs6000_split_multireg_move): Update for new strict argument.
3521 (mem_operand_gpr): Disallow all non-offsettable addresses.
3522 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3524 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3527 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3528 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3529 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3532 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3533 decl, use remap_type if we want to use the type.
3535 2018-03-09 Martin Sebor <msebor@redhat.com>
3537 PR tree-optimization/84526
3538 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3540 (builtin_access::generic_overlap): Be prepared to handle non-array
3543 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3545 PR rtl-optimization/84682
3546 * lra-constraints.c (process_address_1): Check is_address flag
3547 for address constraints.
3548 (process_alt_operands): Likewise.
3549 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3550 preprocess_constraints.
3551 * recog.h (preprocess_constraints): Add oploc parameter.
3553 * recog.c (preprocess_constraints): Test address_operand for
3554 CT_ADDRESS constraints.
3556 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3559 * lra-assigns.c (assign_by_spills): Return a flag of reload
3560 assignment failure. Do not process the reload assignment
3561 failures. Do not spill other reload pseudos if they has the same
3563 (lra_assign): Add a return arg. Set up from the result of
3564 assign_by_spills call.
3565 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3566 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3567 usage_insns if it is not NULL.
3568 (spill_hard_reg_in_range): New function.
3569 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3570 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3571 function prototypes.
3572 (lra_assign): Change prototype.
3573 * lra.c (lra): Add code to deal with fails by splitting hard reg
3576 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3579 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3580 Accept complain bool parameter. Only emit errors if it is true.
3581 (arm_parse_cpu_option_name): Likewise.
3582 (arm_target_thumb_only): Adjust callers of the above.
3583 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3584 prototype to take a default true bool parameter.
3585 (arm_parse_arch_option_name): Likewise.
3587 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3588 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3592 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3593 * configure: Regenerate.
3594 * configure.ac ("linker --version-script option"): New.
3595 ("linker soname option"): New.
3597 2018-03-09 Richard Biener <rguenther@suse.de>
3599 PR tree-optimization/84775
3600 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3601 immediate uses of predicate stmts and mark them modified.
3604 PR tree-optimization/84178
3605 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3607 (version_loop_for_if_conversion): Delay update_ssa call.
3608 (tree_if_conversion): Delay update_ssa until after predicate
3611 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3614 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3615 when the function accesses prior frames.
3617 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3620 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3621 gen_llsym, otherwise call maybe_gen_llsym.
3624 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3625 has ',' character inside of it.
3627 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3630 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3631 as clobbering CC_REGNUM.
3633 2018-03-08 Richard Biener <rguenther@suse.de>
3636 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3637 (follow_copies_to_constant): Do not follow SSA names registered
3640 2018-03-08 Richard Biener <rguenther@suse.de>
3642 PR tree-optimization/84178
3643 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3645 (version_loop_for_if_conversion): Delay update_ssa call.
3646 (tree_if_conversion): Delay update_ssa until after predicate
3649 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3651 PR tree-optimization/84178
3652 * tree-if-conv.c (release_bb_predicate): Remove the
3653 the assertion that the stmts have NULL use_ops.
3654 Discard the statements, asserting that they haven't
3655 yet been added to a BB.
3657 2018-03-08 Richard Biener <rguenther@suse.de>
3659 PR tree-optimization/84746
3660 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3661 (phi_translate): Pass in destination ANTIC_OUT set.
3662 (phi_translate_1): Likewise. For a simplified result lookup
3663 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3664 (phi_translate_set): Adjust.
3665 (do_pre_regular_insertion): Likewise.
3666 (do_pre_partial_partial_insertion): Likewise.
3668 2018-03-08 Martin Liska <mliska@suse.cz>
3670 PR gcov-profile/84735
3671 * doc/gcov.texi: Document usage of profile files.
3672 * gcov-io.h: Document changes in the format.
3674 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3678 * dwarf2out.c (struct dw_line_info_table): Update comments for
3680 (FORCE_RESET_NEXT_VIEW): New.
3681 (FORCE_RESETTING_VIEW_P): New.
3682 (RESETTING_VIEW_P): Check for -1 too.
3683 (ZERO_VIEW_P): Likewise.
3684 (new_line_info_table): Force-reset next view.
3685 (dwarf2out_begin_function): Likewise.
3686 (dwarf2out_source_line): Simplify zero_view_p initialization.
3687 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3688 view directly. Omit view when omitting .loc at line 0.
3690 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3692 PR tree-optimization/84740
3693 * tree-switch-conversion.c (process_switch): Call build_constructors
3694 only if info.phi_count is non-zero.
3696 PR tree-optimization/84739
3697 * tree-tailcall.c (find_tail_calls): Check call arguments against
3698 DECL_ARGUMENTS (current_function_decl) rather than
3699 DECL_ARGUMENTS (func) when checking for tail recursion.
3701 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3703 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3704 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3705 Volker Reichelt's entry and add entries for people that perform
3706 GCC fuzzy testing and report numerous bugs.
3708 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3711 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3712 readonly data in sdata, if that is disabled.
3713 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3714 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3715 -mreadonly-in-sdata option.
3717 2018-03-07 Martin Sebor <msebor@redhat.com>
3719 PR tree-optimization/84468
3720 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3721 basic block when looking for nul assignment.
3723 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3726 * except.h (output_function_exception_table): Adjust prototype.
3727 * except.c (output_function_exception_table): Remove FNNAME parameter
3728 and add SECTION parameter. Ouput one part of the table at a time.
3729 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3730 the first part of the exception table and emit unwind directives.
3731 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3732 (i386_pe_seh_cold_init): Likewise.
3733 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3734 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3735 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3736 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3737 * config/i386/winnt.c: Include except.h.
3738 (struct seh_frame_state): Add reg_offset, after_prologue and
3739 in_cold_section fields.
3740 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3741 (i386_pe_seh_cold_init): New function.
3742 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3743 to seh->in_cold_section.
3744 (seh_emit_push): Record the offset of the push.
3745 (seh_emit_save): Record the offet of the save.
3746 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3747 Test seh->after_prologue to disregard the epilogue.
3748 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3749 (i386_pe_end_cold_function): New function.
3751 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3754 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3755 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3758 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3760 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3761 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3764 * multiple_target.c: Include tree-inline.h and intl.h.
3765 (expand_target_clones): Diagnose and fail if node->definition and
3766 !tree_versionable_function_p (node->decl).
3768 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3770 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3772 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3773 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3774 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3776 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3779 * combine.c (try_combine): Use reg_or_subregno instead of handling
3780 just paradoxical SUBREGs and REGs.
3782 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3784 * config/arc/arc.c (arc_finalize_pic): Remove function.
3785 (arc_must_save_register): We use single base PIC register, remove
3786 checks to save/restore the PIC register.
3787 (arc_expand_prologue): Likewise.
3788 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3790 (arc_verify_short): Likewise.
3791 (arc_attr_type): Likewise.
3792 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3793 (walk_stores): Likewise.
3794 (arc_address_cost): Make it static.
3795 (arc_verify_short): Likewise.
3796 (branch_dest): Likewise.
3797 (arc_attr_type): Likewise.
3798 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3799 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3800 (arc_final_prescan_insn): Remove inserting the nops due to
3801 hardware hazards. It is done in reorg step.
3802 (insn_length_variant_t): Remove.
3803 (insn_length_parameters_t): Likewise.
3804 (arc_insn_length_parameters): Likewise.
3805 (arc_get_insn_variants): Likewise.
3806 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3808 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3811 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3814 PR tree-optimization/84687
3815 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3817 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3819 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3821 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3822 Rename to ppc_speculation_barrier.
3823 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3824 __builtin_ppc_speculation_barrier.
3826 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3829 * combine.c (combine_simplify_rtx): Don't try to simplify if
3830 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3833 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3835 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3836 to 32 bytes when compiling for POWER9.
3838 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3841 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3842 regparm >= 3 with no arg reg available also for calls with
3843 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3846 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3848 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3850 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3853 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3855 2018-03-05 Richard Biener <rguenther@suse.de>
3857 PR tree-optimization/84486
3858 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3859 When inserting a __builtin_assume_aligned call set the LHS
3860 SSA name alignment info accordingly.
3862 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3864 PR tree-optimization/84114
3865 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3866 Avoid reassociation of FLOAT_MODE addition.
3868 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3870 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3871 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3872 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3873 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3874 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3875 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3876 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3878 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3879 __builtin_ia32_wbinvd): New builtins.
3880 (SPECIAL_ARGS2): New.
3881 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3882 (SPECIAL_ARGS2): New.
3883 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3884 (ix86_valid_target_attribute_inner_p): Ditto.
3885 (ix86_init_mmx_sse_builtins): Add special_args2.
3886 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3887 TARGET_WBNOINVD_P): New.
3888 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3889 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3890 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3891 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3892 * config/i386/pconfigintrin.h: New file.
3893 * config/i386/wbnoinvdintrin.h: Ditto.
3894 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3895 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3897 2018-03-05 Richard Biener <rguenther@suse.de>
3899 PR tree-optimization/84670
3900 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3902 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3903 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3904 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3905 doesn't grow if all successors (recursively) were visited at least
3908 2018-03-05 Richard Biener <rguenther@suse.de>
3910 PR tree-optimization/84650
3911 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3912 if executed in the loop pipeline.
3914 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3916 * doc/configfiles.texi (Configuration Files): Move info about
3917 conditionalizing $target-protos.h to...
3918 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3919 differs from $target-protos.h.
3921 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3922 Chung-Ju Wu <jasonwucj@gmail.com>
3924 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3925 * config/nds32/nds32-multiple.md (setmemsi): Define.
3926 * config/nds32/nds32-memory-manipulation.c
3927 (nds32_gen_dup_4_byte_to_word_value): New.
3928 (emit_setmem_word_loop): New.
3929 (emit_setmem_byte_loop): New.
3930 (nds32_expand_setmem_loop): New.
3931 (nds32_expand_setmem_loop_v3m): New.
3932 (nds32_expand_setmem_unroll): New.
3933 (nds32_expand_setmem): New.
3935 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3936 Chung-Ju Wu <jasonwucj@gmail.com>
3938 * config/nds32/nds32-memory-manipulation.c
3939 (nds32_emit_load_store): New.
3940 (nds32_emit_post_inc_load_store): New.
3941 (nds32_emit_mem_move): New.
3942 (nds32_emit_mem_move_block): New.
3943 (nds32_expand_movmemsi_loop_unknown_size): New.
3944 (nds32_expand_movmemsi_loop_known_size): New.
3945 (nds32_expand_movmemsi_loop): New.
3946 (nds32_expand_movmemsi_unroll): New.
3947 (nds32_expand_movmemqi): Rename ...
3948 (nds32_expand_movmemsi): ... to this.
3949 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3950 (movmemsi): ... to this.
3951 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3952 (nds32_expand_movmemsi): ... to this.
3954 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3955 Monk Chiang <sh.chiang04@gmail.com>
3956 Chung-Ju Wu <jasonwucj@gmail.com>
3958 * config/nds32/nds32-protos.h
3959 (nds32_expand_load_multiple): New arguments.
3960 (nds32_expand_store_multiple): Ditto.
3961 (nds32_valid_multiple_load_store): Rename ...
3962 (nds32_valid_multiple_load_store_p): ... to this.
3963 * config/nds32/nds32-memory-manipulation.c
3964 (nds32_expand_load_multiple): Refine implementation.
3965 (nds32_expand_store_multiple): Ditto.
3966 * config/nds32/nds32-multiple.md
3967 (load_multiple): Update nds32_expand_load_multiple interface.
3968 (store_multiple): Update nds32_expand_store_multiple interface.
3969 * config/nds32/nds32-predicates.c
3970 (nds32_valid_multiple_load_store): Rename ...
3971 (nds32_valid_multiple_load_store_p): ... to this and refine
3973 * config/nds32/predicates.md
3974 (nds32_load_multiple_and_update_address_operation): New predicate.
3975 (nds32_store_multiple_and_update_address_operation): New predicate.
3977 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3978 Chung-Ju Wu <jasonwucj@gmail.com>
3980 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3981 (combo): New attribute.
3982 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3984 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3986 * config/nds32/nds32.opt: Change -mcmodel= default value.
3988 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3989 Monk Chiang <sh.chiang04@gmail.com>
3990 Chung-Ju Wu <jasonwucj@gmail.com>
3992 * config/nds32/constants.md (unspec_element): New enum.
3993 * config/nds32/constraints.md (Umw): New constraint.
3994 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3995 * config/nds32/nds32-intrinsic.md: Likewise.
3996 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3997 (nds32_valid_smw_lwm_base_p): New.
3998 (nds32_output_smw_single_word): New.
3999 (nds32_output_lmw_single_word): New.
4000 (nds32_expand_unaligned_load): New.
4001 (nds32_expand_unaligned_store): New.
4002 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
4003 (nds32_output_smw_single_word): Declare.
4004 (nds32_output_lmw_single_word): Declare.
4005 (nds32_expand_unaligned_load): Declare.
4006 (nds32_expand_unaligned_store): Declare.
4007 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
4008 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
4009 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
4010 NDS32_BUILTIN_UASTORE_DW.
4011 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
4014 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
4015 Kito Cheng <kito.cheng@gmail.com>
4016 Chung-Ju Wu <jasonwucj@gmail.com>
4018 * config/nds32/nds32-intrinsic.c
4019 (nds32_expand_builtin_null_ftype_reg): Delete.
4020 (nds32_expand_builtin_reg_ftype_imm): Ditto.
4021 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
4022 (nds32_read_argument): New.
4023 (nds32_legitimize_target): Ditto.
4024 (nds32_legitimize_argument): Ditto.
4025 (nds32_check_constant_argument): Ditto.
4026 (nds32_expand_unop_builtin): Ditto.
4027 (nds32_expand_unopimm_builtin): Ditto.
4028 (nds32_expand_binop_builtin): Ditto.
4029 (nds32_builtin_decl_impl): Ditto.
4030 (builtin_description): Ditto.
4031 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
4032 (nds32_init_builtins_impl): Ditto.
4033 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
4034 (nds32_builtin_decl): New.
4035 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
4036 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
4038 2018-03-02 Jeff Law <law@redhat.com>
4040 * reorg.c (stop_search_p): Handle DEBUG_INSN.
4041 (redundant_insn, fill_simple_delay_slots): Likewise.
4042 (fill_slots_from_thread): Likewise.
4043 * resource.c (mark_referenced_resources): Likewise.
4044 (mark_set_resources, find_dead_or_set_registers): Likewise.
4046 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4048 * substring-locations.h (format_warning_va): Formatting fix for
4050 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
4052 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
4053 * substring-locations.c: Include intl.h.
4054 (format_warning_va): Turned into small wrapper around
4055 format_warning_n_va, renamed to ...
4056 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
4057 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
4059 (format_warning_at_substring_n): New function.
4060 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
4061 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
4062 format_warning_at_substring with just a shorter name instead of
4063 const function pointer.
4064 (fmtwarn_n): New function.
4065 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
4066 appropriate, get rid of all the fmtstr temporaries, move conditionals
4067 with G_() wrapped string literals directly into fmtwarn arguments,
4068 cast dir.len to (int), formatting fixes.
4070 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
4072 * doc/invoke.texi: Remove "Cilk Plus" references.
4074 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4075 Richard Biener <rguenther@suse.de>
4078 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
4079 for error or warning attributes if CALL_FROM_THUNK_P is set.
4082 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4085 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
4086 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
4089 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
4090 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
4092 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
4093 instead of -1U in last predictors element's probability member.
4095 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
4098 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
4099 arguments if they are comparable.
4101 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4103 PR tree-optimization/84634
4104 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
4105 masks and masked_loop_p with a single loop_masks, making sure it's
4106 null for bb vectorization.
4108 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
4110 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
4111 (vect_analyze_data_ref_access): Use loop->safe_len rather than
4112 loop->force_vectorize to check whether there is no alias.
4114 2018-03-02 Jakub Jelinek <jakub@redhat.com>
4117 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
4119 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
4121 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
4122 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
4123 instead of a loop around prev_real_insn.
4124 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
4128 * config/i386/i386.c (ix86_print_operand): Use conditional
4129 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
4132 2018-03-02 Richard Biener <rguenther@suse.de>
4134 PR tree-optimization/84427
4135 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
4136 (bitmap_set_subtract_values): Rewrite to handle multiple
4139 (prune_clobbered_mems): Likewise.
4140 (phi_translate): Take edge instead of pred/phiblock.
4141 (phi_translate_1): Likewise.
4142 (phi_translate_set): Likewise. Insert all translated
4143 exprs for a value into the set, keeping possibly multiple
4144 expressions per value.
4145 (compute_antic_aux): Adjust for phi_translate changes.
4146 When intersecting union the expressions and prune those
4147 not in the final value set, keeping possibly multiple
4148 expressions per value. Do not use value-insertion
4149 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
4150 all expressions. Add verification that the value-sets
4151 only shrink during iteration.
4152 (compute_partial_antic_aux): Adjust for the phi_translate changes.
4153 (do_pre_regular_insertion): Likewise.
4154 (do_pre_partial_partial_insertion): Likewise.
4156 2018-03-02 Richard Biener <rguenther@suse.de>
4159 * config/darwin.c (saved_debug_info_level): New static global.
4160 (darwin_asm_lto_start): Disable debug info generation for LTO out.
4161 (darwin_asm_lto_end): Restore debug info generation settings.
4163 2018-03-01 Martin Liska <mliska@suse.cz>
4166 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
4169 2018-03-01 Richard Biener <rguenther@suse.de>
4172 * dwarf2out.c (gen_variable_die): Properly handle late VLA
4173 type annotation with LTO when debug was disabled at compile-time.
4175 2018-03-01 Matthew Fortune <mfortune@gmail.com>
4177 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
4179 (mips_final_postscan_insn): Likewise.
4181 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
4183 PR rtl-optimization/84528
4184 * alias.c (init_alias_target): Add commentary.
4185 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
4186 a unique base value if the frame pointer is not eliminated
4187 to the stack pointer.
4189 2018-03-01 Tom de Vries <tom@codesourcery.com>
4191 PR rtl-optimization/83327
4192 * lra-int.h (hard_regs_spilled_into): Declare.
4193 * lra.c (hard_regs_spilled_into): Define.
4194 (init_reg_info): Init hard_regs_spilled_into.
4195 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
4196 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
4197 (process_bb_lives): Handle hard_regs_spilled_into.
4198 (lra_create_live_ranges_1): Before doing liveness propagation, clear
4199 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
4201 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
4203 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
4204 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
4205 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
4206 * config/rs6000/aix72.h: New file.
4208 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4210 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4211 instead of warning_at with conditional singular and plural messages
4215 * stor-layout.c (update_alignment_for_field): For
4216 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4217 && !DECL_PACKED (field), do the alignment update, just use
4218 only desired_align instead of MAX (type_align, desired_align)
4220 (place_field): Don't do known_align < desired_align handling
4221 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4222 is non-NULL, instead do it after rli->prev_field handling and
4223 only if not within a bitfield word. For DECL_PACKED (field)
4224 use type_align of BITS_PER_UNIT.
4226 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4228 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4229 superfluous parentheses and trailing spaces.
4231 2018-02-28 Richard Biener <rguenther@suse.de>
4233 PR tree-optimization/84584
4234 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4235 SCoPs with fake exit edge.
4237 2018-02-28 Martin Liska <mliska@suse.cz>
4240 * timevar.c (timer::print): Fix format to properly print 100%
4243 2018-02-28 Richard Biener <rguenther@suse.de>
4246 * genmatch.c (capture_info::walk_match): Do not mark
4247 captured expressions without operands as expr_p given
4248 they act more like predicates and should be subject to
4249 "lost tail" side-effect preserving.
4251 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4253 PR rtl-optimization/81611
4254 * auto-inc-dec.c (attempt_change): Move dead note from
4255 mem_insn if it's the next use of regno
4256 (find_address): Take address use of reg holding
4257 non-incremented value. Add parm to limit search to the named
4259 (merge_in_block): Attempt to use a mem insn that is the next
4260 use of the original regno.
4262 2018-02-27 Martin Sebor <msebor@redhat.com>
4265 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4266 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4268 2018-02-27 Martin Sebor <msebor@redhat.com>
4270 PR translation/84207
4271 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4272 n argument to unsigned HOST_WIDE_INT.
4273 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4274 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4275 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4276 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4278 2018-02-27 Richard Biener <rguenther@suse.de>
4280 PR tree-optimization/84512
4281 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4282 Do not use the estimate returned from record_stmt_cost for
4283 the scalar iteration cost but sum properly using add_stmt_cost.
4285 2018-02-27 Richard Biener <rguenther@suse.de>
4287 PR tree-optimization/84466
4288 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4289 Adjust last change to less strictly validate use operands.
4291 2018-02-27 Martin Liska <mliska@suse.cz>
4293 PR gcov-profile/84548
4294 * gcov.c (process_file): Allow partial overlap and consider it
4295 also as group functions.
4296 (output_lines): Properly calculate range of lines for a group.
4298 2018-02-27 Martin Liska <mliska@suse.cz>
4300 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4301 'ggc' suffixes. Change first column width.
4302 (timer::print): Fix formatting of the column.
4304 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4306 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4307 preserve inline entry blocks for the sake of debug inline
4308 entry point markers alone.
4309 (remove_unused_locals): Suggest in comments a better place to
4310 force the preservation of inline entry blocks that are
4311 otherwise unused, but do not preserve them.
4313 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4315 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4317 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4320 * config/i386/constraints.md (Bs): Replace
4321 ix86_indirect_branch_register with
4322 TARGET_INDIRECT_BRANCH_REGISTER.
4324 * config/i386/i386.md (indirect_jump): Likewise.
4325 (tablejump): Likewise.
4326 (*sibcall_memory): Likewise.
4327 (*sibcall_value_memory): Likewise.
4328 Peepholes of indirect call and jump via memory: Likewise.
4329 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4330 (*sibcall_value_GOT_32): Likewise.
4331 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4332 (GOT_memory_operand): Likewise.
4333 (call_insn_operand): Likewise.
4334 (sibcall_insn_operand): Likewise.
4335 (GOT32_symbol_operand): Likewise.
4336 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4338 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4340 PR rtl-optimization/83496
4341 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4342 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4343 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4344 redundant insn, if any.
4345 (relax_delay_slots): Likewise.
4346 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4348 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4350 PR tree-optimization/83965
4351 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4352 that grouped statements are part of a reduction chain. Return
4353 true if the statement is not marked as a reduction itself but
4355 (vect_recog_dot_prod_pattern): Don't check whether the statement
4356 is part of a group here.
4357 (vect_recog_sad_pattern): Likewise.
4358 (vect_recog_widen_sum_pattern): Likewise.
4360 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4363 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4365 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4368 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4370 (ix86_output_indirect_function_return): New prototype.
4371 (ix86_split_simple_return_pop_internal): Likewise.
4372 * config/i386/i386.c (indirect_return_via_cx): New.
4373 (indirect_return_via_cx_bnd): Likewise.
4374 (indirect_thunk_name): Handle return va CX_REG.
4375 (output_indirect_thunk_function): Create alias for
4376 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4377 (ix86_output_indirect_jmp): Remove the bool argument.
4378 (ix86_output_indirect_function_return): New function.
4379 (ix86_split_simple_return_pop_internal): Likewise.
4380 * config/i386/i386.md (*indirect_jump): Don't pass false
4381 to ix86_output_indirect_jmp.
4382 (*tablejump_1): Likewise.
4383 (simple_return_pop_internal): Change it to define_insn_and_split.
4384 Call ix86_split_simple_return_pop_internal to split it for
4386 (simple_return_indirect_internal): Call
4387 ix86_output_indirect_function_return instead of
4388 ix86_output_indirect_jmp.
4390 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4393 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4394 memset and value initialization afterwards.
4396 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4398 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4400 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4403 * common/config/aarch64/aarch64-common.c
4404 (aarch_option_optimization_table[]): Switch
4405 off fomit-frame-pointer
4407 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4408 Chung-Ju Wu <jasonwucj@gmail.com>
4410 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4412 (store_multiple): Ditto.
4414 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4416 * config.gcc: Add --with-cpu support for nds32 target.
4417 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4418 * config/nds32/nds32.opt: Add -mcpu= option.
4420 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4422 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4423 isel=yes): Warn for these deprecated options.
4425 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4427 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4428 ISA_2_5_MASKS_EMBEDDED.
4430 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4432 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4433 p->max as pointers rather than using iterative_hash_expr.
4435 2018-02-23 Carl Love <cel@us.ibm.com>
4437 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4438 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4440 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4441 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4442 P8V_BUILTIN_VEC_VUNSIGNED2.
4444 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4447 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4448 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4450 * lra-constraints.c (curr_insn_transform): Set up
4451 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4453 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4454 * lra-eliminations.c (spill_pseudos): Ditto.
4455 (process_insn_for_elimination): Ditto.
4456 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4457 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4458 LRA_NON_CLOBBERED_ALT.
4460 2018-02-22 Martin Sebor <msebor@redhat.com>
4462 PR tree-optimization/84480
4463 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4464 to maybe_diag_stxncpy_trunc. Call it.
4465 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4466 from gimple_fold_builtin_strcpy. Print inlining stack.
4467 (handle_builtin_stxncpy): Print inlining stack.
4468 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4470 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4473 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4474 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4475 and -fcheck-pointer-bounds are used together.
4476 (indirect_thunk_prefix): New enum.
4477 (indirect_thunk_need_prefix): New function.
4478 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4479 "_nt" instead of "_bnd" for NOTRACK prefix.
4480 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4481 (output_indirect_thunk_function): Likewise.
4483 (ix86_code_end): Update output_indirect_thunk_function calls.
4484 (ix86_output_indirect_branch_via_reg): Replace
4485 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4486 (ix86_output_indirect_branch_via_push): Likewise.
4487 (ix86_output_function_return): Likewise.
4488 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4489 incompatible with -fcf-protection=branch and
4490 -fcheck-pointer-bounds.
4492 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4495 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4496 Change gcc_assert call to output_operand_lossage.
4498 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4500 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4502 2018-02-22 DJ Delorie <dj@redhat.com>
4503 Sebastian Perta <sebastian.perta@renesas.com>
4504 Oleg Endo <olegendo@gcc.gnu.org>
4506 * config/rx/rx.c (rx_rtx_costs): New function.
4507 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4509 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4511 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4513 2018-02-22 Martin Liska <mliska@suse.cz>
4516 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4517 Add "native" as a possible value.
4519 2018-02-22 Martin Liska <mliska@suse.cz>
4522 * config/i386/i386.c (ix86_option_override_internal):
4523 Add "native" as a possible value for -march and -mtune.
4525 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4528 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4529 to all type variants.
4531 PR tree-optimization/84503
4532 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4533 width as info->bitpos + info->bitsize - start.
4534 (merged_store_group::merge_overlapping): Simplify width computation.
4535 (check_no_overlap): New function.
4536 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4537 start + width and last_order of the group, fail if check_no_overlap
4539 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4540 to group if check_no_overlap fails.
4542 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4544 * config/rs6000/altivec.md: Delete contraint arguments to
4545 define_expand, define_split, and define_peephole2, and in
4546 define_insn_and_split if always unused.
4547 * config/rs6000/darwin.md: Ditto.
4548 * config/rs6000/dfp.md: Ditto.
4549 * config/rs6000/rs6000.md: Ditto.
4550 * config/rs6000/sync.md: Ditto.
4551 * config/rs6000/vector.md: Ditto.
4552 * config/rs6000/vsx.md: Ditto.
4554 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4556 * config/rs6000/altivec.md: Write output control strings as braced
4557 blocks instead of double-quoted strings.
4558 * config/rs6000/darwin.md: Ditto.
4559 * config/rs6000/rs6000.md: Ditto.
4560 * config/rs6000/vector.md: Ditto.
4561 * config/rs6000/vsx.md: Ditto.
4563 2018-02-21 Jason Merrill <jason@redhat.com>
4565 PR c++/84314 - ICE with templates and fastcall attribute.
4566 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4568 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4570 * ipa-cp.c (determine_versionability): Fix comment typos.
4572 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4575 * ipa-cp.c (determine_versionability): Do not version functions caling
4578 2018-02-21 Martin Liska <mliska@suse.cz>
4581 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4582 Add "native" as a possible value.
4583 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4584 the macro when native cpu detection is available.
4586 2018-02-21 Martin Liska <mliska@suse.cz>
4589 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4590 Add "native" as a possible value.
4591 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4592 when native cpu detection is available.
4594 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4595 Martin Sebor <msebor@redhat.com>
4597 PR tree-optimization/84478
4598 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4600 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4601 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4602 support which is conservatively correct, for 2 only stay conservative
4603 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4604 argument to the 2 argument get_range_strlen, adjust 6 arg
4605 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4607 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4608 (gimple_fold_builtin_strlen): Pass true as last argument to
4611 2018-02-20 Martin Sebor <msebor@redhat.com>
4614 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4615 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4616 (builtin_memref::builtin_memref): Factor out parts into
4617 set_base_and_offset and call it.
4619 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4622 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4623 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4624 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4625 search at the associated MODE_INT.
4627 2018-02-20 Jeff Law <law@redhat.com>
4630 PR tree-optimization/81592
4632 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4633 for range data rather than using global data.
4634 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4635 range data rather than using global data.
4636 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4637 pass it to children as needed.
4638 (struct directive::fmtresult): Similarly.
4639 (struct directive::set_width): Similarly.
4640 (struct directive::set_precision): Similarly.
4641 (format_integer, format_directive, parse_directive): Similarly.
4642 (format_none): Accept unnamed vr_values parameter.
4643 (format_percent, format_floating, format_character): Similarly.
4644 (format_string, format_plain): Similarly.
4645 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4646 the EVRP range analyzer for range data rather than using global data.
4647 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4648 gimple-ssa-evrp-analyze.h
4649 (class sprintf_dom_walker): Add after_dom_children member function.
4650 Add evrp_range_analyzer member.
4651 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4652 range analyzer as needed.
4653 (sprintf_dom_walker::after_dom_children): New member function.
4654 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4656 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4657 (evrp_range_analyzer::pop_to_marker): Likewise.
4659 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4661 PR tree-optimization/84419
4662 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4663 with the required type if its current type is compatible but
4666 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4669 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4670 after vectorization.
4672 2018-02-20 Martin Liska <mliska@suse.cz>
4675 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4676 possible values if we don't have a hint.
4678 2018-02-20 Martin Liska <mliska@suse.cz>
4682 * final.c (shorten_branches): Build align_tab array with one
4684 * opts.c (finish_options): Add alignment option limit check.
4685 (MAX_CODE_ALIGN): Likewise.
4686 (MAX_CODE_ALIGN_VALUE): Likewise.
4687 * doc/invoke.texi: Document maximum allowed option value for
4688 all -falign-* options.
4690 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4693 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4694 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4695 * var-tracking.c (emit_note_insn_var_location): Remove all references
4696 to NOTE_INSN_CALL_ARG_LOCATION.
4697 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4698 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4699 Use copy_rtx_if_shared.
4700 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4701 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4702 (dwarf2out_var_location): Remove handling of
4703 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4705 * final.c (final_scan_insn): Remove all references to
4706 NOTE_INSN_CALL_ARG_LOCATION.
4707 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4708 before dumping final insns.
4709 * except.c (emit_note_eh_region_end): Remove all references to
4710 NOTE_INSN_CALL_ARG_LOCATION.
4711 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4712 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4713 * config/arc/arc.c (hwloop_optimize): Likewise.
4714 * config/arm/arm.c (create_fix_barrier): Likewise.
4715 * config/s390/s390.c (s390_chunkify_start): Likewise.
4716 * config/sh/sh.c (find_barrier): Likewise.
4717 * config/i386/i386.c (rest_of_insert_endbranch,
4718 ix86_seh_fixup_eh_fallthru): Likewise.
4719 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4720 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4721 * config/frv/frv.c (frv_function_prologue): Likewise.
4722 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4724 (note_outside_basic_block_p): Remove all references to
4725 NOTE_INSN_CALL_ARG_LOCATION.
4726 * gengtype.c (adjust_field_rtx_def): Likewise.
4727 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4729 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4730 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4733 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4736 PR tree-optimization/84452
4737 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4738 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4741 2018-02-19 Martin Liska <mliska@suse.cz>
4744 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4746 2018-02-19 Martin Liska <mliska@suse.cz>
4747 Richard Sandiford <richard.sandiford@linaro.org>
4749 PR tree-optimization/82491
4750 * gimple-fold.c (get_base_constructor): Make earlier bail out
4753 2018-02-19 Carl Love <cel@us.ibm.com>
4755 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4756 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4758 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4759 P8V_BUILTIN_VEC_NEG.
4761 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4763 * config/rl78/rl78.md (movdf): New define expand.
4765 2018-02-19 Martin Liska <mliska@suse.cz>
4768 * doc/invoke.texi: Fix typo.
4769 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4771 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4773 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4774 handle rs6000_single_float and rs6000_double_float specially for
4777 2018-02-16 Jeff Law <law@redhat.com>
4779 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4782 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4785 * ipa-inline.c (inline_small_functions): Fix a typo.
4787 2018-02-16 Nathan Sidwell <nathan@acm.org>
4789 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4791 2018-02-16 Carl Love <cel@us.ibm.com>
4793 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4794 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4795 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4796 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4797 expansion to P8V_BUILTIN_VEC_FLOAT2.
4799 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4801 PR rtl-optimization/70023
4802 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4803 src_regno into account.
4805 2018-02-16 Carl Love <cel@us.ibm.com>
4807 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4808 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4809 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4810 * config/rs6000/rs6000.c: Remove case statements for
4811 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4812 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4813 and P9V_BUILTIN_VEC_VINSERT4B.
4814 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4815 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4816 * config/rs6000/vsx.md:
4817 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4820 2018-02-16 Carl Love <cel@us.ibm.com>
4822 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4824 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4826 * config/rs6000/rs6000-c.c: Add the definitions for
4827 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4828 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4829 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4830 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4831 definition for insert4b and define insn *insert3b_internal.
4832 * doc/extend.texi: Add documentation for vec_extract4b.
4834 2018-02-16 Nathan Sidwell <nathan@acm.org>
4836 * doc/extend.texi (Backwards Compatibility): Mention friend
4837 injection. Note for-scope is deprecated.
4838 * doc/invoke.texi (-ffriend-injection): Deprecate.
4840 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4842 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4843 that moved to I2, also allow destinations that are a paradoxical
4844 subreg (instead of a normal reg).
4846 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4849 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4852 2018-02-16 Richard Biener <rguenther@suse.de>
4854 PR tree-optimization/84037
4855 PR tree-optimization/84016
4857 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4858 Adjust vec_construct for the fact we need additional higher latency
4859 128bit inserts for AVX256 and AVX512 vector builds.
4860 (ix86_add_stmt_cost): Scale vector construction cost for
4863 2018-02-16 Richard Biener <rguenther@suse.de>
4865 PR tree-optimization/84417
4866 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4867 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4868 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4870 2018-02-16 Martin Liska <mliska@suse.cz>
4873 * internal-fn.def (ASAN_CHECK): Set proper flags.
4874 (ASAN_MARK): Likewise.
4876 2018-02-16 Julia Koval <julia.koval@intel.com>
4878 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4879 from PTA_CANNONLAKE.
4881 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4884 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4885 Use ++iter rather than iter++ for std::list iterators.
4886 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4887 defer deleting them until all nodes in the forest are processed. Do
4888 free even leaf nodes. Change to_process into auto_vec.
4891 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4892 * vec.h (vec_default_construct): Use memset instead of placement new
4893 if BROKEN_VALUE_INITIALIZATION is defined.
4894 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4895 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4898 PR rtl-optimization/83723
4899 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4900 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4901 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4903 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4905 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4907 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4909 PR rtl-optimization/81443
4910 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4911 from inner REGs to paradoxical SUBREGs.
4913 2018-02-16 Richard Biener <rguenther@suse.de>
4915 PR tree-optimization/84399
4916 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4917 For operands we can analyze at their definition make sure we can
4918 analyze them at each use as well.
4920 2018-02-16 Richard Biener <rguenther@suse.de>
4922 PR tree-optimization/84190
4923 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4924 volatile accesses if the decl isn't volatile.
4926 2018-02-15 Jason Merrill <jason@redhat.com>
4928 PR c++/84314 - ICE with templates and fastcall attribute.
4929 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4930 TYPE_CANONICAL on an existing type.
4932 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4934 PR tree-optimization/84383
4935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4936 dstoff nor call operand_equal_p if dstbase is NULL.
4938 PR tree-optimization/84334
4939 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4940 also a CONSTANT_CLASS_P, punt.
4942 2018-02-14 Jim Wilson <jimw@sifive.com>
4944 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4945 first SMALL_OPERAND check. New local min_second_step. Move assert
4946 to where locals are set. Add TARGET_RVC support.
4947 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4949 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4951 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4953 2018-02-14 Martin Sebor <msebor@redhat.com>
4955 PR tree-optimization/83698
4956 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4957 arrays constrain the offset range to their bounds.
4958 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4959 (builtin_access::overlap): Avoid setting the size of overlap if it's
4961 (maybe_diag_overlap): Also consider arrays when deciding what values
4962 of offsets to include in diagnostics.
4964 2018-02-14 Martin Sebor <msebor@redhat.com>
4967 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4968 that correspond to the kind of a declaration.
4970 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4973 * config/pa/pa.md: Load address of PIC label using the linkage table
4974 if the label is nonlocal.
4976 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4978 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4979 warning message if user requests -maltivec=be.
4980 * doc/invoke.texi: Document deprecation of -maltivec=be.
4982 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4985 * config/rs6000/rs6000-c.c: Update definitions for
4986 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4987 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4989 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4992 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4993 add _get_ssp intrinsics. Remove argument from
4994 __builtin_ia32_rdssp[d|q].
4995 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4996 * config/i386/i386-builtin.def: Remove argument from
4997 __builtin_ia32_rdssp[d|q].
4998 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4999 ix86_expand_special_args_builtin for _rdssp[d|q].
5000 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
5001 Clear register before usage.
5002 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
5003 Add documentation for new _get_ssp and _inc_ssp intrinsics.
5005 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
5007 PR tree-optimization/84357
5008 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
5009 operand 1 of an ARRAY_REF too.
5011 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
5014 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
5015 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5017 (set_of_reg): New struct.
5018 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
5019 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
5020 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
5022 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
5023 Split into bitclr, bitset, bitinvert patterns if appropriate.
5024 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
5025 use rx_fuse_in_memory_bitop.
5026 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
5027 to named insn, correct maximum insn length.
5029 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
5032 * machmode.def: Define a complex mode for PARTIAL_INT.
5033 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
5035 * doc/rtl.texi: Document CSPImode.
5036 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
5038 (msp430_hard_regno_nregs_with_padding): Likewise.
5040 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
5043 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
5045 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
5047 PR rtl-optimization/84169
5048 * combine.c (try_combine): New variable split_i2i3. Set it to true if
5049 we generated a parallel as new i3 and we split that to new i2 and i3
5050 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
5051 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
5052 those to i2, not i1. Partially rewrite this scan code.
5054 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5057 * stor-layout.c (place_field): For variable length fields, adjust
5058 offset_align afterwards not just based on the field's alignment,
5059 but also on the size.
5062 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
5063 of exps and logs in the use_exp2 case.
5065 2018-02-13 Jeff Law <law@redhat.com>
5067 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
5070 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
5073 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
5077 * common.opt (gas-loc-support, gas-locview-support): New.
5078 (ginline-points, ginternal-reset-location-views): New.
5079 * doc/invoke.texi: Document them. Use @itemx where intended.
5080 (gvariable-location-views): Adjust.
5081 * target.def (reset_location_view): New.
5082 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
5083 (TARGET_RESET_LOCATION_VIEW): New.
5084 * doc/tm.texi: Rebuilt.
5085 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
5086 (dwarf2out_default_as_locview_support): New.
5087 (output_asm_line_debug_info): Use option variables.
5088 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
5089 (output_loc_list): Likewise.
5090 (add_high_low_attributes): Check option variables.
5091 Don't output entry view attribute in strict mode.
5092 (gen_inlined_subroutine_die): Check option variables.
5093 (dwarf2out_inline_entry): Likewise.
5094 (init_sections_and_labels): Likewise.
5095 (dwarf2out_early_finish): Likewise.
5096 (maybe_reset_location_view): New, from...
5097 (dwarf2out_var_location): ... here. Call it.
5098 * debug.h (dwarf2out_default_as_loc_support): Declare.
5099 (dwarf2out_default_as_locview_support): Declare.
5100 * hooks.c (hook_int_rtx_insn_0): New.
5101 * hooks.h (hook_int_rtx_insn_0): Declare.
5102 * toplev.c (process_options): Take -gas-loc-support and
5103 -gas-locview-support from dwarf2out. Enable
5104 -gvariable-location-views by default only with locview
5105 assembler support. Enable -ginternal-reset-location-views by
5106 default only if the target defines the corresponding hook.
5107 Enable -ginline-points by default if location views are
5108 enabled; force it disabled if statement frontiers are
5110 * tree-inline.c (expand_call_inline): Check option variables.
5111 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5113 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
5115 PR tree-optimization/84321
5116 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
5117 handling. Also check whether the anti-range contains any values
5118 that satisfy the mask; switch to a VR_RANGE if not.
5120 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
5123 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
5125 2018-02-13 Martin Jambor <mjambor@suse.cz>
5128 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
5129 of call statements, also set location of a load to a temporary.
5131 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
5133 * config/rl78/rl78.c (add_vector_labels): New function.
5134 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
5135 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
5136 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
5137 which checks that no arguments are passed.
5138 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
5139 * doc/extend.texi: Documentation for the new attribute.
5141 2018-02-13 Andreas Schwab <schwab@suse.de>
5143 * config/riscv/linux.h (CPP_SPEC): Define.
5145 2018-02-13 Jakub Jelinek <jakub@redhat.com>
5148 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
5149 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
5150 OPTION_MASK_ISA_AES as first argument to def_builtin_const
5151 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
5152 instead of OPTION_MASK_ISA_PCLMUL as first argument to
5153 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
5154 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
5155 temporarily for AES and PCLMUL builtins.
5157 PR tree-optimization/84339
5158 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
5159 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
5163 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
5164 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
5165 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
5167 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
5169 * omp-simd-clone.h: New file.
5170 * omp-simd-clone.c: Include omp-simd-clone.h.
5171 (expand_simd_clones): No longer static.
5172 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
5173 cgraph.h and omp-simd-clone.h.
5174 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
5175 (vect_recog_widen_shift_pattern): Formatting fix.
5176 (vect_pattern_recog_1): Don't check optab for calls.
5179 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
5180 operands[2] into a REG before using gen_lowpart on it.
5182 2018-02-12 Jeff Law <law@redhat.com>
5185 * config/sh/sh.c (find_barrier): Consider a sibling call
5188 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
5189 successfully back substituting a reg.
5191 2018-02-12 Richard Biener <rguenther@suse.de>
5193 PR tree-optimization/84037
5194 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
5195 parameter, move visited init to caller.
5196 (vect_slp_analyze_operations): Separate cost from validity
5197 check, initialize visited once for all instances.
5198 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
5200 * tree-vect-stmts.c (vect_model_simple_cost): Make early
5202 (vect_model_promotion_demotion_cost): Likewise.
5203 (vectorizable_bswap): Guard cost modeling with !slp_node
5204 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
5206 (vectorizable_call): Likewise.
5207 (vectorizable_conversion): Likewise.
5208 (vectorizable_assignment): Likewise.
5209 (vectorizable_shift): Likewise.
5210 (vectorizable_operation): Likewise.
5211 (vectorizable_store): Likewise.
5212 (vectorizable_load): Likewise.
5213 (vectorizable_condition): Likewise.
5214 (vectorizable_comparison): Likewise.
5216 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5219 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5220 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5221 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5223 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5226 * params.def (inline-min-speedup): Increase from 8 to 15.
5227 (max-inline-insns-auto): Decrease from 40 to 30.
5228 * ipa-split.c (consider_split): Add some buffer for function to
5229 be considered inlining candidate.
5230 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5233 2018-02-12 Richard Biener <rguenther@suse.de>
5235 PR tree-optimization/84037
5236 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5237 matched stmts if we cannot swap the non-matched ones.
5239 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5241 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5242 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5243 _mm_maskz_scalef_round_ss): New intrinsics.
5244 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5245 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5246 __builtin_ia32_scalefss_round): Remove.
5247 (__builtin_ia32_scalefsd_mask_round,
5248 __builtin_ia32_scalefss_mask_round): New intrinsics.
5249 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5250 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5251 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5252 "<round_constraint>")): Changed to ...
5253 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5254 "<round_scalar_constraint>")): ... this.
5255 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5256 %0, %1, %2<round_op3>}"): Changed to ...
5257 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5258 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5259 %2<round_scalar_mask_op3>}"): ... this.
5260 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5262 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5264 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5265 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5266 (_mm_maskz_sqrt_round_ss): New intrinsics.
5267 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5268 (__builtin_ia32_sqrtsd_mask_round)
5269 (__builtin_ia32_sqrtss_mask_round): New builtins.
5270 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5271 (__builtin_ia32_sqrtss_round): Remove.
5272 (__builtin_ia32_sqrtsd_mask_round)
5273 (__builtin_ia32_sqrtss_mask_round): New builtins.
5274 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5275 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5276 ((match_operand:VF_128 1 "vector_operand"
5277 "xBm,<round_constraint>")): Changed to ...
5278 ((match_operand:VF_128 1 "vector_operand"
5279 "xBm,<round_scalar_constraint>")): ... this.
5280 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5281 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5282 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5283 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5284 %<iptr>1<round_scalar_mask_op3>}): ... this.
5285 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5286 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5288 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5291 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5292 Cast vec_cmpeq result to correct type.
5293 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5294 Cast vec_cmpgt result to correct type.
5296 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5298 * final.c (final_scan_insn_1): Renamed from...
5299 (final_scan_insn): ... this. New wrapper, to recover
5300 seen from the outermost call in recursive ones.
5301 * config/sparc/sparc.c (output_return): Drop seen from call.
5302 (output_sibcall): Likewise.
5303 * config/visium/visium.c (output_branch): Likewise.
5305 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5307 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5310 2018-02-10 Alan Modra <amodra@gmail.com>
5313 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5314 Specify LR as an input.
5316 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5319 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5320 remove_member_access_dummy_vars): New functions.
5321 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5322 lower_omp_1, execute_lower_omp): Use them.
5324 PR rtl-optimization/84308
5325 * shrink-wrap.c (spread_components): Release todo vector.
5327 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5329 PR rtl-optimization/57193
5330 * ira-color.c (struct allocno_color_data): Add member
5331 conflict_allocno_hard_prefs.
5332 (update_conflict_allocno_hard_prefs): New.
5333 (bucket_allocno_compare_func): Add a preference based on
5334 conflict_allocno_hard_prefs.
5335 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5336 (color_allocnos): Remove a dead code. Initiate
5337 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5339 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5342 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5343 constraint from =wa to wa. Avoid a subreg on the output operand,
5344 instead use a pseudo and subreg it in a move.
5345 (p9_xxbrd_<mode>): Changed to ...
5346 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5347 (p9_xxbrd_v2df): New expander.
5348 (p9_xxbrw_<mode>): Changed to ...
5349 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5350 (p9_xxbrw_v4sf): New expander.
5352 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5354 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5356 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5359 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5360 multiply in 32-bit mode.
5361 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5362 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5365 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5367 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5368 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5369 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5370 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5372 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5375 * dwarf2out.c (is_trivial_indirect_ref): New function.
5376 (dwarf2out_late_global_decl): Do not generate a location
5377 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5378 and that are not defined in the current unit.
5380 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5382 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5383 instead of a libcall for UNORDERED.
5385 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5388 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5389 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5391 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5394 * config/s390/s390.c (s390_set_current_function): Invoke
5395 s390_indirect_branch_settings also if fndecl didn't change.
5397 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5399 * config/rs6000/rs6000.md (blockage): Set length to zero.
5401 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5403 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5405 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5408 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5409 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5413 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5414 PARALLEL incoming that failed vt_get_decl_and_offset check.
5417 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5418 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5420 (get_variable_section): For decls in named .bss* sections pass true as
5421 second argument to bss_initializer_p.
5423 2018-02-09 Marek Polacek <polacek@redhat.com>
5424 Jakub Jelinek <jakub@redhat.com>
5427 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5428 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5429 Sync some changes from cxx_fold_indirect_ref.
5431 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5433 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5435 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5436 (BLOCK_INLINE_ENTRY_LABEL): New.
5437 (dwarf2out_var_location): Disregard inline entry markers.
5438 (inline_entry_data): New struct.
5439 (inline_entry_data_hasher): New hashtable type.
5440 (inline_entry_data_hasher::hash): New.
5441 (inline_entry_data_hasher::equal): New.
5442 (inline_entry_data_table): New variable.
5443 (add_high_low_attributes): Add DW_AT_entry_pc and
5444 DW_AT_GNU_entry_view attributes if a pending entry is found
5445 in inline_entry_data_table. Add old entry_pc attribute only
5446 if debug nonbinding markers are disabled.
5447 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5448 markers are enabled.
5449 (block_within_block_p, dwarf2out_inline_entry): New.
5450 (dwarf2out_finish): Check that no entries remained in
5451 inline_entry_data_table.
5452 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5453 (final_scan_insn, notice_source_line): Likewise.
5454 (rest_of_clean_state): Skip inline entry markers.
5455 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5457 * gimple.c (gimple_build_debug_inline_entry): New.
5458 * gimple.h (enum gimple_debug_subcode): Add
5459 GIMPLE_DEBUG_INLINE_ENTRY.
5460 (gimple_build_debug_inline_entry): Declare.
5461 (gimple_debug_inline_entry_p): New.
5462 (gimple_debug_nonbind_marker_p): Adjust.
5463 * insn-notes.def (INLINE_ENTRY): New.
5464 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5465 inline entry marker notes.
5466 (print_insn): Likewise.
5467 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5468 (INSN_DEBUG_MARKER_KIND): Likewise.
5469 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5470 * tree-inline.c (expand_call_inline): Build and insert
5471 debug_inline_entry stmt.
5472 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5473 inline entry blocks early, if nonbind markers are enabled.
5474 (dump_scope_block): Dump fragment info.
5475 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5476 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5477 (gimple_build_debug_inline_entry): New.
5478 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5479 Enable/disable inline entry points too.
5480 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5481 (DEBUG_INSN): Describe inline entry markers.
5483 * common.opt (gvariable-location-views): New.
5484 (gvariable-location-views=incompat5): New.
5485 * config.in: Rebuilt.
5486 * configure: Rebuilt.
5487 * configure.ac: Test assembler for view support.
5488 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5489 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5490 * dwarf2out.c (var_loc_view): New typedef.
5491 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5492 (dwarf2out_locviews_in_attribute): New.
5493 (dwarf2out_locviews_in_loclist): New.
5494 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5495 (enum dw_line_info_opcode): Add LI_adv_address.
5496 (struct dw_line_info_table): Add view.
5497 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5498 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5499 (zero_view_p): New variable.
5500 (ZERO_VIEW_P): New macro.
5501 (output_asm_line_debug_info): New.
5502 (struct var_loc_node): Add view.
5503 (add_AT_view_list, AT_loc_list): New.
5504 (add_var_loc_to_decl): Add view param. Test it against last.
5505 (new_loc_list): Add view params. Record them.
5506 (AT_loc_list_ptr): Handle loc and view lists.
5507 (view_list_to_loc_list_val_node): New.
5508 (print_dw_val): Handle dw_val_class_view_list.
5509 (size_of_die): Likewise.
5510 (value_format): Likewise.
5511 (loc_list_has_views): New.
5512 (gen_llsym): Set vl_symbol too.
5513 (maybe_gen_llsym, skip_loc_list_entry): New.
5514 (dwarf2out_maybe_output_loclist_view_pair): New.
5515 (output_loc_list): Output view list or entries too.
5516 (output_view_list_offset): New.
5517 (output_die): Handle dw_val_class_view_list.
5518 (output_dwarf_version): New.
5519 (output_compilation_unit_header): Use it.
5520 (output_skeleton_debug_sections): Likewise.
5521 (output_rnglists, output_line_info): Likewise.
5522 (output_pubnames, output_aranges): Update version comments.
5523 (output_one_line_info_table): Output view numbers in asm comments.
5524 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5525 Call maybe_gen_llsym.
5526 (loc_list_from_tree_1): Adjust.
5527 (add_AT_location_description): Create view list attribute if
5528 needed, check it's absent otherwise.
5529 (convert_cfa_to_fb_loc_list): Adjust.
5530 (maybe_emit_file): Call output_asm_line_debug_info for test.
5531 (dwarf2out_var_location): Reset views as needed. Precompute
5532 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5533 attribute. Set view.
5534 (new_line_info_table): Reset next view.
5535 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5536 (dwarf2out_source_line): Likewise. Output view resets and labels to
5537 the assembler, or select appropriate line info opcodes.
5538 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5539 (optimize_string_length): Catch it. Adjust.
5540 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5541 dw_val_class_view_list, and remove it if no longer needed.
5542 (hash_loc_list): Hash view numbers.
5543 (loc_list_hasher::equal): Compare them.
5544 (optimize_location_lists): Check whether a view list symbol is
5545 needed, and whether the locview attribute is present, and
5546 whether they match. Remove the locview attribute if no longer
5548 (index_location_lists): Call skip_loc_list_entry for test.
5549 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5550 Use output_dwarf_version.
5551 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5552 (struct dw_val_node): Add val_view_list.
5553 * final.c (SEEN_NEXT_VIEW): New.
5554 (set_next_view_needed): New.
5555 (clear_next_view_needed): New.
5556 (maybe_output_next_view): New.
5557 (final_start_function): Rename to...
5558 (final_start_function_1): ... this. Take pointer to FIRST,
5559 add SEEN parameter. Emit param bindings in the initial view.
5560 (final_start_function): Reintroduce SEEN-less interface.
5561 (final): Rename to...
5562 (final_1): ... this. Take SEEN parameter. Output final pending
5563 next view at the end.
5564 (final): Reintroduce seen-less interface.
5565 (final_scan_insn): Output pending next view before switching
5566 sections or ending a block. Mark the next view as needed when
5567 outputting variable locations. Notify debug backend of section
5568 changes, and of location view changes.
5569 (rest_of_handle_final): Adjust.
5570 * toplev.c (process_options): Autodetect value for debug variable
5571 location views option. Warn on incompat5 without -gdwarf-5.
5572 * doc/invoke.texi (gvariable-location-views): New.
5573 (gvariable-location-views=incompat5): New.
5574 (gno-variable-location-views): New.
5576 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5578 PR tree-optimization/84136
5579 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5580 that the result of find_edge is non-NULL.
5582 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5585 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5586 storing integer register in SImode. Fix cost of 256 and 512
5587 byte aligned SSE register store.
5589 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5591 * config/i386/i386.c (ix86_multiplication_cost): Fix
5592 multiplication cost for TARGET_AVX512DQ.
5594 2018-02-08 Marek Polacek <polacek@redhat.com>
5596 PR tree-optimization/84238
5597 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5600 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5602 PR tree-optimization/84265
5603 * tree-vect-stmts.c (vectorizable_store): Don't treat
5604 VMAT_CONTIGUOUS accesses as grouped.
5605 (vectorizable_load): Likewise.
5607 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5609 PR tree-optimization/81635
5610 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5611 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5612 (test_round_for_mask): New functions.
5613 (wide_int_cc_tests): Call test_round_for_mask.
5614 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5615 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5616 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5617 range returned by get_range_info.
5619 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5622 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5623 * symtab.c: Include builtins.h
5624 (symtab_node::output_to_lto_symbol_table_p): Move here
5625 from lto-streamer-out.c:output_symbol_p.
5626 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5627 (output_symbol_p): Move all logic to symtab.c
5628 (produce_symtab): Update.
5630 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5632 * config/s390/s390-opts.h (enum indirect_branch): Define.
5633 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5634 (s390_indirect_branch_via_thunk)
5635 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5636 (enum s390_indirect_branch_type): Define.
5637 * config/s390/s390.c (struct s390_frame_layout, struct
5638 machine_function): Remove.
5639 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5640 (indirect_branch_table_label_no, indirect_branch_table_name):
5642 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5643 (enum s390_indirect_branch_option): Define.
5644 (s390_return_addr_from_memory): New function.
5645 (s390_handle_string_attribute): New function.
5646 (s390_attribute_table): Add new attribute handler.
5647 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5648 (s390_indirect_branch_via_thunk): New function.
5649 (s390_indirect_branch_via_inline_thunk): New function.
5650 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5651 sibling call optimization for non z10 compiles.
5652 (s390_emit_call): Force indirect branch target to be a single
5653 register. Add r1 clobber for non-z10 compiles.
5654 (s390_emit_epilogue): Emit return jump via return_use expander.
5655 (s390_reorg): Handle JUMP_INSNs as execute targets.
5656 (s390_option_override_internal): Perform validity checks for the
5657 new command line options.
5658 (s390_indirect_branch_attrvalue): New function.
5659 (s390_indirect_branch_settings): New function.
5660 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5661 (s390_output_indirect_thunk_function): New function.
5662 (s390_code_end): Implement target hook.
5663 (s390_case_values_threshold): Implement target hook.
5664 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5666 * config/s390/s390.h (struct s390_frame_layout)
5667 (struct machine_function): Move here from s390.c.
5668 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5669 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5670 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5671 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5672 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5673 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5674 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5675 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5676 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5677 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5678 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5679 (mnemonic attribute): Add values which aren't recognized
5681 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5682 pattern for branch conversion. Fix mnemonic attribute.
5683 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5684 indirect branch via thunk if requested.
5685 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5686 ("*indirect_jump"): Disable for branch conversion using out of
5688 ("indirect_jump_via_thunk<mode>_z10")
5689 ("indirect_jump_via_thunk<mode>")
5690 ("indirect_jump_via_inlinethunk<mode>_z10")
5691 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5692 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5693 ("casesi_jump_via_inlinethunk<mode>_z10")
5694 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5695 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5696 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5697 ("*indirect2_jump"): Disable for branch conversion.
5698 ("casesi_jump"): Turn into expander and expand patterns for branch
5700 ("return_use"): New expander.
5701 ("*return"): Emit return via thunk and rename it to ...
5702 ("*return<mode>"): ... this one.
5703 * config/s390/s390.opt: Add new options and and enum for the
5706 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5708 * lra-constraints.c (match_reload): Unconditionally use
5709 gen_lowpart_SUBREG, rather than selecting between that
5710 and equivalent gen_rtx_SUBREG code.
5712 2018-02-08 Richard Biener <rguenther@suse.de>
5714 PR tree-optimization/84233
5715 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5716 changed flag instead of boguously re-using phi_inserted.
5718 2018-02-08 Martin Jambor <mjambor@suse.cz>
5720 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5721 static local variables.
5723 2018-02-08 Richard Biener <rguenther@suse.de>
5725 PR tree-optimization/84278
5726 * tree-vect-stmts.c (vectorizable_store): When looking for
5727 smaller vector types to perform grouped strided loads/stores
5728 make sure the mode is supported by the target.
5729 (vectorizable_load): Likewise.
5731 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5733 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5734 Increase LDP/STP opportunities by adding adjacent callee-saves.
5736 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5738 PR rtl-optimization/84068
5739 PR rtl-optimization/83459
5740 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5742 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5744 PR tree-optimization/84224
5745 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5746 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5749 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5752 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5753 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5754 count, remove one USE.
5756 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5758 * doc/install.texi (Configuration): Document the
5759 --with-long-double-format={ibm,ieee} PowerPC configuration
5763 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5764 Convert from define_expand to be define_insn_and_split. Rework
5765 float/double/_Float128 conversions to QI/HI/SImode to work with
5766 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5767 conversions to QI/HImode types did a store and then a load to
5768 truncate the value. For conversions to VSX registers, don't split
5769 the insn, instead emit the code directly. Use the code iterator
5770 any_fix to combine signed and unsigned conversions.
5771 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5772 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5773 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5774 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5775 (fix_<mode>di2_hw): Likewise.
5776 (fixuns_<mode>di2_hw): Likewise.
5777 (fix_<mode>si2_hw): Likewise.
5778 (fixuns_<mode>si2_hw): Likewise.
5779 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5780 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5781 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5782 fix<uns>_trunc<SFDF:mode>si2_p8.
5783 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5785 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5786 (fix<uns>_<mode>_mem): Likewise.
5787 (fctiw<u>z_<mode>_mem): Likewise.
5788 (fix<uns>_<mode>_mem): Likewise.
5789 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5790 the register allocator from doing a direct move to the GPRs to do
5791 a store, and instead use the ISA 3.0 store byte/half-word from
5792 vector register instruction. For IEEE 128-bit floating point,
5793 also optimize stores of 32-bit ints.
5794 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5796 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5798 * genextract.c (push_pathstr_operand): New function to support
5800 (walk_rtx): Call push_pathstr_operand.
5801 (print_path): Support [a-zA-Z].
5803 2018-02-07 Richard Biener <rguenther@suse.de>
5805 PR tree-optimization/84037
5806 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5807 (cse_and_gimplify_to_preheader): Declare.
5808 (vect_get_place_in_interleaving_chain): Likewise.
5809 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5811 (_loop_vec_info::~_loop_vec_info): Delete it.
5812 (cse_and_gimplify_to_preheader): New function.
5813 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5814 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5815 (vectorizable_load): Likewise. For grouped stores always base
5816 the IV on the first element.
5817 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5818 condition before gimplifying.
5820 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5822 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5823 *DIV_EXPR and *MOD_EXPR.
5825 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5828 * config/i386/i386.c (ix86_option_override_internal): Mask out
5829 the CF_SET bit when checking -fcf-protection.
5831 2018-02-07 Tom de Vries <tom@codesourcery.com>
5834 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5837 2018-02-07 Richard Biener <rguenther@suse.de>
5839 PR tree-optimization/84204
5840 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5843 PR tree-optimization/84205
5844 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5845 special-case isl_ast_op_zdiv_r.
5847 PR tree-optimization/84223
5848 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5849 Only add conditions from within the region.
5850 (gather_bbs::after_dom_children): Adjust.
5852 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5855 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5856 * config/avr/avr.md: Only post-reload split REG-REG moves if
5857 either register is GENERAL_REG_P.
5859 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5861 PR tree-optimization/84235
5862 * tree-ssa-scopedtables.c
5863 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5864 if the subtraction is performed in floating point type where NaNs are
5865 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5866 build 1. Formatting fix.
5868 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5871 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5872 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5873 and skip it regardless of bb boundaries. Use CALL_P macro,
5874 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5875 unnecessarily, formatting fix.
5877 2018-02-06 Michael Collison <michael.collison@arm.com>
5879 * config/arm/thumb2.md:
5880 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5881 (*thumb_mov_notscc): Ditto.
5883 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5886 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5887 unsigned_fix, not "s".
5889 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5891 * configure.ac (gcc_fn_eh_frame_ro): New function.
5892 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5893 correct .eh_frame permissions.
5894 * configure: Regenerate.
5896 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5898 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5901 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5903 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5904 Display warning message for -mno-speculate-indirect-jumps.
5906 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5908 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5910 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5912 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5914 PR tree-optimization/84225
5915 * tree-eh.c (find_trapping_overflow): Only call
5916 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5918 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5921 * config/i386/i386.c: Reimplement the check of possible options
5922 -mibt/-mshstk conbination. Change error messages.
5923 * doc/invoke.texi: Fix a typo: remove extra '='.
5925 2018-02-06 Marek Polacek <polacek@redhat.com>
5927 PR tree-optimization/84228
5928 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5930 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5933 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5934 emitted arch directives.
5935 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5936 __ARM_FEATURE_COPROC before changing architectures.
5938 2018-02-06 Richard Biener <rguenther@suse.de>
5940 * config/i386/i386.c (print_reg): Fix typo.
5941 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5943 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5945 * configure: Regenerate.
5947 2018-02-05 Martin Sebor <msebor@redhat.com>
5949 PR tree-optimization/83369
5950 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5953 2018-02-05 Martin Liska <mliska@suse.cz>
5955 * doc/invoke.texi: Cherry-pick upstream r323995.
5957 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5959 * ira.c (ira_init_register_move_cost): Adjust comment.
5961 2018-02-05 Martin Liska <mliska@suse.cz>
5963 PR gcov-profile/84137
5964 * doc/gcov.texi: Fix typo in documentation.
5966 2018-02-05 Martin Liska <mliska@suse.cz>
5968 PR gcov-profile/83879
5969 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5970 using dlopen functionality.
5972 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5974 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5975 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5976 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5977 _mm_maskz_range_round_ss): New intrinsics.
5978 (__builtin_ia32_rangesd128_round)
5979 (__builtin_ia32_rangess128_round): Remove.
5980 (__builtin_ia32_rangesd128_mask_round,
5981 __builtin_ia32_rangess128_mask_round): New builtins.
5982 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5983 __builtin_ia32_rangess128_round): Remove.
5984 (__builtin_ia32_rangesd128_mask_round,
5985 __builtin_ia32_rangess128_mask_round): New builtins.
5986 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5987 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5988 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5989 "<round_saeonly_constraint>")): Changed to ...
5990 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5991 "<round_saeonly_scalar_constraint>")): ... this.
5992 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5993 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5994 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5995 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5996 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5998 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
6000 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
6002 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
6003 Remove all values except native, 8540 and 8548.
6005 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
6007 * config/i386/i386.c (ix86_output_function_return): Pass
6008 INVALID_REGNUM, instead of -1, as invalid register number to
6009 indirect_thunk_name and output_indirect_thunk.
6011 2018-02-02 Julia Koval <julia.koval@intel.com>
6013 * config.gcc: Add -march=icelake.
6014 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
6015 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
6016 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
6017 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
6018 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
6019 (processor_target_table): Add icelake.
6020 (ix86_option_override_internal): Handle new PTAs.
6021 (get_builtin_code_for_version): Handle icelake.
6022 (M_INTEL_COREI7_ICELAKE): New.
6023 (fold_builtin_cpu): Handle icelake.
6024 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
6025 * doc/invoke.texi: Add -march=icelake.
6027 2018-02-02 Julia Koval <julia.koval@intel.com>
6029 * config/i386/i386.c (ix86_option_override_internal): Change flags type
6030 to wide_int_bitmask.
6031 * wide-int-bitmask.h: New.
6033 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
6036 * config/i386/i386.md: Replace Pmode with word_mode in
6037 builtin_setjmp_setup and builtin_longjmp to support x32.
6039 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
6043 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
6045 (rs6000_supported_cpu_names): New static variable.
6046 (linux_cpu_translation_table): Likewise.
6047 (elf_platform) <cpu>: Define new static variable and use it.
6048 Translate kernel AT_PLATFORM name to canonical name if needed.
6049 Error if platform name is unknown.
6051 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
6054 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
6056 2018-02-01 Jeff Law <law@redhat.com>
6059 * config/i386/i386.c (release_scratch_register_on_entry): Add new
6060 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
6061 the scratch if RELEASE_VIA_POP is false.
6062 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
6063 If we have to save a temporary register, decrement SIZE appropriately.
6064 Pass new arguments to release_scratch_register_on_entry.
6065 (ix86_adjust_stack_and_probe): Likewise.
6066 (ix86_emit_probe_stack_range): Pass new arguments to
6067 release_scratch_register_on_entry.
6069 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
6071 PR rtl-optimization/84157
6072 * combine.c (change_zero_ext): Use REG_P predicate in
6073 front of HARD_REGISTER_P predicate.
6075 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
6077 * config/avr/avr.c (avr_option_override): Move disabling of
6078 -fdelete-null-pointer-checks to...
6079 * common/config/avr/avr-common.c (avr_option_optimization_table):
6082 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6084 PR tree-optimization/81635
6085 * tree-data-ref.c (split_constant_offset_1): For types that
6086 wrap on overflow, try to use range info to prove that wrapping
6089 2018-02-01 Renlin Li <renlin.li@arm.com>
6092 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
6094 (aarch64_register_move_cost): Likewise.
6095 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
6097 (REG_CLASS_NAMES): Likewise.
6098 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
6099 TAILCALL_ADDR_REGS. Remove IP registers.
6100 * config/aarch64/aarch64.md (Ucs): Update register constraint.
6102 2018-02-01 Richard Biener <rguenther@suse.de>
6104 * domwalk.h (dom_walker::dom_walker): Add additional constructor
6105 for specifying RPO order and allow NULL for that.
6106 * domwalk.c (dom_walker::dom_walker): Likewise.
6107 (dom_walker::walk): Handle NULL RPO order.
6108 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
6110 (rewrite_update_dom_walker): Likewise.
6111 (mark_def_dom_walker): Likewise.
6113 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6115 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
6116 (aarch64_maybe_expand_sve_subreg_move): Declare.
6117 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
6118 * config/aarch64/predicates.md (aarch64_any_register_operand): New
6120 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
6121 that are semantically a reverse operation.
6122 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
6123 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
6124 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
6126 (aarch64_can_change_mode_class): For big-endian, forbid changes
6127 between two SVE modes if they have different element sizes.
6129 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6131 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
6132 the TImode handling for big-endian targets.
6134 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6136 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
6137 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
6139 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
6140 Remove BSWAP handing for big-endian targets and use the form of
6141 LD1RQ appropariate for the mode.
6143 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6145 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
6146 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
6149 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
6152 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
6153 check for operands that need to go through aarch64_sve_reload_be.
6155 2018-02-01 Jakub Jelinek <jakub@redhat.com>
6157 PR tree-optimization/81661
6158 PR tree-optimization/84117
6159 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
6160 * tree-eh.c: Include gimplify.h.
6161 (find_trapping_overflow, replace_trapping_overflow,
6162 rewrite_to_non_trapping_overflow): New functions.
6163 * tree-vect-loop.c: Include tree-eh.h.
6164 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
6165 * tree-data-ref.c: Include tree-eh.h.
6166 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
6168 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
6170 PR rtl-optimization/84123
6171 * combine.c (change_zero_ext): Check if hard register satisfies
6172 can_change_dest_mode before calling gen_lowpart_SUBREG.
6174 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
6177 * ira.c (ira_init_register_move_cost): Remove assert.
6179 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6181 PR rtl-optimization/84071
6182 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
6183 * doc/tm.texi: Regenerate.
6185 2018-01-31 Richard Biener <rguenther@suse.de>
6187 PR tree-optimization/84132
6188 * tree-data-ref.c (analyze_miv_subscript): Properly
6189 check whether evolution_function_is_affine_multivariate_p
6190 before calling gcd_of_steps_may_divide_p.
6192 2018-01-31 Julia Koval <julia.koval@intel.com>
6195 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
6196 * config/i386/i386.md (rdpid_rex64) New.
6197 (rdpid): Make 32bit only.
6199 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
6202 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
6203 an IDENTIFIER_NODE for FUNCTION_TYPE's.
6205 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6208 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6210 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6212 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6214 PR rtl-optimization/84071
6215 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6216 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6218 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6220 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6221 (arc_attribute_table): Add 'aux' attribute.
6222 (arc_in_small_data_p): Consider aux like variables.
6223 (arc_is_aux_reg_p): New function.
6224 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6225 (arc_get_aux_arg): New function.
6226 (prepare_move_operands): Handle aux-register access.
6227 (arc_handle_aux_attribute): New function.
6228 * doc/extend.texi (ARC Variable attributes): Add subsection.
6230 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6232 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6233 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6234 (arc_attribute_table): Add 'uncached' attribute.
6235 (arc_print_operand): Print '.di' flag for uncached memory
6237 (arc_in_small_data_p): Do not consider for small data the uncached
6239 (arc_is_uncached_mem_p): New function.
6240 * config/arc/predicates.md (compact_store_memory_operand): Check
6241 for uncached memory accesses.
6242 (nonvol_nonimm_operand): Likewise.
6243 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6245 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6248 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6249 falign-loops=): Add Optimization flag.
6251 2018-01-30 Jeff Law <law@redhat.com>
6254 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6255 INT_REGISTERS_SAVED. Check it prior to calling
6256 get_scratch_register_on_entry.
6257 (ix86_adjust_stack_and_probe): Similarly.
6258 (ix86_emit_probe_stack_range): Similarly.
6259 (ix86_expand_prologue): Corresponding changes.
6261 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6264 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6265 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6267 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6270 * lra-constraints.c (curr_insn_transform): Process AND in the
6273 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6275 PR rtl-optimization/83986
6276 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6277 dependence against last_pending_memory_flush in addition to
6280 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6282 PR tree-optimization/81611
6283 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6286 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6289 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6292 2018-01-30 Richard Biener <rguenther@suse.de>
6293 Jakub Jelinek <jakub@redhat.com>
6295 PR tree-optimization/84111
6296 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6297 inner loops added during recursion, as they don't have up-to-date
6300 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6303 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6304 (can_inline_edge_by_limits_p): ... here.
6305 (can_early_inline_edge_p, check_callers,
6306 update_caller_keys, update_callee_keys, recursive_inlining,
6307 add_new_edges_to_heap, speculation_useful_p,
6308 inline_small_functions,
6309 inline_small_functions, flatten_function,
6310 inline_to_all_callers_1): Update.
6312 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6314 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6317 2018-01-30 Richard Biener <rguenther@suse.de>
6319 PR tree-optimization/83008
6320 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6321 invariant and constant vector uses in stmts when they need
6324 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6327 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6328 * configure: Regenerate.
6330 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6332 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6334 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6335 Use gen_rtx_REG rather than gen_lowpart.
6337 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6339 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6340 rather than 0 when creating partial subregs.
6342 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6344 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6347 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6350 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6351 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6352 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6353 flags. This restores the settings used before the 2017-07-24.
6354 Turning off pre increment/decrement/modify allows IVOPTS to
6355 optimize DF/SF loops where the index is an int.
6357 2018-01-29 Richard Biener <rguenther@suse.de>
6358 Kelvin Nilsen <kelvin@gcc.gnu.org>
6361 * tree-vect-stmts.c (vectorizable_call): Don't call
6362 targetm.vectorize_builtin_md_vectorized_function if callee is
6365 2018-01-22 Carl Love <cel@us.ibm.com>
6367 * doc/extend.tex: Fix typo in second arg in
6368 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6370 2018-01-29 Richard Biener <rguenther@suse.de>
6372 PR tree-optimization/84086
6373 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6374 (flush_ssaname_freelist): When SSA names were released reset
6375 the SCEV hash table.
6377 2018-01-29 Richard Biener <rguenther@suse.de>
6379 PR tree-optimization/84057
6380 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6381 removed paths when removing edges.
6383 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6385 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6386 -mfunction-return=@var{choice}.
6388 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6391 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6393 (layout::print_source_line): Likewise.
6394 (test_get_line_width_without_trailing_whitespace): Add test cases.
6396 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6399 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6402 2018-01-26 Jim Wilson <jimw@sifive.com>
6404 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6406 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6409 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6411 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6412 and CMP + SUB-immediate -> SUBS.
6414 2018-01-26 Martin Sebor <msebor@redhat.com>
6416 PR tree-optimization/83896
6417 * tree-ssa-strlen.c (get_string_len): Rename...
6418 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6419 Avoid assuming length is constant.
6420 (handle_char_store): Use HOST_WIDE_INT for string length.
6422 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6425 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6426 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6428 2018-01-26 Richard Biener <rguenther@suse.de>
6430 PR rtl-optimization/84003
6431 * dse.c (record_store): Only record redundant stores when
6432 the earlier store aliases at least all accesses the later one does.
6434 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6436 PR rtl-optimization/83985
6437 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6438 REG_CFA_RESTORE insns.
6439 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6440 REG_CFA_RESTORE insns here.
6443 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6444 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6446 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6448 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6449 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6450 (arc_init): Likewise.
6451 (arc_override_options): Likewise.
6452 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6454 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6456 * config/arc/arc.h (TARGET_DBNZ): Define.
6457 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6458 properly set the tune attribute.
6459 (dbnz): Use TARGET_DBNZ guard.
6460 * config/arc/arc.opt (mtune): Add core3 option.
6462 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6464 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6465 recognize new pic like addresses.
6466 (arc_delegitimize_address): Clean up.
6468 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6470 * config/arc/arc-arches.def: Option mrf16 valid for all
6472 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6473 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6474 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6475 * config/arc/arc-tables.opt: Regenerate.
6476 * config/arc/arc.c (arc_conditional_register_usage): Handle
6477 reduced register file case.
6478 (arc_file_start): Set must have build attributes.
6479 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6481 * config/arc/arc.opt (mrf16): Add new option.
6482 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6483 * config/arc/genmultilib.awk: Handle new mrf16 option.
6484 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6485 * config/arc/t-multilib: Regenerate.
6486 * doc/invoke.texi (ARC Options): Document mrf16 option.
6488 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6490 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6491 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6492 (arc_attribute_table): Add 'secure_call' attribute.
6493 (arc_print_operand): Print secure call operand.
6494 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6496 (arc_is_secure_call_p): New function. * config/arc/arc.md
6497 (call_i): Add support for sjli instruction.
6498 (call_value_i): Likewise.
6499 * config/arc/constraints.md (Csc): New constraint.
6501 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6502 John Eric Martin <John.Martin@emmicro-us.com>
6504 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6505 * config/arc/arc.c (_arc_jli_section): New struct.
6506 (arc_jli_section): New type.
6507 (rc_jli_sections): New static variable.
6508 (arc_handle_jli_attribute): New function.
6509 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6510 (arc_file_end): New function.
6511 (TARGET_ASM_FILE_END): Define.
6512 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6513 (arc_add_jli_section): New function.
6514 (jli_call_scan): Likewise.
6515 (arc_reorg): Call jli_call_scan.
6516 (arc_output_addsi): Remove 'S' from printing asm operand.
6517 (arc_is_jli_call_p): New function.
6518 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6520 (movhi_insn): Likewise.
6521 (movsi_insn): Likewise.
6522 (movsi_set_cc_insn): Likewise.
6523 (loadqi_update): Likewise.
6524 (load_zeroextendqisi_update): Likewise.
6525 (load_signextendqisi_update): Likewise.
6526 (loadhi_update): Likewise.
6527 (load_zeroextendhisi_update): Likewise.
6528 (load_signextendhisi_update): Likewise.
6529 (loadsi_update): Likewise.
6530 (loadsf_update): Likewise.
6531 (movsicc_insn): Likewise.
6532 (bset_insn): Likewise.
6533 (bxor_insn): Likewise.
6534 (bclr_insn): Likewise.
6535 (bmsk_insn): Likewise.
6536 (bicsi3_insn): Likewise.
6537 (cmpsi_cc_c_insn): Likewise.
6538 (movsi_ne): Likewise.
6539 (movsi_cond_exec): Likewise.
6540 (clrsbsi2): Likewise.
6551 (call_i): Remove 'S' asm letter, add jli instruction.
6552 (call_value_i): Likewise.
6553 * config/arc/arc.op (mjli-always): New option.
6554 * config/arc/constraints.md (Cji): New constraint.
6555 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6557 (subsf3_fpx): Likewise.
6558 (mulsf3_fpx): Likewise.
6559 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6561 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6563 * doc/invoke.texi (ARC): Document mjli-always option.
6565 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6567 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6568 avoid addition with 0 and use incw and decw where possible.
6570 2018-01-26 Richard Biener <rguenther@suse.de>
6572 PR tree-optimization/81082
6573 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6574 association if it requires casting to unsigned.
6575 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6576 from fold_plusminus_mult_expr to catch important cases late when
6577 range info is available.
6579 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6581 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6582 * configure.ac (hidden_linkonce): New test.
6583 * configure: Regenerate.
6584 * config.in: Regenerate.
6586 2018-01-26 Julia Koval <julia.koval@intel.com>
6588 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6589 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6590 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6591 _mm_mask_bitshuffle_epi64_mask): Fix type.
6592 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6593 USI_FTYPE_V4DI_V4DI_USI): Remove.
6594 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6595 __builtin_ia32_vpshufbitqmb256_mask,
6596 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6597 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6598 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6600 2018-01-26 Alan Modra <amodra@gmail.com>
6603 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6604 UNSPEC_VBPERMQ. Sort other unspecs.
6606 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6608 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6610 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6613 * predict.c (drop_profile): Do not push/pop cfun; update also
6615 (handle_missing_profiles): Fix logic looking for zero profiles.
6617 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6620 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6621 on functions with #pragma omp declare simd or functions with simd
6623 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6624 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6625 Remove trailing \n from warning_at calls.
6627 2018-01-25 Tom de Vries <tom@codesourcery.com>
6630 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6631 for neutered workers.
6633 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6636 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6637 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6639 2018-01-24 Jeff Law <law@redhat.com>
6642 * i386.c (get_probe_interval): Move to earlier point.
6643 (ix86_compute_frame_layout): If -fstack-clash-protection and
6644 the frame is larger than the probe interval, then use pushes
6645 to save registers rather than reg->mem moves.
6646 (ix86_expand_prologue): Remove conditional for int_registers_saved
6649 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6652 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6653 min/max for never referenced object.
6655 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6658 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6660 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6661 attributes from DECL_ATTRIBUTES (decl) without affecting
6662 DECL_ATTRIBUTES (current_function_decl).
6663 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6664 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6666 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6668 PR tree-optimization/83979
6669 * fold-const.c (fold_comparison): Use constant_boolean_node
6670 instead of boolean_{true,false}_node.
6672 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6674 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6677 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6679 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6680 Simplify the clause that sets the length attribute.
6681 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6682 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6683 clause that sets the length attribute.
6684 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6686 2018-01-24 Tom de Vries <tom@codesourcery.com>
6689 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6690 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6691 Add strict parameter.
6692 (prevent_branch_around_nothing): Insert dummy insn between branch to
6693 label and label with no ptx insn inbetween.
6694 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6696 2018-01-24 Tom de Vries <tom@codesourcery.com>
6699 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6700 for neutered threads in warp.
6701 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6703 2018-01-24 Richard Biener <rguenther@suse.de>
6705 PR tree-optimization/83176
6706 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6709 2018-01-24 Richard Biener <rguenther@suse.de>
6711 PR tree-optimization/82819
6712 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6713 code generating pluses that are no-ops in the target precision.
6715 2018-01-24 Richard Biener <rguenther@suse.de>
6718 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6720 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6722 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6723 to merge probabilities.
6724 * predict.c (probably_never_executed): Also mark as cold functions
6725 with global 0 profile and guessed local profile.
6726 * profile-count.c (profile_probability::combine_with_count): New
6728 * profile-count.h (profile_probability::operator*,
6729 profile_probability::operator*=, profile_probability::operator/,
6730 profile_probability::operator/=): Reduce precision to adjusted
6731 and set value to guessed on contradictory divisions.
6732 (profile_probability::combine_with_freq): Remove.
6733 (profile_probability::combine_wiht_count): Declare.
6734 (profile_count::force_nonzero):: Set to adjusted.
6735 (profile_count::probability_in):: Set quality to adjusted.
6736 * tree-ssa-tail-merge.c (replace_block_by): Use
6739 2018-01-23 Andrew Waterman <andrew@sifive.com>
6740 Jim Wilson <jimw@sifive.com>
6742 * config/riscv/riscv.c (riscv_stack_boundary): New.
6743 (riscv_option_override): Set riscv_stack_boundary. Handle
6744 riscv_preferred_stack_boundary_arg.
6745 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6746 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6747 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6748 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6749 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6750 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6752 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6755 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6756 of struct ix86_frame.
6757 (ix86_expand_epilogue): Likewise. Add a local variable for
6758 the reg_save_offset field in struct ix86_frame.
6760 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6762 PR tree-optimization/82604
6763 * tree-loop-distribution.c (enum partition_kind): New enum item
6764 PKIND_PARTIAL_MEMSET.
6765 (partition_builtin_p): Support above new enum item.
6766 (generate_code_for_partition): Ditto.
6767 (compute_access_range): Differentiate cases that equality can be
6768 proven at all loops, the innermost loops or no loops.
6769 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6770 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6771 (finalize_partitions, distribute_loop): Don't fuse partition of
6772 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6773 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6776 2018-01-23 Martin Liska <mliska@suse.cz>
6778 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6779 order to ignore the predictor.
6780 (PRED_POLYMORPHIC_CALL): Likewise.
6781 (PRED_RECURSIVE_CALL): Likewise.
6783 2018-01-23 Martin Liska <mliska@suse.cz>
6785 * tree-profile.c (tree_profiling): Print function header to
6786 aware reader which function we are working on.
6787 * value-prof.c (gimple_find_values_to_profile): Do not print
6788 not interesting value histograms.
6790 2018-01-23 Martin Liska <mliska@suse.cz>
6792 * profile-count.h (enum profile_quality): Add
6793 profile_uninitialized as the first value. Do not number values
6794 as they are zero based.
6795 (profile_count::verify): Update sanity check.
6796 (profile_probability::verify): Likewise.
6798 2018-01-23 Nathan Sidwell <nathan@acm.org>
6800 * doc/invoke.texi (ffor-scope): Deprecate.
6802 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6804 PR tree-optimization/83510
6805 * domwalk.c (set_all_edges_as_executable): New function.
6806 (dom_walker::dom_walker): Convert bool param
6807 "skip_unreachable_blocks" to enum reachability. Move setup of
6808 edge flags to set_all_edges_as_executable and only do it when
6809 reachability is REACHABLE_BLOCKS.
6810 * domwalk.h (enum dom_walker::reachability): New enum.
6811 (dom_walker::dom_walker): Convert bool param
6812 "skip_unreachable_blocks" to enum reachability.
6813 (set_all_edges_as_executable): New decl.
6814 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6815 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6817 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6818 but converting true to REACHABLE_BLOCKS.
6819 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6821 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6822 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6823 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6825 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6826 if check_all_array_refs will be called.
6828 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6830 * tree.c (selftest::test_location_wrappers): Add more test
6833 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6835 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6836 (selftest::test_bit_in_range): Likewise.
6838 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6841 * doc/sourcebuild.texi (vect_float): Say that the selector
6842 only describes the situation when -funsafe-math-optimizations is on.
6843 (vect_float_strict): Document.
6845 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6847 PR tree-optimization/83965
6848 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6849 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6850 instead of checking only for a reduction.
6851 (vect_recog_widen_sum_pattern): Likewise.
6853 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6855 * predict.c (probably_never_executed): Only use precise profile info.
6856 (compute_function_frequency): Skip after inlining hack since we now
6857 have quality checking.
6859 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6861 * profile-count.h (profile_probability::very_unlikely,
6862 profile_probability::unlikely, profile_probability::even): Set
6863 precision to guessed.
6865 2018-01-23 Richard Biener <rguenther@suse.de>
6867 PR tree-optimization/83963
6868 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6869 Properly terminate dominator walk when crossing the exit edge not
6870 when visiting its source block.
6872 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6875 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6876 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6878 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6880 PR tree-optimization/83957
6881 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6882 semicolon after for body surrounded by braces.
6884 PR tree-optimization/83081
6885 * profile-count.h (profile_probability::split): New method.
6886 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6887 Use profile_probability::split.
6888 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6889 when splitting a single conditional jump into 2.
6891 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6893 PR tree-optimization/69452
6894 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6897 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6899 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6900 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6901 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6903 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6905 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6906 * config/rl78/rl78.md (movdi): New define_expand.
6907 * config/rl78/rl78.c (rl78_split_movdi): New function.
6909 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6912 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6914 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6915 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6916 128-bit to produce an UNSPEC move to get the double word with the
6917 signbit and then a shift directly to do signbit.
6918 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6919 implementation with a new version that just does either a direct
6920 move or a regular move. Move memory interface to separate insns.
6921 Move insns so they are next to the expander.
6922 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6923 with signbit move. Split big and little endian case.
6924 (signbit<mode>2_dm_mem_le): Likewise.
6925 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6926 (signbit<mode>2_dm2): Likewise.
6928 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6930 * config/rl78/rl78.md (anddi3): New define_expand.
6932 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6934 * config/rl78/rl78.md (umindi3): New define_expand.
6936 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6938 * config/rl78/rl78.md (smindi3): New define_expand.
6940 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6942 * config/rl78/rl78.md (smaxdi3): New define_expand.
6944 2018-01-22 Carl Love <cel@us.ibm.com>
6946 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6947 LVX_V1TI): Add macro expansion.
6948 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6949 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6950 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6951 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6952 Change check to determine if the instruction is a byte reversing
6953 entry. Fix typo in comment.
6954 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6955 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6956 Add def_builtin calls for new builtins.
6957 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6958 Add define_insn expansion.
6960 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6962 * config/rl78/rl78.md (umaxdi3): New define_expand.
6964 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6966 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6967 for non-QImode registers.
6969 2018-01-22 Richard Biener <rguenther@suse.de>
6971 PR tree-optimization/83963
6972 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6973 including the loop exit block.
6974 (scop_detection::merge_sese): Likewise.
6975 (scop_detection::add_scop): Do it here instead.
6977 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6979 * doc/sourcebuild.texi (arm_softfloat): Document.
6981 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6984 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6985 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6986 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6988 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6989 David Edelsohn <dje.gcc@gmail.com>
6992 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6993 Change "crset eq" to "crset 2".
6994 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6995 (*call_indirect_aix<mode>_nospec): Likewise.
6996 (*call_value_indirect_aix<mode>_nospec): Likewise.
6997 (*call_indirect_elfv2<mode>_nospec): Likewise.
6998 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6999 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
7000 change assembly output from . to $.
7001 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7002 (indirect_jump<mode>_nospec): Change assembly output from . to $.
7003 (*tablejump<mode>_internal1_nospec): Likewise.
7005 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
7008 * config/sh/sh_optimize_sett_clrt.cc:
7009 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
7011 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
7013 PR tree-optimization/83940
7014 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
7015 offset_dt to vect_constant_def rather than vect_unknown_def_type.
7016 (vect_check_load_store_mask): Add a mask_dt_out parameter and
7017 use it to pass back the definition type.
7018 (vect_check_store_rhs): Likewise rhs_dt_out.
7019 (vect_build_gather_load_calls): Add a mask_dt argument and use
7020 it instead of a call to vect_is_simple_use.
7021 (vectorizable_store): Update calls to vect_check_load_store_mask
7022 and vect_check_store_rhs. Use the dt returned by the latter instead
7023 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
7024 instead of calls to vect_is_simple_use. Pass the scalar rather
7025 than the vector operand to vect_is_simple_use when handling
7026 second and subsequent copies of an rhs value.
7027 (vectorizable_load): Update calls to vect_check_load_store_mask
7028 and vect_build_gather_load_calls. Use the cached mask_dt and
7029 gs_info.offset_dt instead of calls to vect_is_simple_use.
7031 2018-01-20 Jakub Jelinek <jakub@redhat.com>
7034 * tree-emutls.c: Include gimplify.h.
7035 (lower_emutls_2): New function.
7036 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
7037 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
7038 it before further processing.
7041 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
7042 UINTVAL (trueop1) instead of INTVAL (op1).
7044 2018-01-19 Jakub Jelinek <jakub@redhat.com>
7048 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
7049 INCOMING_FRAME_SP_OFFSET if not defined.
7050 (scan_trace): Add ENTRY argument. If true and
7051 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
7052 emit a note to adjust the CFA offset.
7053 (create_cfi_notes): Adjust scan_trace callers.
7054 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
7055 INCOMING_FRAME_SP_OFFSET in the CIE.
7056 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
7057 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
7059 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
7060 * doc/tm.texi: Regenerated.
7062 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
7064 PR rtl-optimization/83147
7065 * lra-constraints.c (remove_inheritance_pseudos): Use
7066 lra_substitute_pseudo_within_insn.
7068 2018-01-19 Tom de Vries <tom@codesourcery.com>
7069 Cesar Philippidis <cesar@codesourcery.com>
7072 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
7074 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
7077 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
7078 spaces for function labels.
7080 2018-01-19 Martin Liska <mliska@suse.cz>
7082 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
7083 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
7084 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
7085 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
7086 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
7087 (PRED_CONST_RETURN): Change from 69 to 65.
7088 (PRED_NULL_RETURN): Change from 91 to 71.
7089 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
7090 (PRED_LOOP_GUARD): Change from 66 to 73.
7092 2018-01-19 Martin Liska <mliska@suse.cz>
7094 * predict.c (predict_insn_def): Add new assert.
7095 (struct branch_predictor): Change type to signed integer.
7096 (test_prediction_value_range): Amend test to cover
7098 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
7099 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
7100 (PRED_LOOP_ITERATIONS_MAX): Likewise.
7101 (PRED_LOOP_IV_COMPARE): Likewise.
7102 * predict.h (PROB_UNINITIALIZED): Define new constant.
7104 2018-01-19 Martin Liska <mliska@suse.cz>
7106 * predict.c (dump_prediction): Add new format for
7107 analyze_brprob.py script which is enabled with -details
7109 * profile-count.h (precise_p): New function.
7111 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7113 PR tree-optimization/83922
7114 * tree-vect-loop.c (vect_verify_full_masking): Return false if
7115 there are no statements that need masking.
7116 (vect_active_double_reduction_p): New function.
7117 (vect_analyze_loop_operations): Use it when handling phis that
7118 are not in the loop header.
7120 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
7122 PR tree-optimization/83914
7123 * tree-vect-loop.c (vectorizable_induction): Don't convert
7124 init_expr or apply the peeling adjustment for inductions
7125 that are nested within the vectorized loop.
7127 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7129 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
7132 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7136 * function.h (gimplify_parameters): Add gimple_seq * argument.
7137 * function.c: Include gimple.h and options.h.
7138 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
7139 for the added local temporaries if needed.
7140 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
7141 if there are any parameter cleanups, wrap whole body into a
7142 try/finally with the cleanups.
7144 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
7147 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
7148 Use GET_MODE_CLASS for scalar floating point.
7150 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7154 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
7155 Fix call of call_cgraph_insertion_hooks.
7157 2018-01-18 Martin Sebor <msebor@redhat.com>
7159 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
7161 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
7164 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
7167 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
7170 * common.opt: (-ffile-prefix-map): New option.
7171 * opts.c (common_handle_option): Defer it.
7172 * opts-global.c (handle_common_deferred_options): Handle it.
7173 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
7174 * file-prefix-map.h: New file.
7175 (remap_debug_filename, add_debug_prefix_map): ...here.
7176 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
7177 * final.c (debug_prefix_map, add_debug_prefix_map
7178 remap_debug_filename): Move to...
7179 * file-prefix-map.c: New file.
7180 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
7181 generalize, get rid of alloca(), use strrchr() instead of strchr().
7182 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
7183 Implement in terms of add_prefix_map().
7184 (remap_macro_filename, remap_debug_filename): Implement in term of
7186 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
7187 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
7188 * dbxout.c: Include file-prefix-map.h.
7189 * varasm.c: Likewise.
7190 * vmsdbgout.c: Likewise.
7191 * xcoffout.c: Likewise.
7192 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
7193 * doc/cppopts.texi (-fmacro-prefix-map): Document.
7194 * doc/invoke.texi (-ffile-prefix-map): Document.
7195 (-fdebug-prefix-map): Update description.
7197 2018-01-18 Martin Liska <mliska@suse.cz>
7199 * config/i386/i386.c (indirect_thunk_name): Document that also
7201 (output_indirect_thunk): Document why both instructions
7202 (pause and lfence) are generated.
7204 2018-01-18 Richard Biener <rguenther@suse.de>
7206 PR tree-optimization/83887
7207 * graphite-scop-detection.c
7208 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7209 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7210 (scop_detection::merge_sese): Re-implement with a flood-fill
7211 algorithm that properly finds a SESE region if it exists.
7213 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7216 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7217 pointer_diff optimizations use view_convert instead of convert.
7219 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7221 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7222 Generate different code for -mno-speculate-indirect-jumps.
7223 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7224 (*call_indirect_aix<mode>): Disable for
7225 -mno-speculate-indirect-jumps.
7226 (*call_indirect_aix<mode>_nospec): New define_insn.
7227 (*call_value_indirect_aix<mode>): Disable for
7228 -mno-speculate-indirect-jumps.
7229 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7230 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7231 -mno-speculate-indirect-jumps.
7232 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7234 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7236 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7237 long double type, set the flags for noting the default long double
7238 type, even if we don't pass or return a long double type.
7240 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7243 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7246 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7248 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7249 support for merge[hl].
7250 (fold_mergehl_helper): New helper function.
7251 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7252 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7253 (altivec_vmrglw_direct): Add xxmrglw insn.
7255 2018-01-17 Andrew Waterman <andrew@sifive.com>
7257 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7258 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7260 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7263 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7264 call the lto_location_cache before reading the
7265 DECL_SOURCE_LOCATION of the types.
7267 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7268 Richard Sandiford <richard.sandiford@linaro.org>
7270 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7271 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7272 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7273 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7274 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7276 * config/aarch64/constraints.md (aarch64_movti_operand):
7278 * config/aarch64/predicates.md (Uti): Add new constraint.
7280 2018-01-17 Carl Love <cel@us.ibm.com>
7282 * config/rs6000/vsx.md (define_expand xl_len_r,
7283 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7284 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7286 (define_expand, define_insn): Move the shift left from the
7287 define_insn to the define_expand for lxvl and stxvl instructions.
7288 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7289 and XL_LEN_R definitions to PURE.
7291 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7293 * config/i386/i386.c (indirect_thunk_name): Declare regno
7294 as unsigned int. Compare regno with INVALID_REGNUM.
7295 (output_indirect_thunk): Ditto.
7296 (output_indirect_thunk_function): Ditto.
7297 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7298 in the call to output_indirect_thunk_function.
7300 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7303 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7304 rather than the size of inner_type to determine the stack slot size
7305 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7307 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7310 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7313 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7315 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7316 endian Linux systems to optionally enable multilibs for selecting
7317 the long double type if the user configured an explicit type.
7318 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7319 have no long double multilibs if not defined.
7320 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7321 warn if the user used -mabi={ieee,ibm}longdouble and we built
7322 multilibs for long double.
7323 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7324 appropriate multilib option.
7325 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7327 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7328 for building long double multilibs.
7329 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7331 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7333 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7336 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7338 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7341 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7344 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7347 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7349 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7350 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7352 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7354 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7355 different rtl trees depending on TARGET_64BIT.
7356 (rs6000_gen_lvx): Likewise.
7358 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7360 * config/visium/visium.md (nop): Tweak comment.
7361 (hazard_nop): Likewise.
7363 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7365 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7366 -mspeculate-indirect-jumps.
7367 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7368 for -mno-speculate-indirect-jumps.
7369 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7370 (*call_value_indirect_elfv2<mode>): Disable for
7371 -mno-speculate-indirect-jumps.
7372 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7373 (indirect_jump): Emit different RTL for
7374 -mno-speculate-indirect-jumps.
7375 (*indirect_jump<mode>): Disable for
7376 -mno-speculate-indirect-jumps.
7377 (*indirect_jump<mode>_nospec): New define_insn.
7378 (tablejump): Emit different RTL for
7379 -mno-speculate-indirect-jumps.
7380 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7381 (tablejumpsi_nospec): New define_expand.
7382 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7383 (tablejumpdi_nospec): New define_expand.
7384 (*tablejump<mode>_internal1): Disable for
7385 -mno-speculate-indirect-jumps.
7386 (*tablejump<mode>_internal1_nospec): New define_insn.
7387 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7390 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7392 * caller-save.c (insert_save): Drop unnecessary parameter. All
7395 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7396 Richard Biener <rguenth@suse.de>
7399 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7400 return early, inline manually is_gimple_sizepos. Make sure if we
7401 call gimplify_expr we don't end up with a gimple constant.
7402 * tree.c (variably_modified_type_p): Don't return true for
7403 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7404 * gimplify.h (is_gimple_sizepos): Remove.
7406 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7408 PR tree-optimization/83857
7409 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7410 vectorizable_live_operation for pure SLP statements.
7411 (vectorizable_live_operation): Handle PHIs.
7413 2018-01-16 Richard Biener <rguenther@suse.de>
7415 PR tree-optimization/83867
7416 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7417 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7419 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7422 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7423 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7424 If off is not INTEGER_CST, issue a may not be aligned warning
7425 rather than isn't aligned. Use isn%'t rather than isn't.
7426 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7428 <case MULT_EXPR>: Improve the case when bottom and one of the
7429 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7430 operand, in that case check if the other operand is multiple of
7431 bottom divided by the INTEGER_CST operand.
7433 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7436 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7437 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7438 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7439 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7440 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7441 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7442 (pa_function_arg_size): New function.
7444 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7446 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7447 in a separate statement.
7449 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7451 PR tree-optimization/83847
7452 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7453 group gathers and scatters.
7455 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7457 PR rtl-optimization/86620
7458 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7460 PR rtl-optimization/83213
7461 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7462 to last if both are JUMP_INSNs.
7464 PR tree-optimization/83843
7465 * gimple-ssa-store-merging.c
7466 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7467 store_immediate_info for bswap/nop orig_stores.
7469 2018-01-15 Andrew Waterman <andrew@sifive.com>
7471 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7473 <UDIV>: Increase cost if !TARGET_DIV.
7475 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7477 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7478 (define_attr "cr_logical_3op"): New.
7479 (cceq_ior_compare): Adjust.
7480 (cceq_ior_compare_complement): Adjust.
7481 (*cceq_rev_compare): Adjust.
7482 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7483 (is_cracked_insn): Adjust.
7484 (insn_must_be_first_in_group): Adjust.
7485 * config/rs6000/40x.md: Adjust.
7486 * config/rs6000/440.md: Adjust.
7487 * config/rs6000/476.md: Adjust.
7488 * config/rs6000/601.md: Adjust.
7489 * config/rs6000/603.md: Adjust.
7490 * config/rs6000/6xx.md: Adjust.
7491 * config/rs6000/7450.md: Adjust.
7492 * config/rs6000/7xx.md: Adjust.
7493 * config/rs6000/8540.md: Adjust.
7494 * config/rs6000/cell.md: Adjust.
7495 * config/rs6000/e300c2c3.md: Adjust.
7496 * config/rs6000/e500mc.md: Adjust.
7497 * config/rs6000/e500mc64.md: Adjust.
7498 * config/rs6000/e5500.md: Adjust.
7499 * config/rs6000/e6500.md: Adjust.
7500 * config/rs6000/mpc.md: Adjust.
7501 * config/rs6000/power4.md: Adjust.
7502 * config/rs6000/power5.md: Adjust.
7503 * config/rs6000/power6.md: Adjust.
7504 * config/rs6000/power7.md: Adjust.
7505 * config/rs6000/power8.md: Adjust.
7506 * config/rs6000/power9.md: Adjust.
7507 * config/rs6000/rs64.md: Adjust.
7508 * config/rs6000/titan.md: Adjust.
7510 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7512 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7513 ix86_indirect_branch_register logic.
7515 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7517 * config/i386/constraints.md (Bs): Update
7518 ix86_indirect_branch_register check. Don't check
7519 ix86_indirect_branch_register with GOT_memory_operand.
7521 * config/i386/predicates.md (GOT_memory_operand): Don't check
7522 ix86_indirect_branch_register here.
7523 (GOT32_symbol_operand): Likewise.
7525 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7527 * config/i386/predicates.md (constant_call_address_operand):
7528 Rewrite ix86_indirect_branch_register logic.
7529 (sibcall_insn_operand): Likewise.
7531 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7533 * config/i386/constraints.md (Bs): Replace
7534 ix86_indirect_branch_thunk_register with
7535 ix86_indirect_branch_register.
7537 * config/i386/i386.md (indirect_jump): Likewise.
7538 (tablejump): Likewise.
7539 (*sibcall_memory): Likewise.
7540 (*sibcall_value_memory): Likewise.
7541 Peepholes of indirect call and jump via memory: Likewise.
7542 * config/i386/i386.opt: Likewise.
7543 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7544 (GOT_memory_operand): Likewise.
7545 (call_insn_operand): Likewise.
7546 (sibcall_insn_operand): Likewise.
7547 (GOT32_symbol_operand): Likewise.
7549 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7552 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7553 type rather than type addr's type points to.
7554 (expand_omp_atomic_mutex): Likewise.
7555 (expand_omp_atomic): Likewise.
7557 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7560 * config/i386/i386.c (output_indirect_thunk_function): Use
7561 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7562 for __x86_return_thunk.
7564 2018-01-15 Richard Biener <rguenther@suse.de>
7567 * expmed.c (extract_bit_field_1): Fix typo.
7569 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7572 * config/arm/iterators.md (VF): New mode iterator.
7573 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7574 Remove integer-related logic from pattern.
7575 (neon_vabd<mode>_3): Likewise.
7577 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7580 * common.opt (fstrict-overflow): No longer an alias.
7581 (fwrapv-pointer): New option.
7582 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7583 also for pointer types based on flag_wrapv_pointer.
7584 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7585 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7586 opts->x_flag_wrapv got set.
7587 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7588 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7589 POINTER_TYPE_OVERFLOW_UNDEFINED.
7590 * match.pd: Likewise in address comparison pattern.
7591 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7593 2018-01-15 Richard Biener <rguenther@suse.de>
7596 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7597 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7598 Reset type names to their identifier if their TYPE_DECL doesn't
7599 have linkage (and thus is used for ODR and devirt).
7600 (save_debug_info_for_decl): Remove.
7601 (save_debug_info_for_type): Likewise.
7602 (add_tree_to_fld_list): Adjust.
7603 * tree-pretty-print.c (dump_generic_node): Make dumping of
7604 type names more robust.
7606 2018-01-15 Richard Biener <rguenther@suse.de>
7608 * BASE-VER: Bump to 8.0.1.
7610 2018-01-14 Martin Sebor <msebor@redhat.com>
7613 * builtins.c (check_access): Avoid warning when the no-warning bit
7616 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7618 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7619 * ira-color (allocno_hard_regs_compare): Likewise.
7621 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7624 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7625 Use .pushsection/.popsection.
7627 2018-01-14 Martin Sebor <msebor@redhat.com>
7630 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7632 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7634 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7635 entry from extra_headers.
7636 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7637 extra_headers, make the list bitwise identical to the i?86-*-* one.
7639 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7641 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7642 -mcmodel=large with -mindirect-branch=thunk,
7643 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7644 -mfunction-return=thunk-extern.
7645 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7646 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7647 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7649 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7651 * config/i386/i386.c (print_reg): Print the name of the full
7652 integer register without '%'.
7653 (ix86_print_operand): Handle 'V'.
7654 * doc/extend.texi: Document 'V' modifier.
7656 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7658 * config/i386/constraints.md (Bs): Disallow memory operand for
7659 -mindirect-branch-register.
7661 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7662 (GOT_memory_operand): Likewise.
7663 (call_insn_operand): Likewise.
7664 (sibcall_insn_operand): Likewise.
7665 (GOT32_symbol_operand): Likewise.
7666 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7667 for -mindirect-branch-register.
7668 (tablejump): Likewise.
7669 (*sibcall_memory): Likewise.
7670 (*sibcall_value_memory): Likewise.
7671 Disallow peepholes of indirect call and jump via memory for
7672 -mindirect-branch-register.
7673 (*call_pop): Replace m with Bw.
7674 (*call_value_pop): Likewise.
7675 (*sibcall_pop_memory): Replace m with Bs.
7676 * config/i386/i386.opt (mindirect-branch-register): New option.
7677 * doc/invoke.texi: Document -mindirect-branch-register option.
7679 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7681 * config/i386/i386-protos.h (ix86_output_function_return): New.
7682 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7683 set function_return_type.
7684 (indirect_thunk_name): Add ret_p to indicate thunk for function
7686 (output_indirect_thunk_function): Pass false to
7687 indirect_thunk_name.
7688 (ix86_output_indirect_branch_via_reg): Likewise.
7689 (ix86_output_indirect_branch_via_push): Likewise.
7690 (output_indirect_thunk_function): Create alias for function
7691 return thunk if regno < 0.
7692 (ix86_output_function_return): New function.
7693 (ix86_handle_fndecl_attribute): Handle function_return.
7694 (ix86_attribute_table): Add function_return.
7695 * config/i386/i386.h (machine_function): Add
7696 function_return_type.
7697 * config/i386/i386.md (simple_return_internal): Use
7698 ix86_output_function_return.
7699 (simple_return_internal_long): Likewise.
7700 * config/i386/i386.opt (mfunction-return=): New option.
7701 (indirect_branch): Mention -mfunction-return=.
7702 * doc/extend.texi: Document function_return function attribute.
7703 * doc/invoke.texi: Document -mfunction-return= option.
7705 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7707 * config/i386/i386-opts.h (indirect_branch): New.
7708 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7709 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7710 with local indirect jump when converting indirect call and jump.
7711 (ix86_set_indirect_branch_type): New.
7712 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7713 (indirectlabelno): New.
7714 (indirect_thunk_needed): Likewise.
7715 (indirect_thunk_bnd_needed): Likewise.
7716 (indirect_thunks_used): Likewise.
7717 (indirect_thunks_bnd_used): Likewise.
7718 (INDIRECT_LABEL): Likewise.
7719 (indirect_thunk_name): Likewise.
7720 (output_indirect_thunk): Likewise.
7721 (output_indirect_thunk_function): Likewise.
7722 (ix86_output_indirect_branch_via_reg): Likewise.
7723 (ix86_output_indirect_branch_via_push): Likewise.
7724 (ix86_output_indirect_branch): Likewise.
7725 (ix86_output_indirect_jmp): Likewise.
7726 (ix86_code_end): Call output_indirect_thunk_function if needed.
7727 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7729 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7730 (ix86_attribute_table): Add indirect_branch.
7731 * config/i386/i386.h (machine_function): Add indirect_branch_type
7732 and has_local_indirect_jump.
7733 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7735 (tablejump): Likewise.
7736 (*indirect_jump): Use ix86_output_indirect_jmp.
7737 (*tablejump_1): Likewise.
7738 (simple_return_indirect_internal): Likewise.
7739 * config/i386/i386.opt (mindirect-branch=): New option.
7740 (indirect_branch): New.
7743 (thunk-inline): Likewise.
7744 (thunk-extern): Likewise.
7745 * doc/extend.texi: Document indirect_branch function attribute.
7746 * doc/invoke.texi: Document -mindirect-branch= option.
7748 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7751 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7753 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7755 * ipa-inline.c (want_inline_small_function_p): Return false if
7756 inlining has already failed with CIF_FINAL_ERROR.
7757 (update_caller_keys): Call want_inline_small_function_p before
7759 (update_callee_keys): Likewise.
7761 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7763 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7765 (rs6000_quadword_masked_address_p): Likewise.
7766 (quad_aligned_load_p): Likewise.
7767 (quad_aligned_store_p): Likewise.
7768 (const_load_sequence_p): Add comment to describe the outer-most loop.
7769 (mimic_memory_attributes_and_flags): New function.
7770 (rs6000_gen_stvx): Likewise.
7771 (replace_swapped_aligned_store): Likewise.
7772 (rs6000_gen_lvx): Likewise.
7773 (replace_swapped_aligned_load): Likewise.
7774 (replace_swapped_load_constant): Capitalize argument name in
7775 comment describing this function.
7776 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7777 and stores that access quad-word aligned addresses and replace
7778 with stvx or lvx instructions when appropriate.
7779 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7780 New function prototype.
7781 (rs6000_quadword_masked_address_p): Likewise.
7782 (rs6000_gen_lvx): Likewise.
7783 (rs6000_gen_stvx): Likewise.
7784 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7785 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7786 when memory address is aligned.
7787 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7788 this split to select lvx instruction when memory address is aligned.
7789 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7790 instruction when memory address is aligned.
7791 (*vsx_le_perm_load_v16qi): Likewise.
7792 (four unnamed splitters): Modify to select the stvx instruction
7793 when memory is aligned.
7795 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7797 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7798 which appears in the queue multiple times.
7800 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7801 Alan Hayward <alan.hayward@arm.com>
7802 David Sherwood <david.sherwood@arm.com>
7804 * tree-vectorizer.h (vec_lower_bound): New structure.
7805 (_loop_vec_info): Add check_nonzero and lower_bounds.
7806 (LOOP_VINFO_CHECK_NONZERO): New macro.
7807 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7808 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7809 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7810 fields. Make seg_len the distance travelled, not including the
7812 (dr_direction_indicator): Declare.
7813 (dr_zero_step_indicator): Likewise.
7814 (dr_known_forward_stride_p): Likewise.
7815 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7817 (runtime_alias_check_p): Allow runtime alias checks with
7819 (operator ==): Compare access_size and align.
7820 (prune_runtime_alias_test_list): Rework for new distinction between
7821 the access_size and seg_len.
7822 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7824 (get_segment_min_max): New function.
7825 (create_intersect_range_checks): Use it.
7826 (dr_step_indicator): New function.
7827 (dr_direction_indicator): Likewise.
7828 (dr_zero_step_indicator): Likewise.
7829 (dr_known_forward_stride_p): Likewise.
7830 * tree-loop-distribution.c (data_ref_segment_size): Return
7831 DR_STEP * (niters - 1).
7832 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7834 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7835 (vect_preserves_scalar_order_p): New function, split out from...
7836 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7837 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7838 (vect_vfa_access_size): New function.
7839 (vect_vfa_align): Likewise.
7840 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7841 (dump_lower_bound): New function.
7842 (vect_check_lower_bound): Likewise.
7843 (vect_small_gap_p): Likewise.
7844 (vectorizable_with_step_bound_p): Likewise.
7845 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7846 depencies if the vectorization factor is 1. Convert the checks
7847 for nonzero steps into checks on the bounds of DR_STEP. Try using
7848 a bunds check for variable steps if the minimum required step is
7849 relatively small. Update calls to the dr_with_seg_len
7850 constructor and to vect_compile_time_alias.
7851 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7853 (vect_loop_versioning): Call it.
7854 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7856 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7859 Alan Hayward <alan.hayward@arm.com>
7860 David Sherwood <david.sherwood@arm.com>
7862 * doc/sourcebuild.texi (vect_scatter_store): Document.
7863 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7865 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7867 * genopinit.c (main): Add supports_vec_scatter_store and
7868 supports_vec_scatter_store_cached to target_optabs.
7869 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7870 IFN_MASK_SCATTER_STORE.
7871 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7873 * internal-fn.h (internal_store_fn_p): Declare.
7874 (internal_fn_stored_value_index): Likewise.
7875 * internal-fn.c (scatter_store_direct): New macro.
7876 (expand_scatter_store_optab_fn): New function.
7877 (direct_scatter_store_optab_supported_p): New macro.
7878 (internal_store_fn_p): New function.
7879 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7880 IFN_MASK_SCATTER_STORE.
7881 (internal_fn_mask_index): Likewise.
7882 (internal_fn_stored_value_index): New function.
7883 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7885 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7886 * optabs-query.c (supports_vec_scatter_store_p): New function.
7887 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7888 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7889 true for scatter stores.
7890 (vect_gather_scatter_fn_p): Handle scatter stores too.
7891 (vect_check_gather_scatter): Consider using scatter stores if
7892 supports_vec_scatter_store_p.
7893 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7895 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7896 internal_fn_stored_value_index.
7897 (check_load_store_masking): Handle scatter stores too.
7898 (vect_get_store_rhs): Make public.
7899 (vectorizable_call): Use internal_store_fn_p.
7900 (vectorizable_store): Handle scatter store internal functions.
7901 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7902 when deciding whether the end of the group has been reached.
7903 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7904 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7905 (mask_scatter_store<mode>): New insns.
7907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7908 Alan Hayward <alan.hayward@arm.com>
7909 David Sherwood <david.sherwood@arm.com>
7911 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7912 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7913 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7915 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7916 Use vect_truncate_gather_scatter_offset if we can't treat the
7917 operation as a normal gather load or scatter store.
7918 (get_group_load_store_type): Take the gather_scatter_info
7919 as argument. Try using a gather load or scatter store for
7920 single-element groups.
7921 (get_load_store_type): Update calls to get_group_load_store_type
7922 and vect_use_strided_gather_scatters_p.
7924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7925 Alan Hayward <alan.hayward@arm.com>
7926 David Sherwood <david.sherwood@arm.com>
7928 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7929 optional tree argument.
7930 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7932 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7933 but continue to use the current value as a fallback.
7934 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7935 to compare the updates.
7936 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7937 (get_load_store_type): Use it when handling a strided access.
7938 (vect_get_strided_load_store_ops): New function.
7939 (vect_get_data_ptr_increment): Likewise.
7940 (vectorizable_load): Handle strided gather loads. Always pass
7941 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7944 Alan Hayward <alan.hayward@arm.com>
7945 David Sherwood <david.sherwood@arm.com>
7947 * doc/md.texi (gather_load@var{m}): Document.
7948 (mask_gather_load@var{m}): Likewise.
7949 * genopinit.c (main): Add supports_vec_gather_load and
7950 supports_vec_gather_load_cached to target_optabs.
7951 * optabs-tree.c (init_tree_optimization_optabs): Use
7952 ggc_cleared_alloc to allocate target_optabs.
7953 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7954 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7956 * internal-fn.h (internal_load_fn_p): Declare.
7957 (internal_gather_scatter_fn_p): Likewise.
7958 (internal_fn_mask_index): Likewise.
7959 (internal_gather_scatter_fn_supported_p): Likewise.
7960 * internal-fn.c (gather_load_direct): New macro.
7961 (expand_gather_load_optab_fn): New function.
7962 (direct_gather_load_optab_supported_p): New macro.
7963 (direct_internal_fn_optab): New function.
7964 (internal_load_fn_p): Likewise.
7965 (internal_gather_scatter_fn_p): Likewise.
7966 (internal_fn_mask_index): Likewise.
7967 (internal_gather_scatter_fn_supported_p): Likewise.
7968 * optabs-query.c (supports_at_least_one_mode_p): New function.
7969 (supports_vec_gather_load_p): Likewise.
7970 * optabs-query.h (supports_vec_gather_load_p): Declare.
7971 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7972 and memory_type field.
7973 (NUM_PATTERNS): Bump to 15.
7974 * tree-vect-data-refs.c: Include internal-fn.h.
7975 (vect_gather_scatter_fn_p): New function.
7976 (vect_describe_gather_scatter_call): Likewise.
7977 (vect_check_gather_scatter): Try using internal functions for
7978 gather loads. Recognize existing calls to a gather load function.
7979 (vect_analyze_data_refs): Consider using gather loads if
7980 supports_vec_gather_load_p.
7981 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7982 (vect_get_gather_scatter_offset_type): Likewise.
7983 (vect_convert_mask_for_vectype): Likewise.
7984 (vect_add_conversion_to_patterm): Likewise.
7985 (vect_try_gather_scatter_pattern): Likewise.
7986 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7987 (vect_vect_recog_func_ptrs): Add it.
7988 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7989 internal_fn_mask_index and internal_gather_scatter_fn_p.
7990 (check_load_store_masking): Take the gather_scatter_info as an
7991 argument and handle gather loads.
7992 (vect_get_gather_scatter_ops): New function.
7993 (vectorizable_call): Check internal_load_fn_p.
7994 (vectorizable_load): Likewise. Handle gather load internal
7996 (vectorizable_store): Update call to check_load_store_masking.
7997 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7998 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7999 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
8000 (aarch64_gather_scale_operand_d): New predicates.
8001 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
8002 (mask_gather_load<mode>): New insns.
8004 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8005 Alan Hayward <alan.hayward@arm.com>
8006 David Sherwood <david.sherwood@arm.com>
8008 * optabs.def (fold_left_plus_optab): New optab.
8009 * doc/md.texi (fold_left_plus_@var{m}): Document.
8010 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
8011 * internal-fn.c (fold_left_direct): Define.
8012 (expand_fold_left_optab_fn): Likewise.
8013 (direct_fold_left_optab_supported_p): Likewise.
8014 * fold-const-call.c (fold_const_fold_left): New function.
8015 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
8016 * tree-parloops.c (valid_reduction_p): New function.
8017 (gather_scalar_reductions): Use it.
8018 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
8019 (vect_finish_replace_stmt): Declare.
8020 * tree-vect-loop.c (fold_left_reduction_fn): New function.
8021 (needs_fold_left_reduction_p): New function, split out from...
8022 (vect_is_simple_reduction): ...here. Accept reductions that
8023 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
8024 (vect_force_simple_reduction): Also store the reduction type in
8025 the assignment's STMT_VINFO_REDUC_TYPE.
8026 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
8027 (merge_with_identity): New function.
8028 (vect_expand_fold_left): Likewise.
8029 (vectorize_fold_left_reduction): Likewise.
8030 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
8031 scalar phi in place for it. Check for target support and reject
8032 cases that would reassociate the operation. Defer the transform
8033 phase to vectorize_fold_left_reduction.
8034 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
8035 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
8036 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
8038 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8040 * tree-if-conv.c (predicate_mem_writes): Remove redundant
8041 call to ifc_temp_var.
8043 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8044 Alan Hayward <alan.hayward@arm.com>
8045 David Sherwood <david.sherwood@arm.com>
8047 * target.def (legitimize_address_displacement): Take the original
8048 offset as a poly_int.
8049 * targhooks.h (default_legitimize_address_displacement): Update
8051 * targhooks.c (default_legitimize_address_displacement): Likewise.
8052 * doc/tm.texi: Regenerate.
8053 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
8054 as an argument, moving assert of ad->disp == ad->disp_term to...
8055 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
8056 Try calling targetm.legitimize_address_displacement before expanding
8057 the address rather than afterwards, and adjust for the new interface.
8058 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
8059 Match the new hook interface. Handle SVE addresses.
8060 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
8063 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8065 * Makefile.in (OBJS): Add early-remat.o.
8066 * target.def (select_early_remat_modes): New hook.
8067 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
8068 * doc/tm.texi: Regenerate.
8069 * targhooks.h (default_select_early_remat_modes): Declare.
8070 * targhooks.c (default_select_early_remat_modes): New function.
8071 * timevar.def (TV_EARLY_REMAT): New timevar.
8072 * passes.def (pass_early_remat): New pass.
8073 * tree-pass.h (make_pass_early_remat): Declare.
8074 * early-remat.c: New file.
8075 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
8077 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
8079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8080 Alan Hayward <alan.hayward@arm.com>
8081 David Sherwood <david.sherwood@arm.com>
8083 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
8084 vfm1 with a bound_epilog parameter.
8085 (vect_do_peeling): Update calls accordingly, and move the prologue
8086 call earlier in the function. Treat the base bound_epilog as 0 for
8087 fully-masked loops and retain vf - 1 for other loops. Add 1 to
8088 this base when peeling for gaps.
8089 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
8090 with fully-masked loops.
8091 (vect_estimate_min_profitable_iters): Handle the single peeled
8092 iteration in that case.
8094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8095 Alan Hayward <alan.hayward@arm.com>
8096 David Sherwood <david.sherwood@arm.com>
8098 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
8099 single-element interleaving even if the size is not a power of 2.
8100 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
8101 accesses for single-element interleaving if the group size is
8104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8105 Alan Hayward <alan.hayward@arm.com>
8106 David Sherwood <david.sherwood@arm.com>
8108 * doc/md.texi (fold_extract_last_@var{m}): Document.
8109 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
8110 * optabs.def (fold_extract_last_optab): New optab.
8111 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
8112 * internal-fn.c (fold_extract_direct): New macro.
8113 (expand_fold_extract_optab_fn): Likewise.
8114 (direct_fold_extract_optab_supported_p): Likewise.
8115 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
8116 * tree-vect-loop.c (vect_model_reduction_cost): Handle
8117 EXTRACT_LAST_REDUCTION.
8118 (get_initial_def_for_reduction): Do not create an initial vector
8119 for EXTRACT_LAST_REDUCTION reductions.
8120 (vectorizable_reduction): Leave the scalar phi in place for
8121 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
8122 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
8123 epilogue code for EXTRACT_LAST_REDUCTION and defer the
8124 transform phase to vectorizable_condition.
8125 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
8127 (vect_finish_stmt_generation): ...here.
8128 (vect_finish_replace_stmt): New function.
8129 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
8130 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
8132 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
8134 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8135 Alan Hayward <alan.hayward@arm.com>
8136 David Sherwood <david.sherwood@arm.com>
8138 * doc/md.texi (extract_last_@var{m}): Document.
8139 * optabs.def (extract_last_optab): New optab.
8140 * internal-fn.def (EXTRACT_LAST): New internal function.
8141 * internal-fn.c (cond_unary_direct): New macro.
8142 (expand_cond_unary_optab_fn): Likewise.
8143 (direct_cond_unary_optab_supported_p): Likewise.
8144 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
8145 loops using EXTRACT_LAST.
8146 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
8147 (extract_last_<mode>): ...this optab.
8148 (vec_extract<mode><Vel>): Update accordingly.
8150 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8151 Alan Hayward <alan.hayward@arm.com>
8152 David Sherwood <david.sherwood@arm.com>
8154 * target.def (empty_mask_is_expensive): New hook.
8155 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
8156 * doc/tm.texi: Regenerate.
8157 * targhooks.h (default_empty_mask_is_expensive): Declare.
8158 * targhooks.c (default_empty_mask_is_expensive): New function.
8159 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
8160 if the target says that empty masks are expensive.
8161 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
8163 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
8165 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8166 Alan Hayward <alan.hayward@arm.com>
8167 David Sherwood <david.sherwood@arm.com>
8169 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
8170 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
8171 (vect_use_loop_mask_for_alignment_p): New function.
8172 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
8173 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
8174 niters_skip argument. Make sure that the first niters_skip elements
8175 of the first iteration are inactive.
8176 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
8177 Update call to vect_set_loop_masks_directly.
8178 (get_misalign_in_elems): New function, split out from...
8179 (vect_gen_prolog_loop_niters): ...here.
8180 (vect_update_init_of_dr): Take a code argument that specifies whether
8181 the adjustment should be added or subtracted.
8182 (vect_update_init_of_drs): Likewise.
8183 (vect_prepare_for_masked_peels): New function.
8184 (vect_do_peeling): Skip prologue peeling if we're using a mask
8185 instead. Update call to vect_update_inits_of_drs.
8186 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8188 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
8189 alignment. Do not include the number of peeled iterations in
8190 the minimum threshold in that case.
8191 (vectorizable_induction): Adjust the start value down by
8192 LOOP_VINFO_MASK_SKIP_NITERS iterations.
8193 (vect_transform_loop): Call vect_prepare_for_masked_peels.
8194 Take the number of skipped iterations into account when calculating
8196 * tree-vect-stmts.c (vect_gen_while_not): New function.
8198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8199 Alan Hayward <alan.hayward@arm.com>
8200 David Sherwood <david.sherwood@arm.com>
8202 * doc/sourcebuild.texi (vect_fully_masked): Document.
8203 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
8205 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8207 (vect_analyze_loop_2): ...here. Don't check the vectorization
8208 factor against the number of loop iterations if the loop is
8211 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8212 Alan Hayward <alan.hayward@arm.com>
8213 David Sherwood <david.sherwood@arm.com>
8215 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8216 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8217 (dump_groups): Update accordingly.
8218 (iv_use::mem_type): New member variable.
8219 (address_p): New function.
8220 (record_use): Add a mem_type argument and initialize the new
8222 (record_group_use): Add a mem_type argument. Use address_p.
8223 Remove obsolete null checks of base_object. Update call to record_use.
8224 (find_interesting_uses_op): Update call to record_group_use.
8225 (find_interesting_uses_cond): Likewise.
8226 (find_interesting_uses_address): Likewise.
8227 (get_mem_type_for_internal_fn): New function.
8228 (find_address_like_use): Likewise.
8229 (find_interesting_uses_stmt): Try find_address_like_use before
8230 calling find_interesting_uses_op.
8231 (addr_offset_valid_p): Use the iv mem_type field as the type
8232 of the addressed memory.
8233 (add_autoinc_candidates): Likewise.
8234 (get_address_cost): Likewise.
8235 (split_small_address_groups_p): Use address_p.
8236 (split_address_groups): Likewise.
8237 (add_iv_candidate_for_use): Likewise.
8238 (autoinc_possible_for_pair): Likewise.
8239 (rewrite_groups): Likewise.
8240 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8241 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8242 (get_alias_ptr_type_for_ptr_address): New function.
8243 (rewrite_use_address): Rewrite address uses in calls that were
8244 identified by find_address_like_use.
8246 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8247 Alan Hayward <alan.hayward@arm.com>
8248 David Sherwood <david.sherwood@arm.com>
8250 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8252 * gimple-expr.h (is_gimple_addressable: Likewise.
8253 * gimple-expr.c (is_gimple_address): Likewise.
8254 * internal-fn.c (expand_call_mem_ref): New function.
8255 (expand_mask_load_optab_fn): Use it.
8256 (expand_mask_store_optab_fn): Likewise.
8258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8259 Alan Hayward <alan.hayward@arm.com>
8260 David Sherwood <david.sherwood@arm.com>
8262 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8263 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8264 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8265 (cond_umax@var{mode}): Document.
8266 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8267 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8268 (cond_umin_optab, cond_umax_optab): New optabs.
8269 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8270 (COND_IOR, COND_XOR): New internal functions.
8271 * internal-fn.h (get_conditional_internal_fn): Declare.
8272 * internal-fn.c (cond_binary_direct): New macro.
8273 (expand_cond_binary_optab_fn): Likewise.
8274 (direct_cond_binary_optab_supported_p): Likewise.
8275 (get_conditional_internal_fn): New function.
8276 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8277 Cope with reduction statements that are vectorized as calls rather
8279 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8280 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8281 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8282 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8283 (UNSPEC_COND_EOR): New unspecs.
8284 (optab): Add mappings for them.
8285 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8286 (sve_int_op, sve_fp_op): New int attributes.
8288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8289 Alan Hayward <alan.hayward@arm.com>
8290 David Sherwood <david.sherwood@arm.com>
8292 * optabs.def (while_ult_optab): New optab.
8293 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8294 * internal-fn.def (WHILE_ULT): New internal function.
8295 * internal-fn.h (direct_internal_fn_supported_p): New override
8296 that takes two types as argument.
8297 * internal-fn.c (while_direct): New macro.
8298 (expand_while_optab_fn): New function.
8299 (convert_optab_supported_p): Likewise.
8300 (direct_while_optab_supported_p): New macro.
8301 * wide-int.h (wi::udiv_ceil): New function.
8302 * tree-vectorizer.h (rgroup_masks): New structure.
8303 (vec_loop_masks): New typedef.
8304 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8306 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8307 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8308 (vect_max_vf): New function.
8309 (slpeel_make_loop_iterate_ntimes): Delete.
8310 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8311 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8312 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8313 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8314 internal-fn.h, stor-layout.h and optabs-query.h.
8315 (vect_set_loop_mask): New function.
8316 (add_preheader_seq): Likewise.
8317 (add_header_seq): Likewise.
8318 (interleave_supported_p): Likewise.
8319 (vect_maybe_permute_loop_masks): Likewise.
8320 (vect_set_loop_masks_directly): Likewise.
8321 (vect_set_loop_condition_masked): Likewise.
8322 (vect_set_loop_condition_unmasked): New function, split out from
8323 slpeel_make_loop_iterate_ntimes.
8324 (slpeel_make_loop_iterate_ntimes): Rename to..
8325 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8326 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8327 (vect_do_peeling): Update call accordingly.
8328 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8330 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8331 mask_compare_type, can_fully_mask_p and fully_masked_p.
8332 (release_vec_loop_masks): New function.
8333 (_loop_vec_info): Use it to free the loop masks.
8334 (can_produce_all_loop_masks_p): New function.
8335 (vect_get_max_nscalars_per_iter): Likewise.
8336 (vect_verify_full_masking): Likewise.
8337 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8338 retries, and free the mask rgroups before retrying. Check loop-wide
8339 reasons for disallowing fully-masked loops. Make the final decision
8340 about whether use a fully-masked loop or not.
8341 (vect_estimate_min_profitable_iters): Do not assume that peeling
8342 for the number of iterations will be needed for fully-masked loops.
8343 (vectorizable_reduction): Disable fully-masked loops.
8344 (vectorizable_live_operation): Likewise.
8345 (vect_halve_mask_nunits): New function.
8346 (vect_double_mask_nunits): Likewise.
8347 (vect_record_loop_mask): Likewise.
8348 (vect_get_loop_mask): Likewise.
8349 (vect_transform_loop): Handle the case in which the final loop
8350 iteration might handle a partial vector. Call vect_set_loop_condition
8351 instead of slpeel_make_loop_iterate_ntimes.
8352 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8353 (check_load_store_masking): New function.
8354 (prepare_load_store_mask): Likewise.
8355 (vectorizable_store): Handle fully-masked loops.
8356 (vectorizable_load): Likewise.
8357 (supportable_widening_operation): Use vect_halve_mask_nunits for
8359 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8360 (vect_gen_while): New function.
8361 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8362 (aarch64_uqdec<mode>): New insn.
8364 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8365 Alan Hayward <alan.hayward@arm.com>
8366 David Sherwood <david.sherwood@arm.com>
8368 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8369 (reduc_xor_scal_optab): New optabs.
8370 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8371 (reduc_xor_scal_@var{m}): Document.
8372 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8373 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8375 * fold-const-call.c (fold_const_call): Handle them.
8376 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8377 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8378 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8379 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8380 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8381 (UNSPEC_XORV): New unspecs.
8382 (optab): Add entries for them.
8383 (BITWISEV): New int iterator.
8384 (bit_reduc_op): New int attributes.
8386 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8387 Alan Hayward <alan.hayward@arm.com>
8388 David Sherwood <david.sherwood@arm.com>
8390 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8391 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8392 * optabs.def (vec_shl_insert_optab): New optab.
8393 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8394 (duplicate_and_interleave): Likewise.
8395 * tree-vect-loop.c: Include internal-fn.h.
8396 (neutral_op_for_slp_reduction): New function, split out from
8397 get_initial_defs_for_reduction.
8398 (get_initial_def_for_reduction): Handle option 2 for variable-length
8399 vectors by loading the neutral value into a vector and then shifting
8400 the initial value into element 0.
8401 (get_initial_defs_for_reduction): Replace the code argument with
8402 the neutral value calculated by neutral_op_for_slp_reduction.
8403 Use gimple_build_vector for constant-length vectors.
8404 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8405 but the first group_size elements have a neutral value.
8406 Use duplicate_and_interleave otherwise.
8407 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8408 Update call to get_initial_defs_for_reduction. Handle SLP
8409 reductions for variable-length vectors by creating one vector
8410 result for each scalar result, with the elements associated
8411 with other scalar results stubbed out with the neutral value.
8412 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8413 Require IFN_VEC_SHL_INSERT for double reductions on
8414 variable-length vectors, or SLP reductions that have
8415 a neutral value. Require can_duplicate_and_interleave_p
8416 support for variable-length unchained SLP reductions if there
8417 is no neutral value, such as for MIN/MAX reductions. Also require
8418 the number of vector elements to be a multiple of the number of
8419 SLP statements when doing variable-length unchained SLP reductions.
8420 Update call to vect_create_epilog_for_reduction.
8421 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8422 and remove initial values.
8423 (duplicate_and_interleave): Make public.
8424 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8425 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8428 Alan Hayward <alan.hayward@arm.com>
8429 David Sherwood <david.sherwood@arm.com>
8431 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8432 (can_duplicate_and_interleave_p): New function.
8433 (vect_get_and_check_slp_defs): Take the vector of statements
8434 rather than just the current one. Remove excess parentheses.
8435 Restriction rejectinon of vect_constant_def and vect_external_def
8436 for variable-length vectors to boolean types, or types for which
8437 can_duplicate_and_interleave_p is false.
8438 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8439 (duplicate_and_interleave): New function.
8440 (vect_get_constant_vectors): Use gimple_build_vector for
8441 constant-length vectors and suitable variable-length constant
8442 vectors. Use duplicate_and_interleave for other variable-length
8443 vectors. Don't defer the update when inserting new statements.
8445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8446 Alan Hayward <alan.hayward@arm.com>
8447 David Sherwood <david.sherwood@arm.com>
8449 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8450 min_profitable_iters doesn't go negative.
8452 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8453 Alan Hayward <alan.hayward@arm.com>
8454 David Sherwood <david.sherwood@arm.com>
8456 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8457 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8458 * optabs.def (vec_mask_load_lanes_optab): New optab.
8459 (vec_mask_store_lanes_optab): Likewise.
8460 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8461 (MASK_STORE_LANES): Likewise.
8462 * internal-fn.c (mask_load_lanes_direct): New macro.
8463 (mask_store_lanes_direct): Likewise.
8464 (expand_mask_load_optab_fn): Handle masked operations.
8465 (expand_mask_load_lanes_optab_fn): New macro.
8466 (expand_mask_store_optab_fn): Handle masked operations.
8467 (expand_mask_store_lanes_optab_fn): New macro.
8468 (direct_mask_load_lanes_optab_supported_p): Likewise.
8469 (direct_mask_store_lanes_optab_supported_p): Likewise.
8470 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8472 (vect_load_lanes_supported): Likewise.
8473 * tree-vect-data-refs.c (strip_conversion): New function.
8474 (can_group_stmts_p): Likewise.
8475 (vect_analyze_data_ref_accesses): Use it instead of checking
8476 for a pair of assignments.
8477 (vect_store_lanes_supported): Take a masked_p parameter.
8478 (vect_load_lanes_supported): Likewise.
8479 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8480 vect_store_lanes_supported and vect_load_lanes_supported.
8481 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8482 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8483 parameter. Don't allow gaps for masked accesses.
8484 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8485 and vect_load_lanes_supported.
8486 (get_load_store_type): Take a masked_p parameter and update
8487 call to get_group_load_store_type.
8488 (vectorizable_store): Update call to get_load_store_type.
8489 Handle IFN_MASK_STORE_LANES.
8490 (vectorizable_load): Update call to get_load_store_type.
8491 Handle IFN_MASK_LOAD_LANES.
8493 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8494 Alan Hayward <alan.hayward@arm.com>
8495 David Sherwood <david.sherwood@arm.com>
8497 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8499 * config/aarch64/aarch64-protos.h
8500 (aarch64_sve_struct_memory_operand_p): Declare.
8501 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8502 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8503 (VPRED, vpred): Handle SVE structure modes.
8504 * config/aarch64/constraints.md (Utx): New constraint.
8505 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8506 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8507 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8508 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8509 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8510 structure modes. Split into pieces after RA.
8511 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8512 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8514 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8515 SVE structure modes.
8516 (aarch64_classify_address): Likewise.
8517 (sizetochar): Move earlier in file.
8518 (aarch64_print_operand): Handle SVE register lists.
8519 (aarch64_array_mode): New function.
8520 (aarch64_sve_struct_memory_operand_p): Likewise.
8521 (TARGET_ARRAY_MODE): Redefine.
8523 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8524 Alan Hayward <alan.hayward@arm.com>
8525 David Sherwood <david.sherwood@arm.com>
8527 * target.def (array_mode): New target hook.
8528 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8529 * doc/tm.texi: Regenerate.
8530 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8531 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8532 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8534 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8538 Alan Hayward <alan.hayward@arm.com>
8539 David Sherwood <david.sherwood@arm.com>
8541 * fold-const.c (fold_binary_loc): Check the argument types
8542 rather than the result type when testing for a vector operation.
8544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8546 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8547 * doc/tm.texi: Regenerate.
8549 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8550 Alan Hayward <alan.hayward@arm.com>
8551 David Sherwood <david.sherwood@arm.com>
8553 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8554 (sve): Document new AArch64 extension.
8555 * doc/md.texi (w): Extend the description of the AArch64
8556 constraint to include SVE vectors.
8557 (Upl, Upa): Document new AArch64 predicate constraints.
8558 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8560 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8561 (msve-vector-bits=): New option.
8562 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8563 SVE when these are disabled.
8564 (sve): New extension.
8565 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8566 modes. Adjust their number of units based on aarch64_sve_vg.
8567 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8568 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8569 aarch64_addr_query_type.
8570 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8571 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8572 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8573 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8574 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8575 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8576 (aarch64_simd_imm_zero_p): Delete.
8577 (aarch64_check_zero_based_sve_index_immediate): Declare.
8578 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8579 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8580 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8581 (aarch64_sve_float_mul_immediate_p): Likewise.
8582 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8584 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8585 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8586 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8587 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8588 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8589 (aarch64_regmode_natural_size): Likewise.
8590 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8591 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8593 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8594 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8595 for VG and the SVE predicate registers.
8596 (V_ALIASES): Add a "z"-prefixed alias.
8597 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8598 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8599 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8600 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8601 (REG_CLASS_NAMES): Add entries for them.
8602 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8603 and the predicate registers.
8604 (aarch64_sve_vg): Declare.
8605 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8606 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8607 (REGMODE_NATURAL_SIZE): Define.
8608 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8610 * config/aarch64/aarch64.c: Include cfgrtl.h.
8611 (simd_immediate_info): Add a constructor for series vectors,
8612 and an associated step field.
8613 (aarch64_sve_vg): New variable.
8614 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8615 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8616 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8617 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8618 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8619 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8620 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8621 (aarch64_get_mask_mode): New functions.
8622 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8623 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8624 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8625 predicate modes and predicate registers. Explicitly restrict
8626 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8627 to store a vector mode if it is recognized by
8628 aarch64_classify_vector_mode.
8629 (aarch64_regmode_natural_size): New function.
8630 (aarch64_hard_regno_caller_save_mode): Return the original mode
8632 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8633 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8634 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8635 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8637 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8638 does not overlap dest if the function is frame-related. Handle
8640 (aarch64_split_add_offset): New function.
8641 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8642 them aarch64_add_offset.
8643 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8644 and update call to aarch64_sub_sp.
8645 (aarch64_add_cfa_expression): New function.
8646 (aarch64_expand_prologue): Pass extra temporary registers to the
8647 functions above. Handle the case in which we need to emit new
8648 DW_CFA_expressions for registers that were originally saved
8649 relative to the stack pointer, but now have to be expressed
8650 relative to the frame pointer.
8651 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8653 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8654 IP0 and IP1 values for SVE frames.
8655 (aarch64_expand_vec_series): New function.
8656 (aarch64_expand_sve_widened_duplicate): Likewise.
8657 (aarch64_expand_sve_const_vector): Likewise.
8658 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8659 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8660 into the register, rather than emitting a SET directly.
8661 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8662 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8663 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8664 (offset_9bit_signed_scaled_p): New functions.
8665 (aarch64_replicate_bitmask_imm): New function.
8666 (aarch64_bitmask_imm): Use it.
8667 (aarch64_cannot_force_const_mem): Reject expressions involving
8668 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8669 (aarch64_classify_index): Handle SVE indices, by requiring
8670 a plain register index with a scale that matches the element size.
8671 (aarch64_classify_address): Handle SVE addresses. Assert that
8672 the mode of the address is VOIDmode or an integer mode.
8673 Update call to aarch64_classify_symbol.
8674 (aarch64_classify_symbolic_expression): Update call to
8675 aarch64_classify_symbol.
8676 (aarch64_const_vec_all_in_range_p): New function.
8677 (aarch64_print_vector_float_operand): Likewise.
8678 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8679 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8680 and the FP immediates 1.0 and 0.5.
8681 (aarch64_print_address_internal): Handle SVE addresses.
8682 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8683 (aarch64_regno_regclass): Handle predicate registers.
8684 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8686 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8687 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8688 (aarch64_convert_sve_vector_bits): New function.
8689 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8690 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8692 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8693 Handle SVE vector and predicate modes. Accept VL-based constants
8694 that need only one temporary register, and VL offsets that require
8695 no temporary registers.
8696 (aarch64_conditional_register_usage): Mark the predicate registers
8697 as fixed if SVE isn't available.
8698 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8699 Return true for SVE vector and predicate modes.
8700 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8701 rather than an unsigned int. Handle SVE modes.
8702 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8704 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8706 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8707 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8708 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8709 (aarch64_sve_float_mul_immediate_p): New functions.
8710 (aarch64_sve_valid_immediate): New function.
8711 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8712 Explicitly reject structure modes. Check for INDEX constants.
8713 Handle PTRUE and PFALSE constants.
8714 (aarch64_check_zero_based_sve_index_immediate): New function.
8715 (aarch64_simd_imm_zero_p): Delete.
8716 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8717 vector modes. Accept constants in the range of CNT[BHWD].
8718 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8719 ask for an Advanced SIMD mode.
8720 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8721 (aarch64_simd_vector_alignment): Handle SVE predicates.
8722 (aarch64_vectorize_preferred_vector_alignment): New function.
8723 (aarch64_simd_vector_alignment_reachable): Use it instead of
8725 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8726 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8728 (MAX_VECT_LEN): Delete.
8729 (expand_vec_perm_d): Add a vec_flags field.
8730 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8731 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8732 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8734 (aarch64_evpc_rev): Rename to...
8735 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8736 (aarch64_evpc_rev_global): New function.
8737 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8738 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8740 (aarch64_evpc_sve_tbl): New function.
8741 (aarch64_expand_vec_perm_const_1): Update after rename of
8742 aarch64_evpc_rev. Handle SVE permutes too, trying
8743 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8744 than aarch64_evpc_tbl.
8745 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8746 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8747 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8748 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8749 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8750 (aarch64_expand_sve_vcond): New functions.
8751 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8752 of aarch64_vector_mode_p.
8753 (aarch64_dwarf_poly_indeterminate_value): New function.
8754 (aarch64_compute_pressure_classes): Likewise.
8755 (aarch64_can_change_mode_class): Likewise.
8756 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8757 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8758 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8759 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8760 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8761 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8762 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8763 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8765 (Dn, Dl, Dr): Accept const as well as const_vector.
8766 (Dz): Likewise. Compare against CONST0_RTX.
8767 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8768 of "vector" where appropriate.
8769 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8770 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8771 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8772 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8773 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8774 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8775 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8776 (v_int_equiv): Extend to SVE modes.
8777 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8779 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8780 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8781 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8782 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8783 (SVE_COND_FP_CMP): New int iterators.
8784 (perm_hilo): Handle the new unpack unspecs.
8785 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8787 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8788 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8789 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8790 (aarch64_equality_operator, aarch64_constant_vector_operand)
8791 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8792 (aarch64_sve_nonimmediate_operand): Likewise.
8793 (aarch64_sve_general_operand): Likewise.
8794 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8795 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8796 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8797 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8798 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8799 (aarch64_sve_float_arith_immediate): Likewise.
8800 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8801 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8802 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8803 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8804 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8805 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8806 (aarch64_sve_float_arith_operand): Likewise.
8807 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8808 (aarch64_sve_float_mul_operand): Likewise.
8809 (aarch64_sve_vec_perm_operand): Likewise.
8810 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8811 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8812 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8813 as well as const_vector.
8814 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8815 in file. Use CONST0_RTX and CONSTM1_RTX.
8816 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8817 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8818 Use aarch64_simd_imm_zero.
8819 * config/aarch64/aarch64-sve.md: New file.
8820 * config/aarch64/aarch64.md: Include it.
8821 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8822 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8823 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8824 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8825 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8826 (sve): New attribute.
8827 (enabled): Disable instructions with the sve attribute unless
8829 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8830 aarch64_expand_mov_immediate.
8831 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8832 CNT[BHSD] immediates.
8833 (movti): Split CONST_POLY_INT moves into two halves.
8834 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8835 Split additions that need a temporary here if the destination
8836 is the stack pointer.
8837 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8838 (*add<mode>3_poly_1): New instruction.
8839 (set_clobber_cc): New expander.
8841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8843 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8844 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8845 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8846 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8847 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8848 Change innermode from fixed_mode_size to machine_mode.
8849 (simplify_subreg): Update call accordingly. Handle a constant-sized
8850 subreg of a variable-length CONST_VECTOR.
8852 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8853 Alan Hayward <alan.hayward@arm.com>
8854 David Sherwood <david.sherwood@arm.com>
8856 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8857 (add_offset_to_base): New function, split out from...
8858 (create_mem_ref): ...here. When handling a scale other than 1,
8859 check first whether the address is valid without the offset.
8860 Add it into the base if so, leaving the index and scale as-is.
8862 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8865 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8866 fold_for_warn before checking if arg2 is INTEGER_CST.
8868 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8870 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8871 (store_multiple_operation): Delete.
8872 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8873 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8874 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8875 guarded by TARGET_STRING.
8876 (rs6000_output_load_multiple): Delete.
8877 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8878 OPTION_MASK_STRING / TARGET_STRING handling.
8879 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8880 (const rs6000_opt_masks) <"string">: Change mask to 0.
8881 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8882 (MASK_STRING): Delete.
8883 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8885 (load_multiple): Delete.
8892 (store_multiple): Delete.
8899 (movmemsi_8reg): Delete.
8900 (corresponding unnamed define_insn): Delete.
8901 (movmemsi_6reg): Delete.
8902 (corresponding unnamed define_insn): Delete.
8903 (movmemsi_4reg): Delete.
8904 (corresponding unnamed define_insn): Delete.
8905 (movmemsi_2reg): Delete.
8906 (corresponding unnamed define_insn): Delete.
8907 (movmemsi_1reg): Delete.
8908 (corresponding unnamed define_insn): Delete.
8909 * config/rs6000/rs6000.opt (mno-string): New.
8910 (mstring): Replace by deprecation warning stub.
8911 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8913 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8915 * regrename.c (regrename_do_replace): If replacing the same
8916 reg multiple times, try to reuse last created gen_raw_REG.
8919 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8920 main to workaround a bug in GDB.
8922 2018-01-12 Tom de Vries <tom@codesourcery.com>
8925 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8927 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8929 PR rtl-optimization/80481
8930 * ira-color.c (get_cap_member): New function.
8931 (allocnos_conflict_by_live_ranges_p): Use it.
8932 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8933 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8935 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8938 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8939 (*saddl_se_1): Ditto.
8941 (*ssubl_se_1): Ditto.
8943 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8945 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8946 rather than wi::to_widest for DR_INITs.
8947 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8948 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8949 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8951 (vect_analyze_group_access_1): Note that here.
8953 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8955 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8956 polynomial type sizes.
8958 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8960 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8961 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8962 (gimple_add_tmp_var): Likewise.
8964 2018-01-12 Martin Liska <mliska@suse.cz>
8966 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8967 (gimple_alloc_sizes): Likewise.
8968 (dump_gimple_statistics): Use PRIu64 in printf format.
8969 * gimple.h: Change uint64_t to int.
8971 2018-01-12 Martin Liska <mliska@suse.cz>
8973 * tree-core.h: Use uint64_t instead of int.
8974 * tree.c (tree_node_counts): Likewise.
8975 (tree_node_sizes): Likewise.
8976 (dump_tree_statistics): Use PRIu64 in printf format.
8978 2018-01-12 Martin Liska <mliska@suse.cz>
8980 * Makefile.in: As qsort_chk is implemented in vec.c, add
8981 vec.o to linkage of gencfn-macros.
8982 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8983 passing the info to record_node_allocation_statistics.
8984 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8986 * ggc-common.c (struct ggc_usage): Add operator== and use
8987 it in operator< and compare function.
8988 * mem-stats.h (struct mem_usage): Likewise.
8989 * vec.c (struct vec_usage): Remove operator< and compare
8990 function. Can be simply inherited.
8992 2018-01-12 Martin Jambor <mjambor@suse.cz>
8995 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8996 * tree-ssa-math-opts.c: Include domwalk.h.
8997 (convert_mult_to_fma_1): New function.
8998 (fma_transformation_info): New type.
8999 (fma_deferring_state): Likewise.
9000 (cancel_fma_deferring): New function.
9001 (result_of_phi): Likewise.
9002 (last_fma_candidate_feeds_initial_phi): Likewise.
9003 (convert_mult_to_fma): Added deferring logic, split actual
9004 transformation to convert_mult_to_fma_1.
9005 (math_opts_dom_walker): New type.
9006 (math_opts_dom_walker::after_dom_children): New method, body moved
9007 here from pass_optimize_widening_mul::execute, added deferring logic
9009 (pass_optimize_widening_mul::execute): Moved most of code to
9010 math_opts_dom_walker::after_dom_children.
9011 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
9012 * config/i386/i386.c (ix86_option_override_internal): Added
9013 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
9015 2018-01-12 Richard Biener <rguenther@suse.de>
9018 * dwarf2out.c (gen_variable_die): Do not reset old_die for
9019 inline instance vars.
9021 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
9024 * config/rx/rx.c (rx_is_restricted_memory_address):
9027 2018-01-12 Richard Biener <rguenther@suse.de>
9029 PR tree-optimization/80846
9030 * target.def (split_reduction): New target hook.
9031 * targhooks.c (default_split_reduction): New function.
9032 * targhooks.h (default_split_reduction): Declare.
9033 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
9034 target requests first reduce vectors by combining low and high
9036 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
9037 (get_vectype_for_scalar_type_and_size): Export.
9038 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
9039 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
9040 * doc/tm.texi: Regenerate.
9041 * config/i386/i386.c (ix86_split_reduction): Implement
9042 TARGET_VECTORIZE_SPLIT_REDUCTION.
9044 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9047 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
9048 in PIC mode except for TARGET_VXWORKS_RTP.
9049 * config/sparc/sparc.c: Include cfgrtl.h.
9050 (TARGET_INIT_PIC_REG): Define.
9051 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
9052 (sparc_pic_register_p): New predicate.
9053 (sparc_legitimate_address_p): Use it.
9054 (sparc_legitimize_pic_address): Likewise.
9055 (sparc_delegitimize_address): Likewise.
9056 (sparc_mode_dependent_address_p): Likewise.
9057 (gen_load_pcrel_sym): Remove 4th parameter.
9058 (load_got_register): Adjust call to above. Remove obsolete stuff.
9059 (sparc_expand_prologue): Do not call load_got_register here.
9060 (sparc_flat_expand_prologue): Likewise.
9061 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
9062 (sparc_use_pseudo_pic_reg): New function.
9063 (sparc_init_pic_reg): Likewise.
9064 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
9065 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
9067 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
9069 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
9070 Add item for branch_cost.
9072 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
9074 PR rtl-optimization/83565
9075 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
9076 not extend the result to a larger mode for rotate operations.
9077 (num_sign_bit_copies1): Likewise.
9079 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9082 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
9084 Use values-Xc.o for -pedantic.
9085 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
9087 2018-01-12 Martin Liska <mliska@suse.cz>
9090 * ipa-devirt.c (final_warning_record::grow_type_warnings):
9092 (possible_polymorphic_call_targets): Use it.
9093 (ipa_devirt): Likewise.
9095 2018-01-12 Martin Liska <mliska@suse.cz>
9097 * profile-count.h (enum profile_quality): Use 0 as invalid
9098 enum value of profile_quality.
9100 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
9102 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
9103 -mext-string options.
9105 2018-01-12 Richard Biener <rguenther@suse.de>
9107 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
9108 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
9109 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
9111 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
9113 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
9115 * configure.ac (--with-long-double-format): Add support for the
9116 configuration option to change the default long double format on
9118 * config.gcc (powerpc*-linux*-*): Likewise.
9119 * configure: Regenerate.
9120 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
9121 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
9122 used without modification.
9124 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9126 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
9127 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
9128 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
9129 MISC_BUILTIN_SPEC_BARRIER.
9130 (rs6000_init_builtins): Likewise.
9131 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
9133 (speculation_barrier): New define_insn.
9134 * doc/extend.texi: Document __builtin_speculation_barrier.
9136 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9139 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
9140 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
9141 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
9143 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
9144 integral modes instead of "ss" and "sd".
9145 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
9146 vectors with 32-bit and 64-bit elements.
9147 (vecdupssescalarmodesuffix): New mode attribute.
9148 (vec_dup<mode>): Use it.
9150 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
9153 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
9154 frame if argument is passed on stack.
9156 2018-01-11 Jakub Jelinek <jakub@redhat.com>
9159 * ree.c (combine_reaching_defs): Optimize also
9160 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
9161 reg2=any_extend(exp); reg1=reg2;, formatting fix.
9163 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9166 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
9168 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
9171 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
9172 after they are computed.
9174 2018-01-11 Bin Cheng <bin.cheng@arm.com>
9176 PR tree-optimization/83695
9177 * gimple-loop-linterchange.cc
9178 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
9179 reset cached scev information after interchange.
9180 (pass_linterchange::execute): Remove call to scev_reset_htab.
9182 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9184 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
9185 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
9186 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
9187 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
9188 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
9189 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
9190 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
9191 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
9192 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
9193 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
9194 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
9195 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
9196 (V_lane_reg): Likewise.
9197 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
9199 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
9200 (vfmal_lane_low<mode>_intrinsic,
9201 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
9202 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
9203 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
9204 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
9205 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
9206 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9208 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9210 * config/arm/arm-cpus.in (fp16fml): New feature.
9211 (ALL_SIMD): Add fp16fml.
9212 (armv8.2-a): Add fp16fml as an option.
9213 (armv8.3-a): Likewise.
9214 (armv8.4-a): Add fp16fml as part of fp16.
9215 * config/arm/arm.h (TARGET_FP16FML): Define.
9216 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9218 * config/arm/arm-modes.def (V2HF): Define.
9219 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9220 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9221 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9222 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9223 vfmsl_low, vfmsl_high): New set of builtins.
9224 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9225 (vfml_op): New code attribute.
9226 (VFMLHALVES): New int iterator.
9227 (VFML, VFMLSEL): New mode attributes.
9228 (V_reg): Define mapping for V2HF.
9229 (V_hi, V_lo): New mode attributes.
9230 (VF_constraint): Likewise.
9231 (vfml_half, vfml_half_selector): New int attributes.
9232 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9234 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9235 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9237 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9238 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9239 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9240 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9242 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9243 Document new effective target and option set.
9245 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9247 * config/arm/arm-cpus.in (armv8_4): New feature.
9248 (ARMv8_4a): New fgroup.
9249 (armv8.4-a): New arch.
9250 * config/arm/arm-tables.opt: Regenerate.
9251 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9252 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9253 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9254 Add matching rules for -march=armv8.4-a and extensions.
9255 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9257 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9260 * config/rx/rx.md (BW): New mode attribute.
9261 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9263 2018-01-11 Richard Biener <rguenther@suse.de>
9265 PR tree-optimization/83435
9266 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9267 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9268 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9270 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9271 Alan Hayward <alan.hayward@arm.com>
9272 David Sherwood <david.sherwood@arm.com>
9274 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9276 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9277 (aarch64_print_address_internal): Use it to check for a zero offset.
9279 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9280 Alan Hayward <alan.hayward@arm.com>
9281 David Sherwood <david.sherwood@arm.com>
9283 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9284 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9285 Return a poly_int64 rather than a HOST_WIDE_INT.
9286 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9287 rather than a HOST_WIDE_INT.
9288 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9289 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9290 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9291 final_offset from HOST_WIDE_INT to poly_int64.
9292 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9293 to_constant when getting the number of units in an Advanced SIMD
9295 (aarch64_builtin_vectorized_function): Check for a constant number
9297 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9299 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9300 attribute instead of GET_MODE_NUNITS.
9301 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9302 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9303 GET_MODE_SIZE for fixed-size registers.
9304 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9305 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9306 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9307 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9308 (aarch64_print_operand, aarch64_print_address_internal)
9309 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9310 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9311 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9312 Handle polynomial GET_MODE_SIZE.
9313 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9314 wider than SImode without modification.
9315 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9316 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9317 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9318 passing and returning SVE modes.
9319 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9320 rather than GEN_INT.
9321 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9322 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9323 (aarch64_allocate_and_probe_stack_space): Likewise.
9324 (aarch64_layout_frame): Cope with polynomial offsets.
9325 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9326 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9328 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9329 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9330 poly_int64 rather than a HOST_WIDE_INT.
9331 (aarch64_get_separate_components, aarch64_process_components)
9332 (aarch64_expand_prologue, aarch64_expand_epilogue)
9333 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9334 (aarch64_anchor_offset): New function, split out from...
9335 (aarch64_legitimize_address): ...here.
9336 (aarch64_builtin_vectorization_cost): Handle polynomial
9337 TYPE_VECTOR_SUBPARTS.
9338 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9340 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9341 number of elements from the PARALLEL rather than the mode.
9342 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9343 rather than GET_MODE_BITSIZE.
9344 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9345 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9346 (aarch64_expand_vec_perm_const_1): Handle polynomial
9347 d->perm.length () and d->perm elements.
9348 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9349 Apply to_constant to d->perm elements.
9350 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9351 polynomial CONST_VECTOR_NUNITS.
9352 (aarch64_move_pointer): Take amount as a poly_int64 rather
9354 (aarch64_progress_pointer): Avoid temporary variable.
9355 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9356 the mode attribute instead of GET_MODE.
9358 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9359 Alan Hayward <alan.hayward@arm.com>
9360 David Sherwood <david.sherwood@arm.com>
9362 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9363 x exists before using it.
9364 (aarch64_add_constant_internal): Rename to...
9365 (aarch64_add_offset_1): ...this. Replace regnum with separate
9366 src and dest rtxes. Handle the case in which they're different,
9367 including when the offset is zero. Replace scratchreg with an rtx.
9368 Use 2 additions if there is no spare register into which we can
9369 move a 16-bit constant.
9370 (aarch64_add_constant): Delete.
9371 (aarch64_add_offset): Replace reg with separate src and dest
9372 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9373 Use aarch64_add_offset_1.
9374 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9375 an rtx rather than an int. Take the delta as a poly_int64
9376 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9377 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9378 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9379 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9380 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9382 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9383 aarch64_add_constant.
9385 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9387 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9388 Use scalar_float_mode.
9390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9392 * config/aarch64/aarch64-simd.md
9393 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9394 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9395 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9396 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9397 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9398 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9399 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9400 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9401 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9402 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9404 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9407 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9408 targ_options->x_arm_arch_string is non NULL.
9410 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9412 * config/aarch64/aarch64.h
9413 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9415 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9418 * expmed.c (emit_store_flag_force): Swap if const op0
9419 and change VOIDmode to mode of op0.
9421 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9423 PR rtl-optimization/83761
9424 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9425 than bytes to mode_for_size.
9427 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9430 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9431 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9434 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9437 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9438 when in layout mode.
9439 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9440 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9443 2018-01-10 Michael Collison <michael.collison@arm.com>
9445 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9446 * config/aarch64/aarch64-option-extension.def: Add
9447 AARCH64_OPT_EXTENSION of 'fp16fml'.
9448 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9449 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9450 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9451 * config/aarch64/constraints.md (Ui7): New constraint.
9452 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9453 (VFMLA_SEL_W): Ditto.
9456 (VFMLA16_LOW): New int iterator.
9457 (VFMLA16_HIGH): Ditto.
9458 (UNSPEC_FMLAL): New unspec.
9459 (UNSPEC_FMLSL): Ditto.
9460 (UNSPEC_FMLAL2): Ditto.
9461 (UNSPEC_FMLSL2): Ditto.
9462 (f16mac): New code attribute.
9463 * config/aarch64/aarch64-simd-builtins.def
9464 (aarch64_fmlal_lowv2sf): Ditto.
9465 (aarch64_fmlsl_lowv2sf): Ditto.
9466 (aarch64_fmlalq_lowv4sf): Ditto.
9467 (aarch64_fmlslq_lowv4sf): Ditto.
9468 (aarch64_fmlal_highv2sf): Ditto.
9469 (aarch64_fmlsl_highv2sf): Ditto.
9470 (aarch64_fmlalq_highv4sf): Ditto.
9471 (aarch64_fmlslq_highv4sf): Ditto.
9472 (aarch64_fmlal_lane_lowv2sf): Ditto.
9473 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9474 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9475 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9476 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9477 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9478 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9479 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9480 (aarch64_fmlal_lane_highv2sf): Ditto.
9481 (aarch64_fmlsl_lane_highv2sf): Ditto.
9482 (aarch64_fmlal_laneq_highv2sf): Ditto.
9483 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9484 (aarch64_fmlalq_lane_highv4sf): Ditto.
9485 (aarch64_fmlsl_lane_highv4sf): Ditto.
9486 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9487 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9488 * config/aarch64/aarch64-simd.md:
9489 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9490 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9491 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9492 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9493 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9494 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9495 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9496 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9497 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9498 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9499 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9500 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9501 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9502 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9503 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9504 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9505 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9506 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9507 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9508 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9509 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9510 (vfmlsl_low_u32): Ditto.
9511 (vfmlalq_low_u32): Ditto.
9512 (vfmlslq_low_u32): Ditto.
9513 (vfmlal_high_u32): Ditto.
9514 (vfmlsl_high_u32): Ditto.
9515 (vfmlalq_high_u32): Ditto.
9516 (vfmlslq_high_u32): Ditto.
9517 (vfmlal_lane_low_u32): Ditto.
9518 (vfmlsl_lane_low_u32): Ditto.
9519 (vfmlal_laneq_low_u32): Ditto.
9520 (vfmlsl_laneq_low_u32): Ditto.
9521 (vfmlalq_lane_low_u32): Ditto.
9522 (vfmlslq_lane_low_u32): Ditto.
9523 (vfmlalq_laneq_low_u32): Ditto.
9524 (vfmlslq_laneq_low_u32): Ditto.
9525 (vfmlal_lane_high_u32): Ditto.
9526 (vfmlsl_lane_high_u32): Ditto.
9527 (vfmlal_laneq_high_u32): Ditto.
9528 (vfmlsl_laneq_high_u32): Ditto.
9529 (vfmlalq_lane_high_u32): Ditto.
9530 (vfmlslq_lane_high_u32): Ditto.
9531 (vfmlalq_laneq_high_u32): Ditto.
9532 (vfmlslq_laneq_high_u32): Ditto.
9533 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9534 (AARCH64_FL_FOR_ARCH8_4): New.
9535 (AARCH64_ISA_F16FML): New ISA flag.
9536 (TARGET_F16FML): New feature flag for fp16fml.
9537 (doc/invoke.texi): Document new fp16fml option.
9539 2018-01-10 Michael Collison <michael.collison@arm.com>
9541 * config/aarch64/aarch64-builtins.c:
9542 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9543 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9544 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9545 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9546 (AARCH64_ISA_SHA3): New ISA flag.
9547 (TARGET_SHA3): New feature flag for sha3.
9548 * config/aarch64/iterators.md (sha512_op): New int attribute.
9549 (CRYPTO_SHA512): New int iterator.
9550 (UNSPEC_SHA512H): New unspec.
9551 (UNSPEC_SHA512H2): Ditto.
9552 (UNSPEC_SHA512SU0): Ditto.
9553 (UNSPEC_SHA512SU1): Ditto.
9554 * config/aarch64/aarch64-simd-builtins.def
9555 (aarch64_crypto_sha512hqv2di): New builtin.
9556 (aarch64_crypto_sha512h2qv2di): Ditto.
9557 (aarch64_crypto_sha512su0qv2di): Ditto.
9558 (aarch64_crypto_sha512su1qv2di): Ditto.
9559 (aarch64_eor3qv8hi): Ditto.
9560 (aarch64_rax1qv2di): Ditto.
9561 (aarch64_xarqv2di): Ditto.
9562 (aarch64_bcaxqv8hi): Ditto.
9563 * config/aarch64/aarch64-simd.md:
9564 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9565 (aarch64_crypto_sha512su0qv2di): Ditto.
9566 (aarch64_crypto_sha512su1qv2di): Ditto.
9567 (aarch64_eor3qv8hi): Ditto.
9568 (aarch64_rax1qv2di): Ditto.
9569 (aarch64_xarqv2di): Ditto.
9570 (aarch64_bcaxqv8hi): Ditto.
9571 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9572 (vsha512h2q_u64): Ditto.
9573 (vsha512su0q_u64): Ditto.
9574 (vsha512su1q_u64): Ditto.
9575 (veor3q_u16): Ditto.
9576 (vrax1q_u64): Ditto.
9578 (vbcaxq_u16): Ditto.
9579 * config/arm/types.md (crypto_sha512): New type attribute.
9580 (crypto_sha3): Ditto.
9581 (doc/invoke.texi): Document new sha3 option.
9583 2018-01-10 Michael Collison <michael.collison@arm.com>
9585 * config/aarch64/aarch64-builtins.c:
9586 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9587 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9588 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9589 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9590 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9591 (AARCH64_ISA_SM4): New ISA flag.
9592 (TARGET_SM4): New feature flag for sm4.
9593 * config/aarch64/aarch64-simd-builtins.def
9594 (aarch64_sm3ss1qv4si): Ditto.
9595 (aarch64_sm3tt1aq4si): Ditto.
9596 (aarch64_sm3tt1bq4si): Ditto.
9597 (aarch64_sm3tt2aq4si): Ditto.
9598 (aarch64_sm3tt2bq4si): Ditto.
9599 (aarch64_sm3partw1qv4si): Ditto.
9600 (aarch64_sm3partw2qv4si): Ditto.
9601 (aarch64_sm4eqv4si): Ditto.
9602 (aarch64_sm4ekeyqv4si): Ditto.
9603 * config/aarch64/aarch64-simd.md:
9604 (aarch64_sm3ss1qv4si): Ditto.
9605 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9606 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9607 (aarch64_sm4eqv4si): Ditto.
9608 (aarch64_sm4ekeyqv4si): Ditto.
9609 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9610 (sm3part_op): Ditto.
9611 (CRYPTO_SM3TT): Ditto.
9612 (CRYPTO_SM3PART): Ditto.
9613 (UNSPEC_SM3SS1): New unspec.
9614 (UNSPEC_SM3TT1A): Ditto.
9615 (UNSPEC_SM3TT1B): Ditto.
9616 (UNSPEC_SM3TT2A): Ditto.
9617 (UNSPEC_SM3TT2B): Ditto.
9618 (UNSPEC_SM3PARTW1): Ditto.
9619 (UNSPEC_SM3PARTW2): Ditto.
9620 (UNSPEC_SM4E): Ditto.
9621 (UNSPEC_SM4EKEY): Ditto.
9622 * config/aarch64/constraints.md (Ui2): New constraint.
9623 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9624 * config/arm/types.md (crypto_sm3): New type attribute.
9625 (crypto_sm4): Ditto.
9626 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9627 (vsm3tt1aq_u32): Ditto.
9628 (vsm3tt1bq_u32): Ditto.
9629 (vsm3tt2aq_u32): Ditto.
9630 (vsm3tt2bq_u32): Ditto.
9631 (vsm3partw1q_u32): Ditto.
9632 (vsm3partw2q_u32): Ditto.
9633 (vsm4eq_u32): Ditto.
9634 (vsm4ekeyq_u32): Ditto.
9635 (doc/invoke.texi): Document new sm4 option.
9637 2018-01-10 Michael Collison <michael.collison@arm.com>
9639 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9640 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9641 (AARCH64_FL_FOR_ARCH8_4): New.
9642 (AARCH64_FL_V8_4): New flag.
9643 (doc/invoke.texi): Document new armv8.4-a option.
9645 2018-01-10 Michael Collison <michael.collison@arm.com>
9647 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9648 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9649 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9650 * config/aarch64/aarch64-option-extension.def: Add
9651 AARCH64_OPT_EXTENSION of 'sha2'.
9652 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9653 (crypto): Disable sha2 and aes if crypto disabled.
9654 (crypto): Enable aes and sha2 if enabled.
9655 (simd): Disable sha2 and aes if simd disabled.
9656 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9658 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9659 (TARGET_SHA2): New feature flag for sha2.
9660 (TARGET_AES): New feature flag for aes.
9661 * config/aarch64/aarch64-simd.md:
9662 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9663 conditional on TARGET_AES.
9664 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9665 (aarch64_crypto_sha1hsi): Make pattern conditional
9667 (aarch64_crypto_sha1hv4si): Ditto.
9668 (aarch64_be_crypto_sha1hv4si): Ditto.
9669 (aarch64_crypto_sha1su1v4si): Ditto.
9670 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9671 (aarch64_crypto_sha1su0v4si): Ditto.
9672 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9673 (aarch64_crypto_sha256su0v4si): Ditto.
9674 (aarch64_crypto_sha256su1v4si): Ditto.
9675 (doc/invoke.texi): Document new aes and sha2 options.
9677 2018-01-10 Martin Sebor <msebor@redhat.com>
9679 PR tree-optimization/83781
9680 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9683 2018-01-11 Martin Sebor <msebor@gmail.com>
9684 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9686 PR tree-optimization/83501
9687 PR tree-optimization/81703
9689 * tree-ssa-strlen.c (get_string_cst): Rename...
9690 (get_string_len): ...to this. Handle global constants.
9691 (handle_char_store): Adjust.
9693 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9694 Jim Wilson <jimw@sifive.com>
9696 * config/riscv/riscv-protos.h (riscv_output_return): New.
9697 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9698 (riscv_attribute_table, riscv_output_return),
9699 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9700 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9701 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9703 (riscv_expand_prologue): Add early return for naked function.
9704 (riscv_expand_epilogue): Likewise.
9705 (riscv_function_ok_for_sibcall): Return false for naked function.
9706 (riscv_set_current_function): New.
9707 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9708 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9709 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9710 * doc/extend.texi (RISC-V Function Attributes): New.
9712 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9714 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9715 check for 128-bit long double before checking TCmode.
9716 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9717 128-bit long doubles before checking TFmode or TCmode.
9718 (FLOAT128_IBM_P): Likewise.
9720 2018-01-10 Martin Sebor <msebor@redhat.com>
9722 PR tree-optimization/83671
9723 * builtins.c (c_strlen): Unconditionally return zero for the empty
9725 Use -Warray-bounds for warnings.
9726 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9727 for non-constant array indices with COMPONENT_REF, arrays of
9728 arrays, and pointers to arrays.
9729 (gimple_fold_builtin_strlen): Determine and set length range for
9730 non-constant character arrays.
9732 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9735 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9738 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9740 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9742 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9745 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9746 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9747 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9748 indexed_or_indirect_operand predicate.
9749 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9750 (*vsx_le_perm_load_v8hi): Likewise.
9751 (*vsx_le_perm_load_v16qi): Likewise.
9752 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9753 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9754 (*vsx_le_perm_store_v8hi): Likewise.
9755 (*vsx_le_perm_store_v16qi): Likewise.
9756 (eight unnamed splitters): Likewise.
9758 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9760 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9761 * config/rs6000/emmintrin.h: Likewise.
9762 * config/rs6000/mmintrin.h: Likewise.
9763 * config/rs6000/xmmintrin.h: Likewise.
9765 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9768 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9770 * tree.c (tree_nop_conversion): Return true for location wrapper
9772 (maybe_wrap_with_location): New function.
9773 (selftest::check_strip_nops): New function.
9774 (selftest::test_location_wrappers): New function.
9775 (selftest::tree_c_tests): Call it.
9776 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9777 (maybe_wrap_with_location): New decl.
9778 (EXPR_LOCATION_WRAPPER_P): New macro.
9779 (location_wrapper_p): New inline function.
9780 (tree_strip_any_location_wrapper): New inline function.
9782 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9785 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9786 stack_realign_offset for the largest alignment of stack slot
9788 (ix86_find_max_used_stack_alignment): New function.
9789 (ix86_finalize_stack_frame_flags): Use it. Set
9790 max_used_stack_alignment if we don't realign stack.
9791 * config/i386/i386.h (machine_function): Add
9792 max_used_stack_alignment.
9794 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9796 * config/arm/arm.opt (-mbranch-cost): New option.
9797 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9800 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9803 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9804 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9806 2018-01-10 Richard Biener <rguenther@suse.de>
9809 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9810 early out so it also covers the case where we have a non-NULL
9813 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9815 PR tree-optimization/83753
9816 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9817 for non-strided grouped accesses if the number of elements is 1.
9819 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9822 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9823 * i386.h (TARGET_USE_GATHER): Define.
9824 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9826 2018-01-10 Martin Liska <mliska@suse.cz>
9829 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9830 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9832 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9833 CLEANUP_NO_PARTITIONING is not set.
9835 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9837 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9838 for vectors, as a partial revert of r254296.
9839 * rtl.h (const_vec_p): Delete.
9840 (const_vec_duplicate_p): Don't test for vector CONSTs.
9841 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9842 * expmed.c (make_tree): Likewise.
9845 * common.md (E, F): Use CONSTANT_P instead of checking for
9847 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9848 checking for CONST_VECTOR.
9850 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9853 * predict.c (force_edge_cold): Handle in more sane way edges
9856 2018-01-09 Carl Love <cel@us.ibm.com>
9858 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9860 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9861 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9862 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9863 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9864 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9865 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9866 * config/rs6000/rs6000-protos.h: Add extern defition for
9867 rs6000_generate_float2_double_code.
9868 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9870 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9871 (float2_v2df): Add define_expand.
9873 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9876 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9877 op_mode in the force_to_mode call.
9879 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9881 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9882 instead of checking each element individually.
9883 (aarch64_evpc_uzp): Likewise.
9884 (aarch64_evpc_zip): Likewise.
9885 (aarch64_evpc_ext): Likewise.
9886 (aarch64_evpc_rev): Likewise.
9887 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9888 instead of checking each element individually. Return true without
9890 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9891 whether all selected elements come from the same input, instead of
9892 checking each element individually. Remove calls to gen_rtx_REG,
9893 start_sequence and end_sequence and instead assert that no rtl is
9896 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9898 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9899 order of HIGH and CONST checks.
9901 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9903 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9904 if the destination isn't an SSA_NAME.
9906 2018-01-09 Richard Biener <rguenther@suse.de>
9908 PR tree-optimization/83668
9909 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9911 (canonicalize_loop_form): ... here, renamed from ...
9912 (canonicalize_loop_closed_ssa_form): ... this and amended to
9913 swap successor edges for loop exit blocks to make us use
9914 the RPO order we need for initial schedule generation.
9916 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9918 PR tree-optimization/64811
9919 * match.pd: When optimizing comparisons with Inf, avoid
9920 introducing or losing exceptions from comparisons with NaN.
9922 2018-01-09 Martin Liska <mliska@suse.cz>
9925 * asan.c (shadow_mem_size): Add gcc_assert.
9927 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9929 Don't save registers in main().
9932 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9933 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9934 * config/avr/avr.c (avr_set_current_function): Don't error if
9935 naked, OS_task or OS_main are specified at the same time.
9936 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9938 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9940 * common/config/avr/avr-common.c (avr_option_optimization_table):
9941 Switch on -mmain-is-OS_task for optimizing compilations.
9943 2018-01-09 Richard Biener <rguenther@suse.de>
9945 PR tree-optimization/83572
9946 * graphite.c: Include cfganal.h.
9947 (graphite_transform_loops): Connect infinite loops to exit
9948 and remove fake edges at the end.
9950 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9952 * ipa-inline.c (edge_badness): Revert accidental checkin.
9954 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9957 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9958 symbols; not inline clones.
9960 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9963 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9964 hard registers. Formatting fixes.
9966 PR preprocessor/83722
9967 * gcc.c (try_generate_repro): Pass
9968 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9969 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9972 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9973 Kito Cheng <kito.cheng@gmail.com>
9975 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9976 (riscv_leaf_function_p): Delete.
9977 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9979 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9981 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9983 (do_ifelse): New function.
9984 (do_isel): New function.
9985 (do_sub3): New function.
9986 (do_add3): New function.
9987 (do_load_mask_compare): New function.
9988 (do_overlap_load_compare): New function.
9989 (expand_compare_loop): New function.
9990 (expand_block_compare): Call expand_compare_loop() when appropriate.
9991 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9993 (-mblock-compare-inline-loop-limit): New option.
9995 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9998 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9999 Reverse order of second and third operands in first alternative.
10000 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
10001 of first and second elements in UNSPEC_VPERMR vector.
10002 (altivec_expand_vec_perm_le): Likewise.
10004 2018-01-08 Jeff Law <law@redhat.com>
10006 PR rtl-optimizatin/81308
10007 * tree-switch-conversion.c (cfg_altered): New file scoped static.
10008 (process_switch): If group_case_labels makes a change, then set
10010 (pass_convert_switch::execute): If a switch is converted, then
10011 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
10013 PR rtl-optimization/81308
10014 * recog.c (split_all_insns): Conditionally cleanup the CFG after
10017 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
10019 PR target/83663 - Revert r255946
10020 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
10021 generation for cases where splatting a value is not useful.
10022 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
10023 across a vec_duplicate and a paradoxical subreg forming a vector
10024 mode to a vec_concat.
10026 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10028 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
10029 -march=armv8.3-a variants.
10030 * config/arm/t-multilib: Likewise.
10031 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
10033 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10035 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
10037 (cceq_ior_compare_complement): Give it a name so I can use it, and
10038 change boolean_or_operator predicate to boolean_operator so it can
10039 be used to generate a crand.
10040 (eqne): New code iterator.
10041 (bd/bd_neg): New code_attrs.
10042 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
10043 a single define_insn.
10044 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
10045 decrement (bdnzt/bdnzf/bdzt/bdzf).
10046 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
10047 with the new names of the branch decrement patterns, and added the
10048 names of the branch decrement conditional patterns.
10050 2018-01-08 Richard Biener <rguenther@suse.de>
10052 PR tree-optimization/83563
10053 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
10056 2018-01-08 Richard Biener <rguenther@suse.de>
10058 PR middle-end/83713
10059 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
10061 2018-01-08 Richard Biener <rguenther@suse.de>
10063 PR tree-optimization/83685
10064 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
10065 references to abnormals.
10067 2018-01-08 Richard Biener <rguenther@suse.de>
10070 * dwarf2out.c (output_indirect_strings): Handle empty
10071 skeleton_debug_str_hash.
10072 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
10074 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10076 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
10077 (emit_store_direct): Likewise.
10078 (arc_trampoline_adjust_address): Likewise.
10079 (arc_asm_trampoline_template): New function.
10080 (arc_initialize_trampoline): Use asm_trampoline_template.
10081 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
10082 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
10083 * config/arc/arc.md (flush_icache): Delete pattern.
10085 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
10087 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
10088 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
10091 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10094 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
10095 by not USED_FOR_TARGET.
10096 (make_pass_resolve_sw_modes): Likewise.
10098 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
10100 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
10103 2018-01-08 Richard Biener <rguenther@suse.de>
10105 PR middle-end/83580
10106 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
10108 2018-01-08 Richard Biener <rguenther@suse.de>
10110 PR middle-end/83517
10111 * match.pd ((t * 2) / 2) -> t): Add missing :c.
10113 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
10115 PR middle-end/81897
10116 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
10117 basic blocks with a small number of successors.
10118 (convert_control_dep_chain_into_preds): Improve handling of
10120 (dump_predicates): Split apart into...
10121 (dump_pred_chain): ...here...
10122 (dump_pred_info): ...and here.
10123 (can_one_predicate_be_invalidated_p): Add debugging printfs.
10124 (can_chain_union_be_invalidated_p): Improve check for invalidation
10126 (uninit_uses_cannot_happen): Avoid unnecessary if
10127 convert_control_dep_chain_into_preds yielded nothing.
10129 2018-01-06 Martin Sebor <msebor@redhat.com>
10131 PR tree-optimization/83640
10132 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
10133 subtracting negative offset from size.
10134 (builtin_access::overlap): Adjust offset bounds of the access to fall
10135 within the size of the object if possible.
10137 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
10139 PR rtl-optimization/83699
10140 * expmed.c (extract_bit_field_1): Restrict the vector usage of
10141 extract_bit_field_as_subreg to cases in which the extracted
10142 value is also a vector.
10144 * lra-constraints.c (process_alt_operands): Test for the equivalence
10145 substitutions when detecting a possible reload cycle.
10147 2018-01-06 Jakub Jelinek <jakub@redhat.com>
10150 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
10151 by default if flag_selective_schedling{,2}. Formatting fixes.
10153 PR rtl-optimization/83682
10154 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
10155 if it has non-VECTOR_MODE element mode.
10156 (vec_duplicate_p): Likewise.
10158 PR middle-end/83694
10159 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
10160 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
10162 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10165 * config/i386/i386-builtin.def
10166 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10167 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
10168 Require also OPTION_MASK_ISA_AVX512F in addition to
10169 OPTION_MASK_ISA_GFNI.
10170 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
10171 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
10172 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
10173 to OPTION_MASK_ISA_GFNI.
10174 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
10175 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
10176 OPTION_MASK_ISA_AVX512BW.
10177 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
10178 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
10179 addition to OPTION_MASK_ISA_GFNI.
10180 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
10181 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
10182 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
10183 to OPTION_MASK_ISA_GFNI.
10184 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
10185 a requirement for all ISAs rather than any of them with a few
10187 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
10189 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
10190 bitmasks to be enabled with 3 exceptions, instead of requiring any
10191 enabled ISA with lots of exceptions.
10192 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
10193 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
10194 Change avx512bw in isa attribute to avx512f.
10195 * config/i386/sgxintrin.h: Add license boilerplate.
10196 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
10197 to __AVX512F__ and __AVX512VL to __AVX512VL__.
10198 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
10199 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
10201 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
10202 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
10203 temporarily sse2 rather than sse if not enabled already.
10206 * config/i386/sse.md (VI248_VLBW): Rename to ...
10207 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10208 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10209 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10210 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10211 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10212 mode iterator instead of VI248_VLBW.
10214 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10216 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10217 (record_modified): Skip clobbers; add debug output.
10218 (param_change_prob): Use sreal frequencies.
10220 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10222 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10223 punt for user-aligned variables.
10225 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10227 * tree-chrec.c (chrec_contains_symbols): Return true for
10230 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10233 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10234 of (x|y) == x for BICS pattern.
10236 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10238 PR tree-optimization/83605
10239 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10240 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10243 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10245 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10246 * config/epiphany/rtems.h: New file.
10248 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10249 Uros Bizjak <ubizjak@gmail.com>
10252 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10253 QIreg_operand instead of register_operand predicate.
10254 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10255 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10256 comments instead of -fmitigate[-_]rop.
10258 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10261 * cgraphunit.c (symbol_table::compile): Switch to text_section
10262 before calling assembly_start debug hook.
10263 * run-rtl-passes.c (run_rtl_passes): Likewise.
10266 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10268 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10269 range_int_cst_p rather than !symbolic_range_p before calling
10270 extract_range_from_multiplicative_op_1.
10272 2018-01-04 Jeff Law <law@redhat.com>
10274 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10275 redundant test in assertion.
10277 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10279 * doc/rtl.texi: Document machine_mode wrapper classes.
10281 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10283 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10284 using tree_to_uhwi.
10286 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10288 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10289 the VEC_PERM_EXPR fold to fail.
10291 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10294 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10295 to switched_sections.
10297 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10300 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10301 test for d.testing.
10303 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10306 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10307 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10309 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10312 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10313 is BLKmode and bitpos not zero or mode change is needed.
10315 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10318 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10321 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10324 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10325 instead of MULT rtx. Update all corresponding splitters.
10326 (*saddl_se): Ditto.
10327 (*ssub<modesuffix>): Ditto.
10328 (*ssubl_se): Ditto.
10329 (*cmp_sadd_di): Update split patterns.
10330 (*cmp_sadd_si): Ditto.
10331 (*cmp_sadd_sidi): Ditto.
10332 (*cmp_ssub_di): Ditto.
10333 (*cmp_ssub_si): Ditto.
10334 (*cmp_ssub_sidi): Ditto.
10335 * config/alpha/predicates.md (const23_operand): New predicate.
10336 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10337 Look for ASHIFT, not MULT inner operand.
10338 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10340 2018-01-04 Martin Liska <mliska@suse.cz>
10342 PR gcov-profile/83669
10343 * gcov.c (output_intermediate_file): Add version to intermediate
10345 * doc/gcov.texi: Document new field 'version' in intermediate
10346 file format. Fix location of '-k' option of gcov command.
10348 2018-01-04 Martin Liska <mliska@suse.cz>
10351 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10353 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10355 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10357 2018-01-03 Martin Sebor <msebor@redhat.com>
10359 PR tree-optimization/83655
10360 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10361 checking calls with invalid arguments.
10363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10365 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10366 (vectorizable_mask_load_store): Delete.
10367 (vectorizable_call): Return false for masked loads and stores.
10368 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10369 instead of gimple_assign_rhs1.
10370 (vectorizable_load): Handle IFN_MASK_LOAD.
10371 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10375 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10377 (vectorizable_mask_load_store): ...here.
10378 (vectorizable_load): ...and here.
10380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10382 * tree-vect-stmts.c (vect_build_all_ones_mask)
10383 (vect_build_zero_merge_argument): New functions, split out from...
10384 (vectorizable_load): ...here.
10386 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10388 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10390 (vectorizable_mask_load_store): ...here.
10391 (vectorizable_store): ...and here.
10393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10395 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10397 (vectorizable_mask_load_store): ...here.
10399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10401 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10402 (vect_model_store_cost): Take a vec_load_store_type instead of a
10404 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10405 (vect_model_store_cost): Take a vec_load_store_type instead of a
10407 (vectorizable_mask_load_store): Update accordingly.
10408 (vectorizable_store): Likewise.
10409 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10413 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10414 IFN_MASK_LOAD calls here rather than...
10415 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10418 Alan Hayward <alan.hayward@arm.com>
10419 David Sherwood <david.sherwood@arm.com>
10421 * expmed.c (extract_bit_field_1): For vector extracts,
10422 fall back to extract_bit_field_as_subreg if vec_extract
10425 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10426 Alan Hayward <alan.hayward@arm.com>
10427 David Sherwood <david.sherwood@arm.com>
10429 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10430 they are variable or constant sized.
10431 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10432 slots for constant-sized data.
10434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10435 Alan Hayward <alan.hayward@arm.com>
10436 David Sherwood <david.sherwood@arm.com>
10438 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10439 handling COND_EXPRs with boolean comparisons, try to find a better
10440 basis for the mask type than the boolean itself.
10442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10444 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10445 is calculated and how it can be overridden.
10446 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10447 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10449 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10453 Alan Hayward <alan.hayward@arm.com>
10454 David Sherwood <david.sherwood@arm.com>
10456 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10457 Remove the mode argument.
10458 (aarch64_simd_valid_immediate): Remove the mode and inverse
10460 * config/aarch64/iterators.md (bitsize): New iterator.
10461 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10462 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10463 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10464 aarch64_simd_valid_immediate.
10465 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10466 (aarch64_reg_or_bic_imm): Likewise.
10467 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10468 with an insn_type enum and msl with a modifier_type enum.
10469 Replace element_width with a scalar_mode. Change the shift
10470 to unsigned int. Add constructors for scalar_float_mode and
10471 scalar_int_mode elements.
10472 (aarch64_vect_float_const_representable_p): Delete.
10473 (aarch64_can_const_movi_rtx_p)
10474 (aarch64_simd_scalar_immediate_valid_for_move)
10475 (aarch64_simd_make_constant): Update call to
10476 aarch64_simd_valid_immediate.
10477 (aarch64_advsimd_valid_immediate_hs): New function.
10478 (aarch64_advsimd_valid_immediate): Likewise.
10479 (aarch64_simd_valid_immediate): Remove mode and inverse
10480 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10481 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10482 and aarch64_float_const_representable_p on the result.
10483 (aarch64_output_simd_mov_immediate): Remove mode argument.
10484 Update call to aarch64_simd_valid_immediate and use of
10485 simd_immediate_info.
10486 (aarch64_output_scalar_simd_mov_immediate): Update call
10489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10490 Alan Hayward <alan.hayward@arm.com>
10491 David Sherwood <david.sherwood@arm.com>
10493 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10494 (mode_nunits): Likewise CONST_MODE_NUNITS.
10495 * machmode.def (ADJUST_NUNITS): Document.
10496 * genmodes.c (mode_data::need_nunits_adj): New field.
10497 (blank_mode): Update accordingly.
10498 (adj_nunits): New variable.
10499 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10501 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10502 listed in adj_nunits.
10503 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10504 listed in adj_nunits. Don't emit case statements for such modes.
10505 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10506 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10507 nothing if adj_nunits is nonnull.
10508 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10509 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10510 (emit_mode_fbit): Update use of print_maybe_const_decl.
10511 (emit_move_size): Likewise. Treat the array as non-const
10513 (emit_mode_adjustments): Handle adj_nunits.
10515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10517 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10518 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10519 (VECTOR_MODES): Use it.
10520 (make_vector_modes): Take the prefix as an argument.
10522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10523 Alan Hayward <alan.hayward@arm.com>
10524 David Sherwood <david.sherwood@arm.com>
10526 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10527 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10528 for MODE_VECTOR_BOOL.
10529 * machmode.def (VECTOR_BOOL_MODE): Document.
10530 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10531 (make_vector_bool_mode): New function.
10532 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10534 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10535 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10537 * stor-layout.c (int_mode_for_mode): Likewise.
10538 * tree.c (build_vector_type_for_mode): Likewise.
10539 * varasm.c (output_constant_pool_2): Likewise.
10540 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10541 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10542 for MODE_VECTOR_BOOL.
10543 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10544 of mode class checks.
10545 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10546 instead of a list of mode class checks.
10547 (expand_vector_scalar_condition): Likewise.
10548 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10551 Alan Hayward <alan.hayward@arm.com>
10552 David Sherwood <david.sherwood@arm.com>
10554 * machmode.h (mode_size): Change from unsigned short to
10556 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10557 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10558 or if measurement_type is not polynomial.
10559 (fixed_size_mode::includes_p): Check for constant-sized modes.
10560 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10561 return a poly_uint16 rather than an unsigned short.
10562 (emit_mode_size): Change the type of mode_size from unsigned short
10563 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10564 (emit_mode_adjustments): Cope with polynomial vector sizes.
10565 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10567 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10569 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10570 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10571 * caller-save.c (setup_save_areas): Likewise.
10572 (replace_reg_with_saved_mem): Likewise.
10573 * calls.c (emit_library_call_value_1): Likewise.
10574 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10575 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10576 (gen_lowpart_for_combine): Likewise.
10577 * convert.c (convert_to_integer_1): Likewise.
10578 * cse.c (equiv_constant, cse_insn): Likewise.
10579 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10580 (cselib_subst_to_values): Likewise.
10581 * dce.c (word_dce_process_block): Likewise.
10582 * df-problems.c (df_word_lr_mark_ref): Likewise.
10583 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10584 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10585 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10586 (rtl_for_decl_location): Likewise.
10587 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10588 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10589 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10590 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10591 (expand_expr_real_1): Likewise.
10592 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10593 (pad_below): Likewise.
10594 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10595 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10596 * ira.c (get_subreg_tracking_sizes): Likewise.
10597 * ira-build.c (ira_create_allocno_objects): Likewise.
10598 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10599 (ira_sort_regnos_for_alter_reg): Likewise.
10600 * ira-costs.c (record_operand_costs): Likewise.
10601 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10602 (resolve_simple_move): Likewise.
10603 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10604 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10605 (lra_constraints): Likewise.
10606 (CONST_POOL_OK_P): Reject variable-sized modes.
10607 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10608 (add_pseudo_to_slot, lra_spill): Likewise.
10609 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10610 * optabs-query.c (get_best_extraction_insn): Likewise.
10611 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10612 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10613 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10614 * recog.c (offsettable_address_addr_space_p): Likewise.
10615 * regcprop.c (maybe_mode_change): Likewise.
10616 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10617 * regrename.c (build_def_use): Likewise.
10618 * regstat.c (dump_reg_info): Likewise.
10619 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10620 (find_reloads, find_reloads_subreg_address): Likewise.
10621 * reload1.c (eliminate_regs_1): Likewise.
10622 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10623 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10624 (simplify_binary_operation_1, simplify_subreg): Likewise.
10625 * targhooks.c (default_function_arg_padding): Likewise.
10626 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10627 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10628 (verify_gimple_assign_ternary): Likewise.
10629 * tree-inline.c (estimate_move_cost): Likewise.
10630 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10631 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10632 (get_address_cost_ainc): Likewise.
10633 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10634 (vect_supportable_dr_alignment): Likewise.
10635 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10636 (vectorizable_reduction): Likewise.
10637 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10638 (vectorizable_operation, vectorizable_load): Likewise.
10639 * tree.c (build_same_sized_truth_vector_type): Likewise.
10640 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10641 * var-tracking.c (emit_note_insn_var_location): Likewise.
10642 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10643 (ADDR_VEC_ALIGN): Likewise.
10645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10646 Alan Hayward <alan.hayward@arm.com>
10647 David Sherwood <david.sherwood@arm.com>
10649 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10651 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10652 or if measurement_type is polynomial.
10653 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10654 * combine.c (make_extraction): Likewise.
10655 * dse.c (find_shift_sequence): Likewise.
10656 * dwarf2out.c (mem_loc_descriptor): Likewise.
10657 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10658 (extract_bit_field, extract_low_bits): Likewise.
10659 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10660 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10661 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10662 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10663 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10664 * reload.c (find_reloads): Likewise.
10665 * reload1.c (alter_reg): Likewise.
10666 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10667 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10668 * tree-if-conv.c (predicate_mem_writes): Likewise.
10669 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10670 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10671 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10672 * valtrack.c (dead_debug_insert_temp): Likewise.
10673 * varasm.c (mergeable_constant_section): Likewise.
10674 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10677 Alan Hayward <alan.hayward@arm.com>
10678 David Sherwood <david.sherwood@arm.com>
10680 * expr.c (expand_assignment): Cope with polynomial mode sizes
10681 when assigning to a CONCAT.
10683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10684 Alan Hayward <alan.hayward@arm.com>
10685 David Sherwood <david.sherwood@arm.com>
10687 * machmode.h (mode_precision): Change from unsigned short to
10689 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10691 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10692 or if measurement_type is not polynomial.
10693 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10694 in which the mode is already known to be a scalar_int_mode.
10695 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10696 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10698 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10699 for GET_MODE_PRECISION.
10700 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10701 for GET_MODE_PRECISION.
10702 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10704 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10705 (expand_field_assignment, make_extraction): Likewise.
10706 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10707 (get_last_value): Likewise.
10708 * convert.c (convert_to_integer_1): Likewise.
10709 * cse.c (cse_insn): Likewise.
10710 * expr.c (expand_expr_real_1): Likewise.
10711 * lra-constraints.c (simplify_operand_subreg): Likewise.
10712 * optabs-query.c (can_atomic_load_p): Likewise.
10713 * optabs.c (expand_atomic_load): Likewise.
10714 (expand_atomic_store): Likewise.
10715 * ree.c (combine_reaching_defs): Likewise.
10716 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10717 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10718 * tree.h (type_has_mode_precision_p): Likewise.
10719 * ubsan.c (instrument_si_overflow): Likewise.
10721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10722 Alan Hayward <alan.hayward@arm.com>
10723 David Sherwood <david.sherwood@arm.com>
10725 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10726 polynomial numbers of units.
10727 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10728 (valid_vector_subparts_p): New function.
10729 (build_vector_type): Remove temporary shim and take the number
10730 of units as a poly_uint64 rather than an int.
10731 (build_opaque_vector_type): Take the number of units as a
10732 poly_uint64 rather than an int.
10733 * tree.c (build_vector_from_ctor): Handle polynomial
10734 TYPE_VECTOR_SUBPARTS.
10735 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10736 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10737 (build_vector_from_val): If the number of units is variable,
10738 use build_vec_duplicate_cst for constant operands and
10739 VEC_DUPLICATE_EXPR otherwise.
10740 (make_vector_type): Remove temporary is_constant ().
10741 (build_vector_type, build_opaque_vector_type): Take the number of
10742 units as a poly_uint64 rather than an int.
10743 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10745 * cfgexpand.c (expand_debug_expr): Likewise.
10746 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10747 (store_constructor, expand_expr_real_1): Likewise.
10748 (const_scalar_mask_from_tree): Likewise.
10749 * fold-const-call.c (fold_const_reduction): Likewise.
10750 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10751 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10752 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10753 (fold_relational_const): Likewise.
10754 (native_interpret_vector): Likewise. Change the size from an
10755 int to an unsigned int.
10756 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10757 TYPE_VECTOR_SUBPARTS.
10758 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10759 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10760 duplicating a non-constant operand into a variable-length vector.
10761 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10762 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10763 * ipa-icf.c (sem_variable::equals): Likewise.
10764 * match.pd: Likewise.
10765 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10766 * print-tree.c (print_node): Likewise.
10767 * stor-layout.c (layout_type): Likewise.
10768 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10769 * tree-cfg.c (verify_gimple_comparison): Likewise.
10770 (verify_gimple_assign_binary): Likewise.
10771 (verify_gimple_assign_ternary): Likewise.
10772 (verify_gimple_assign_single): Likewise.
10773 * tree-pretty-print.c (dump_generic_node): Likewise.
10774 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10775 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10776 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10777 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10778 (vect_shift_permute_load_chain): Likewise.
10779 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10780 (expand_vector_condition, optimize_vector_constructor): Likewise.
10781 (lower_vec_perm, get_compute_type): Likewise.
10782 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10783 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10784 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10785 (vect_recog_mask_conversion_pattern): Likewise.
10786 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10787 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10788 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10789 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10790 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10791 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10792 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10793 (supportable_widening_operation): Likewise.
10794 (supportable_narrowing_operation): Likewise.
10795 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10797 * varasm.c (output_constant): Likewise.
10799 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10800 Alan Hayward <alan.hayward@arm.com>
10801 David Sherwood <david.sherwood@arm.com>
10803 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10804 so that both the length == 3 and length != 3 cases set up their
10805 own permute vectors. Add comments explaining why we know the
10806 number of elements is constant.
10807 (vect_permute_load_chain): Likewise.
10809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10810 Alan Hayward <alan.hayward@arm.com>
10811 David Sherwood <david.sherwood@arm.com>
10813 * machmode.h (mode_nunits): Change from unsigned char to
10815 (ONLY_FIXED_SIZE_MODES): New macro.
10816 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10817 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10818 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10820 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10821 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10822 or if measurement_type is not polynomial.
10823 * genmodes.c (ZERO_COEFFS): New macro.
10824 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10826 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10827 Use ZERO_COEFFS when emitting initializers.
10828 * data-streamer.h (bp_pack_poly_value): New function.
10829 (bp_unpack_poly_value): Likewise.
10830 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10831 for GET_MODE_NUNITS.
10832 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10833 for GET_MODE_NUNITS.
10834 * tree.c (make_vector_type): Remove temporary shim and make
10835 the real function take the number of units as a poly_uint64
10836 rather than an int.
10837 (build_vector_type_for_mode): Handle polynomial nunits.
10838 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10839 * emit-rtl.c (const_vec_series_p_1): Likewise.
10840 (gen_rtx_CONST_VECTOR): Likewise.
10841 * fold-const.c (test_vec_duplicate_folding): Likewise.
10842 * genrecog.c (validate_pattern): Likewise.
10843 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10844 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10845 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10846 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10847 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10848 * rtlanal.c (subreg_get_info): Likewise.
10849 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10850 (vect_grouped_load_supported): Likewise.
10851 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10852 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10853 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10854 (simplify_const_unary_operation, simplify_binary_operation_1)
10855 (simplify_const_binary_operation, simplify_ternary_operation)
10856 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10857 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10858 instead of CONST_VECTOR_NUNITS.
10859 * varasm.c (output_constant_pool_2): Likewise.
10860 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10861 explicit-encoded elements in the XVEC for variable-length vectors.
10863 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10865 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10868 Alan Hayward <alan.hayward@arm.com>
10869 David Sherwood <david.sherwood@arm.com>
10871 * coretypes.h (fixed_size_mode): Declare.
10872 (fixed_size_mode_pod): New typedef.
10873 * builtins.h (target_builtins::x_apply_args_mode)
10874 (target_builtins::x_apply_result_mode): Change type to
10875 fixed_size_mode_pod.
10876 * builtins.c (apply_args_size, apply_result_size, result_vector)
10877 (expand_builtin_apply_args_1, expand_builtin_apply)
10878 (expand_builtin_return): Update accordingly.
10880 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10882 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10883 * cselib.c (cselib_hash_rtx): Likewise.
10884 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10885 CONST_VECTOR encoding.
10887 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10888 Jeff Law <law@redhat.com>
10891 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10892 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10893 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10894 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10897 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10898 explicitly probe *sp in a noreturn function if there were any callee
10899 register saves or frame pointer is needed.
10901 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10904 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10905 BLKmode for ternary, binary or unary expressions.
10908 * var-tracking.c (delete_vta_debug_insn): New inline function.
10909 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10910 insns from get_insns () to NULL instead of each bb separately.
10911 Use delete_vta_debug_insn. No longer static.
10912 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10913 delete_vta_debug_insns callers.
10914 * rtl.h (delete_vta_debug_insns): Declare.
10915 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10916 instead of variable_tracking_main.
10918 2018-01-03 Martin Sebor <msebor@redhat.com>
10920 PR tree-optimization/83603
10921 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10922 arguments past the endof the argument list in functions declared
10923 without a prototype.
10924 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10925 Avoid checking when arguments are null.
10927 2018-01-03 Martin Sebor <msebor@redhat.com>
10930 * doc/extend.texi (attribute const): Fix a typo.
10931 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10932 issuing -Wsuggest-attribute for void functions.
10934 2018-01-03 Martin Sebor <msebor@redhat.com>
10936 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10937 offset_int::from instead of wide_int::to_shwi.
10938 (maybe_diag_overlap): Remove assertion.
10939 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10940 * gimple-ssa-sprintf.c (format_directive): Same.
10941 (parse_directive): Same.
10942 (sprintf_dom_walker::compute_format_length): Same.
10943 (try_substitute_return_value): Same.
10945 2018-01-03 Jeff Law <law@redhat.com>
10947 PR middle-end/83654
10948 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10949 non-constant residual for zero at runtime and avoid probing in
10950 that case. Reorganize code for trailing problem to mirror handling
10953 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10955 PR tree-optimization/83501
10956 * tree-ssa-strlen.c (get_string_cst): New.
10957 (handle_char_store): Call get_string_cst.
10959 2018-01-03 Martin Liska <mliska@suse.cz>
10961 PR tree-optimization/83593
10962 * tree-ssa-strlen.c: Include tree-cfg.h.
10963 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10964 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10965 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10967 (strlen_dom_walker::before_dom_children): Call
10968 gimple_purge_dead_eh_edges. Dump tranformation with details
10970 (strlen_dom_walker::before_dom_children): Update call by adding
10971 new argument cleanup_eh.
10972 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10974 2018-01-03 Martin Liska <mliska@suse.cz>
10977 * cif-code.def (VARIADIC_THUNK): New enum value.
10978 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10981 2018-01-03 Jan Beulich <jbeulich@suse.com>
10983 * sse.md (mov<mode>_internal): Tighten condition for when to use
10984 vmovdqu<ssescalarsize> for TI and OI modes.
10986 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10988 Update copyright years.
10990 2018-01-03 Martin Liska <mliska@suse.cz>
10993 * ipa-visibility.c (function_and_variable_visibility): Skip
10994 functions with noipa attribure.
10996 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10998 * gcc.c (process_command): Update copyright notice dates.
10999 * gcov-dump.c (print_version): Ditto.
11000 * gcov.c (print_version): Ditto.
11001 * gcov-tool.c (print_version): Ditto.
11002 * gengtype.c (create_file): Ditto.
11003 * doc/cpp.texi: Bump @copying's copyright year.
11004 * doc/cppinternals.texi: Ditto.
11005 * doc/gcc.texi: Ditto.
11006 * doc/gccint.texi: Ditto.
11007 * doc/gcov.texi: Ditto.
11008 * doc/install.texi: Ditto.
11009 * doc/invoke.texi: Ditto.
11011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11013 * vector-builder.h (vector_builder::m_full_nelts): Change from
11014 unsigned int to poly_uint64.
11015 (vector_builder::full_nelts): Update prototype accordingly.
11016 (vector_builder::new_vector): Likewise.
11017 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
11018 (vector_builder::operator ==): Likewise.
11019 (vector_builder::finalize): Likewise.
11020 * int-vector-builder.h (int_vector_builder::int_vector_builder):
11021 Take the number of elements as a poly_uint64 rather than an
11023 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
11024 from unsigned int to poly_uint64.
11025 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
11026 (vec_perm_indices::new_vector): Likewise.
11027 (vec_perm_indices::length): Likewise.
11028 (vec_perm_indices::nelts_per_input): Likewise.
11029 (vec_perm_indices::input_nelts): Likewise.
11030 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
11031 number of elements per input as a poly_uint64 rather than an
11032 unsigned int. Use the original encoding for variable-length
11033 vectors, rather than clamping each individual element.
11034 For the second and subsequent elements in each pattern,
11035 clamp the step and base before clamping their sum.
11036 (vec_perm_indices::series_p): Handle polynomial element counts.
11037 (vec_perm_indices::all_in_range_p): Likewise.
11038 (vec_perm_indices_to_tree): Likewise.
11039 (vec_perm_indices_to_rtx): Likewise.
11040 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
11041 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
11042 (tree_vector_builder::new_binary_operation): Handle polynomial
11043 element counts. Return false if we need to know the number
11044 of elements at compile time.
11045 * fold-const.c (fold_vec_perm): Punt if the number of elements
11046 isn't known at compile time.
11048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11050 * vec-perm-indices.h (vec_perm_builder): Change element type
11051 from HOST_WIDE_INT to poly_int64.
11052 (vec_perm_indices::element_type): Update accordingly.
11053 (vec_perm_indices::clamp): Handle polynomial element_types.
11054 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11055 (vec_perm_indices::all_in_range_p): Likewise.
11056 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
11058 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
11059 polynomial vec_perm_indices element types.
11060 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
11061 * fold-const.c (fold_vec_perm): Likewise.
11062 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
11063 * tree-vect-generic.c (lower_vec_perm): Likewise.
11064 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11065 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
11066 element type to HOST_WIDE_INT.
11068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11069 Alan Hayward <alan.hayward@arm.com>
11070 David Sherwood <david.sherwood@arm.com>
11072 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
11073 rather than an int. Use plus_constant.
11074 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
11075 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
11077 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11078 Alan Hayward <alan.hayward@arm.com>
11079 David Sherwood <david.sherwood@arm.com>
11081 * calls.c (emit_call_1, expand_call): Change struct_value_size from
11082 a HOST_WIDE_INT to a poly_int64.
11084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11085 Alan Hayward <alan.hayward@arm.com>
11086 David Sherwood <david.sherwood@arm.com>
11088 * calls.c (load_register_parameters): Cope with polynomial
11089 mode sizes. Require a constant size for BLKmode parameters
11090 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
11091 forces a parameter to be padded at the lsb end in order to
11092 fill a complete number of words, require the parameter size
11093 to be ordered wrt UNITS_PER_WORD.
11095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11096 Alan Hayward <alan.hayward@arm.com>
11097 David Sherwood <david.sherwood@arm.com>
11099 * reload1.c (spill_stack_slot_width): Change element type
11100 from unsigned int to poly_uint64_pod.
11101 (alter_reg): Treat mode sizes as polynomial.
11103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11104 Alan Hayward <alan.hayward@arm.com>
11105 David Sherwood <david.sherwood@arm.com>
11107 * reload.c (complex_word_subreg_p): New function.
11108 (reload_inner_reg_of_subreg, push_reload): Use it.
11110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11111 Alan Hayward <alan.hayward@arm.com>
11112 David Sherwood <david.sherwood@arm.com>
11114 * lra-constraints.c (process_alt_operands): Reject matched
11115 operands whose sizes aren't ordered.
11116 (match_reload): Refer to this check here.
11118 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11119 Alan Hayward <alan.hayward@arm.com>
11120 David Sherwood <david.sherwood@arm.com>
11122 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
11123 that the mode size is in the set {1, 2, 4, 8, 16}.
11125 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11126 Alan Hayward <alan.hayward@arm.com>
11127 David Sherwood <david.sherwood@arm.com>
11129 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
11130 Use plus_constant instead of gen_rtx_PLUS.
11132 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11133 Alan Hayward <alan.hayward@arm.com>
11134 David Sherwood <david.sherwood@arm.com>
11136 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
11137 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
11138 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
11139 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
11140 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
11141 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
11142 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
11143 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
11144 * config/i386/i386.c (ix86_push_rounding): ...this new function.
11145 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
11147 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
11148 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
11149 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
11150 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
11151 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
11152 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
11153 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
11154 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
11155 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
11156 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
11158 * expr.c (emit_move_resolve_push): Treat the input and result
11159 of PUSH_ROUNDING as a poly_int64.
11160 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
11161 (emit_push_insn): Likewise.
11162 * lra-eliminations.c (mark_not_eliminable): Likewise.
11163 * recog.c (push_operand): Likewise.
11164 * reload1.c (elimination_effects): Likewise.
11165 * rtlanal.c (nonzero_bits1): Likewise.
11166 * calls.c (store_one_arg): Likewise. Require the padding to be
11167 known at compile time.
11169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11170 Alan Hayward <alan.hayward@arm.com>
11171 David Sherwood <david.sherwood@arm.com>
11173 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
11174 Use plus_constant instead of gen_rtx_PLUS.
11176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11177 Alan Hayward <alan.hayward@arm.com>
11178 David Sherwood <david.sherwood@arm.com>
11180 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
11181 rather than an int.
11183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11184 Alan Hayward <alan.hayward@arm.com>
11185 David Sherwood <david.sherwood@arm.com>
11187 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
11188 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
11189 via stack temporaries. Treat the mode size as polynomial too.
11191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11192 Alan Hayward <alan.hayward@arm.com>
11193 David Sherwood <david.sherwood@arm.com>
11195 * expr.c (expand_expr_real_2): When handling conversions involving
11196 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
11197 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
11198 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
11206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11207 Alan Hayward <alan.hayward@arm.com>
11208 David Sherwood <david.sherwood@arm.com>
11210 * combine.c (can_change_dest_mode): Handle polynomial
11211 REGMODE_NATURAL_SIZE.
11212 * expmed.c (store_bit_field_1): Likewise.
11213 * expr.c (store_constructor): Likewise.
11214 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11215 and polynomial REGMODE_NATURAL_SIZE.
11216 (gen_lowpart_common): Likewise.
11217 * reginfo.c (record_subregs_of_mode): Likewise.
11218 * rtlanal.c (read_modify_subreg_p): Likewise.
11220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11221 Alan Hayward <alan.hayward@arm.com>
11222 David Sherwood <david.sherwood@arm.com>
11224 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11225 numbers of 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 * match.pd: Cope with polynomial numbers of vector elements.
11233 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11234 Alan Hayward <alan.hayward@arm.com>
11235 David Sherwood <david.sherwood@arm.com>
11237 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11238 in a POINTER_PLUS_EXPR.
11240 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11241 Alan Hayward <alan.hayward@arm.com>
11242 David Sherwood <david.sherwood@arm.com>
11244 * omp-simd-clone.c (simd_clone_subparts): New function.
11245 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11246 (ipa_simd_modify_function_body): Likewise.
11248 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11249 Alan Hayward <alan.hayward@arm.com>
11250 David Sherwood <david.sherwood@arm.com>
11252 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11253 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11254 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11255 (expand_vector_condition, vector_element): Likewise.
11256 (subparts_gt): New function.
11257 (get_compute_type): Use subparts_gt.
11258 (count_type_subparts): Delete.
11259 (expand_vector_operations_1): Use subparts_gt instead of
11260 count_type_subparts.
11262 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11263 Alan Hayward <alan.hayward@arm.com>
11264 David Sherwood <david.sherwood@arm.com>
11266 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11267 (vect_compile_time_alias): ...this new function. Do the calculation
11268 on poly_ints rather than trees.
11269 (vect_prune_runtime_alias_test_list): Update call accordingly.
11271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11272 Alan Hayward <alan.hayward@arm.com>
11273 David Sherwood <david.sherwood@arm.com>
11275 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11277 (vect_schedule_slp_instance): Likewise.
11279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11280 Alan Hayward <alan.hayward@arm.com>
11281 David Sherwood <david.sherwood@arm.com>
11283 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11284 constant and extern definitions for variable-length vectors.
11285 (vect_get_constant_vectors): Note that the number of units
11286 is known to be constant.
11288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11289 Alan Hayward <alan.hayward@arm.com>
11290 David Sherwood <david.sherwood@arm.com>
11292 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11293 of units as polynomial. Choose between WIDE and NARROW based
11296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11297 Alan Hayward <alan.hayward@arm.com>
11298 David Sherwood <david.sherwood@arm.com>
11300 * tree-vect-stmts.c (simd_clone_subparts): New function.
11301 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11304 Alan Hayward <alan.hayward@arm.com>
11305 David Sherwood <david.sherwood@arm.com>
11307 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11308 vectors as polynomial. Use build_index_vector for
11309 IFN_GOMP_SIMD_LANE.
11311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11312 Alan Hayward <alan.hayward@arm.com>
11313 David Sherwood <david.sherwood@arm.com>
11315 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11316 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11317 for variable-length vectors.
11318 (vectorizable_mask_load_store): Treat the number of units as
11319 polynomial, asserting that it is constant if the condition has
11320 already been enforced.
11321 (vectorizable_store, vectorizable_load): Likewise.
11323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11324 Alan Hayward <alan.hayward@arm.com>
11325 David Sherwood <david.sherwood@arm.com>
11327 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11328 of units as polynomial. Punt if we can't tell at compile time
11329 which vector contains the final result.
11331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11332 Alan Hayward <alan.hayward@arm.com>
11333 David Sherwood <david.sherwood@arm.com>
11335 * tree-vect-loop.c (vectorizable_induction): Treat the number
11336 of units as polynomial. Punt on SLP inductions. Use an integer
11337 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11338 cast of such a series for variable-length floating-point
11341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11342 Alan Hayward <alan.hayward@arm.com>
11343 David Sherwood <david.sherwood@arm.com>
11345 * tree.h (build_index_vector): Declare.
11346 * tree.c (build_index_vector): New function.
11347 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11348 of units as polynomial, forcibly converting it to a constant if
11349 vectorizable_reduction has already enforced the condition.
11350 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11351 to create a {1,2,3,...} vector.
11352 (vectorizable_reduction): Treat the number of units as polynomial.
11353 Choose vectype_in based on the largest scalar element size rather
11354 than the smallest number of units. Enforce the restrictions
11357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11358 Alan Hayward <alan.hayward@arm.com>
11359 David Sherwood <david.sherwood@arm.com>
11361 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11362 number of units as polynomial.
11364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11365 Alan Hayward <alan.hayward@arm.com>
11366 David Sherwood <david.sherwood@arm.com>
11368 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11369 * target.def (autovectorize_vector_sizes): Return the vector sizes
11370 by pointer, using vector_sizes rather than a bitmask.
11371 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11372 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11373 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11375 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11376 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11377 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11378 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11379 * omp-general.c (omp_max_vf): Likewise.
11380 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11381 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11382 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11383 * tree-vect-slp.c (vect_slp_bb): Likewise.
11384 * doc/tm.texi: Regenerate.
11385 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11387 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11388 the vector size as a poly_uint64 rather than an unsigned int.
11389 (current_vector_size): Change from an unsigned int to a poly_uint64.
11390 (get_vectype_for_scalar_type): Update accordingly.
11391 * tree.h (build_truth_vector_type): Take the size and number of
11392 units as a poly_uint64 rather than an unsigned int.
11393 (build_vector_type): Add a temporary overload that takes
11394 the number of units as a poly_uint64 rather than an unsigned int.
11395 * tree.c (make_vector_type): Likewise.
11396 (build_truth_vector_type): Take the number of units as a poly_uint64
11397 rather than an unsigned int.
11399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11400 Alan Hayward <alan.hayward@arm.com>
11401 David Sherwood <david.sherwood@arm.com>
11403 * target.def (get_mask_mode): Take the number of units and length
11404 as poly_uint64s rather than unsigned ints.
11405 * targhooks.h (default_get_mask_mode): Update accordingly.
11406 * targhooks.c (default_get_mask_mode): Likewise.
11407 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11408 * doc/tm.texi: Regenerate.
11410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11411 Alan Hayward <alan.hayward@arm.com>
11412 David Sherwood <david.sherwood@arm.com>
11414 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11415 * omp-general.c (omp_max_vf): Likewise.
11416 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11417 (expand_omp_simd): Handle polynomial safelen.
11418 * omp-low.c (omplow_simd_context): Add a default constructor.
11419 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11420 (lower_rec_simd_input_clauses): Update accordingly.
11421 (lower_rec_input_clauses): Likewise.
11423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11424 Alan Hayward <alan.hayward@arm.com>
11425 David Sherwood <david.sherwood@arm.com>
11427 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11428 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11429 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11430 (vect_analyze_slp_cost): Likewise.
11431 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11432 (vect_model_load_cost): Likewise.
11434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11435 Alan Hayward <alan.hayward@arm.com>
11436 David Sherwood <david.sherwood@arm.com>
11438 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11439 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11440 from an unsigned int * to a poly_uint64_pod *.
11441 (calculate_unrolling_factor): New function.
11442 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11445 Alan Hayward <alan.hayward@arm.com>
11446 David Sherwood <david.sherwood@arm.com>
11448 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11449 from an unsigned int to a poly_uint64.
11450 (_loop_vec_info::slp_unrolling_factor): Likewise.
11451 (_loop_vec_info::vectorization_factor): Change from an int
11453 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11454 (vect_get_num_vectors): New function.
11455 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11456 (vect_get_num_copies): Use vect_get_num_vectors.
11457 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11458 to an unsigned int *.
11459 (vect_analyze_data_refs): Change min_vf from an int * to a
11461 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11462 than an unsigned HOST_WIDE_INT.
11463 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11464 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11465 to an unsigned int *.
11466 (vect_analyze_data_ref_dependences): Likewise.
11467 (vect_compute_data_ref_alignment): Handle polynomial vf.
11468 (vect_enhance_data_refs_alignment): Likewise.
11469 (vect_prune_runtime_alias_test_list): Likewise.
11470 (vect_shift_permute_load_chain): Likewise.
11471 (vect_supportable_dr_alignment): Likewise.
11472 (dependence_distance_ge_vf): Take the vectorization factor as a
11473 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11474 (vect_analyze_data_refs): Change min_vf from an int * to a
11476 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11477 vfm1 as a poly_uint64 rather than an int. Make the same change
11478 for the returned bound_scalar.
11479 (vect_gen_vector_loop_niters): Handle polynomial vf.
11480 (vect_do_peeling): Likewise. Update call to
11481 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11482 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11484 * tree-vect-loop.c (vect_determine_vectorization_factor)
11485 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11486 (vect_get_known_peeling_cost): Likewise.
11487 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11488 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11489 (vect_transform_loop): Likewise. Use the lowest possible VF when
11490 updating the upper bounds of the loop.
11491 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11492 rather than an int.
11493 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11494 polynomial unroll factors.
11495 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11496 (vect_make_slp_decision): Likewise.
11497 (vect_supported_load_permutation_p): Likewise, and polynomial
11499 (vect_analyze_slp_cost): Handle polynomial vf.
11500 (vect_slp_analyze_node_operations): Likewise.
11501 (vect_slp_analyze_bb_1): Likewise.
11502 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11503 than an unsigned HOST_WIDE_INT.
11504 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11505 (vectorizable_load): Handle polynomial vf.
11506 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11508 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11511 Alan Hayward <alan.hayward@arm.com>
11512 David Sherwood <david.sherwood@arm.com>
11514 * match.pd: Handle bit operations involving three constants
11515 and try to fold one pair.
11517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11519 * tree-vect-loop-manip.c: Include gimple-fold.h.
11520 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11521 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11522 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11523 Add a path that uses a step of VF instead of 1, but disable it
11525 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11526 and niters_no_overflow parameters. Update calls to
11527 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11528 Create a new SSA name if the latter choses to use a ste other
11529 than zero, and return it via niters_vector_mult_vf_var.
11530 * tree-vect-loop.c (vect_transform_loop): Update calls to
11531 vect_do_peeling, vect_gen_vector_loop_niters and
11532 slpeel_make_loop_iterate_ntimes.
11533 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11534 (vect_gen_vector_loop_niters): Update declarations after above changes.
11536 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11538 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11539 128-bit round to integer instructions.
11540 (ceil<mode>2): Likewise.
11541 (btrunc<mode>2): Likewise.
11542 (round<mode>2): Likewise.
11544 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11546 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11547 unaligned VSX load/store on P8/P9.
11548 (expand_block_clear): Allow the use of unaligned VSX
11549 load/store on P8/P9.
11551 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11553 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11555 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11556 swap associated with both a load and a store.
11558 2018-01-02 Andrew Waterman <andrew@sifive.com>
11560 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11561 * config/riscv/riscv.md (clear_cache): Use it.
11563 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11565 * web.c: Remove out-of-date comment.
11567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11569 * expr.c (fixup_args_size_notes): Check that any existing
11570 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11571 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11572 (emit_single_push_insn): ...here.
11574 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11576 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11577 (const_vector_encoded_nelts): New function.
11578 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11579 (const_vector_int_elt, const_vector_elt): Declare.
11580 * emit-rtl.c (const_vector_int_elt_1): New function.
11581 (const_vector_elt): Likewise.
11582 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11583 of CONST_VECTOR_ELT.
11585 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11587 * expr.c: Include rtx-vector-builder.h.
11588 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11589 directly on the tree encoding.
11590 (const_vector_from_tree): Likewise.
11591 * optabs.c: Include rtx-vector-builder.h.
11592 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11593 sequence of "u" values.
11594 * vec-perm-indices.c: Include rtx-vector-builder.h.
11595 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11596 directly on the vec_perm_indices encoding.
11598 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11600 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11601 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11602 * rtx-vector-builder.h: New file.
11603 * rtx-vector-builder.c: Likewise.
11604 * rtl.h (rtx_def::u2): Add a const_vector field.
11605 (CONST_VECTOR_NPATTERNS): New macro.
11606 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11607 (CONST_VECTOR_DUPLICATE_P): Likewise.
11608 (CONST_VECTOR_STEPPED_P): Likewise.
11609 (CONST_VECTOR_ENCODED_ELT): Likewise.
11610 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11611 (unwrap_const_vec_duplicate): Likewise.
11612 (const_vec_series_p): Check for a non-duplicated vector encoding.
11613 Say that the function only returns true for integer vectors.
11614 * emit-rtl.c: Include rtx-vector-builder.h.
11615 (gen_const_vec_duplicate_1): Delete.
11616 (gen_const_vector): Call gen_const_vec_duplicate instead of
11617 gen_const_vec_duplicate_1.
11618 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11619 (gen_const_vec_duplicate): Use rtx_vector_builder.
11620 (gen_const_vec_series): Likewise.
11621 (gen_rtx_CONST_VECTOR): Likewise.
11622 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11623 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11624 Build a new vector rather than modifying a CONST_VECTOR in-place.
11625 (handle_special_swappables): Update call accordingly.
11626 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11627 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11628 Build a new vector rather than modifying a CONST_VECTOR in-place.
11629 (handle_special_swappables): Update call accordingly.
11631 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11633 * simplify-rtx.c (simplify_const_binary_operation): Use
11634 CONST_VECTOR_ELT instead of XVECEXP.
11636 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11638 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11639 the selector elements to be different from the data elements
11640 if the selector is a VECTOR_CST.
11641 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11642 ssizetype for the selector.
11644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11646 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11647 before testing each element individually.
11648 * tree-vect-generic.c (lower_vec_perm): Likewise.
11650 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11652 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11653 * selftest-run-tests.c (selftest::run_tests): Call it.
11654 * vector-builder.h (vector_builder::operator ==): New function.
11655 (vector_builder::operator !=): Likewise.
11656 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11657 (vec_perm_indices::all_from_input_p): New function.
11658 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11659 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11660 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11661 instead of reading the VECTOR_CST directly. Detect whether both
11662 vector inputs are the same before constructing the vec_perm_indices,
11663 and update the number of inputs argument accordingly. Use the
11664 utility functions added above. Only construct sel2 if we need to.
11666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11668 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11669 the broadcast of the low byte.
11670 (expand_mult_highpart): Use an explicit encoding for the permutes.
11671 * optabs-query.c (can_mult_highpart_p): Likewise.
11672 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11673 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11674 (vectorizable_bswap): Likewise.
11675 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11676 explicit encoding for the power-of-2 permutes.
11677 (vect_permute_store_chain): Likewise.
11678 (vect_grouped_load_supported): Likewise.
11679 (vect_permute_load_chain): Likewise.
11681 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11683 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11684 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11685 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11686 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11687 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11688 (vect_gen_perm_mask_any): Likewise.
11690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11692 * int-vector-builder.h: New file.
11693 * vec-perm-indices.h: Include int-vector-builder.h.
11694 (vec_perm_indices): Redefine as an int_vector_builder.
11695 (auto_vec_perm_indices): Delete.
11696 (vec_perm_builder): Redefine as a stand-alone class.
11697 (vec_perm_indices::vec_perm_indices): New function.
11698 (vec_perm_indices::clamp): Likewise.
11699 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11700 (vec_perm_indices::new_vector): New function.
11701 (vec_perm_indices::new_expanded_vector): Update for new
11702 vec_perm_indices class.
11703 (vec_perm_indices::rotate_inputs): New function.
11704 (vec_perm_indices::all_in_range_p): Operate directly on the
11705 encoded form, without computing elided elements.
11706 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11707 encoding. Update for new vec_perm_indices class.
11708 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11709 the given vec_perm_builder.
11710 (expand_vec_perm_var): Update vec_perm_builder constructor.
11711 (expand_mult_highpart): Use vec_perm_builder instead of
11712 auto_vec_perm_indices.
11713 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11714 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11715 or double series encoding as appropriate.
11716 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11717 vec_perm_indices instead of auto_vec_perm_indices.
11718 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11719 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11720 (vect_permute_store_chain): Likewise.
11721 (vect_grouped_load_supported): Likewise.
11722 (vect_permute_load_chain): Likewise.
11723 (vect_shift_permute_load_chain): Likewise.
11724 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11725 (vect_transform_slp_perm_load): Likewise.
11726 (vect_schedule_slp_instance): Likewise.
11727 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11728 (vectorizable_mask_load_store): Likewise.
11729 (vectorizable_bswap): Likewise.
11730 (vectorizable_store): Likewise.
11731 (vectorizable_load): Likewise.
11732 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11733 vec_perm_indices instead of auto_vec_perm_indices. Use
11734 tree_to_vec_perm_builder to read the vector from a tree.
11735 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11736 vec_perm_builder instead of a vec_perm_indices.
11737 (have_whole_vector_shift): Use vec_perm_builder and
11738 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11739 truncation to calc_vec_perm_mask_for_shift.
11740 (vect_create_epilog_for_reduction): Likewise.
11741 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11742 from auto_vec_perm_indices to vec_perm_indices.
11743 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11744 instead of changing individual elements.
11745 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11746 the vector in d.perm.
11747 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11748 from auto_vec_perm_indices to vec_perm_indices.
11749 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11750 instead of changing individual elements.
11751 (arm_vectorize_vec_perm_const): Use new_vector to install
11752 the vector in d.perm.
11753 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11754 Update vec_perm_builder constructor.
11755 (rs6000_expand_interleave): Likewise.
11756 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11757 (rs6000_expand_interleave): Likewise.
11759 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11761 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11762 to qimode could truncate the indices.
11763 * optabs.c (expand_vec_perm_var): Likewise.
11765 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11767 * Makefile.in (OBJS): Add vec-perm-indices.o.
11768 * vec-perm-indices.h: New file.
11769 * vec-perm-indices.c: Likewise.
11770 * target.h (vec_perm_indices): Replace with a forward class
11772 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11773 * optabs.h: Include vec-perm-indices.h.
11774 (expand_vec_perm): Delete.
11775 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11776 (expand_vec_perm_const): Declare.
11777 * target.def (vec_perm_const_ok): Replace with...
11778 (vec_perm_const): ...this new hook.
11779 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11780 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11781 * doc/tm.texi: Regenerate.
11782 * optabs.def (vec_perm_const): Delete.
11783 * doc/md.texi (vec_perm_const): Likewise.
11784 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11785 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11786 expand_vec_perm for constant permutation vectors. Assert that
11787 the mode of variable permutation vectors is the integer equivalent
11788 of the mode that is being permuted.
11789 * optabs-query.h (selector_fits_mode_p): Declare.
11790 * optabs-query.c: Include vec-perm-indices.h.
11791 (selector_fits_mode_p): New function.
11792 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11793 is defined, instead of checking whether the vec_perm_const_optab
11794 exists. Use targetm.vectorize.vec_perm_const instead of
11795 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11796 fit in the vector mode before using a variable permute.
11797 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11798 vec_perm_indices instead of an rtx.
11799 (expand_vec_perm): Replace with...
11800 (expand_vec_perm_const): ...this new function. Take the selector
11801 as a vec_perm_indices rather than an rtx. Also take the mode of
11802 the selector. Update call to shift_amt_for_vec_perm_mask.
11803 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11804 Use vec_perm_indices::new_expanded_vector to expand the original
11805 selector into bytes. Check whether the indices fit in the vector
11806 mode before using a variable permute.
11807 (expand_vec_perm_var): Make global.
11808 (expand_mult_highpart): Use expand_vec_perm_const.
11809 * fold-const.c: Includes vec-perm-indices.h.
11810 * tree-ssa-forwprop.c: Likewise.
11811 * tree-vect-data-refs.c: Likewise.
11812 * tree-vect-generic.c: Likewise.
11813 * tree-vect-loop.c: Likewise.
11814 * tree-vect-slp.c: Likewise.
11815 * tree-vect-stmts.c: Likewise.
11816 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11818 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11819 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11820 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11821 (aarch64_vectorize_vec_perm_const): ...this new function.
11822 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11823 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11824 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11825 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11826 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11827 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11828 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11830 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11831 check for NEON modes.
11832 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11833 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11834 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11835 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11837 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11838 the old VEC_PERM_CONST conditions.
11839 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11840 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11841 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11842 (ia64_vectorize_vec_perm_const_ok): Merge into...
11843 (ia64_vectorize_vec_perm_const): ...this new function.
11844 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11845 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11846 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11847 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11848 * config/mips/mips.c (mips_expand_vec_perm_const)
11849 (mips_vectorize_vec_perm_const_ok): Merge into...
11850 (mips_vectorize_vec_perm_const): ...this new function.
11851 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11852 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11853 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11854 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11855 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11856 (rs6000_expand_vec_perm_const): Delete.
11857 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11859 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11860 (altivec_expand_vec_perm_const_le): Take each operand individually.
11861 Operate on constant selectors rather than rtxes.
11862 (altivec_expand_vec_perm_const): Likewise. Update call to
11863 altivec_expand_vec_perm_const_le.
11864 (rs6000_expand_vec_perm_const): Delete.
11865 (rs6000_vectorize_vec_perm_const_ok): Delete.
11866 (rs6000_vectorize_vec_perm_const): New function.
11867 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11868 an element count and rtx array.
11869 (rs6000_expand_extract_even): Update call accordingly.
11870 (rs6000_expand_interleave): Likewise.
11871 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11872 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11873 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11874 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11875 (rs6000_expand_vec_perm_const): Delete.
11876 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11877 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11878 (altivec_expand_vec_perm_const_le): Take each operand individually.
11879 Operate on constant selectors rather than rtxes.
11880 (altivec_expand_vec_perm_const): Likewise. Update call to
11881 altivec_expand_vec_perm_const_le.
11882 (rs6000_expand_vec_perm_const): Delete.
11883 (rs6000_vectorize_vec_perm_const_ok): Delete.
11884 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11885 reference to the SPE evmerge intructions.
11886 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11887 an element count and rtx array.
11888 (rs6000_expand_extract_even): Update call accordingly.
11889 (rs6000_expand_interleave): Likewise.
11890 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11891 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11893 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11895 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11897 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11898 vector mode and that that mode matches the mode of the data
11900 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11901 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11902 directly using expand_vec_perm_1 when forcing selectors into
11904 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11906 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11908 * optabs-query.h (can_vec_perm_p): Delete.
11909 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11910 * optabs-query.c (can_vec_perm_p): Split into...
11911 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11912 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11913 particular selector is valid.
11914 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11915 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11916 (vect_grouped_load_supported): Likewise.
11917 (vect_shift_permute_load_chain): Likewise.
11918 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11919 (vect_transform_slp_perm_load): Likewise.
11920 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11921 (vectorizable_bswap): Likewise.
11922 (vect_gen_perm_mask_checked): Likewise.
11923 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11924 implementations of variable permutation vectors into account
11925 when deciding which selector to use.
11926 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11927 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11928 with a false third argument.
11929 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11930 to test whether the constant selector is valid and can_vec_perm_var_p
11931 to test whether a variable selector is valid.
11933 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11935 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11936 * optabs-query.c (can_vec_perm_p): Likewise.
11937 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11938 instead of vec_perm_indices.
11939 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11940 (vect_gen_perm_mask_checked): Likewise,
11941 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11942 (vect_gen_perm_mask_checked): Likewise,
11944 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11946 * optabs-query.h (qimode_for_vec_perm): Declare.
11947 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11948 (qimode_for_vec_perm): ...this new function.
11949 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11951 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11953 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11954 does not have a conditional at the top.
11956 2018-01-02 Richard Biener <rguenther@suse.de>
11958 * ipa-inline.c (big_speedup_p): Fix expression.
11960 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11963 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11966 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11970 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11971 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11972 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11973 cond_taken_branch_cost 3->4.
11975 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11977 PR tree-optimization/83581
11978 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11979 TODO_cleanup_cfg if any changes have been made.
11981 PR middle-end/83608
11982 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11983 convert_modes if target mode has the right side, but different mode
11986 PR middle-end/83609
11987 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11988 last argument when extracting from CONCAT. If either from_real or
11989 from_imag is NULL, use expansion through memory. If result is not
11990 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11991 the parts directly to inner mode, if even that fails, use expansion
11994 PR middle-end/83623
11995 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11996 check for bswap in mode rather than HImode and use that in expand_unop
11999 Copyright (C) 2018 Free Software Foundation, Inc.
12001 Copying and distribution of this file, with or without modification,
12002 are permitted in any medium without royalty provided the copyright
12003 notice and this notice are preserved.