1 2018-04-30 Jason Merrill <jason@redhat.com>
3 PR c++/61982 - dead stores to destroyed objects.
4 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
7 2018-04-30 Jason Merrill <jason@redhat.com>
9 * tree.c (build_clobber): New.
11 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
13 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15 * diagnostic-show-locus.c (layout::layout): Update for
16 location_get_source_line returning a char_span.
17 (struct char_span): Move to input.h.
18 (struct correction): Update for fields in char_span becoming
20 (struct source_line): Update for location_get_source_line
21 returning a char_span.
22 (layout::print_line): Likewise.
23 * edit-context.c (edited_file::print_content): Likewise.
24 (edited_file::print_diff_hunk): Likewise.
25 (edited_file::print_run_of_changed_lines): Likewise.
26 (edited_file::get_num_lines): Likewise.
27 (edited_line::edited_line): Likewise.
28 * final.c (asm_show_source): Likewise.
29 * input.c (location_get_source_line): Convert return type
30 from const char * to char_span, losing the final "line_len"
32 (dump_location_info): Update for the above.
33 (get_substring_ranges_for_loc): Likewise. Use a char_span
34 when handling the literal within the line.
35 (test_reading_source_line): Update for location_get_source_line
36 returning a char_span.
37 * input.h (class char_span): Move here from
38 diagnostic-show-locus.c, converting from a struct to a class.
39 Make data members private.
40 (char_span::operator bool): New.
41 (char_span::length): New.
42 (char_span::get_buffer): New.
43 (char_span::operator[]): New.
44 (char_span::subspan): Make const.
45 (char_span::xstrdup): New.
46 (location_get_source_line): Convert return type from const char *
47 to char_span, losing the final "line_size" param.
49 2018-04-30 Jan Hubicka <jh@suse.cz>
51 * lto-wrapper.c (ltrans_priorities): New static var.
53 (run_gcc): Read priorities and if doing parallel build order
56 2018-04-30 David Malcolm <dmalcolm@redhat.com>
58 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
60 2018-04-30 Richard Biener <rguenther@suse.de>
62 * tree-cfg.c (verify_address): Remove base argument, add
63 flag whether to check TREE_ADDRESSABLE and do that.
64 (verify_expr): Remove.
65 (verify_types_in_gimple_reference): Add pieces from verify_expr.
66 (verify_gimple_assign_single): Likewise.
67 (verify_gimple_switch): Likewise.
68 (verify_expr_location_1): Dereference tp once. Add (disabled)
69 piece from verify_expr.
70 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
72 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
74 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
76 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
78 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
79 (small_data_pattern): Likewise.
80 (arc_rewrite_small_data): Likewise.
81 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
82 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
83 (get_symbol_alignment): New function.
84 (legitimate_small_data_address_p): Likewise.
85 (legitimate_scaled_address): Update, call
86 legitimate_small_data_address_p.
87 (output_sdata): New static variable.
88 (arc_print_operand): Update how we handle small data operands.
89 (arc_print_operand_address): Likewise.
90 (arc_legitimate_address_p): Update, use
91 legitimate_small_data_address_p.
92 (arc_rewrite_small_data_p): Remove.
93 (arc_rewrite_small_data_1): Likewise.
94 (arc_rewrite_small_data): Likewise.
95 (small_data_pattern): Likewise.
96 (compact_sda_memory_operand): Update to use
97 legitimate_small_data_address_p and get_symbol_alignment.
98 (prepare_move_operands): Don't rewite sdata pattern.
99 (prepare_extend_operands): Remove.
100 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
102 (zero_extendqisi2): Likewise.
103 (zero_extendhisi2): Likewise.
104 (extendqihi2): Likewise.
105 (extendqisi2): Likewise.
106 (extendhisi2): Likewise.
110 * config/arc/constraints.md (Usd): Change it to memory constraint.
112 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
114 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
115 as source of std instructions.
116 * config/arc/arc.md (movsi_insn): Update pattern predicate to
117 allow 6-bit constants as source for store instructions.
118 (movdi_insn): Update instruction pattern to allow 6-bit constants
119 as source for store instructions.
121 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
123 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
125 2018-04-30 Nathan Sidwell <nathan@acm.org>
126 Sandra Loosemore <sandra@codesourcery.com>
128 * dumpfile.c (dump_open): Allow '-' for stdout.
129 * doc/invoke.texi (Developer Options): Document dump filename
130 determination early. Document stdin/stdout selection.
132 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
134 Microblaze Target: PIC data text relative
136 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
137 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
139 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
140 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
141 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
142 New addressing mode for data-text relative position indepenedent code.
143 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
144 'ADDRESS_SYMBOLIC_TXT_REL'.
145 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
146 (microblaze_legitimate_pic_operand): Exclude function calls from
147 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
148 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
150 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
151 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
152 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
153 for 'address + offset'.
154 (microblaze_expand_prologue): Add new function prologue call for
156 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
157 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
158 table in case of TARGET_PIC_DATA_TEXT_REL.
159 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
160 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
161 Add new macros 'UNSPEC_TEXT',
162 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
163 + exclude function calls from 'UNSPEC_PLT' in case of data text
165 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
166 new target hook for generating address diff vector tables in case of
168 * doc/tm.texi : Regenerate.
169 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
170 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
171 of addr diff vector generation.
172 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
173 target hook definition.
174 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
175 Add default function for generate_pic_addr_diff_vec -> flag_pic.
176 * doc/invoke.texi (Add new pic option): Add new microblaze pic
177 option for data text relative.
179 2018-04-30 Richard Biener <rguenther@suse.de>
181 * tree-chrec.h (evolution_function_is_constant_p): Remove
183 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
185 2018-04-30 Richard Biener <rguenther@suse.de>
188 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
190 2018-04-30 Richard Biener <rguenther@suse.de>
192 PR tree-optimization/28364
193 PR tree-optimization/85275
194 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
195 copying first exit test.
197 2018-04-28 Mark Wielaard <mark@klomp.org>
199 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
201 (dwarf_AT): Handle DW_AT_addr_base.
202 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
204 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
207 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
208 (*ashl<dwi>3_doubleword_mask_1): Ditto.
209 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
210 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
212 2018-04-28 Richard Biener <rguenther@suse.de>
214 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
215 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
216 to reflect use. Only add interesting stmts.
218 2018-04-27 Martin Jambor <mjambor@suse.cz>
221 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
222 the jump function allows for passing through aggregate values.
224 2018-04-27 David Malcolm <dmalcolm@redhat.com>
226 * input.h (in_system_header_at): Convert from macro to inline
228 (from_macro_expansion_at): Likewise.
229 (from_macro_definition_at): Likewise.
231 2018-04-27 Jeff Law <law@redhat.com>
233 * config.gcc: Mark tile* targets as deprecated/obsolete.
235 2018-04-27 Richard Biener <rguenther@suse.de>
237 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
240 2018-04-27 Richard Biener <rguenther@suse.de>
242 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
244 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
246 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
247 with Yd constraint. Set "preferred_for_speed" attribute from
248 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
250 (*movdi_internal): Ditto.
251 (movti_interunit splitters): Remove
252 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
253 (movdi_interunit splitters): Ditto.
254 * config/i386/constraints.md (Ye): Remove.
255 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
257 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
260 * config/aarch64/constraints.md (Usg): Limit to 31.
263 2018-04-27 Jakub Jelinek <jakub@redhat.com>
265 PR tree-optimization/85529
266 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
267 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
268 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
269 zero extension or masking of the MSB bit.
270 (optimize_range_tests): Add FIRST_BB argument, pass it through
271 to optimize_range_tests_var_bound.
272 (maybe_optimize_range_tests, reassociate_bb): Adjust
273 optimize_range_tests callers.
275 2018-04-26 Richard Biener <rguenther@suse.de>
276 Jakub Jelinek <jakub@redhat.com>
278 * cgraph.h (symbol_table): Just declare debug method here.
279 * symtab.c (symbol_table::debug): Define.
281 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
283 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
285 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
287 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
288 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
289 (*movdi_internal): Substitute Yi and Yj constraint with x
290 and Ym and Yn constraint with y constraint. Update "isa"
291 attribute and set "preferred_for_speed" attribute from
292 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
293 (*movsi_internal): Ditto.
294 (*movdf_internal): Ditto.
295 (*movsf_internal): Ditto.
296 (*zero_extendsidi2): Ditto.
297 * config/i386/sse.md (vec_set<mode>_0): Ditto.
298 (sse2_loadld): Ditto.
299 (*vec_extract<ssevecmodelower>_0): Ditto.
300 (*vec_extractv4si_0_zext_sse4): Ditto.
301 (vec_concatv2di): Ditto.
302 (*vec_dup<mode>): Ditto.
303 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
304 * config/i386/constraints.md (Yi): Remove.
309 2018-04-26 Nathan Sidwell <nathan@acm.org>
311 * dumpfile.c (dump_open): New.
312 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
313 (dump_finish): Detect stdio/stderr by value not name.
315 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
317 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
319 2018-04-26 Tom de Vries <tom@codesourcery.com>
322 * config/nvptx/nvptx.c (verify_neutering_jumps)
323 (verify_neutering_labels): New function
324 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
326 2018-04-26 Tom de Vries <tom@codesourcery.com>
329 * config/nvptx/nvptx.c (needs_neutering_p): New function.
330 (nvptx_single): Use needs_neutering_p to skip over insns that do not
333 2018-04-26 Richard Biener <rguenther@suse.de>
334 Tom de Vries <tom@codesourcery.com>
337 * lto-streamer-out.c (output_function): Fixup loops if required to match
338 discovery done in the reader.
340 2018-04-26 Richard Biener <rguenther@suse.de>
342 PR tree-optimization/85116
343 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
344 have a loop exit from the single latch predecessor. Remove
345 case of header with just condition.
346 (ch_base::copy_headers): Exclude infinite loops from any
348 (pass_ch::execute): Record exits.
350 2018-04-26 Richard Biener <rguenther@suse.de>
352 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
353 prologue cost vector and pass it to vect_get_load_cost.
354 (vect_get_peeling_costs_all_drs): Likewise.
355 (vect_peeling_hash_get_lowest_cost): Likewise.
356 (vect_enhance_data_refs_alignment): Likewise.
358 2018-04-26 Richard Biener <rguenther@suse.de>
361 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
362 checking of integer<->pointer conversions.
363 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
364 sign-/zero-extending pointer types.
365 (expand_omp_for_static_chunk): Likewise.
367 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
368 Jean Lee <xiaoyur347@gmail.com>
370 * config/mips/mips.c (mips_asan_shadow_offset): New function.
371 (TARGET_ASAN_SHADOW_OFFSET): Define.
372 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
373 true for -fsanitize=address.
375 2018-04-25 Mark Wielaard <mark@klomp.org>
377 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
380 2018-04-25 Jakub Jelinek <jakub@redhat.com>
382 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
383 than "alu", remove explicit "memory" and "imm_disp" attributes.
384 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
387 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
388 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
391 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
394 * config/i386/i386.c (ix86_expand_builtin): Change memory
395 operand to XI, extend p0 to Pmode.
396 * config/i386/i386.md: Change unspec volatile and operand
397 1 mode to XI, change operand 0 mode to P.
399 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
401 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
402 GET_MODE_MASK before any checking.
403 (nds32_can_use_bset_p): Likewise.
404 (nds32_can_use_btgl_p): Likewise.
406 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
408 * config/nds32/nds32-doubleword.md: New define_split pattern for
409 illegal register number.
411 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
413 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
415 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
417 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
419 2018-04-25 Richard Biener <rguenther@suse.de>
421 * lto-streamer.h (LTO_major_version): Bump to 8.
423 2018-04-25 Jakub Jelinek <jakub@redhat.com>
425 * BASE-VER: Set to 9.0.0.
427 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
429 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
430 in __abskf2 and __powikf2.
432 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
435 * config/aarch64/constraints.md (Usg, Usj): New constraints.
436 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
437 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
438 Use the above on operand 2. Reindent.
439 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
441 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
444 * common/config/i386/i386-common.c (ix86_handle_option): Don't
446 * config/i386/i386.opt (mcet): Removed.
447 * doc/install.texi: Remove -mcet documentation.
448 * doc/invoke.texi: Likewise.
450 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
453 * doc/install.texi: Remove -mcet from bootstrap-cet.
455 2018-04-24 Jakub Jelinek <jakub@redhat.com>
458 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
459 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
463 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
464 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
465 containing a CONST_VECTOR.
467 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
469 * doc/install.texi: Update newlib dependency for nvptx.
471 2018-04-24 Jakub Jelinek <jakub@redhat.com>
474 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
475 instead of INTVAL when shifting x left.
477 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
479 PR tree-optimization/85478
480 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
481 vect_grouped_store_supported for single element vectors.
483 2018-04-24 Richard Biener <rguenther@suse.de>
486 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
487 load cost increase to the case of non-constant step.
489 2018-04-24 Jakub Jelinek <jakub@redhat.com>
492 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
493 destination if any_malformed_asm.
495 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
498 * expr.c (store_field): In the bitfield case, if the value comes from
499 a function call and is returned in registers by means of a PARALLEL,
500 do not change the mode of the temporary unless BLKmode and VOIDmode.
502 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
504 PR rtl-optimization/85423
505 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
506 dependencies to debug insns when the previous insn is non-debug.
508 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
510 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
511 enums into a single definition.
512 (fls): Fix predicates and printing.
515 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
517 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
518 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
519 and short u6 immediate.
520 (check_if_valid_sleep_operand): Remove.
521 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
523 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
525 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
526 flag_always_save_lp condition.
527 * config/nds32/nds32.opt (malways-save-lp): New option.
529 2018-04-22 Shiva Chen <shiva0217@gmail.com>
531 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
532 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
533 * config/nds32/nds32.h
534 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
535 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
537 2018-04-22 Shiva Chen <shiva0217@gmail.com>
539 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
540 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
542 2018-04-22 Shiva Chen <shiva0217@gmail.com>
543 Chung-Ju Wu <jasonwucj@gmail.com>
545 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
547 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
548 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
550 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
552 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
554 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
556 * config/nds32/nds32-protos.h (nds32_data_alignment,
557 nds32_local_alignment): Declare.
558 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
559 nds32_local_alignment): New functions.
560 (TARGET_CONSTANT_ALIGNMENT): Define.
561 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
563 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
565 * config/nds32/nds32.c
566 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
567 (TARGET_MODES_TIEABLE_P): Likewise.
569 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
571 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
574 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
575 Chung-Ju Wu <jasonwucj@gmail.com>
577 * config/nds32/constants.md (unspec_volatile_element): Add enum values
578 for unaligned access.
579 * config/nds32/nds32-intrinsic.c: Implementation of expanding
581 * config/nds32/nds32-intrinsic.md: Likewise.
582 * config/nds32/nds32_intrinsic.h: Likewise.
583 * config/nds32/nds32.h (nds32_builtins): Likewise.
584 * config/nds32/nds32.opt (munaligned-access): New option.
585 * config/nds32/nds32.c (nds32_asm_file_start): Display
586 flag_unaligned_access status.
588 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
590 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
591 -mno-relax is present.
592 * config/riscv/linux.h (LINK_SPEC): Ditto.
594 2018-04-20 Martin Sebor <msebor@redhat.com>
597 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
599 (gimple_fold_builtin_stxcpy_chk): Same.
600 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
602 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
605 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
606 __powikf2 when long double is IEEE 128-bit.
608 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
610 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
611 step to make sure stack always aligned.
613 2018-04-20 Carl Love <cel@us.ibm.com>
616 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
619 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
620 Tom de Vries <tom@codesourcery.com>
623 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
624 Emit insns for calls too.
625 (nvptx_find_par): Always look for worker-level predecessor insn.
626 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
628 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
629 (nvptx_process_pars): Propagate frames for calls.
631 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
634 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
636 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
637 (ix86_handle_option): Don't handle OPT_mibt.
638 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
640 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
642 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
643 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
644 (ix86_target_macros): Define __CET__ with flag_cf_protection
646 * config/i386/i386.c (isa2_opts): Remove -mibt.
647 * config/i386/i386.h (TARGET_IBT): Removed.
648 (TARGET_IBT_P): Likewise.
649 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
650 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
651 * config/i386/i386.opt (mcet): Update help message.
654 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
655 -mcet as an alias for -mshstk.
657 2018-04-20 Richard Biener <rguenther@suse.de>
660 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
661 complexity by forcing a single use of the multiply operand.
663 2018-04-20 Martin Jambor <mjambor@suse.cz>
666 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
667 recursion dependency to only apply to non-clones.
669 2018-04-20 Martin Jambor <mjambor@suse.cz>
672 * ipa-cp.c (create_specialized_node): Check that clones of
673 self-recursive edges exist during IPA-CP.
675 2018-04-19 Toon Moene <toon@moene.org>
677 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
680 2018-04-19 Jakub Jelinek <jakub@redhat.com>
682 PR tree-optimization/85467
683 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
684 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
685 VECTOR_CST element to type.
687 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
690 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
691 * config/i386/i386.md (builtin_setjmp_setup): Removed.
692 (builtin_longjmp): Likewise.
693 (save_stack_nonlocal): New pattern.
694 (restore_stack_nonlocal): Likewise.
696 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
699 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
700 Replace ASM_OUTPUT_LABEL with fprintf.
702 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
705 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
706 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
707 * config/i386/i386-c.c (ix86_target_macros_internal): Also
708 define __IBT__ and __SHSTK__ for -fcf-protection.
709 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
711 (ix86_trampoline_init): Likewise.
712 (x86_output_mi_thunk): Likewise.
713 (ix86_notrack_prefixed_insn_p): Likewise.
714 (ix86_option_override_internal): Don't disallow -fcf-protection.
715 * config/i386/i386.md (rdssp<mode>): Also enable for
717 (incssp<mode>): Likewise.
718 (nop_endbr): Likewise.
719 * config/i386/i386.opt (mcet): Change help message to built-in
723 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
724 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
725 enable CET built-in functions.
727 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
729 * common/config/i386/i386-common.c
730 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
731 OPTION_MASK_ISA_MOVDIRI_UNSET,
732 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
733 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
734 * config.gcc (movdirintrin.h): New header.
735 * config/i386/cpuid.h (bit_MOVDIRI,
736 bit_MOVDIR64B): New bits.
737 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
739 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
740 (VOID, PVOID, PCVOID)): New function types.
741 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
742 __builtin_ia32_directstoreu_u64,
743 __builtin_ia32_movdir64b): New builtins.
744 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
745 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
747 (ix86_valid_target_attribute_inner_p): Ditto.
748 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
749 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
750 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
751 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
752 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
753 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
754 (movdiri<mode>, movdir64b_<mode>): New.
755 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
756 * config/i386/immintrin.h: Include movdirintrin.h.
757 * config/i386/movdirintrin.h: New file.
758 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
760 2018-04-19 Richard Biener <rguenther@suse.de>
763 * cfg.c (clear_bb_flags): When loop state says we have
764 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
766 2018-04-19 Richard Biener <rguenther@suse.de>
768 PR tree-optimization/84737
769 * tree-vect-data-refs.c (vect_copy_ref_info): New function
770 copying restrict info.
771 (vect_setup_realignment): Use it.
772 * tree-vectorizer.h (vect_copy_ref_info): Declare.
773 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
774 the first DR to all generated stores.
775 (vectorizable_load): Likewise for loads.
777 2018-04-19 Jakub Jelinek <jakub@redhat.com>
779 PR tree-optimization/85446
780 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
781 the integral and pointer types to have the same precision.
783 * doc/install.texi: Document --disable-cet being the default and
786 2018-04-18 Martin Liska <mliska@suse.cz>
788 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
791 2018-04-18 Martin Liska <mliska@suse.cz>
794 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
797 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
798 arguments if they are comparable.
800 2018-04-18 Martin Liska <mliska@suse.cz>
803 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
806 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
809 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
812 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
813 ENDBR after calling __morestack.
815 2018-04-18 David Malcolm <dmalcolm@redhat.com>
818 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
819 by using gcc_base_ver to generate a gcc_driver_version, and use
820 it when generating GCC_DRIVER_NAME.
821 * configure: Regenerate.
823 2018-04-18 Jakub Jelinek <jakub@redhat.com>
826 * config.gcc: Obsolete powerpc*-*-*spe*.
828 2018-04-17 Jakub Jelinek <jakub@redhat.com>
831 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
832 (stabstr_D): Change type of unum from unsigned int to
833 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
836 2018-04-17 Jim Wilson <jimw@sifive.com>
839 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
840 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
841 Set arg_pointer_offset after using pretend_args_size.
843 2018-04-17 Jakub Jelinek <jakub@redhat.com>
845 PR rtl-optimization/85431
846 * dse.c (record_store): Ignore zero width stores.
849 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
850 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
851 __builtin_stack_restore rather than after it.
852 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
853 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
854 argument instead of virtual_dynamic_stack_rtx.
856 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
858 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
860 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
861 Add note to error message to explain internal mapping of overloaded
862 built-in function name to non-overloaded built-in function name.
863 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
866 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
869 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
870 where the inputs overlap with the output.
872 2018-04-17 Jakub Jelinek <jakub@redhat.com>
874 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
875 (=v, v) alternative and explicit "memory" attribute.
876 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
877 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
879 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
880 "sselog1" type instead of "sselog".
881 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
882 "sselog". Remove explicit "memory" attribute.
883 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
884 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
886 (vec_extract_hi_v32hi): Merge all alternatives into one, use
887 "sselog1" type instead of "sselog". Remove explicit "memory"
889 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
890 use "sselog1" type instead of "sselog". Remove explicit "memory"
892 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
893 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
895 (vec_extract_hi_v64qi): Merge all alternatives into one, use
896 "sselog1" type instead of "sselog". Remove explicit "memory"
898 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
899 use "sselog1" type instead of "sselog". Remove explicit "memory"
903 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
906 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
909 2018-04-17 Martin Jambor <mjambor@suse.cz>
912 * ipa-cp.c (create_specialized_node): Call
913 expand_all_artificial_thunks if necessary.
915 2018-04-17 Martin Liska <mliska@suse.cz>
918 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
919 in message, remote space in between '_G' and '('.
921 2018-04-17 Jakub Jelinek <jakub@redhat.com>
924 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
925 avx512f_vmcmp<mode>3<round_saeonly_name>,
926 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
927 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
928 avx512f_rndscale<mode><round_saeonly_name>,
929 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
930 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
931 Use %<iptr>2 instead of %2 for -masm=intel.
932 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
933 avx512f_vcvttss2usi<round_saeonly_name>,
934 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
936 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
937 avx512f_vcvttsd2usi<round_saeonly_name>,
938 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
939 Use %q1 instead of %1 for -masm=intel.
940 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
941 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
942 of %3 for -masm=intel.
943 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
945 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
947 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
949 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
950 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
952 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
954 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
955 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
956 %g1 and one with %0 and %1.
957 (avx512er_vmrcp28<mode><round_saeonly_name>,
958 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
960 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
961 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
962 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
963 of %0 and %{%4%} for -masm=intel.
964 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
965 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
966 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
967 order of %0 and %{%5%}%{z%} for -masm=intel.
969 2018-04-17 Jan Hubicka <jh@suse.cz>
972 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
974 2018-04-17 Martin Liska <mliska@suse.cz>
977 * multiple_target.c (create_dispatcher_calls): Set apostrophes
978 for target_clone error message. Make default implementation
979 clone to be a local declaration.
980 (separate_attrs): Add new argument and check for an empty
982 (expand_target_clones): Handle it.
983 (ipa_target_clone): Make redirection just for target_clones
986 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
987 Tom de Vries <tom@codesourcery.com>
990 * omp-expand.c (expand_oacc_for): Add dummy false branch for
991 tiled basic blocks without omp continue statements.
993 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
996 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
997 vec_extract expression as having side effects to make sure it gets
1000 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
1003 * config/i386/i386.c (get_builtin_code_for_version): Check
1006 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
1009 * gcc/config.gcc: Support "skylake".
1010 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
1012 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
1013 (processor_target_table): Add "skylake".
1014 (ix86_option_override_internal): Add "skylake".
1015 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
1016 PROCESSOR_CANNONLAKE.
1017 (get_builtin_code_for_version): Fix priority for
1018 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
1019 PROCESSOR_SKYLAKE-AVX512.
1020 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
1021 (processor_type): Add PROCESSOR_SKYLAKE.
1023 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
1024 Jason Merrill <jason@redhat.com>
1027 * convert.c (convert_to_integer_1): Use direct recursion for
1028 enumeral types and types with a precision less than the number
1029 of bits in their mode.
1031 2018-04-16 Julia Koval <julia.koval@intel.com>
1034 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
1035 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
1037 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
1040 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
1041 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
1042 and -mno-direct-move.
1044 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
1047 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
1048 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
1049 Ensure negative shifts result in {0}.
1051 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
1053 PR rtl-optimization/79916
1054 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
1055 regs (if any) to define how to gnerate SD moves when LRA is in
1058 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1060 PR rtl-optimization/85393
1061 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1062 * except.c (expand_dw2_landing_pad_for_region): Make static.
1063 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1064 a label and unconditional jump to old_bb, rather than
1065 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1068 PR rtl-optimization/85376
1069 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1070 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1071 instead of a specific value.
1073 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1074 Bin Cheng <bin.cheng@arm.com>
1076 PR tree-optimization/82965
1077 PR tree-optimization/83991
1078 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1079 by_profile_only parameter.
1080 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1081 information if the loop was predicted to iterate too many times.
1082 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1084 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1087 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1090 2018-04-13 Martin Liska <mliska@suse.cz>
1091 Jakub Jelinek <jakub@redhat.com>
1094 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1095 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1096 * builtins.c (expand_builtin_memory_copy_args): Use
1097 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1098 handle dest_addr == pc_rtx.
1100 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1103 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1104 asked to not generate direct moves.
1105 (fix_trunc<mode>si2_stfiwx): Similar.
1106 (fix_trunc<mode>si2_internal): Similar.
1108 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1111 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1112 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1113 lookup if dest in some wider mode is known to be const0_rtx and
1114 if so, record permanent equivalence for it to be ZERO_EXTEND of
1115 the narrower mode destination.
1117 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1119 * lto-streamer-out.c (output_function): Revert 259346.
1120 * omp-expand.c (expand_oacc_for): Likewise.
1122 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1124 PR rtl-optimization/85354
1125 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1126 * sel-sched.c (sel_global_init): ... here.
1128 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1131 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1132 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1133 mode for PE-COFF targets.
1134 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1135 (i386_pe_asm_lto_end): Likewise.
1136 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1137 (TARGET_ASM_LTO_END): Likewise.
1138 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1139 (i386_pe_asm_lto_start): New function.
1140 (i386_pe_asm_lto_end): Likewise.
1142 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1143 Richard Biener <rguenther@suse.de>
1146 * lto-streamer-out.c (output_function): Fix CFG loop state before
1148 * omp-expand.c (expand_oacc_for): Handle calls to internal
1149 functions like regular functions.
1151 2018-04-12 Richard Biener <rguenther@suse.de>
1154 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1155 for the early LTO debug to properly generate references to it
1156 during DIE emission. Do not re-use that for the skeleton for
1158 (dwarf2out_early_finish): Likewise.
1160 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1163 * config/i386/sse.md
1164 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1165 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1166 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1167 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1168 and output is a reg, avoid creating invalid lowpart subreg, but
1169 instead split into a 512-bit move. Don't split if not AVX512VL,
1170 input is xmm16+ reg and output is a mem.
1171 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1172 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1173 xmm16+ reg and output is a mem.
1175 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1177 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1178 also for flag_dwarf2_cfi_asm.
1180 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1182 PR rtl-optimization/85342
1183 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1184 a bool scalar var inside of the loop instead. Don't try to update
1185 recog_data.operand after failed apply_change_group.
1187 2018-04-12 Tom de Vries <tom@codesourcery.com>
1190 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1191 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1192 array with flexible array member as array without given dimension.
1193 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1194 argument for undefined param to true.
1196 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1199 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1200 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1201 from PowerPC section.
1202 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1203 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1204 help text that is too long.
1205 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1206 help text that is too long.
1207 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1208 help text that is too long.
1210 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1212 * config/alpha/alpha.md (stack_probe_internal): Rename
1213 from "probe_stack". Update all callers.
1215 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1217 PR rtl-optimization/84566
1218 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1219 sched_macro_fuse_insns.
1221 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1224 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1225 (compute_block_dependences): ... from here.
1227 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1229 PR tree-optimization/85331
1230 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1231 from int to HOST_WIDE_INT.
1233 2018-04-11 Martin Jambor <mjambor@suse.cz>
1236 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1237 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1238 not the same as the source val.
1239 (cgraph_edge_brings_value_p): New parameter.
1240 (gather_edges_for_value): Pass destination value to
1241 cgraph_edge_brings_value_p.
1242 (perhaps_add_new_callers): Likewise.
1243 (get_info_about_necessary_edges): Likewise and exclude values brought
1244 only by self-recursive edges.
1245 (create_specialized_node): Redirect only clones of self-calling edges.
1246 (+self_recursive_pass_through_p): New function.
1247 (find_more_scalar_values_for_callers_subset): Use it.
1248 (find_aggregate_values_for_callers_subset): Likewise.
1249 (known_aggs_to_agg_replacement_list): Removed.
1250 (decide_whether_version_node): Re-calculate known constants for all
1251 remaining context clones.
1253 2018-04-11 Richard Biener <rguenther@suse.de>
1256 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1257 from early DWARF output.
1258 (dwarf2out_early_finish): Output line info unconditionally into
1259 early DWARF and add reference to it.
1261 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1264 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1265 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1266 other than V2DFmode using iptr mode attribute.
1267 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1269 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1271 PR rtl-optimization/84659
1272 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1274 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1277 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1279 (output_loc_list): Pass address of a dummy size variable even in the
1280 locview handling loop.
1281 (index_location_lists): Add comment on why skip_loc_list_entry can't
1284 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1287 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1290 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1293 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1294 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1295 and -mstring-compare-inline-limit.
1297 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1300 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1301 for stack clash protection in a register whenever we need it to be in
1304 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1306 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1307 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1309 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1312 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1314 (mlong-double-): Ditto.
1315 * config/rs6000/sysv4.opt (msdata=): Ditto.
1316 (mtls-size=): Ditto.
1318 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1320 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1321 erroneous entries for
1322 "vector int vec_ldl (int, long int *)", and
1323 "vector unsigned int vec_ldl (int, unsigned long int *)".
1324 Add comments and entries for
1325 "vector bool char vec_ldl (int, bool char *)",
1326 "vector bool short vec_ldl (int, bool short *)",
1327 "vector bool int vec_ldl (int, bool int *)",
1328 "vector bool long long vec_ldl (int, bool long long *)",
1329 "vector pixel vec_ldl (int, pixel *)",
1330 "vector long long vec_ldl (int, long long *)",
1331 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1332 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1333 type tree bool_long_long_type_node and correct definition of
1334 bool_V2DI_type_node to make reference to this new type tree.
1335 (rs6000_mangle_type): Replace erroneous reference to
1336 bool_long_type_node with bool_long_long_type_node.
1337 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1338 comments to emphasize sign distinctions for char and int types and
1339 replace RS6000_BTI_bool_long constant with
1340 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1341 use of RS6000_BTI_pixel.
1342 (bool_long_type_node): Remove this macro definition.
1343 (bool_long_long_type_node): New macro definition
1345 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1347 PR rtl-optimization/85300
1348 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1349 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1350 simplify_unary_operation fails.
1352 2018-04-10 Martin Liska <mliska@suse.cz>
1354 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1355 cgraph_edge and ipa_ref.
1357 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1361 * config/i386/sse.md
1362 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1363 computation of the VEC_MERGE selector from mask.
1364 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1365 Fix decoding of the VEC_MERGE selector into mask.
1367 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1369 PR tree-optimization/85286
1370 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1372 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1374 * final.c (final_1): Set insn_last_address as well as
1375 insn_current_address.
1377 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1380 * explow.c (emit_stack_probe): Call validize_mem on memory location
1381 before passing it to gen_probe_stack. Create address operand and
1382 legitimize it for the probe_stack_address case.
1384 2018-04-09 Jan Hubicka <jh@suse.cz>
1387 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1388 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1389 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1391 (free_lang_data): Rebuild type inheritance graph.
1393 2018-04-09 Martin Sebor <msebor@redhat.com>
1395 * invoke.texi (-finline-small-functions): Mention other optimization
1397 (-findirect-inlining, -fpartial-inlining): Same.
1398 (-finline-functions-called-once): Same.
1399 (-freorder-blocks-and-partition): Same.
1401 2018-04-09 Jan Hubicka <jh@suse.cz>
1404 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1405 jumps; choose last target that matches the criteria (i.e.
1406 no partition changes for non-crossing jumps).
1407 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1408 support for redirecting crossing jumps to non-crossing.
1410 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1412 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1413 also for naked functions.
1415 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1417 * config/arc/arc.md (add_shift): New pattern.
1418 (add_shift2): Likewise.
1419 (sub_shift): Likewise.
1420 (sub_shift_cmp0_noout): Likewise.
1421 (compare_si_ashiftsi): Likewise.
1422 (xbfu_cmp0_noout): New combine pattern.
1423 (xbfu_cmp0"): Likewise.
1424 (movsi_set_cc_insn): Place the predicable variant first.
1425 (commutative_binary_cmp0_noout): Remove clobber.
1426 (commutative_binary_cmp0): New pattern.
1427 (noncommutative_binary_cmp0): Likewise.
1428 (noncommutative_binary_cmp0_noout): Likewise.
1429 (noncommutative_binary_comparison_result_used): Removed.
1430 (rsub_cmp0): New pattern.
1431 (rsub_cmp0_noout): Likewise.
1432 (extzvsi): Changed, keep only meaningful variants.
1433 (SQH, SEZ): New iterators.
1434 (SQH_postfix): New mode attribute.
1435 (SEZ_prefix): New code attribute.
1436 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1437 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1438 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1440 (noncommutative_operator): Check the availability of barrel
1443 2018-04-09 Richard Biener <rguenther@suse.de>
1445 PR tree-optimization/85284
1446 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1447 Only use the niter constraining form of simple_iv when the exit
1450 2018-04-09 Tom de Vries <tom@codesourcery.com>
1453 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1454 (define_expand "*memory_barrier"): New define_expand.
1455 (define_insn "memory_barrier"): New insn.
1457 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1459 PR rtl-optimization/80463
1460 PR rtl-optimization/83972
1461 PR rtl-optimization/83480
1463 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1464 correct producer for the insn.
1465 (tidy_control_flow): Fixup seqnos in case of debug insns.
1467 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1469 PR rtl-optimization/83913
1471 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1472 different sched-times when merging exprs.
1474 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1476 PR rtl-optimization/83962
1478 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1479 tidy_fallthru_edge and tidy_control_flow.
1481 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1483 PR rtl-optimization/83530
1485 * sel-sched.c (force_next_insn): New global variable.
1486 (remove_insn_for_debug): When force_next_insn is true, also leave only
1487 next insn in the ready list.
1488 (sel_sched_region): When the region wasn't scheduled, make another pass
1489 over it with force_next_insn set to 1.
1491 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1493 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1495 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1496 for interrupt control.
1497 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1498 functions for interrupt control.
1499 * config/nds32/nds32-intrinsic.md: Likewise.
1500 * config/nds32/nds32_intrinsic.h: Likewise.
1501 * config/nds32/nds32.h (nds32_builtins): Likewise.
1503 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1505 * config/nds32/nds32.c (nds32_init_machine_status,
1506 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1507 strict_aligned_p field.
1508 (nds32_expand_to_rtl_hook): New function.
1509 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1510 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1512 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1513 Chung-Ju Wu <jasonwucj@gmail.com>
1515 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1516 * config/nds32/nds32-n7.md: New file.
1517 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1518 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1520 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1521 * config/nds32/nds32.md (pipeline_model): Add n7.
1522 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1523 * config/nds32/pipelines.md: Include n7 settings.
1525 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1526 Chung-Ju Wu <jasonwucj@gmail.com>
1528 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1529 * config/nds32/nds32-e8.md: New file.
1530 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1531 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1533 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1534 * config/nds32/nds32.md (pipeline_model): Add e8.
1535 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1536 * config/nds32/pipelines.md: Include e8 settings.
1538 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1539 Chung-Ju Wu <jasonwucj@gmail.com>
1541 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1542 * config/nds32/nds32-n8.md: New file.
1543 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1544 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1546 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1547 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1548 * config/nds32/nds32.md (pipeline_model): Add n8.
1549 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1550 * config/nds32/pipelines.md: Include n8 settings.
1552 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1553 Chung-Ju Wu <jasonwucj@gmail.com>
1555 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1556 * config/nds32/nds32-n9-2r1w.md: New file.
1557 * config/nds32/nds32-n9-3r2w.md: New file.
1558 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1559 nds32_register_ports): New or modify for cpu n9.
1560 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1562 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1563 * config/nds32/nds32-utils.c: New file.
1564 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1565 TARGET_MUL_SLOW): Define.
1566 * config/nds32/nds32.md (pipeline_model): New attribute.
1567 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1568 New options that support cpu n9.
1569 * config/nds32/pipelines.md: Include n9 settings.
1570 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1572 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1574 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1575 information if necessary.
1576 (output_cond_branch_compare_zero): Likewise.
1577 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1578 (nds32_target_alignment): Refine for alignment.
1579 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1580 (FUNCTION_BOUNDARY): Modify.
1581 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1583 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1585 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1587 * config/nds32/constants.md (unspec_volatile_element): Add values for
1588 TLB operation and data prefetch.
1589 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1590 functions for TLB operation and data prefetch.
1591 * config/nds32/nds32-intrinsic.md: Likewise.
1592 * config/nds32/nds32_intrinsic.h: Likewise.
1593 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1594 (nds32_print_operand): Likewise.
1595 * config/nds32/nds32.h (nds32_builtins): Likewise.
1597 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1598 Andrew Pinski <pinsika@gcc.gnu.org>
1601 * match.pd: Use constant_boolean_node of correct type instead of
1602 boolean_true_node or boolean_false_node for simplifying
1603 pointer comparisons to zero.
1605 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1607 PR tree-optimization/80021
1608 * tree.c (verify_type_variant): Make error call in verify_variant_match
1609 translatable and remove final full stop.
1611 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1613 * config/nds32/constants.md (unspec_volatile_element): Add
1614 UNSPEC_VOLATILE_EH_RETURN.
1615 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1616 nds32_output_stack_pop): Support dwarf exception handling process.
1617 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1618 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1619 exception handling process.
1620 (nds32_compute_stack_frame): Likewise.
1621 (nds32_return_addr_rtx): Likewise.
1622 (nds32_initial_elimination_offset): Likewise.
1623 (nds32_expand_prologue): Likewise.
1624 (nds32_expand_epilogue): Likewise.
1625 (nds32_dynamic_chain_address): New function.
1626 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1628 (DYNAMIC_CHAIN_ADDRESS): Define.
1629 (EH_RETURN_DATA_REGNO): Define.
1630 (EH_RETURN_STACKADJ_RTX): Define.
1631 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1632 patterns for dwarf exception handling.
1634 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1636 * config/nds32/nds32.h: Clean up obsolete macros.
1638 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1640 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1641 Add enum values for particular instructions.
1642 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1643 particular intrinsic functions.
1644 * config/nds32/nds32-intrinsic.md: Likewise.
1645 * config/nds32/nds32_intrinsic.h: Likewise.
1646 * config/nds32/nds32.h (nds32_builtins): Likewise.
1647 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1648 (btst, ave): New patterns for particular instructions.
1650 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1652 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1653 Add enum values for atomic load/store and memory sync.
1654 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1656 * config/nds32/nds32-intrinsic.md: Likewise.
1657 * config/nds32/nds32_intrinsic.h: Likewise.
1658 * config/nds32/nds32.h (nds32_builtins): Likewise.
1660 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1662 PR tree-optimization/85257
1663 * fold-const.c (native_encode_vector): If not all elts could fit
1664 and off is -1, return 0 rather than offset.
1665 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1666 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1667 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1668 adjust buffer in native_interpret_expr call.
1670 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1672 * config/nds32/constants.md (unspec_volatile_element): Add cache
1673 control enum values.
1674 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1675 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1676 * config/nds32/nds32.c (nds32_cctl_names): New.
1677 (nds32_print_operand): Handle cache control register names.
1678 * config/nds32/nds32.h (nds32_builtins): New enum values.
1679 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1681 * config/nds32/nds32.md (type): Add mmu.
1682 * config/nds32/pipelines.md (simple_insn): Add mmu.
1684 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1686 * config/nds32/nds32.md (type): Remove call.
1687 * config/nds32/pipelines.md (simple_insn): Likewise.
1689 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1691 * config/nds32/constants.md (unspec_volatile_element): Add
1692 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1693 UNSPEC_VOLATILE_FMFCFG.
1694 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1695 description for fmfcfg and fmfcsr.
1696 (bdesc_1arg): Add fmtcsr.
1697 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1698 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1699 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1700 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1701 unspec_fmfcfg): New patterns.
1702 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1703 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1704 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1705 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1706 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1707 __nds32__fmfcfg): Define.
1709 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1711 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1712 intrinsic register names.
1713 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1714 intrinsic register enum values and macros.
1716 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1718 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1719 for load/store addressing form.
1720 (nds32_print_operand_address): Likewise.
1722 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1725 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1726 based on LABEL_REF. Remove useless assertion.
1727 (pic_address_needs_scratch): Fix formatting.
1728 (sparc_legitimize_pic_address): Minor tweaks.
1729 (sparc_delegitimize_address): Adjust assertion accordingly.
1730 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1731 into symbolic_operand.
1732 (movsi_high_pic_label_ref): Likewise.
1733 (movsi_lo_sum_pic_label_ref): Likewise.
1734 (movdi_pic_label_ref): Likewise.
1735 (movdi_high_pic_label_ref): Likewise.
1736 (movdi_lo_sum_pic_label_ref): Likewise.
1738 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1740 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1741 custom LIB_SPEC setup.
1743 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1744 Kito Cheng <kito.cheng@gmail.com>
1746 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1747 * config/riscv/freebsd.h: New.
1749 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1751 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1752 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1755 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1756 Kito Cheng <kito.cheng@gmail.com>
1758 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1759 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1760 * config/nds32/nds32-protos.h (nds32_output_call,
1761 nds32_output_return): Declare.
1762 * config/nds32/nds32.md: Refine all the call and return patterns.
1764 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1767 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1768 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1770 PR rtl-optimization/84872
1771 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1772 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1775 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1777 * expr.c (copy_blkmode_to_reg): Revert 254862.
1778 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1780 2018-04-06 Richard Biener <rguenther@suse.de>
1783 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1784 after seeing a component reference with an adjacent field. Treat
1785 refs to arrays at struct end of external decls similar to
1786 refs to unconstrained commons.
1788 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1791 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1792 look through SAVE_EXPRs with non-side-effects argument. Adjust
1794 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1797 2018-04-06 Richard Biener <rguenther@suse.de>
1800 * alias.c (find_base_term): New wrapper around find_base_term
1801 unwinding CSELIB_VAL_PTR changes.
1802 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1805 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1807 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1809 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1810 constant definitions.
1811 ("nop"): lr 0,0 -> nopr r0
1812 ("nop_lr0", "nop_lr1"): New insn definitions.
1814 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1816 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1817 NDS32_V3PUSH_AVAILABLE_P macro.
1819 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1820 Chung-Ju Wu <jasonwucj@gmail.com>
1822 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1823 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1824 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1825 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1826 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1827 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1828 * config/nds32/constraints.md: New constraints and checking for hard
1829 float configuration.
1830 * config/nds32/iterators.md: New mode iterator and attribute for hard
1831 float configuration.
1832 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1834 * config/nds32/nds32-fpu.md: New file.
1835 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1836 deal with hard float code generation.
1837 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1839 (abi_type, float_reg_number): New enum type.
1840 * config/nds32/nds32-predicates.c: New predicates for hard float.
1841 * config/nds32/nds32-protos.h: Declare functions for hard float.
1842 * config/nds32/nds32.c: Implementation for hard float configuration.
1843 * config/nds32/nds32.h: Definitions for hard float configuration.
1844 * config/nds32/nds32.md: Include hard float machine description and
1845 modify patterns for hard float configuration.
1846 * config/nds32/nds32.opt: New options for hard float configuration.
1847 * config/nds32/predicates.md: New predicates for hard float
1850 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1852 * common/config/nds32/nds32-common.c
1853 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1855 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1858 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1859 CONSTRUCTOR_ELT (ctor, ...)->value.
1861 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1864 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1866 2018-04-05 Tom de Vries <tom@codesourcery.com>
1869 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1872 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1873 Kito Cheng <kito.cheng@gmail.com>
1875 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1876 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1877 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1878 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1880 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1881 Kito Cheng <kito.cheng@gmail.com>
1883 * config/nds32/constraints.md (Ufe): New memory constraint.
1884 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1885 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1886 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1888 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1889 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1891 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1893 * config/nds32/nds32.md: Use optimize_size in the condition for
1894 alu-shift instructions.
1896 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1898 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1900 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1902 * config/nds32/nds32.md (negsi2): Refine pattern.
1904 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1905 Chung-Ju Wu <jasonwucj@gmail.com>
1907 * config/nds32/iterators.md (shift_rotate): New code iterator.
1908 (shift): New code attribute.
1909 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1910 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1911 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1912 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1913 bit-wise operations.
1914 (andsi3, *andsi3): Ditto.
1915 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1916 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1917 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1918 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1919 nds32_ior_operand, nds32_xor_operand): New predicates.
1921 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1923 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1924 (addsi3, subsi3): ... this.
1926 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1928 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1930 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1932 * config/nds32/nds32.md: Adjust indention.
1934 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1936 * config/nds32/nds32.md (feature): New attribute.
1938 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1940 * config/nds32/nds32.md (subtype): New attribute.
1942 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1945 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1946 expansion to perform a bitwise AND of the argument followed by a
1947 boolean negation of the result.
1949 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1951 PR rtl-optimization/84878
1952 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1953 the basic block. Assert the use reference is not artificial and that
1954 it has an associated insn.
1956 2018-04-04 Michael Matz <matz@suse.de>
1958 * builtins.c (compute_objsize): Pass correct operand
1959 to array_at_struct_end_p.
1961 2018-04-04 Richard Biener <rguenther@suse.de>
1964 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1965 from contexts for DINFO_LEVEL_TERSE and below.
1967 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1969 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1970 resiter_operand condition.
1971 * config/nds32/nds32.md (*move<mode>): Ditto.
1973 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1974 Monk Chiang <sh.chiang04@gmail.com>
1976 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1978 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1980 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1982 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1983 Kito Cheng <kito.cheng@gmail.com>
1985 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1986 nds32_cond_code_str, output_cond_branch,
1987 output_cond_branch_compare_zero, nds32_expand_cbranch,
1988 nds32_expand_cstore, nds32_expand_movcc,
1989 nds32_output_cbranchsi4_equality_zero,
1990 nds32_output_cbranchsi4_equality_reg,
1991 nds32_output_cbranchsi4_equality_reg_or_const_int,
1992 nds32_output_cbranchsi4_greater_less_zero: New functions.
1993 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1994 nds32_expand_cstore, nds32_expand_movcc,
1995 nds32_output_cbranchsi4_equality_zero,
1996 nds32_output_cbranchsi4_equality_reg,
1997 nds32_output_cbranchsi4_equality_reg_or_const_int,
1998 nds32_output_cbranchsi4_greater_less_zero): Declare.
1999 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
2000 nds32_rimm11s_operand): New predicates.
2001 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
2002 * config/nds32/nds32.md: Rewrite all the branch and conditional move
2005 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
2007 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
2008 * config/nds32/nds32.md: Ditto.
2009 * config/nds32/pipelines.md: Ditto.
2011 2018-04-04 Richard Biener <rguenther@suse.de>
2013 PR tree-optimization/85168
2014 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
2015 propagating abnormals.
2017 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2019 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
2021 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
2022 Kito Cheng <kito.cheng@gmail.com>
2024 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
2025 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
2026 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
2027 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
2028 * config/nds32/nds32.md (sibcall_internal): New.
2029 (sibcall_register): Remove.
2030 (sibcall_immediate): Remove.
2031 (sibcall_value_internal): New.
2032 (sibcall_value_register): Remove.
2033 (sibcall_value_immediate): Remove.
2034 * config/nds32/predicates.md (nds32_general_register_operand): New.
2035 (nds32_call_address_operand): New.
2037 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2039 PR rtl-optimization/85167
2040 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
2041 bb_defs if *split_p, instead preinitialize it to NULL.
2043 PR tree-optimization/85156
2044 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
2045 evaluating the argument multiple times.
2047 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
2049 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
2051 (_mm_cvtpd_ps): Likewise.
2052 (_mm_cvttpd_epi32): Likewise.
2053 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
2054 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
2055 vector, pixel, and bool following altivec.h include.
2057 2018-04-03 Martin Sebor <msebor@redhat.com>
2059 * doc/extend.texi (Common Function Attributes): Clarify.
2060 (const attribute): Likewise.
2061 (pure attribute): Likewise.
2063 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2066 * config/i386/i386.c (ix86_expand_vector_set): Use
2067 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2069 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2071 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2072 instructions when changing rounding bits to preserve precision bits
2073 in the x87 control word.
2075 2018-04-03 Martin Liska <mliska@suse.cz>
2077 PR tree-optimization/82491
2078 * rtl.h (strip_offset_and_add): Replace += suboffset with
2079 poly_uint64 () + suboffset.
2081 2018-03-29 Martin Liska <mliska@suse.cz>
2082 Martin Jambor <mjambor@suse.cz>
2085 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2086 param_type is not an integral or pointer type.
2088 2018-04-03 Richard Biener <rguenther@suse.de>
2090 * sese.h (recompute_all_dominators): Remove.
2092 2018-04-02 Martin Sebor <msebor@redhat.com>
2094 * doc/invoke.texi (-Wrestrict): Fix typos.
2096 2018-04-02 Jim Wilson <jimw@sifive.com>
2098 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2099 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2100 (<optab>di3, <optab>si3_extend): Likewise.
2101 (<optab>si3_mask, <optab>si3_mask_1): New.
2102 (<optab>di3_mask, <optab>di3_mask_1): New.
2103 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2104 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2105 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2107 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2109 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2112 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2114 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2115 (nds32_canonicalize_comparison): New function.
2117 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2118 Kito Cheng <kito.cheng@gmail.com>
2119 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2121 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2122 * config/nds32/constants.md (unspec_volatile_element): Add
2123 UNSPEC_VOLATILE_RELAX_GROUP.
2124 * config/nds32/nds32-relax-opt.c: New file.
2125 * config/nds32/nds32-predicates.c
2126 (nds32_symbol_load_store_p): New function.
2127 * config/nds32/nds32-protos.h
2128 (nds32_symbol_load_store_p): Declare function.
2129 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2130 * config/nds32/nds32.c
2131 (nds32_register_pass): New function to register pass.
2132 (nds32_register_passes): New function to register passes.
2133 * config/nds32/nds32.md (relax_group): New pattern.
2134 * config/nds32/nds32.opt (mrelax-hint): New option.
2135 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2137 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2139 * config/nds32/t-nds32: Modify files dependency.
2141 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2143 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2144 (PROFILE_HOOK): Define its implementation.
2146 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2148 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2149 type and 32-bit size.
2151 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2154 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2155 (V_128_256): New mode iterator.
2156 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2157 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2158 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2160 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2161 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2163 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2166 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2167 NaN inputs correctly.
2169 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2172 * config/rs6000/vsx.md (??r): New mode attribute.
2173 (*vsx_mov<mode>_64bit): Use it.
2174 (*vsx_mov<mode>_32bit): Likewise.
2176 2018-03-30 Martin Sebor <msebor@redhat.com>
2178 PR tree-optimization/84818
2179 * builtins.c (check_access): Use warning_n.
2181 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2184 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2186 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2189 2018-03-30 Julia Koval <julia.koval@intel.com>
2192 * x86-tune.def (movx, partial_reg_dependency): Enable for
2195 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2198 * lra-constraints.c (process_alt_operands): Move setting
2199 this_alternative_matches below.
2201 2018-03-29 Martin Liska <mliska@suse.cz>
2204 * doc/invoke.texi: Document how LTO works with debug info.
2205 Describe auto-load support of binutils. Mention 'x86-64'
2206 as valid option value of -march option.
2208 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2210 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2213 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2214 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2215 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2218 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2221 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2222 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2223 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2224 for RS6000_BTM_POWERPC64.
2225 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2226 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2227 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2232 2018-03-28 Carl Love <cel@us.ibm.com>
2235 2017-09-27 Carl Love <cel@us.ibm.com>
2237 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2238 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2239 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2242 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2244 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2245 instead of __vector bool.
2246 (_mm_max_pu8): Likewise.
2247 (_mm_min_pi16): Likewise.
2249 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2252 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2253 (DIVWEUO): Likewise.
2255 (DIVDEUO): Likewise.
2256 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2257 DIVWEUO and DIVDEUO.
2258 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2259 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2260 (div_extend): Likewise.
2261 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2263 (__builtin_divweuo): Likewise.
2264 (__builtin_divdeo): Likewise.
2265 (__builtin_divdeuo): Likewise.
2267 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2270 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2271 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2273 PR tree-optimization/82004
2274 * gimple-match-head.c (optimize_pow_to_exp): New function.
2275 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2276 Don't fold to exp if optimize_pow_to_exp is false.
2278 2018-03-28 Martin Liska <mliska@suse.cz>
2281 * calls.c (initialize_argument_information): Fix trailing space.
2282 * common.opt: Fix typo and provide better explanation for
2283 -fsanitize-coverage option.
2284 * config/i386/i386.opt: Fix typo.
2286 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2287 Martin Liska <mliska@suse.cz>
2290 * gimplify.c (asan_poison_variable): Don't do the check for
2291 gimplify_omp_ctxp here.
2292 (gimplify_decl_expr): Do it here.
2293 (gimplify_target_expr): Likewise.
2295 2018-03-28 Martin Liska <mliska@suse.cz>
2298 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2299 chkp_type_bounds_count if MPX is not enabled.
2301 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2303 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2305 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2308 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2309 function to create the function decl for complex long double
2310 multiply and divide for -mabi=ieeelongdouble.
2311 (init_float128_ieee): Call it.
2313 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2316 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2317 -fcf-protection=branch -mibt.
2318 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2320 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2323 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2325 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2328 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2329 extern array declarations.
2331 2018-03-27 Richard Biener <rguenther@suse.de>
2334 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2335 explicit single_use checks.
2337 2018-03-27 Richard Biener <rguenther@suse.de>
2339 PR tree-optimization/85082
2340 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2343 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2345 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2346 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2347 Turn on fasynchronous-unwind-tables and funwind-tables.
2349 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2352 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2353 (*bmi_blsr_<mode>_ccz): Ditto.
2355 2018-03-26 Tom de Vries <tom@codesourcery.com>
2357 PR tree-optimization/85063
2358 * omp-general.c (offloading_function_p): New function. Factor out
2360 * omp-offload.c (pass_omp_target_link::gate): ... here.
2361 * omp-general.h (offloading_function_p): Declare.
2362 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2363 with attribute omp declare target for offloading functions.
2365 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2367 PR tree-optimization/84005
2368 * tree-data-ref.h (get_base_for_alignment): Declare.
2369 * tree-data-ref.c (get_base_for_alignment_1): New function.
2370 (get_base_for_alignment): Likewise.
2371 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2372 get_base_for_alignment to find a suitable base object, instead
2373 of always using drb->base_address.
2375 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2378 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2379 known size by default.
2381 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2384 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2385 and non BLKmode operands.
2387 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2390 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2391 Clean up attributes.
2393 2018-03-23 Richard Biener <rguenther@suse.de>
2396 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2397 we are going to emit early debug for LTO.
2399 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2402 * function.c (match_asm_constraints_1): Don't optimize if input
2403 doesn't satisfy general_operand predicate for output's mode.
2406 * alias.c (write_dependence_p): Don't require for x_canonicalized
2407 non-VOIDmode if x has VOIDmode.
2410 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2411 just don't try to optimize it rather than assert it never happens.
2413 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2415 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2416 macro expansions for definition of ST_INTERNAL_<mode> and
2417 LD_INTERNAL_<mode> builtins.
2418 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2420 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2422 (altivec_expand_st_builtin): Likewise.
2423 (altivec_expand_builtin): Remove calls to deleted functions.
2424 (rs6000_address_for_altivec): Delete this function.
2425 * config/rs6000/vector.md: Remove expands for
2426 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2428 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2431 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2432 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2433 re-computing once computed.
2434 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2435 (arm_init_machine_status): Initialize
2436 machine->static_chain_stack_bytes.
2438 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2441 * doc/extend.texi: Add four new prototypes for vec_ld.
2442 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2443 definitions for more logical presentation.
2444 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2445 entries for V1TI variants of __builtin_altivec_ld builtin.
2446 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2447 handling of V1TI variant of LVX icode pattern.
2448 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2449 (rs6000_gimple_fold_builtin): Likewise.
2450 (altivec_init_builtins): Add code to define
2451 __builtin_altivec_lvx_v1ti function.
2453 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2456 * function.c (match_asm_constraints_1): Don't do the optimization
2457 if input isn't a REG, SUBREG, MEM or constant.
2459 2018-03-22 Tom de Vries <tom@codesourcery.com>
2461 PR tree-optimization/84956
2462 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2463 bb_has_abnormal_pred.
2465 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2468 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2469 DECL_INITIAL (decl) to decl at the end.
2470 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2473 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2475 * doc/extend.texi (__builtin_tgmath): Document when complex
2476 integer types are treated as _Complex _Float64.
2478 2018-03-21 Tom de Vries <tom@codesourcery.com>
2480 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2482 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2484 PR tree-optimization/84960
2485 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2486 if it is ENTRY block, move them into single succ of ENTRY in that case.
2488 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2490 PR tree-optimization/84811
2491 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2492 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2493 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2494 (known_subrange_p): Update accordingly. Cast each value involved
2495 in the size comparison, rather than casting the result of the
2498 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2500 PR tree-optimization/84982
2501 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2502 by flipping the least significant bit rather than all bits from
2503 bitpos to bitpos + bitsize - 1.
2505 2018-03-21 Nathan Sidwell <nathan@acm.org>
2507 * doc/extend.texi (Deprecated Features): Remove mention of
2508 long-deleted deprecations.
2510 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2513 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2514 * configure: Regenerate.
2516 2018-03-21 Tom de Vries <tom@codesourcery.com>
2518 PR tree-optimization/83126
2519 * tree-parloops.c (num_phis): New function.
2520 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2522 2018-03-21 Nathan Sidwell <nathan@acm.org>
2524 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2525 mention anon-struct/union members and trailing attributes.
2527 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2529 PR tree-optimization/84969
2530 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2531 builtin memset partitions if they set different rhs values.
2533 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2535 PR rtl-optimization/84989
2536 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2537 VEC_DUPLICATE with scalar result mode.
2539 2018-03-21 Martin Liska <mliska@suse.cz>
2542 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2543 not intended return statement.
2545 2018-03-21 Martin Liska <mliska@suse.cz>
2548 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2549 (chkp_find_bound_slots_1): Limit number of iterations.
2551 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2554 * Minor grammar fixes for x86 options.
2556 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2559 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2560 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2562 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2565 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2566 (altivec_lvx_<mode>_1op): Likewise.
2567 (altivec_stvx_<mode>_2op): Likewise.
2568 (altivec_stvx_<mode>_1op): Likewise.
2569 (altivec_lvx_<VM2:mode>): New define_expand.
2570 (altivec_stvx_<VM2:mode>): Likewise.
2571 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2572 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2573 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2574 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2575 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2576 (rs6000_gen_lvx): Likewise.
2577 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2578 (altivec_expand_stv_builtin): Likewise.
2579 (altivec_expand_builtin): Likewise.
2580 * config/rs6000/vector.md: Likewise.
2582 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2585 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2588 2018-03-20 Richard Biener <rguenther@suse.de>
2591 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2592 sign-conversions as zero, fall back to standard scalar_stmt
2595 2018-03-20 Martin Liska <mliska@suse.cz>
2598 * predict.c (rebuild_frequencies): Handle case when we have
2599 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2601 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2604 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2605 flag_section_anchors.
2606 * varasm.c (use_blocks_for_decl_p): Remove hack for
2607 dw2_force_const_mem.
2610 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2612 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2613 be created, use lowpart_subreg of operands[0] rather than operands[0]
2615 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2616 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2617 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2618 and n constraint instead of aarch64_shift_imm_di and Usd.
2619 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2620 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2622 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2625 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2626 to favor GPR over NEON registers.
2627 (<shift>di3_neon): Likewise.
2629 2018-03-20 Tom de Vries <tom@codesourcery.com>
2632 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2633 (nvptx_process_pars): Emit bar.sync asap and alap.
2635 2018-03-20 Tom de Vries <tom@codesourcery.com>
2638 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2639 seen_label if seen_label is already set.
2641 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2644 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2645 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2646 Use 1U instead of 1. Formatting fixes.
2649 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2650 instead of TREE_TYPE (s1) for the return value.
2652 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2654 PR tree-optimization/84946
2655 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2656 bitsize + bitsize in poly_uint64 rather than poly_int64.
2659 * dwarf2asm.c: Include fold-const.c.
2660 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2661 of decl rather than decl itself.
2663 PR rtl-optimization/84643
2664 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2666 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2669 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2670 calling assemble_variable.
2672 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2675 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2676 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2678 2018-03-19 Jim Wilson <jimw@sifive.com>
2681 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2682 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2683 (riscv_first_stack_step): Likewise.
2684 (riscv_option_override): Use STACK_BOUNDARY instead of
2686 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2688 (BIGGEST_ALIGNMENT): Set to 128.
2689 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2690 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2693 2018-03-19 Richard Biener <rguenther@suse.de>
2695 PR tree-optimization/84933
2696 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2697 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2699 2018-03-19 Richard Biener <rguenther@suse.de>
2701 PR tree-optimization/84859
2702 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2703 (cond_if_else_store_replacement): Perform sinking operation on
2704 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2705 Generalize what a BB with a single eligible store is.
2707 2018-03-19 Richard Biener <rguenther@suse.de>
2709 PR tree-optimization/84929
2710 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2711 chrec_is_positive against non-chrec arg.
2713 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2716 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2718 2018-03-18 Martin Liska <mliska@suse.cz>
2720 PR rtl-optimization/84635
2721 * regrename.c (build_def_use): Use matches_mode only when
2724 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2726 PR tree-optimization/84913
2727 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2728 vectorize chains of COND_EXPRs.
2730 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2732 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2734 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2736 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2738 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2740 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2742 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2743 Kito Cheng <kito.cheng@gmail.com>
2745 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2746 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2747 (nds32_adjust_reg_alloc_order): New function.
2748 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2750 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2752 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2753 nds32_print_operand, nds32_print_operand_address): Use
2754 HOST_WIDE_INT_PRINT_DEC instead.
2756 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2758 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2760 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2763 * config/i386/i386.c (initial_ix86_tune_features,
2764 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2766 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2767 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2768 rather than 1u << ix86_tune. Formatting fix.
2769 (ix86_option_override_internal): Change ix86_arch_mask from
2770 unsigned int to unsigned HOST_WIDE_INT, initialize to
2771 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2772 (ix86_function_specific_restore): Likewise.
2774 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2777 * postreload.c (reload_combine_recognize_pattern): Perform
2778 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2779 truncate_int_for_mode the result for the destination's mode.
2782 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2783 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2785 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2788 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2791 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2792 regno_allocno_class_array and sorted_pseudos.
2793 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2794 insns where regno is used.
2796 2018-03-16 Martin Liska <mliska@suse.cz>
2799 * multiple_target.c (create_dispatcher_calls): Redirect
2800 reference in the symbol table.
2802 2018-03-16 Martin Liska <mliska@suse.cz>
2805 * multiple_target.c (create_dispatcher_calls): Redirect also
2808 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2812 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2813 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2814 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2816 2018-03-16 Julia Koval <julia.koval@intel.com>
2818 * doc/invoke.texi (Skylake Server): Add CLWB.
2819 Cannonlake): Remove CLWB.
2821 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2823 PR tree-optimization/84841
2824 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2826 (FLOAT_ONE_CONST_TYPE): Define.
2827 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2828 (sort_by_operand_rank): Put entries with higher constant_type last
2829 rather than first to match comments.
2831 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2833 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2836 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2839 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2840 check and use address of target always.
2842 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2845 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2846 (indirect_thunk_bnd_needed): Likewise.
2847 (indirect_thunks_used): Likewise.
2848 (indirect_thunks_bnd_used): Likewise.
2849 (indirect_return_needed): New.
2850 (indirect_return_bnd_needed): Likewise.
2851 (output_indirect_thunk_function): Add a bool argument for
2853 (output_indirect_thunk_function): Don't generate alias for
2854 function return thunk.
2855 (ix86_code_end): Call output_indirect_thunk_function to generate
2856 function return thunks.
2857 (ix86_output_function_return): Set indirect_return_bnd_needed
2858 and indirect_return_needed instead of indirect_thunk_bnd_needed
2859 and indirect_thunk_needed.
2861 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2863 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2864 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2865 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2867 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2868 Paul Hua <paul.hua.gm@gmail.com>
2871 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2873 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2875 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2876 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2879 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2882 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2883 instead of GET_MODE_SIZE when comparing Units.
2885 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2888 * varasm.c (hash_section): Return an unchangeble hash value
2889 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2890 Return !aarch64_can_use_per_function_literal_pools_p ().
2892 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2895 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2896 rather than address of cmode as last argument to prepare_cmp_insn.
2898 2018-03-15 Julia Koval <julia.koval@intel.com>
2900 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2901 F_AVX512VNNI, F_AVX512BITALG): New.
2903 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2906 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2907 insn for floating-point loads and stores.
2909 2018-03-14 Carl Love <cel@us.ibm.com>
2911 * config/rs6000/rs6000-c.c: Add macro definitions for
2912 ALTIVEC_BUILTIN_VEC_PERMXOR.
2913 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2914 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2915 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2916 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2918 * config/doc/extend.texi: Add prototypes for vec_permxor.
2920 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2923 * diagnostic-show-locus.c (class layout_point): Convert m_line
2924 from int to linenum_type.
2925 (line_span::comparator): Use linenum "compare" function when
2926 comparing line numbers.
2927 (test_line_span): New function.
2928 (layout_range::contains_point): Convert param "row" from int to
2930 (layout_range::intersects_line_p): Likewise.
2931 (layout::will_show_line_p): Likewise.
2932 (layout::print_source_line): Likewise.
2933 (layout::should_print_annotation_line_p): Likewise.
2934 (layout::print_annotation_line): Likewise.
2935 (layout::print_leading_fixits): Likewise.
2936 (layout::annotation_line_showed_range_p): Likewise.
2937 (struct line_corrections): Likewise for field m_row.
2938 (line_corrections::line_corrections): Likewise for param "row".
2939 (layout::print_trailing_fixits): Likewise.
2940 (layout::get_state_at_point): Likewise.
2941 (layout::get_x_bound_for_row): Likewise.
2942 (layout::print_line): Likewise.
2943 (diagnostic_show_locus): Likewise for locals "last_line" and
2945 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2946 * input.c (selftest::test_linenum_comparisons): New function.
2947 (selftest::input_c_tests): Call it.
2948 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2949 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2950 * selftest.h (ASSERT_GT): New macro.
2951 (ASSERT_GT_AT): New macro.
2952 (ASSERT_LT): New macro.
2953 (ASSERT_LT_AT): New macro.
2955 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2957 PR rtl-optimization/84780
2958 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2960 2018-03-14 Martin Liska <mliska@suse.cz>
2962 * tree.c (record_node_allocation_statistics): Use
2963 get_stats_node_kind.
2964 (get_stats_node_kind): New function extracted from
2965 record_node_allocation_statistics.
2966 (free_node): Use get_stats_node_kind.
2968 2018-03-14 Richard Biener <rguenther@suse.de>
2970 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2971 that the value-set of ANTIC_IN doesn't grow.
2974 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2976 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2977 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2979 2018-03-14 Julia Koval <julia.koval@intel.com>
2981 * config.gcc (icelake-client, icelake-server): New.
2983 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2984 (initial_ix86_arch_features): Ditto.
2985 (PTA_SKYLAKE): Add SGX.
2986 (PTA_ICELAKE): Remove.
2987 (PTA_ICELAKE_CLIENT): New.
2988 (PTA_ICELAKE_SERVER): New.
2989 (ix86_option_override_internal): Split up icelake on icelake client and
2991 (get_builtin_code_for_version): Ditto.
2992 (fold_builtin_cpu): Ditto.
2993 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2994 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2995 * config/i386/i386.h (processor_type): Ditto.
2996 * doc/invoke.texi: Ditto.
2998 2018-03-14 Jakub Jelinek <jakub@redhat.com>
3001 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
3002 INTEGER_CST offset, add it together with bitpos / 8 and
3003 sign extend based on POINTER_SIZE.
3007 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
3010 * config/i386/constraints.md (Yc): New register constraint.
3011 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
3012 Use Yc constraint for alternative 2 of operand 0. Remove
3013 preferred_for_speed attribute.
3015 2018-03-14 Richard Biener <rguenther@suse.de>
3017 PR tree-optimization/84830
3018 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
3019 with the old one to avoid oscillations.
3021 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
3024 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
3026 (assign_by_spills): Return a flag of reload assignment failure.
3027 Do not process the reload assignment failures. Do not spill other
3028 reload pseudos if they has the same reg class. Update n if
3030 (lra_assign): Add a return arg. Set up from the result of
3031 assign_by_spills call.
3032 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3033 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3034 usage_insns if it is not NULL.
3035 (spill_hard_reg_in_range): New function.
3036 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3037 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3038 function prototypes.
3039 (lra_assign): Change prototype.
3040 * lra.c (lra): Add code to deal with fails by splitting hard reg
3043 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
3045 * config/riscv/riscv.opt (mrelax): New option.
3046 * config/riscv/riscv.c (riscv_file_start): Emit ".option
3047 "norelax" when riscv_mrelax is disabled.
3048 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
3050 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
3053 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
3054 reassociation for int modes.
3056 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3058 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3059 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3061 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3062 * config/aarch64/aarch64-sve.md
3063 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3064 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3065 (*extend<mode><Vwide>2): Rename to...
3066 (aarch64_sve_extend<mode><Vwide>2): ...this.
3067 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3068 renaming the old pattern to...
3069 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3071 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3072 define_expand, renaming the old pattern to...
3073 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3074 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3075 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3076 account when deciding which SVE instruction the optab should use.
3077 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3079 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3081 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3082 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3083 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3084 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3085 (tlsdesc_small_<mode>): Turn a define_expand and use
3086 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3087 (tlsdesc_small_advsimd_<mode>): ...this.
3088 (tlsdesc_small_sve_<mode>): New pattern.
3090 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3092 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3093 (UNSPEC_UMUL_HIGHPART): New constants.
3094 (MUL_HIGHPART): New int iteraor.
3095 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3096 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3098 (*<su>mul<mode>3_highpart): New define_insn.
3100 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3103 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3106 2018-03-13 Martin Liska <mliska@suse.cz>
3109 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3111 (sem_item_optimizer::~sem_item_optimizer): Release it.
3112 (sem_item_optimizer::merge_classes): Register variable aliases.
3113 (sem_item_optimizer::fixup_pt_set): New function.
3114 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3115 * ipa-icf.h: Declare new variables and functions.
3117 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3120 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3121 integer_pow2p@2 and test integer_pow2p in condition.
3122 (A < 0 ? C : 0): Similarly for @1.
3125 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3126 characters starting at p contain '\0' character, don't look beyond
3130 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3131 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3134 * reg-stack.c (change_stack): Change update_end var from int to
3135 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3136 also call set_block_for_insn on the newly added insns and rescan.
3139 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3140 on the last operand.
3143 * tree.c (stabilize_reference_1): Return save_expr (e) for
3144 STATEMENT_LIST even if it doesn't have side-effects.
3146 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3148 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3150 2018-03-12 Renlin Li <renlin.li@arm.com>
3152 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3153 aarch64_output_scalar_simd_mov_immediate.
3155 2018-03-12 Martin Sebor <msebor@redhat.com>
3157 PR tree-optimization/83456
3158 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3159 for perfectly overlapping calls to memcpy.
3160 (gimple_fold_builtin_memory_chk): Same.
3161 (gimple_fold_builtin_strcpy): Handle no-warning.
3162 (gimple_fold_builtin_stxcpy_chk): Same.
3163 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3165 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3167 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3168 parameter. Use it for SFmode.
3169 (rs6000_function_arg_advance_1): Adjust.
3170 (rs6000_function_arg): Adjust.
3171 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3173 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3175 PR rtl-optimization/84169
3176 PR rtl-optimization/84780
3177 * combine.c (can_combine_p): Check for a 2-insn combination whether
3178 the destination register is used between the two insns, too.
3180 2018-03-12 Richard Biener <rguenther@suse.de>
3182 PR tree-optimization/84803
3183 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3184 for refs DR analysis didn't process.
3186 2018-03-12 Richard Biener <rguenther@suse.de>
3188 PR tree-optimization/84777
3189 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3190 force-vectorize loops ignore whether we are optimizing for size.
3192 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3194 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3195 (TARGET_MD_ASM_ADJUST): Define.
3197 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3198 Kito Cheng <kito.cheng@gmail.com>
3199 Chung-Ju Wu <jasonwucj@gmail.com>
3201 * config/nds32/nds32.c (nds32_compute_stack_frame,
3202 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3203 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3204 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3205 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3206 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3207 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3208 * config/nds32/nds32.md (prologue, epilogue): Use macro
3209 NDS32_V3PUSH_AVAILABLE_P to do checking.
3211 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3214 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3215 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3216 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3217 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3218 addition of most attributes on !orig_type_die or the attribute not
3219 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3221 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3222 Chung-Ju Wu <jasonwucj@gmail.com>
3224 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3226 * config/nds32/nds32.opt (mvh): New option.
3228 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3229 Chung-Ju Wu <jasonwucj@gmail.com>
3231 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3233 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3234 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3237 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3238 Chung-Ju Wu <jasonwucj@gmail.com>
3240 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3242 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3243 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3245 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3246 Kito Cheng <kito.cheng@gmail.com>
3247 Chung-Ju Wu <jasonwucj@gmail.com>
3249 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3250 UNSPEC_FFMISM and UNSPEC_FLMISM.
3251 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3252 for ffb, ffmism and flmism.
3253 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3254 (unspec_ffmism): Ditto.
3255 (unspec_flmism): Ditto.
3256 (nds32_expand_builtin_impl): Check if string extension is available.
3257 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3258 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3260 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3263 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3266 * lra-assigns.c (assign_by_spills): Return a flag of reload
3267 assignment failure. Do not process the reload assignment
3268 failures. Do not spill other reload pseudos if they has the same
3270 (lra_assign): Add a return arg. Set up from the result of
3271 assign_by_spills call.
3272 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3273 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3274 usage_insns if it is not NULL.
3275 (spill_hard_reg_in_range): New function.
3276 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3277 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3278 function prototypes.
3279 (lra_assign): Change prototype.
3280 * lra.c (lra): Add code to deal with fails by splitting hard reg
3283 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3286 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3288 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3291 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3292 (dw_val_node): Add val_symbolic_view.
3293 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3294 (symview_upper_bound): New.
3295 (new_line_info_table): Initialize symviews_since_reset.
3296 (dwarf2out_source_line): Count symviews_since_reset and set
3297 symview_upper_bound.
3298 (dw_val_equal_p): Handle symview.
3299 (add_AT_symview): New.
3300 (print_dw_val): Handle symview.
3301 (attr_checksum, attr_checksum_ordered): Likewise.
3302 (same_dw_val_p, size_of_die): Likewise.
3303 (value_format, output_die): Likewise.
3304 (add_high_low_attributes): Use add_AT_symview for entry_view.
3305 (dwarf2out_finish): Reset symview_upper_bound, clear
3308 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3311 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3312 Add strict argument and use it.
3313 (rs6000_split_multireg_move): Update for new strict argument.
3314 (mem_operand_gpr): Disallow all non-offsettable addresses.
3315 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3317 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3320 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3321 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3322 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3325 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3326 decl, use remap_type if we want to use the type.
3328 2018-03-09 Martin Sebor <msebor@redhat.com>
3330 PR tree-optimization/84526
3331 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3333 (builtin_access::generic_overlap): Be prepared to handle non-array
3336 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3338 PR rtl-optimization/84682
3339 * lra-constraints.c (process_address_1): Check is_address flag
3340 for address constraints.
3341 (process_alt_operands): Likewise.
3342 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3343 preprocess_constraints.
3344 * recog.h (preprocess_constraints): Add oploc parameter.
3346 * recog.c (preprocess_constraints): Test address_operand for
3347 CT_ADDRESS constraints.
3349 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3352 * lra-assigns.c (assign_by_spills): Return a flag of reload
3353 assignment failure. Do not process the reload assignment
3354 failures. Do not spill other reload pseudos if they has the same
3356 (lra_assign): Add a return arg. Set up from the result of
3357 assign_by_spills call.
3358 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3359 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3360 usage_insns if it is not NULL.
3361 (spill_hard_reg_in_range): New function.
3362 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3363 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3364 function prototypes.
3365 (lra_assign): Change prototype.
3366 * lra.c (lra): Add code to deal with fails by splitting hard reg
3369 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3372 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3373 Accept complain bool parameter. Only emit errors if it is true.
3374 (arm_parse_cpu_option_name): Likewise.
3375 (arm_target_thumb_only): Adjust callers of the above.
3376 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3377 prototype to take a default true bool parameter.
3378 (arm_parse_arch_option_name): Likewise.
3380 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3381 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3385 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3386 * configure: Regenerate.
3387 * configure.ac ("linker --version-script option"): New.
3388 ("linker soname option"): New.
3390 2018-03-09 Richard Biener <rguenther@suse.de>
3392 PR tree-optimization/84775
3393 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3394 immediate uses of predicate stmts and mark them modified.
3397 PR tree-optimization/84178
3398 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3400 (version_loop_for_if_conversion): Delay update_ssa call.
3401 (tree_if_conversion): Delay update_ssa until after predicate
3404 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3407 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3408 when the function accesses prior frames.
3410 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3413 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3414 gen_llsym, otherwise call maybe_gen_llsym.
3417 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3418 has ',' character inside of it.
3420 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3423 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3424 as clobbering CC_REGNUM.
3426 2018-03-08 Richard Biener <rguenther@suse.de>
3429 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3430 (follow_copies_to_constant): Do not follow SSA names registered
3433 2018-03-08 Richard Biener <rguenther@suse.de>
3435 PR tree-optimization/84178
3436 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3438 (version_loop_for_if_conversion): Delay update_ssa call.
3439 (tree_if_conversion): Delay update_ssa until after predicate
3442 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3444 PR tree-optimization/84178
3445 * tree-if-conv.c (release_bb_predicate): Remove the
3446 the assertion that the stmts have NULL use_ops.
3447 Discard the statements, asserting that they haven't
3448 yet been added to a BB.
3450 2018-03-08 Richard Biener <rguenther@suse.de>
3452 PR tree-optimization/84746
3453 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3454 (phi_translate): Pass in destination ANTIC_OUT set.
3455 (phi_translate_1): Likewise. For a simplified result lookup
3456 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3457 (phi_translate_set): Adjust.
3458 (do_pre_regular_insertion): Likewise.
3459 (do_pre_partial_partial_insertion): Likewise.
3461 2018-03-08 Martin Liska <mliska@suse.cz>
3463 PR gcov-profile/84735
3464 * doc/gcov.texi: Document usage of profile files.
3465 * gcov-io.h: Document changes in the format.
3467 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3471 * dwarf2out.c (struct dw_line_info_table): Update comments for
3473 (FORCE_RESET_NEXT_VIEW): New.
3474 (FORCE_RESETTING_VIEW_P): New.
3475 (RESETTING_VIEW_P): Check for -1 too.
3476 (ZERO_VIEW_P): Likewise.
3477 (new_line_info_table): Force-reset next view.
3478 (dwarf2out_begin_function): Likewise.
3479 (dwarf2out_source_line): Simplify zero_view_p initialization.
3480 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3481 view directly. Omit view when omitting .loc at line 0.
3483 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3485 PR tree-optimization/84740
3486 * tree-switch-conversion.c (process_switch): Call build_constructors
3487 only if info.phi_count is non-zero.
3489 PR tree-optimization/84739
3490 * tree-tailcall.c (find_tail_calls): Check call arguments against
3491 DECL_ARGUMENTS (current_function_decl) rather than
3492 DECL_ARGUMENTS (func) when checking for tail recursion.
3494 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3496 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3497 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3498 Volker Reichelt's entry and add entries for people that perform
3499 GCC fuzzy testing and report numerous bugs.
3501 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3504 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3505 readonly data in sdata, if that is disabled.
3506 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3507 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3508 -mreadonly-in-sdata option.
3510 2018-03-07 Martin Sebor <msebor@redhat.com>
3512 PR tree-optimization/84468
3513 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3514 basic block when looking for nul assignment.
3516 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3519 * except.h (output_function_exception_table): Adjust prototype.
3520 * except.c (output_function_exception_table): Remove FNNAME parameter
3521 and add SECTION parameter. Ouput one part of the table at a time.
3522 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3523 the first part of the exception table and emit unwind directives.
3524 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3525 (i386_pe_seh_cold_init): Likewise.
3526 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3527 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3528 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3529 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3530 * config/i386/winnt.c: Include except.h.
3531 (struct seh_frame_state): Add reg_offset, after_prologue and
3532 in_cold_section fields.
3533 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3534 (i386_pe_seh_cold_init): New function.
3535 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3536 to seh->in_cold_section.
3537 (seh_emit_push): Record the offset of the push.
3538 (seh_emit_save): Record the offet of the save.
3539 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3540 Test seh->after_prologue to disregard the epilogue.
3541 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3542 (i386_pe_end_cold_function): New function.
3544 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3547 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3548 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3551 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3553 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3554 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3557 * multiple_target.c: Include tree-inline.h and intl.h.
3558 (expand_target_clones): Diagnose and fail if node->definition and
3559 !tree_versionable_function_p (node->decl).
3561 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3563 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3565 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3566 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3567 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3569 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3572 * combine.c (try_combine): Use reg_or_subregno instead of handling
3573 just paradoxical SUBREGs and REGs.
3575 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3577 * config/arc/arc.c (arc_finalize_pic): Remove function.
3578 (arc_must_save_register): We use single base PIC register, remove
3579 checks to save/restore the PIC register.
3580 (arc_expand_prologue): Likewise.
3581 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3583 (arc_verify_short): Likewise.
3584 (arc_attr_type): Likewise.
3585 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3586 (walk_stores): Likewise.
3587 (arc_address_cost): Make it static.
3588 (arc_verify_short): Likewise.
3589 (branch_dest): Likewise.
3590 (arc_attr_type): Likewise.
3591 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3592 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3593 (arc_final_prescan_insn): Remove inserting the nops due to
3594 hardware hazards. It is done in reorg step.
3595 (insn_length_variant_t): Remove.
3596 (insn_length_parameters_t): Likewise.
3597 (arc_insn_length_parameters): Likewise.
3598 (arc_get_insn_variants): Likewise.
3599 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3601 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3604 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3607 PR tree-optimization/84687
3608 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3610 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3612 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3614 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3615 Rename to ppc_speculation_barrier.
3616 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3617 __builtin_ppc_speculation_barrier.
3619 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3622 * combine.c (combine_simplify_rtx): Don't try to simplify if
3623 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3626 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3628 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3629 to 32 bytes when compiling for POWER9.
3631 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3634 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3635 regparm >= 3 with no arg reg available also for calls with
3636 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3639 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3641 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3643 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3646 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3648 2018-03-05 Richard Biener <rguenther@suse.de>
3650 PR tree-optimization/84486
3651 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3652 When inserting a __builtin_assume_aligned call set the LHS
3653 SSA name alignment info accordingly.
3655 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3657 PR tree-optimization/84114
3658 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3659 Avoid reassociation of FLOAT_MODE addition.
3661 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3663 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3664 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3665 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3666 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3667 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3668 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3669 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3671 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3672 __builtin_ia32_wbinvd): New builtins.
3673 (SPECIAL_ARGS2): New.
3674 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3675 (SPECIAL_ARGS2): New.
3676 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3677 (ix86_valid_target_attribute_inner_p): Ditto.
3678 (ix86_init_mmx_sse_builtins): Add special_args2.
3679 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3680 TARGET_WBNOINVD_P): New.
3681 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3682 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3683 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3684 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3685 * config/i386/pconfigintrin.h: New file.
3686 * config/i386/wbnoinvdintrin.h: Ditto.
3687 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3688 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3690 2018-03-05 Richard Biener <rguenther@suse.de>
3692 PR tree-optimization/84670
3693 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3695 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3696 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3697 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3698 doesn't grow if all successors (recursively) were visited at least
3701 2018-03-05 Richard Biener <rguenther@suse.de>
3703 PR tree-optimization/84650
3704 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3705 if executed in the loop pipeline.
3707 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3709 * doc/configfiles.texi (Configuration Files): Move info about
3710 conditionalizing $target-protos.h to...
3711 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3712 differs from $target-protos.h.
3714 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3715 Chung-Ju Wu <jasonwucj@gmail.com>
3717 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3718 * config/nds32/nds32-multiple.md (setmemsi): Define.
3719 * config/nds32/nds32-memory-manipulation.c
3720 (nds32_gen_dup_4_byte_to_word_value): New.
3721 (emit_setmem_word_loop): New.
3722 (emit_setmem_byte_loop): New.
3723 (nds32_expand_setmem_loop): New.
3724 (nds32_expand_setmem_loop_v3m): New.
3725 (nds32_expand_setmem_unroll): New.
3726 (nds32_expand_setmem): New.
3728 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3729 Chung-Ju Wu <jasonwucj@gmail.com>
3731 * config/nds32/nds32-memory-manipulation.c
3732 (nds32_emit_load_store): New.
3733 (nds32_emit_post_inc_load_store): New.
3734 (nds32_emit_mem_move): New.
3735 (nds32_emit_mem_move_block): New.
3736 (nds32_expand_movmemsi_loop_unknown_size): New.
3737 (nds32_expand_movmemsi_loop_known_size): New.
3738 (nds32_expand_movmemsi_loop): New.
3739 (nds32_expand_movmemsi_unroll): New.
3740 (nds32_expand_movmemqi): Rename ...
3741 (nds32_expand_movmemsi): ... to this.
3742 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3743 (movmemsi): ... to this.
3744 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3745 (nds32_expand_movmemsi): ... to this.
3747 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3748 Monk Chiang <sh.chiang04@gmail.com>
3749 Chung-Ju Wu <jasonwucj@gmail.com>
3751 * config/nds32/nds32-protos.h
3752 (nds32_expand_load_multiple): New arguments.
3753 (nds32_expand_store_multiple): Ditto.
3754 (nds32_valid_multiple_load_store): Rename ...
3755 (nds32_valid_multiple_load_store_p): ... to this.
3756 * config/nds32/nds32-memory-manipulation.c
3757 (nds32_expand_load_multiple): Refine implementation.
3758 (nds32_expand_store_multiple): Ditto.
3759 * config/nds32/nds32-multiple.md
3760 (load_multiple): Update nds32_expand_load_multiple interface.
3761 (store_multiple): Update nds32_expand_store_multiple interface.
3762 * config/nds32/nds32-predicates.c
3763 (nds32_valid_multiple_load_store): Rename ...
3764 (nds32_valid_multiple_load_store_p): ... to this and refine
3766 * config/nds32/predicates.md
3767 (nds32_load_multiple_and_update_address_operation): New predicate.
3768 (nds32_store_multiple_and_update_address_operation): New predicate.
3770 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3771 Chung-Ju Wu <jasonwucj@gmail.com>
3773 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3774 (combo): New attribute.
3775 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3777 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3779 * config/nds32/nds32.opt: Change -mcmodel= default value.
3781 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3782 Monk Chiang <sh.chiang04@gmail.com>
3783 Chung-Ju Wu <jasonwucj@gmail.com>
3785 * config/nds32/constants.md (unspec_element): New enum.
3786 * config/nds32/constraints.md (Umw): New constraint.
3787 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3788 * config/nds32/nds32-intrinsic.md: Likewise.
3789 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3790 (nds32_valid_smw_lwm_base_p): New.
3791 (nds32_output_smw_single_word): New.
3792 (nds32_output_lmw_single_word): New.
3793 (nds32_expand_unaligned_load): New.
3794 (nds32_expand_unaligned_store): New.
3795 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3796 (nds32_output_smw_single_word): Declare.
3797 (nds32_output_lmw_single_word): Declare.
3798 (nds32_expand_unaligned_load): Declare.
3799 (nds32_expand_unaligned_store): Declare.
3800 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3801 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3802 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3803 NDS32_BUILTIN_UASTORE_DW.
3804 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3807 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3808 Kito Cheng <kito.cheng@gmail.com>
3809 Chung-Ju Wu <jasonwucj@gmail.com>
3811 * config/nds32/nds32-intrinsic.c
3812 (nds32_expand_builtin_null_ftype_reg): Delete.
3813 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3814 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3815 (nds32_read_argument): New.
3816 (nds32_legitimize_target): Ditto.
3817 (nds32_legitimize_argument): Ditto.
3818 (nds32_check_constant_argument): Ditto.
3819 (nds32_expand_unop_builtin): Ditto.
3820 (nds32_expand_unopimm_builtin): Ditto.
3821 (nds32_expand_binop_builtin): Ditto.
3822 (nds32_builtin_decl_impl): Ditto.
3823 (builtin_description): Ditto.
3824 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3825 (nds32_init_builtins_impl): Ditto.
3826 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3827 (nds32_builtin_decl): New.
3828 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3829 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3831 2018-03-02 Jeff Law <law@redhat.com>
3833 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3834 (redundant_insn, fill_simple_delay_slots): Likewise.
3835 (fill_slots_from_thread): Likewise.
3836 * resource.c (mark_referenced_resources): Likewise.
3837 (mark_set_resources, find_dead_or_set_registers): Likewise.
3839 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3841 * substring-locations.h (format_warning_va): Formatting fix for
3843 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3845 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3846 * substring-locations.c: Include intl.h.
3847 (format_warning_va): Turned into small wrapper around
3848 format_warning_n_va, renamed to ...
3849 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3850 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3852 (format_warning_at_substring_n): New function.
3853 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3854 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3855 format_warning_at_substring with just a shorter name instead of
3856 const function pointer.
3857 (fmtwarn_n): New function.
3858 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3859 appropriate, get rid of all the fmtstr temporaries, move conditionals
3860 with G_() wrapped string literals directly into fmtwarn arguments,
3861 cast dir.len to (int), formatting fixes.
3863 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3865 * doc/invoke.texi: Remove "Cilk Plus" references.
3867 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3868 Richard Biener <rguenther@suse.de>
3871 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3872 for error or warning attributes if CALL_FROM_THUNK_P is set.
3875 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3878 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3879 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3882 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3883 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3885 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3886 instead of -1U in last predictors element's probability member.
3888 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3891 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3892 arguments if they are comparable.
3894 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3896 PR tree-optimization/84634
3897 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3898 masks and masked_loop_p with a single loop_masks, making sure it's
3899 null for bb vectorization.
3901 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3903 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3904 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3905 loop->force_vectorize to check whether there is no alias.
3907 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3910 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3912 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3914 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3915 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3916 instead of a loop around prev_real_insn.
3917 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3921 * config/i386/i386.c (ix86_print_operand): Use conditional
3922 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3925 2018-03-02 Richard Biener <rguenther@suse.de>
3927 PR tree-optimization/84427
3928 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3929 (bitmap_set_subtract_values): Rewrite to handle multiple
3932 (prune_clobbered_mems): Likewise.
3933 (phi_translate): Take edge instead of pred/phiblock.
3934 (phi_translate_1): Likewise.
3935 (phi_translate_set): Likewise. Insert all translated
3936 exprs for a value into the set, keeping possibly multiple
3937 expressions per value.
3938 (compute_antic_aux): Adjust for phi_translate changes.
3939 When intersecting union the expressions and prune those
3940 not in the final value set, keeping possibly multiple
3941 expressions per value. Do not use value-insertion
3942 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3943 all expressions. Add verification that the value-sets
3944 only shrink during iteration.
3945 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3946 (do_pre_regular_insertion): Likewise.
3947 (do_pre_partial_partial_insertion): Likewise.
3949 2018-03-02 Richard Biener <rguenther@suse.de>
3952 * config/darwin.c (saved_debug_info_level): New static global.
3953 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3954 (darwin_asm_lto_end): Restore debug info generation settings.
3956 2018-03-01 Martin Liska <mliska@suse.cz>
3959 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3962 2018-03-01 Richard Biener <rguenther@suse.de>
3965 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3966 type annotation with LTO when debug was disabled at compile-time.
3968 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3970 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3972 (mips_final_postscan_insn): Likewise.
3974 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3976 PR rtl-optimization/84528
3977 * alias.c (init_alias_target): Add commentary.
3978 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3979 a unique base value if the frame pointer is not eliminated
3980 to the stack pointer.
3982 2018-03-01 Tom de Vries <tom@codesourcery.com>
3984 PR rtl-optimization/83327
3985 * lra-int.h (hard_regs_spilled_into): Declare.
3986 * lra.c (hard_regs_spilled_into): Define.
3987 (init_reg_info): Init hard_regs_spilled_into.
3988 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3989 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3990 (process_bb_lives): Handle hard_regs_spilled_into.
3991 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3992 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3994 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3996 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3997 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3998 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3999 * config/rs6000/aix72.h: New file.
4001 2018-02-28 Jakub Jelinek <jakub@redhat.com>
4003 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
4004 instead of warning_at with conditional singular and plural messages
4008 * stor-layout.c (update_alignment_for_field): For
4009 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
4010 && !DECL_PACKED (field), do the alignment update, just use
4011 only desired_align instead of MAX (type_align, desired_align)
4013 (place_field): Don't do known_align < desired_align handling
4014 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
4015 is non-NULL, instead do it after rli->prev_field handling and
4016 only if not within a bitfield word. For DECL_PACKED (field)
4017 use type_align of BITS_PER_UNIT.
4019 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
4021 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
4022 superfluous parentheses and trailing spaces.
4024 2018-02-28 Richard Biener <rguenther@suse.de>
4026 PR tree-optimization/84584
4027 * graphite-scop-detection.c (scop_detection::add_scop): Discard
4028 SCoPs with fake exit edge.
4030 2018-02-28 Martin Liska <mliska@suse.cz>
4033 * timevar.c (timer::print): Fix format to properly print 100%
4036 2018-02-28 Richard Biener <rguenther@suse.de>
4039 * genmatch.c (capture_info::walk_match): Do not mark
4040 captured expressions without operands as expr_p given
4041 they act more like predicates and should be subject to
4042 "lost tail" side-effect preserving.
4044 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
4046 PR rtl-optimization/81611
4047 * auto-inc-dec.c (attempt_change): Move dead note from
4048 mem_insn if it's the next use of regno
4049 (find_address): Take address use of reg holding
4050 non-incremented value. Add parm to limit search to the named
4052 (merge_in_block): Attempt to use a mem insn that is the next
4053 use of the original regno.
4055 2018-02-27 Martin Sebor <msebor@redhat.com>
4058 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4059 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4061 2018-02-27 Martin Sebor <msebor@redhat.com>
4063 PR translation/84207
4064 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4065 n argument to unsigned HOST_WIDE_INT.
4066 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4067 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4068 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4069 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4071 2018-02-27 Richard Biener <rguenther@suse.de>
4073 PR tree-optimization/84512
4074 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4075 Do not use the estimate returned from record_stmt_cost for
4076 the scalar iteration cost but sum properly using add_stmt_cost.
4078 2018-02-27 Richard Biener <rguenther@suse.de>
4080 PR tree-optimization/84466
4081 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4082 Adjust last change to less strictly validate use operands.
4084 2018-02-27 Martin Liska <mliska@suse.cz>
4086 PR gcov-profile/84548
4087 * gcov.c (process_file): Allow partial overlap and consider it
4088 also as group functions.
4089 (output_lines): Properly calculate range of lines for a group.
4091 2018-02-27 Martin Liska <mliska@suse.cz>
4093 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4094 'ggc' suffixes. Change first column width.
4095 (timer::print): Fix formatting of the column.
4097 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4099 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4100 preserve inline entry blocks for the sake of debug inline
4101 entry point markers alone.
4102 (remove_unused_locals): Suggest in comments a better place to
4103 force the preservation of inline entry blocks that are
4104 otherwise unused, but do not preserve them.
4106 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4108 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4110 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4113 * config/i386/constraints.md (Bs): Replace
4114 ix86_indirect_branch_register with
4115 TARGET_INDIRECT_BRANCH_REGISTER.
4117 * config/i386/i386.md (indirect_jump): Likewise.
4118 (tablejump): Likewise.
4119 (*sibcall_memory): Likewise.
4120 (*sibcall_value_memory): Likewise.
4121 Peepholes of indirect call and jump via memory: Likewise.
4122 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4123 (*sibcall_value_GOT_32): Likewise.
4124 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4125 (GOT_memory_operand): Likewise.
4126 (call_insn_operand): Likewise.
4127 (sibcall_insn_operand): Likewise.
4128 (GOT32_symbol_operand): Likewise.
4129 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4131 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4133 PR rtl-optimization/83496
4134 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4135 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4136 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4137 redundant insn, if any.
4138 (relax_delay_slots): Likewise.
4139 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4141 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4143 PR tree-optimization/83965
4144 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4145 that grouped statements are part of a reduction chain. Return
4146 true if the statement is not marked as a reduction itself but
4148 (vect_recog_dot_prod_pattern): Don't check whether the statement
4149 is part of a group here.
4150 (vect_recog_sad_pattern): Likewise.
4151 (vect_recog_widen_sum_pattern): Likewise.
4153 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4156 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4158 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4161 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4163 (ix86_output_indirect_function_return): New prototype.
4164 (ix86_split_simple_return_pop_internal): Likewise.
4165 * config/i386/i386.c (indirect_return_via_cx): New.
4166 (indirect_return_via_cx_bnd): Likewise.
4167 (indirect_thunk_name): Handle return va CX_REG.
4168 (output_indirect_thunk_function): Create alias for
4169 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4170 (ix86_output_indirect_jmp): Remove the bool argument.
4171 (ix86_output_indirect_function_return): New function.
4172 (ix86_split_simple_return_pop_internal): Likewise.
4173 * config/i386/i386.md (*indirect_jump): Don't pass false
4174 to ix86_output_indirect_jmp.
4175 (*tablejump_1): Likewise.
4176 (simple_return_pop_internal): Change it to define_insn_and_split.
4177 Call ix86_split_simple_return_pop_internal to split it for
4179 (simple_return_indirect_internal): Call
4180 ix86_output_indirect_function_return instead of
4181 ix86_output_indirect_jmp.
4183 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4186 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4187 memset and value initialization afterwards.
4189 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4191 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4193 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4196 * common/config/aarch64/aarch64-common.c
4197 (aarch_option_optimization_table[]): Switch
4198 off fomit-frame-pointer
4200 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4201 Chung-Ju Wu <jasonwucj@gmail.com>
4203 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4205 (store_multiple): Ditto.
4207 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4209 * config.gcc: Add --with-cpu support for nds32 target.
4210 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4211 * config/nds32/nds32.opt: Add -mcpu= option.
4213 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4215 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4216 isel=yes): Warn for these deprecated options.
4218 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4220 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4221 ISA_2_5_MASKS_EMBEDDED.
4223 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4225 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4226 p->max as pointers rather than using iterative_hash_expr.
4228 2018-02-23 Carl Love <cel@us.ibm.com>
4230 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4231 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4233 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4234 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4235 P8V_BUILTIN_VEC_VUNSIGNED2.
4237 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4240 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4241 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4243 * lra-constraints.c (curr_insn_transform): Set up
4244 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4246 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4247 * lra-eliminations.c (spill_pseudos): Ditto.
4248 (process_insn_for_elimination): Ditto.
4249 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4250 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4251 LRA_NON_CLOBBERED_ALT.
4253 2018-02-22 Martin Sebor <msebor@redhat.com>
4255 PR tree-optimization/84480
4256 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4257 to maybe_diag_stxncpy_trunc. Call it.
4258 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4259 from gimple_fold_builtin_strcpy. Print inlining stack.
4260 (handle_builtin_stxncpy): Print inlining stack.
4261 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4263 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4266 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4267 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4268 and -fcheck-pointer-bounds are used together.
4269 (indirect_thunk_prefix): New enum.
4270 (indirect_thunk_need_prefix): New function.
4271 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4272 "_nt" instead of "_bnd" for NOTRACK prefix.
4273 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4274 (output_indirect_thunk_function): Likewise.
4276 (ix86_code_end): Update output_indirect_thunk_function calls.
4277 (ix86_output_indirect_branch_via_reg): Replace
4278 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4279 (ix86_output_indirect_branch_via_push): Likewise.
4280 (ix86_output_function_return): Likewise.
4281 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4282 incompatible with -fcf-protection=branch and
4283 -fcheck-pointer-bounds.
4285 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4288 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4289 Change gcc_assert call to output_operand_lossage.
4291 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4293 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4295 2018-02-22 DJ Delorie <dj@redhat.com>
4296 Sebastian Perta <sebastian.perta@renesas.com>
4297 Oleg Endo <olegendo@gcc.gnu.org>
4299 * config/rx/rx.c (rx_rtx_costs): New function.
4300 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4302 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4304 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4306 2018-02-22 Martin Liska <mliska@suse.cz>
4309 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4310 Add "native" as a possible value.
4312 2018-02-22 Martin Liska <mliska@suse.cz>
4315 * config/i386/i386.c (ix86_option_override_internal):
4316 Add "native" as a possible value for -march and -mtune.
4318 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4321 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4322 to all type variants.
4324 PR tree-optimization/84503
4325 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4326 width as info->bitpos + info->bitsize - start.
4327 (merged_store_group::merge_overlapping): Simplify width computation.
4328 (check_no_overlap): New function.
4329 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4330 start + width and last_order of the group, fail if check_no_overlap
4332 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4333 to group if check_no_overlap fails.
4335 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4337 * config/rs6000/altivec.md: Delete contraint arguments to
4338 define_expand, define_split, and define_peephole2, and in
4339 define_insn_and_split if always unused.
4340 * config/rs6000/darwin.md: Ditto.
4341 * config/rs6000/dfp.md: Ditto.
4342 * config/rs6000/rs6000.md: Ditto.
4343 * config/rs6000/sync.md: Ditto.
4344 * config/rs6000/vector.md: Ditto.
4345 * config/rs6000/vsx.md: Ditto.
4347 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4349 * config/rs6000/altivec.md: Write output control strings as braced
4350 blocks instead of double-quoted strings.
4351 * config/rs6000/darwin.md: Ditto.
4352 * config/rs6000/rs6000.md: Ditto.
4353 * config/rs6000/vector.md: Ditto.
4354 * config/rs6000/vsx.md: Ditto.
4356 2018-02-21 Jason Merrill <jason@redhat.com>
4358 PR c++/84314 - ICE with templates and fastcall attribute.
4359 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4361 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4363 * ipa-cp.c (determine_versionability): Fix comment typos.
4365 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4368 * ipa-cp.c (determine_versionability): Do not version functions caling
4371 2018-02-21 Martin Liska <mliska@suse.cz>
4374 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4375 Add "native" as a possible value.
4376 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4377 the macro when native cpu detection is available.
4379 2018-02-21 Martin Liska <mliska@suse.cz>
4382 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4383 Add "native" as a possible value.
4384 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4385 when native cpu detection is available.
4387 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4388 Martin Sebor <msebor@redhat.com>
4390 PR tree-optimization/84478
4391 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4393 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4394 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4395 support which is conservatively correct, for 2 only stay conservative
4396 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4397 argument to the 2 argument get_range_strlen, adjust 6 arg
4398 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4400 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4401 (gimple_fold_builtin_strlen): Pass true as last argument to
4404 2018-02-20 Martin Sebor <msebor@redhat.com>
4407 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4408 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4409 (builtin_memref::builtin_memref): Factor out parts into
4410 set_base_and_offset and call it.
4412 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4415 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4416 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4417 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4418 search at the associated MODE_INT.
4420 2018-02-20 Jeff Law <law@redhat.com>
4423 PR tree-optimization/81592
4425 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4426 for range data rather than using global data.
4427 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4428 range data rather than using global data.
4429 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4430 pass it to children as needed.
4431 (struct directive::fmtresult): Similarly.
4432 (struct directive::set_width): Similarly.
4433 (struct directive::set_precision): Similarly.
4434 (format_integer, format_directive, parse_directive): Similarly.
4435 (format_none): Accept unnamed vr_values parameter.
4436 (format_percent, format_floating, format_character): Similarly.
4437 (format_string, format_plain): Similarly.
4438 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4439 the EVRP range analyzer for range data rather than using global data.
4440 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4441 gimple-ssa-evrp-analyze.h
4442 (class sprintf_dom_walker): Add after_dom_children member function.
4443 Add evrp_range_analyzer member.
4444 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4445 range analyzer as needed.
4446 (sprintf_dom_walker::after_dom_children): New member function.
4447 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4449 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4450 (evrp_range_analyzer::pop_to_marker): Likewise.
4452 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4454 PR tree-optimization/84419
4455 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4456 with the required type if its current type is compatible but
4459 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4462 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4463 after vectorization.
4465 2018-02-20 Martin Liska <mliska@suse.cz>
4468 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4469 possible values if we don't have a hint.
4471 2018-02-20 Martin Liska <mliska@suse.cz>
4475 * final.c (shorten_branches): Build align_tab array with one
4477 * opts.c (finish_options): Add alignment option limit check.
4478 (MAX_CODE_ALIGN): Likewise.
4479 (MAX_CODE_ALIGN_VALUE): Likewise.
4480 * doc/invoke.texi: Document maximum allowed option value for
4481 all -falign-* options.
4483 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4486 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4487 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4488 * var-tracking.c (emit_note_insn_var_location): Remove all references
4489 to NOTE_INSN_CALL_ARG_LOCATION.
4490 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4491 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4492 Use copy_rtx_if_shared.
4493 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4494 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4495 (dwarf2out_var_location): Remove handling of
4496 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4498 * final.c (final_scan_insn): Remove all references to
4499 NOTE_INSN_CALL_ARG_LOCATION.
4500 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4501 before dumping final insns.
4502 * except.c (emit_note_eh_region_end): Remove all references to
4503 NOTE_INSN_CALL_ARG_LOCATION.
4504 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4505 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4506 * config/arc/arc.c (hwloop_optimize): Likewise.
4507 * config/arm/arm.c (create_fix_barrier): Likewise.
4508 * config/s390/s390.c (s390_chunkify_start): Likewise.
4509 * config/sh/sh.c (find_barrier): Likewise.
4510 * config/i386/i386.c (rest_of_insert_endbranch,
4511 ix86_seh_fixup_eh_fallthru): Likewise.
4512 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4513 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4514 * config/frv/frv.c (frv_function_prologue): Likewise.
4515 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4517 (note_outside_basic_block_p): Remove all references to
4518 NOTE_INSN_CALL_ARG_LOCATION.
4519 * gengtype.c (adjust_field_rtx_def): Likewise.
4520 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4522 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4523 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4526 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4529 PR tree-optimization/84452
4530 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4531 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4534 2018-02-19 Martin Liska <mliska@suse.cz>
4537 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4539 2018-02-19 Martin Liska <mliska@suse.cz>
4540 Richard Sandiford <richard.sandiford@linaro.org>
4542 PR tree-optimization/82491
4543 * gimple-fold.c (get_base_constructor): Make earlier bail out
4546 2018-02-19 Carl Love <cel@us.ibm.com>
4548 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4549 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4551 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4552 P8V_BUILTIN_VEC_NEG.
4554 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4556 * config/rl78/rl78.md (movdf): New define expand.
4558 2018-02-19 Martin Liska <mliska@suse.cz>
4561 * doc/invoke.texi: Fix typo.
4562 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4564 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4566 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4567 handle rs6000_single_float and rs6000_double_float specially for
4570 2018-02-16 Jeff Law <law@redhat.com>
4572 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4575 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4578 * ipa-inline.c (inline_small_functions): Fix a typo.
4580 2018-02-16 Nathan Sidwell <nathan@acm.org>
4582 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4584 2018-02-16 Carl Love <cel@us.ibm.com>
4586 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4587 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4588 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4589 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4590 expansion to P8V_BUILTIN_VEC_FLOAT2.
4592 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4594 PR rtl-optimization/70023
4595 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4596 src_regno into account.
4598 2018-02-16 Carl Love <cel@us.ibm.com>
4600 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4601 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4602 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4603 * config/rs6000/rs6000.c: Remove case statements for
4604 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4605 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4606 and P9V_BUILTIN_VEC_VINSERT4B.
4607 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4608 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4609 * config/rs6000/vsx.md:
4610 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4613 2018-02-16 Carl Love <cel@us.ibm.com>
4615 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4617 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4619 * config/rs6000/rs6000-c.c: Add the definitions for
4620 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4621 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4622 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4623 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4624 definition for insert4b and define insn *insert3b_internal.
4625 * doc/extend.texi: Add documentation for vec_extract4b.
4627 2018-02-16 Nathan Sidwell <nathan@acm.org>
4629 * doc/extend.texi (Backwards Compatibility): Mention friend
4630 injection. Note for-scope is deprecated.
4631 * doc/invoke.texi (-ffriend-injection): Deprecate.
4633 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4635 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4636 that moved to I2, also allow destinations that are a paradoxical
4637 subreg (instead of a normal reg).
4639 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4642 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4645 2018-02-16 Richard Biener <rguenther@suse.de>
4647 PR tree-optimization/84037
4648 PR tree-optimization/84016
4650 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4651 Adjust vec_construct for the fact we need additional higher latency
4652 128bit inserts for AVX256 and AVX512 vector builds.
4653 (ix86_add_stmt_cost): Scale vector construction cost for
4656 2018-02-16 Richard Biener <rguenther@suse.de>
4658 PR tree-optimization/84417
4659 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4660 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4661 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4663 2018-02-16 Martin Liska <mliska@suse.cz>
4666 * internal-fn.def (ASAN_CHECK): Set proper flags.
4667 (ASAN_MARK): Likewise.
4669 2018-02-16 Julia Koval <julia.koval@intel.com>
4671 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4672 from PTA_CANNONLAKE.
4674 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4677 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4678 Use ++iter rather than iter++ for std::list iterators.
4679 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4680 defer deleting them until all nodes in the forest are processed. Do
4681 free even leaf nodes. Change to_process into auto_vec.
4684 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4685 * vec.h (vec_default_construct): Use memset instead of placement new
4686 if BROKEN_VALUE_INITIALIZATION is defined.
4687 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4688 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4691 PR rtl-optimization/83723
4692 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4693 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4694 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4696 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4698 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4700 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4702 PR rtl-optimization/81443
4703 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4704 from inner REGs to paradoxical SUBREGs.
4706 2018-02-16 Richard Biener <rguenther@suse.de>
4708 PR tree-optimization/84399
4709 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4710 For operands we can analyze at their definition make sure we can
4711 analyze them at each use as well.
4713 2018-02-16 Richard Biener <rguenther@suse.de>
4715 PR tree-optimization/84190
4716 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4717 volatile accesses if the decl isn't volatile.
4719 2018-02-15 Jason Merrill <jason@redhat.com>
4721 PR c++/84314 - ICE with templates and fastcall attribute.
4722 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4723 TYPE_CANONICAL on an existing type.
4725 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4727 PR tree-optimization/84383
4728 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4729 dstoff nor call operand_equal_p if dstbase is NULL.
4731 PR tree-optimization/84334
4732 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4733 also a CONSTANT_CLASS_P, punt.
4735 2018-02-14 Jim Wilson <jimw@sifive.com>
4737 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4738 first SMALL_OPERAND check. New local min_second_step. Move assert
4739 to where locals are set. Add TARGET_RVC support.
4740 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4742 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4744 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4746 2018-02-14 Martin Sebor <msebor@redhat.com>
4748 PR tree-optimization/83698
4749 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4750 arrays constrain the offset range to their bounds.
4751 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4752 (builtin_access::overlap): Avoid setting the size of overlap if it's
4754 (maybe_diag_overlap): Also consider arrays when deciding what values
4755 of offsets to include in diagnostics.
4757 2018-02-14 Martin Sebor <msebor@redhat.com>
4760 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4761 that correspond to the kind of a declaration.
4763 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4766 * config/pa/pa.md: Load address of PIC label using the linkage table
4767 if the label is nonlocal.
4769 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4771 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4772 warning message if user requests -maltivec=be.
4773 * doc/invoke.texi: Document deprecation of -maltivec=be.
4775 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4778 * config/rs6000/rs6000-c.c: Update definitions for
4779 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4780 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4782 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4785 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4786 add _get_ssp intrinsics. Remove argument from
4787 __builtin_ia32_rdssp[d|q].
4788 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4789 * config/i386/i386-builtin.def: Remove argument from
4790 __builtin_ia32_rdssp[d|q].
4791 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4792 ix86_expand_special_args_builtin for _rdssp[d|q].
4793 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4794 Clear register before usage.
4795 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4796 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4798 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4800 PR tree-optimization/84357
4801 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4802 operand 1 of an ARRAY_REF too.
4804 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4807 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4808 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4810 (set_of_reg): New struct.
4811 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4812 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4813 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4815 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4816 Split into bitclr, bitset, bitinvert patterns if appropriate.
4817 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4818 use rx_fuse_in_memory_bitop.
4819 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4820 to named insn, correct maximum insn length.
4822 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4825 * machmode.def: Define a complex mode for PARTIAL_INT.
4826 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4828 * doc/rtl.texi: Document CSPImode.
4829 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4831 (msp430_hard_regno_nregs_with_padding): Likewise.
4833 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4836 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4838 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4840 PR rtl-optimization/84169
4841 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4842 we generated a parallel as new i3 and we split that to new i2 and i3
4843 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4844 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4845 those to i2, not i1. Partially rewrite this scan code.
4847 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4850 * stor-layout.c (place_field): For variable length fields, adjust
4851 offset_align afterwards not just based on the field's alignment,
4852 but also on the size.
4855 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4856 of exps and logs in the use_exp2 case.
4858 2018-02-13 Jeff Law <law@redhat.com>
4860 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4863 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4866 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4870 * common.opt (gas-loc-support, gas-locview-support): New.
4871 (ginline-points, ginternal-reset-location-views): New.
4872 * doc/invoke.texi: Document them. Use @itemx where intended.
4873 (gvariable-location-views): Adjust.
4874 * target.def (reset_location_view): New.
4875 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4876 (TARGET_RESET_LOCATION_VIEW): New.
4877 * doc/tm.texi: Rebuilt.
4878 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4879 (dwarf2out_default_as_locview_support): New.
4880 (output_asm_line_debug_info): Use option variables.
4881 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4882 (output_loc_list): Likewise.
4883 (add_high_low_attributes): Check option variables.
4884 Don't output entry view attribute in strict mode.
4885 (gen_inlined_subroutine_die): Check option variables.
4886 (dwarf2out_inline_entry): Likewise.
4887 (init_sections_and_labels): Likewise.
4888 (dwarf2out_early_finish): Likewise.
4889 (maybe_reset_location_view): New, from...
4890 (dwarf2out_var_location): ... here. Call it.
4891 * debug.h (dwarf2out_default_as_loc_support): Declare.
4892 (dwarf2out_default_as_locview_support): Declare.
4893 * hooks.c (hook_int_rtx_insn_0): New.
4894 * hooks.h (hook_int_rtx_insn_0): Declare.
4895 * toplev.c (process_options): Take -gas-loc-support and
4896 -gas-locview-support from dwarf2out. Enable
4897 -gvariable-location-views by default only with locview
4898 assembler support. Enable -ginternal-reset-location-views by
4899 default only if the target defines the corresponding hook.
4900 Enable -ginline-points by default if location views are
4901 enabled; force it disabled if statement frontiers are
4903 * tree-inline.c (expand_call_inline): Check option variables.
4904 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4906 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4908 PR tree-optimization/84321
4909 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4910 handling. Also check whether the anti-range contains any values
4911 that satisfy the mask; switch to a VR_RANGE if not.
4913 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4916 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4918 2018-02-13 Martin Jambor <mjambor@suse.cz>
4921 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4922 of call statements, also set location of a load to a temporary.
4924 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4926 * config/rl78/rl78.c (add_vector_labels): New function.
4927 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4928 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4929 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4930 which checks that no arguments are passed.
4931 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4932 * doc/extend.texi: Documentation for the new attribute.
4934 2018-02-13 Andreas Schwab <schwab@suse.de>
4936 * config/riscv/linux.h (CPP_SPEC): Define.
4938 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4941 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4942 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4943 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4944 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4945 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4946 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4947 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4948 temporarily for AES and PCLMUL builtins.
4950 PR tree-optimization/84339
4951 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4952 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4956 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4957 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4958 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4960 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4962 * omp-simd-clone.h: New file.
4963 * omp-simd-clone.c: Include omp-simd-clone.h.
4964 (expand_simd_clones): No longer static.
4965 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4966 cgraph.h and omp-simd-clone.h.
4967 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4968 (vect_recog_widen_shift_pattern): Formatting fix.
4969 (vect_pattern_recog_1): Don't check optab for calls.
4972 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4973 operands[2] into a REG before using gen_lowpart on it.
4975 2018-02-12 Jeff Law <law@redhat.com>
4978 * config/sh/sh.c (find_barrier): Consider a sibling call
4981 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4982 successfully back substituting a reg.
4984 2018-02-12 Richard Biener <rguenther@suse.de>
4986 PR tree-optimization/84037
4987 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4988 parameter, move visited init to caller.
4989 (vect_slp_analyze_operations): Separate cost from validity
4990 check, initialize visited once for all instances.
4991 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4993 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4995 (vect_model_promotion_demotion_cost): Likewise.
4996 (vectorizable_bswap): Guard cost modeling with !slp_node
4997 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4999 (vectorizable_call): Likewise.
5000 (vectorizable_conversion): Likewise.
5001 (vectorizable_assignment): Likewise.
5002 (vectorizable_shift): Likewise.
5003 (vectorizable_operation): Likewise.
5004 (vectorizable_store): Likewise.
5005 (vectorizable_load): Likewise.
5006 (vectorizable_condition): Likewise.
5007 (vectorizable_comparison): Likewise.
5009 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
5012 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
5013 (ASAN_MARK): Fix fnspec to account for return value, change pointer
5014 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
5016 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5019 * params.def (inline-min-speedup): Increase from 8 to 15.
5020 (max-inline-insns-auto): Decrease from 40 to 30.
5021 * ipa-split.c (consider_split): Add some buffer for function to
5022 be considered inlining candidate.
5023 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
5026 2018-02-12 Richard Biener <rguenther@suse.de>
5028 PR tree-optimization/84037
5029 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
5030 matched stmts if we cannot swap the non-matched ones.
5032 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5034 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
5035 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
5036 _mm_maskz_scalef_round_ss): New intrinsics.
5037 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
5038 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
5039 __builtin_ia32_scalefss_round): Remove.
5040 (__builtin_ia32_scalefsd_mask_round,
5041 __builtin_ia32_scalefss_mask_round): New intrinsics.
5042 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
5043 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
5044 ((match_operand:VF_128 2 "<round_nimm_predicate>"
5045 "<round_constraint>")): Changed to ...
5046 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
5047 "<round_scalar_constraint>")): ... this.
5048 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
5049 %0, %1, %2<round_op3>}"): Changed to ...
5050 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
5051 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
5052 %2<round_scalar_mask_op3>}"): ... this.
5053 * config/i386/subst.md (round_scalar_nimm_predicate): New.
5055 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
5057 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
5058 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5059 (_mm_maskz_sqrt_round_ss): New intrinsics.
5060 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5061 (__builtin_ia32_sqrtsd_mask_round)
5062 (__builtin_ia32_sqrtss_mask_round): New builtins.
5063 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5064 (__builtin_ia32_sqrtss_round): Remove.
5065 (__builtin_ia32_sqrtsd_mask_round)
5066 (__builtin_ia32_sqrtss_mask_round): New builtins.
5067 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5068 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5069 ((match_operand:VF_128 1 "vector_operand"
5070 "xBm,<round_constraint>")): Changed to ...
5071 ((match_operand:VF_128 1 "vector_operand"
5072 "xBm,<round_scalar_constraint>")): ... this.
5073 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5074 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5075 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5076 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5077 %<iptr>1<round_scalar_mask_op3>}): ... this.
5078 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5079 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5081 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5084 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5085 Cast vec_cmpeq result to correct type.
5086 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5087 Cast vec_cmpgt result to correct type.
5089 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5091 * final.c (final_scan_insn_1): Renamed from...
5092 (final_scan_insn): ... this. New wrapper, to recover
5093 seen from the outermost call in recursive ones.
5094 * config/sparc/sparc.c (output_return): Drop seen from call.
5095 (output_sibcall): Likewise.
5096 * config/visium/visium.c (output_branch): Likewise.
5098 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5100 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5103 2018-02-10 Alan Modra <amodra@gmail.com>
5106 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5107 Specify LR as an input.
5109 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5112 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5113 remove_member_access_dummy_vars): New functions.
5114 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5115 lower_omp_1, execute_lower_omp): Use them.
5117 PR rtl-optimization/84308
5118 * shrink-wrap.c (spread_components): Release todo vector.
5120 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5122 PR rtl-optimization/57193
5123 * ira-color.c (struct allocno_color_data): Add member
5124 conflict_allocno_hard_prefs.
5125 (update_conflict_allocno_hard_prefs): New.
5126 (bucket_allocno_compare_func): Add a preference based on
5127 conflict_allocno_hard_prefs.
5128 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5129 (color_allocnos): Remove a dead code. Initiate
5130 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5132 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5135 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5136 constraint from =wa to wa. Avoid a subreg on the output operand,
5137 instead use a pseudo and subreg it in a move.
5138 (p9_xxbrd_<mode>): Changed to ...
5139 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5140 (p9_xxbrd_v2df): New expander.
5141 (p9_xxbrw_<mode>): Changed to ...
5142 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5143 (p9_xxbrw_v4sf): New expander.
5145 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5147 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5149 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5152 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5153 multiply in 32-bit mode.
5154 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5155 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5158 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5160 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5161 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5162 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5163 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5165 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5168 * dwarf2out.c (is_trivial_indirect_ref): New function.
5169 (dwarf2out_late_global_decl): Do not generate a location
5170 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5171 and that are not defined in the current unit.
5173 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5175 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5176 instead of a libcall for UNORDERED.
5178 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5181 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5182 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5184 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5187 * config/s390/s390.c (s390_set_current_function): Invoke
5188 s390_indirect_branch_settings also if fndecl didn't change.
5190 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5192 * config/rs6000/rs6000.md (blockage): Set length to zero.
5194 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5196 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5198 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5201 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5202 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5206 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5207 PARALLEL incoming that failed vt_get_decl_and_offset check.
5210 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5211 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5213 (get_variable_section): For decls in named .bss* sections pass true as
5214 second argument to bss_initializer_p.
5216 2018-02-09 Marek Polacek <polacek@redhat.com>
5217 Jakub Jelinek <jakub@redhat.com>
5220 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5221 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5222 Sync some changes from cxx_fold_indirect_ref.
5224 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5226 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5228 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5229 (BLOCK_INLINE_ENTRY_LABEL): New.
5230 (dwarf2out_var_location): Disregard inline entry markers.
5231 (inline_entry_data): New struct.
5232 (inline_entry_data_hasher): New hashtable type.
5233 (inline_entry_data_hasher::hash): New.
5234 (inline_entry_data_hasher::equal): New.
5235 (inline_entry_data_table): New variable.
5236 (add_high_low_attributes): Add DW_AT_entry_pc and
5237 DW_AT_GNU_entry_view attributes if a pending entry is found
5238 in inline_entry_data_table. Add old entry_pc attribute only
5239 if debug nonbinding markers are disabled.
5240 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5241 markers are enabled.
5242 (block_within_block_p, dwarf2out_inline_entry): New.
5243 (dwarf2out_finish): Check that no entries remained in
5244 inline_entry_data_table.
5245 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5246 (final_scan_insn, notice_source_line): Likewise.
5247 (rest_of_clean_state): Skip inline entry markers.
5248 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5250 * gimple.c (gimple_build_debug_inline_entry): New.
5251 * gimple.h (enum gimple_debug_subcode): Add
5252 GIMPLE_DEBUG_INLINE_ENTRY.
5253 (gimple_build_debug_inline_entry): Declare.
5254 (gimple_debug_inline_entry_p): New.
5255 (gimple_debug_nonbind_marker_p): Adjust.
5256 * insn-notes.def (INLINE_ENTRY): New.
5257 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5258 inline entry marker notes.
5259 (print_insn): Likewise.
5260 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5261 (INSN_DEBUG_MARKER_KIND): Likewise.
5262 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5263 * tree-inline.c (expand_call_inline): Build and insert
5264 debug_inline_entry stmt.
5265 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5266 inline entry blocks early, if nonbind markers are enabled.
5267 (dump_scope_block): Dump fragment info.
5268 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5269 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5270 (gimple_build_debug_inline_entry): New.
5271 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5272 Enable/disable inline entry points too.
5273 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5274 (DEBUG_INSN): Describe inline entry markers.
5276 * common.opt (gvariable-location-views): New.
5277 (gvariable-location-views=incompat5): New.
5278 * config.in: Rebuilt.
5279 * configure: Rebuilt.
5280 * configure.ac: Test assembler for view support.
5281 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5282 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5283 * dwarf2out.c (var_loc_view): New typedef.
5284 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5285 (dwarf2out_locviews_in_attribute): New.
5286 (dwarf2out_locviews_in_loclist): New.
5287 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5288 (enum dw_line_info_opcode): Add LI_adv_address.
5289 (struct dw_line_info_table): Add view.
5290 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5291 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5292 (zero_view_p): New variable.
5293 (ZERO_VIEW_P): New macro.
5294 (output_asm_line_debug_info): New.
5295 (struct var_loc_node): Add view.
5296 (add_AT_view_list, AT_loc_list): New.
5297 (add_var_loc_to_decl): Add view param. Test it against last.
5298 (new_loc_list): Add view params. Record them.
5299 (AT_loc_list_ptr): Handle loc and view lists.
5300 (view_list_to_loc_list_val_node): New.
5301 (print_dw_val): Handle dw_val_class_view_list.
5302 (size_of_die): Likewise.
5303 (value_format): Likewise.
5304 (loc_list_has_views): New.
5305 (gen_llsym): Set vl_symbol too.
5306 (maybe_gen_llsym, skip_loc_list_entry): New.
5307 (dwarf2out_maybe_output_loclist_view_pair): New.
5308 (output_loc_list): Output view list or entries too.
5309 (output_view_list_offset): New.
5310 (output_die): Handle dw_val_class_view_list.
5311 (output_dwarf_version): New.
5312 (output_compilation_unit_header): Use it.
5313 (output_skeleton_debug_sections): Likewise.
5314 (output_rnglists, output_line_info): Likewise.
5315 (output_pubnames, output_aranges): Update version comments.
5316 (output_one_line_info_table): Output view numbers in asm comments.
5317 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5318 Call maybe_gen_llsym.
5319 (loc_list_from_tree_1): Adjust.
5320 (add_AT_location_description): Create view list attribute if
5321 needed, check it's absent otherwise.
5322 (convert_cfa_to_fb_loc_list): Adjust.
5323 (maybe_emit_file): Call output_asm_line_debug_info for test.
5324 (dwarf2out_var_location): Reset views as needed. Precompute
5325 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5326 attribute. Set view.
5327 (new_line_info_table): Reset next view.
5328 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5329 (dwarf2out_source_line): Likewise. Output view resets and labels to
5330 the assembler, or select appropriate line info opcodes.
5331 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5332 (optimize_string_length): Catch it. Adjust.
5333 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5334 dw_val_class_view_list, and remove it if no longer needed.
5335 (hash_loc_list): Hash view numbers.
5336 (loc_list_hasher::equal): Compare them.
5337 (optimize_location_lists): Check whether a view list symbol is
5338 needed, and whether the locview attribute is present, and
5339 whether they match. Remove the locview attribute if no longer
5341 (index_location_lists): Call skip_loc_list_entry for test.
5342 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5343 Use output_dwarf_version.
5344 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5345 (struct dw_val_node): Add val_view_list.
5346 * final.c (SEEN_NEXT_VIEW): New.
5347 (set_next_view_needed): New.
5348 (clear_next_view_needed): New.
5349 (maybe_output_next_view): New.
5350 (final_start_function): Rename to...
5351 (final_start_function_1): ... this. Take pointer to FIRST,
5352 add SEEN parameter. Emit param bindings in the initial view.
5353 (final_start_function): Reintroduce SEEN-less interface.
5354 (final): Rename to...
5355 (final_1): ... this. Take SEEN parameter. Output final pending
5356 next view at the end.
5357 (final): Reintroduce seen-less interface.
5358 (final_scan_insn): Output pending next view before switching
5359 sections or ending a block. Mark the next view as needed when
5360 outputting variable locations. Notify debug backend of section
5361 changes, and of location view changes.
5362 (rest_of_handle_final): Adjust.
5363 * toplev.c (process_options): Autodetect value for debug variable
5364 location views option. Warn on incompat5 without -gdwarf-5.
5365 * doc/invoke.texi (gvariable-location-views): New.
5366 (gvariable-location-views=incompat5): New.
5367 (gno-variable-location-views): New.
5369 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5371 PR tree-optimization/84136
5372 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5373 that the result of find_edge is non-NULL.
5375 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5378 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5379 storing integer register in SImode. Fix cost of 256 and 512
5380 byte aligned SSE register store.
5382 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5384 * config/i386/i386.c (ix86_multiplication_cost): Fix
5385 multiplication cost for TARGET_AVX512DQ.
5387 2018-02-08 Marek Polacek <polacek@redhat.com>
5389 PR tree-optimization/84238
5390 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5393 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5395 PR tree-optimization/84265
5396 * tree-vect-stmts.c (vectorizable_store): Don't treat
5397 VMAT_CONTIGUOUS accesses as grouped.
5398 (vectorizable_load): Likewise.
5400 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5402 PR tree-optimization/81635
5403 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5404 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5405 (test_round_for_mask): New functions.
5406 (wide_int_cc_tests): Call test_round_for_mask.
5407 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5408 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5409 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5410 range returned by get_range_info.
5412 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5415 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5416 * symtab.c: Include builtins.h
5417 (symtab_node::output_to_lto_symbol_table_p): Move here
5418 from lto-streamer-out.c:output_symbol_p.
5419 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5420 (output_symbol_p): Move all logic to symtab.c
5421 (produce_symtab): Update.
5423 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5425 * config/s390/s390-opts.h (enum indirect_branch): Define.
5426 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5427 (s390_indirect_branch_via_thunk)
5428 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5429 (enum s390_indirect_branch_type): Define.
5430 * config/s390/s390.c (struct s390_frame_layout, struct
5431 machine_function): Remove.
5432 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5433 (indirect_branch_table_label_no, indirect_branch_table_name):
5435 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5436 (enum s390_indirect_branch_option): Define.
5437 (s390_return_addr_from_memory): New function.
5438 (s390_handle_string_attribute): New function.
5439 (s390_attribute_table): Add new attribute handler.
5440 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5441 (s390_indirect_branch_via_thunk): New function.
5442 (s390_indirect_branch_via_inline_thunk): New function.
5443 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5444 sibling call optimization for non z10 compiles.
5445 (s390_emit_call): Force indirect branch target to be a single
5446 register. Add r1 clobber for non-z10 compiles.
5447 (s390_emit_epilogue): Emit return jump via return_use expander.
5448 (s390_reorg): Handle JUMP_INSNs as execute targets.
5449 (s390_option_override_internal): Perform validity checks for the
5450 new command line options.
5451 (s390_indirect_branch_attrvalue): New function.
5452 (s390_indirect_branch_settings): New function.
5453 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5454 (s390_output_indirect_thunk_function): New function.
5455 (s390_code_end): Implement target hook.
5456 (s390_case_values_threshold): Implement target hook.
5457 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5459 * config/s390/s390.h (struct s390_frame_layout)
5460 (struct machine_function): Move here from s390.c.
5461 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5462 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5463 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5464 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5465 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5466 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5467 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5468 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5469 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5470 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5471 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5472 (mnemonic attribute): Add values which aren't recognized
5474 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5475 pattern for branch conversion. Fix mnemonic attribute.
5476 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5477 indirect branch via thunk if requested.
5478 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5479 ("*indirect_jump"): Disable for branch conversion using out of
5481 ("indirect_jump_via_thunk<mode>_z10")
5482 ("indirect_jump_via_thunk<mode>")
5483 ("indirect_jump_via_inlinethunk<mode>_z10")
5484 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5485 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5486 ("casesi_jump_via_inlinethunk<mode>_z10")
5487 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5488 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5489 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5490 ("*indirect2_jump"): Disable for branch conversion.
5491 ("casesi_jump"): Turn into expander and expand patterns for branch
5493 ("return_use"): New expander.
5494 ("*return"): Emit return via thunk and rename it to ...
5495 ("*return<mode>"): ... this one.
5496 * config/s390/s390.opt: Add new options and and enum for the
5499 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5501 * lra-constraints.c (match_reload): Unconditionally use
5502 gen_lowpart_SUBREG, rather than selecting between that
5503 and equivalent gen_rtx_SUBREG code.
5505 2018-02-08 Richard Biener <rguenther@suse.de>
5507 PR tree-optimization/84233
5508 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5509 changed flag instead of boguously re-using phi_inserted.
5511 2018-02-08 Martin Jambor <mjambor@suse.cz>
5513 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5514 static local variables.
5516 2018-02-08 Richard Biener <rguenther@suse.de>
5518 PR tree-optimization/84278
5519 * tree-vect-stmts.c (vectorizable_store): When looking for
5520 smaller vector types to perform grouped strided loads/stores
5521 make sure the mode is supported by the target.
5522 (vectorizable_load): Likewise.
5524 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5526 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5527 Increase LDP/STP opportunities by adding adjacent callee-saves.
5529 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5531 PR rtl-optimization/84068
5532 PR rtl-optimization/83459
5533 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5535 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5537 PR tree-optimization/84224
5538 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5539 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5542 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5545 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5546 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5547 count, remove one USE.
5549 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5551 * doc/install.texi (Configuration): Document the
5552 --with-long-double-format={ibm,ieee} PowerPC configuration
5556 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5557 Convert from define_expand to be define_insn_and_split. Rework
5558 float/double/_Float128 conversions to QI/HI/SImode to work with
5559 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5560 conversions to QI/HImode types did a store and then a load to
5561 truncate the value. For conversions to VSX registers, don't split
5562 the insn, instead emit the code directly. Use the code iterator
5563 any_fix to combine signed and unsigned conversions.
5564 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5565 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5566 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5567 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5568 (fix_<mode>di2_hw): Likewise.
5569 (fixuns_<mode>di2_hw): Likewise.
5570 (fix_<mode>si2_hw): Likewise.
5571 (fixuns_<mode>si2_hw): Likewise.
5572 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5573 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5574 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5575 fix<uns>_trunc<SFDF:mode>si2_p8.
5576 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5578 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5579 (fix<uns>_<mode>_mem): Likewise.
5580 (fctiw<u>z_<mode>_mem): Likewise.
5581 (fix<uns>_<mode>_mem): Likewise.
5582 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5583 the register allocator from doing a direct move to the GPRs to do
5584 a store, and instead use the ISA 3.0 store byte/half-word from
5585 vector register instruction. For IEEE 128-bit floating point,
5586 also optimize stores of 32-bit ints.
5587 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5589 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5591 * genextract.c (push_pathstr_operand): New function to support
5593 (walk_rtx): Call push_pathstr_operand.
5594 (print_path): Support [a-zA-Z].
5596 2018-02-07 Richard Biener <rguenther@suse.de>
5598 PR tree-optimization/84037
5599 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5600 (cse_and_gimplify_to_preheader): Declare.
5601 (vect_get_place_in_interleaving_chain): Likewise.
5602 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5604 (_loop_vec_info::~_loop_vec_info): Delete it.
5605 (cse_and_gimplify_to_preheader): New function.
5606 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5607 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5608 (vectorizable_load): Likewise. For grouped stores always base
5609 the IV on the first element.
5610 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5611 condition before gimplifying.
5613 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5615 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5616 *DIV_EXPR and *MOD_EXPR.
5618 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5621 * config/i386/i386.c (ix86_option_override_internal): Mask out
5622 the CF_SET bit when checking -fcf-protection.
5624 2018-02-07 Tom de Vries <tom@codesourcery.com>
5627 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5630 2018-02-07 Richard Biener <rguenther@suse.de>
5632 PR tree-optimization/84204
5633 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5636 PR tree-optimization/84205
5637 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5638 special-case isl_ast_op_zdiv_r.
5640 PR tree-optimization/84223
5641 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5642 Only add conditions from within the region.
5643 (gather_bbs::after_dom_children): Adjust.
5645 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5648 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5649 * config/avr/avr.md: Only post-reload split REG-REG moves if
5650 either register is GENERAL_REG_P.
5652 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5654 PR tree-optimization/84235
5655 * tree-ssa-scopedtables.c
5656 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5657 if the subtraction is performed in floating point type where NaNs are
5658 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5659 build 1. Formatting fix.
5661 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5664 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5665 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5666 and skip it regardless of bb boundaries. Use CALL_P macro,
5667 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5668 unnecessarily, formatting fix.
5670 2018-02-06 Michael Collison <michael.collison@arm.com>
5672 * config/arm/thumb2.md:
5673 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5674 (*thumb_mov_notscc): Ditto.
5676 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5679 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5680 unsigned_fix, not "s".
5682 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5684 * configure.ac (gcc_fn_eh_frame_ro): New function.
5685 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5686 correct .eh_frame permissions.
5687 * configure: Regenerate.
5689 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5691 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5694 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5696 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5697 Display warning message for -mno-speculate-indirect-jumps.
5699 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5701 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5703 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5705 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5707 PR tree-optimization/84225
5708 * tree-eh.c (find_trapping_overflow): Only call
5709 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5711 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5714 * config/i386/i386.c: Reimplement the check of possible options
5715 -mibt/-mshstk conbination. Change error messages.
5716 * doc/invoke.texi: Fix a typo: remove extra '='.
5718 2018-02-06 Marek Polacek <polacek@redhat.com>
5720 PR tree-optimization/84228
5721 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5723 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5726 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5727 emitted arch directives.
5728 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5729 __ARM_FEATURE_COPROC before changing architectures.
5731 2018-02-06 Richard Biener <rguenther@suse.de>
5733 * config/i386/i386.c (print_reg): Fix typo.
5734 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5736 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5738 * configure: Regenerate.
5740 2018-02-05 Martin Sebor <msebor@redhat.com>
5742 PR tree-optimization/83369
5743 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5746 2018-02-05 Martin Liska <mliska@suse.cz>
5748 * doc/invoke.texi: Cherry-pick upstream r323995.
5750 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5752 * ira.c (ira_init_register_move_cost): Adjust comment.
5754 2018-02-05 Martin Liska <mliska@suse.cz>
5756 PR gcov-profile/84137
5757 * doc/gcov.texi: Fix typo in documentation.
5759 2018-02-05 Martin Liska <mliska@suse.cz>
5761 PR gcov-profile/83879
5762 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5763 using dlopen functionality.
5765 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5767 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5768 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5769 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5770 _mm_maskz_range_round_ss): New intrinsics.
5771 (__builtin_ia32_rangesd128_round)
5772 (__builtin_ia32_rangess128_round): Remove.
5773 (__builtin_ia32_rangesd128_mask_round,
5774 __builtin_ia32_rangess128_mask_round): New builtins.
5775 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5776 __builtin_ia32_rangess128_round): Remove.
5777 (__builtin_ia32_rangesd128_mask_round,
5778 __builtin_ia32_rangess128_mask_round): New builtins.
5779 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5780 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5781 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5782 "<round_saeonly_constraint>")): Changed to ...
5783 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5784 "<round_saeonly_scalar_constraint>")): ... this.
5785 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5786 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5787 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5788 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5789 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5791 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5793 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5795 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5796 Remove all values except native, 8540 and 8548.
5798 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5800 * config/i386/i386.c (ix86_output_function_return): Pass
5801 INVALID_REGNUM, instead of -1, as invalid register number to
5802 indirect_thunk_name and output_indirect_thunk.
5804 2018-02-02 Julia Koval <julia.koval@intel.com>
5806 * config.gcc: Add -march=icelake.
5807 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5808 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5809 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5810 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5811 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5812 (processor_target_table): Add icelake.
5813 (ix86_option_override_internal): Handle new PTAs.
5814 (get_builtin_code_for_version): Handle icelake.
5815 (M_INTEL_COREI7_ICELAKE): New.
5816 (fold_builtin_cpu): Handle icelake.
5817 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5818 * doc/invoke.texi: Add -march=icelake.
5820 2018-02-02 Julia Koval <julia.koval@intel.com>
5822 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5823 to wide_int_bitmask.
5824 * wide-int-bitmask.h: New.
5826 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5829 * config/i386/i386.md: Replace Pmode with word_mode in
5830 builtin_setjmp_setup and builtin_longjmp to support x32.
5832 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5836 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5838 (rs6000_supported_cpu_names): New static variable.
5839 (linux_cpu_translation_table): Likewise.
5840 (elf_platform) <cpu>: Define new static variable and use it.
5841 Translate kernel AT_PLATFORM name to canonical name if needed.
5842 Error if platform name is unknown.
5844 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5847 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5849 2018-02-01 Jeff Law <law@redhat.com>
5852 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5853 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5854 the scratch if RELEASE_VIA_POP is false.
5855 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5856 If we have to save a temporary register, decrement SIZE appropriately.
5857 Pass new arguments to release_scratch_register_on_entry.
5858 (ix86_adjust_stack_and_probe): Likewise.
5859 (ix86_emit_probe_stack_range): Pass new arguments to
5860 release_scratch_register_on_entry.
5862 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5864 PR rtl-optimization/84157
5865 * combine.c (change_zero_ext): Use REG_P predicate in
5866 front of HARD_REGISTER_P predicate.
5868 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5870 * config/avr/avr.c (avr_option_override): Move disabling of
5871 -fdelete-null-pointer-checks to...
5872 * common/config/avr/avr-common.c (avr_option_optimization_table):
5875 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5877 PR tree-optimization/81635
5878 * tree-data-ref.c (split_constant_offset_1): For types that
5879 wrap on overflow, try to use range info to prove that wrapping
5882 2018-02-01 Renlin Li <renlin.li@arm.com>
5885 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5887 (aarch64_register_move_cost): Likewise.
5888 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5890 (REG_CLASS_NAMES): Likewise.
5891 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5892 TAILCALL_ADDR_REGS. Remove IP registers.
5893 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5895 2018-02-01 Richard Biener <rguenther@suse.de>
5897 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5898 for specifying RPO order and allow NULL for that.
5899 * domwalk.c (dom_walker::dom_walker): Likewise.
5900 (dom_walker::walk): Handle NULL RPO order.
5901 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5903 (rewrite_update_dom_walker): Likewise.
5904 (mark_def_dom_walker): Likewise.
5906 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5908 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5909 (aarch64_maybe_expand_sve_subreg_move): Declare.
5910 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5911 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5913 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5914 that are semantically a reverse operation.
5915 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5916 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5917 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5919 (aarch64_can_change_mode_class): For big-endian, forbid changes
5920 between two SVE modes if they have different element sizes.
5922 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5924 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5925 the TImode handling for big-endian targets.
5927 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5929 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5930 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5932 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5933 Remove BSWAP handing for big-endian targets and use the form of
5934 LD1RQ appropariate for the mode.
5936 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5938 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5939 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5942 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5945 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5946 check for operands that need to go through aarch64_sve_reload_be.
5948 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5950 PR tree-optimization/81661
5951 PR tree-optimization/84117
5952 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5953 * tree-eh.c: Include gimplify.h.
5954 (find_trapping_overflow, replace_trapping_overflow,
5955 rewrite_to_non_trapping_overflow): New functions.
5956 * tree-vect-loop.c: Include tree-eh.h.
5957 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5958 * tree-data-ref.c: Include tree-eh.h.
5959 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5961 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5963 PR rtl-optimization/84123
5964 * combine.c (change_zero_ext): Check if hard register satisfies
5965 can_change_dest_mode before calling gen_lowpart_SUBREG.
5967 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5970 * ira.c (ira_init_register_move_cost): Remove assert.
5972 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5974 PR rtl-optimization/84071
5975 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5976 * doc/tm.texi: Regenerate.
5978 2018-01-31 Richard Biener <rguenther@suse.de>
5980 PR tree-optimization/84132
5981 * tree-data-ref.c (analyze_miv_subscript): Properly
5982 check whether evolution_function_is_affine_multivariate_p
5983 before calling gcd_of_steps_may_divide_p.
5985 2018-01-31 Julia Koval <julia.koval@intel.com>
5988 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5989 * config/i386/i386.md (rdpid_rex64) New.
5990 (rdpid): Make 32bit only.
5992 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5995 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5996 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5998 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6001 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
6003 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
6005 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
6007 PR rtl-optimization/84071
6008 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
6009 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
6011 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6013 * config/arc/arc.c (arc_handle_aux_attribute): New function.
6014 (arc_attribute_table): Add 'aux' attribute.
6015 (arc_in_small_data_p): Consider aux like variables.
6016 (arc_is_aux_reg_p): New function.
6017 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
6018 (arc_get_aux_arg): New function.
6019 (prepare_move_operands): Handle aux-register access.
6020 (arc_handle_aux_attribute): New function.
6021 * doc/extend.texi (ARC Variable attributes): Add subsection.
6023 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
6025 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
6026 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
6027 (arc_attribute_table): Add 'uncached' attribute.
6028 (arc_print_operand): Print '.di' flag for uncached memory
6030 (arc_in_small_data_p): Do not consider for small data the uncached
6032 (arc_is_uncached_mem_p): New function.
6033 * config/arc/predicates.md (compact_store_memory_operand): Check
6034 for uncached memory accesses.
6035 (nonvol_nonimm_operand): Likewise.
6036 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
6038 2018-01-31 Jakub Jelinek <jakub@redhat.com>
6041 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
6042 falign-loops=): Add Optimization flag.
6044 2018-01-30 Jeff Law <law@redhat.com>
6047 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
6048 INT_REGISTERS_SAVED. Check it prior to calling
6049 get_scratch_register_on_entry.
6050 (ix86_adjust_stack_and_probe): Similarly.
6051 (ix86_emit_probe_stack_range): Similarly.
6052 (ix86_expand_prologue): Corresponding changes.
6054 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6057 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
6058 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6060 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6063 * lra-constraints.c (curr_insn_transform): Process AND in the
6066 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6068 PR rtl-optimization/83986
6069 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6070 dependence against last_pending_memory_flush in addition to
6073 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6075 PR tree-optimization/81611
6076 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6079 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6082 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6085 2018-01-30 Richard Biener <rguenther@suse.de>
6086 Jakub Jelinek <jakub@redhat.com>
6088 PR tree-optimization/84111
6089 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6090 inner loops added during recursion, as they don't have up-to-date
6093 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6096 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6097 (can_inline_edge_by_limits_p): ... here.
6098 (can_early_inline_edge_p, check_callers,
6099 update_caller_keys, update_callee_keys, recursive_inlining,
6100 add_new_edges_to_heap, speculation_useful_p,
6101 inline_small_functions,
6102 inline_small_functions, flatten_function,
6103 inline_to_all_callers_1): Update.
6105 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6107 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6110 2018-01-30 Richard Biener <rguenther@suse.de>
6112 PR tree-optimization/83008
6113 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6114 invariant and constant vector uses in stmts when they need
6117 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6120 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6121 * configure: Regenerate.
6123 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6125 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6127 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6128 Use gen_rtx_REG rather than gen_lowpart.
6130 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6132 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6133 rather than 0 when creating partial subregs.
6135 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6137 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6140 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6143 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6144 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6145 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6146 flags. This restores the settings used before the 2017-07-24.
6147 Turning off pre increment/decrement/modify allows IVOPTS to
6148 optimize DF/SF loops where the index is an int.
6150 2018-01-29 Richard Biener <rguenther@suse.de>
6151 Kelvin Nilsen <kelvin@gcc.gnu.org>
6154 * tree-vect-stmts.c (vectorizable_call): Don't call
6155 targetm.vectorize_builtin_md_vectorized_function if callee is
6158 2018-01-22 Carl Love <cel@us.ibm.com>
6160 * doc/extend.tex: Fix typo in second arg in
6161 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6163 2018-01-29 Richard Biener <rguenther@suse.de>
6165 PR tree-optimization/84086
6166 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6167 (flush_ssaname_freelist): When SSA names were released reset
6168 the SCEV hash table.
6170 2018-01-29 Richard Biener <rguenther@suse.de>
6172 PR tree-optimization/84057
6173 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6174 removed paths when removing edges.
6176 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6178 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6179 -mfunction-return=@var{choice}.
6181 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6184 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6186 (layout::print_source_line): Likewise.
6187 (test_get_line_width_without_trailing_whitespace): Add test cases.
6189 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6192 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6195 2018-01-26 Jim Wilson <jimw@sifive.com>
6197 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6199 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6202 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6204 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6205 and CMP + SUB-immediate -> SUBS.
6207 2018-01-26 Martin Sebor <msebor@redhat.com>
6209 PR tree-optimization/83896
6210 * tree-ssa-strlen.c (get_string_len): Rename...
6211 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6212 Avoid assuming length is constant.
6213 (handle_char_store): Use HOST_WIDE_INT for string length.
6215 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6218 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6219 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6221 2018-01-26 Richard Biener <rguenther@suse.de>
6223 PR rtl-optimization/84003
6224 * dse.c (record_store): Only record redundant stores when
6225 the earlier store aliases at least all accesses the later one does.
6227 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6229 PR rtl-optimization/83985
6230 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6231 REG_CFA_RESTORE insns.
6232 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6233 REG_CFA_RESTORE insns here.
6236 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6237 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6239 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6241 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6242 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6243 (arc_init): Likewise.
6244 (arc_override_options): Likewise.
6245 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6247 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6249 * config/arc/arc.h (TARGET_DBNZ): Define.
6250 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6251 properly set the tune attribute.
6252 (dbnz): Use TARGET_DBNZ guard.
6253 * config/arc/arc.opt (mtune): Add core3 option.
6255 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6257 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6258 recognize new pic like addresses.
6259 (arc_delegitimize_address): Clean up.
6261 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6263 * config/arc/arc-arches.def: Option mrf16 valid for all
6265 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6266 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6267 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6268 * config/arc/arc-tables.opt: Regenerate.
6269 * config/arc/arc.c (arc_conditional_register_usage): Handle
6270 reduced register file case.
6271 (arc_file_start): Set must have build attributes.
6272 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6274 * config/arc/arc.opt (mrf16): Add new option.
6275 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6276 * config/arc/genmultilib.awk: Handle new mrf16 option.
6277 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6278 * config/arc/t-multilib: Regenerate.
6279 * doc/invoke.texi (ARC Options): Document mrf16 option.
6281 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6283 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6284 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6285 (arc_attribute_table): Add 'secure_call' attribute.
6286 (arc_print_operand): Print secure call operand.
6287 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6289 (arc_is_secure_call_p): New function. * config/arc/arc.md
6290 (call_i): Add support for sjli instruction.
6291 (call_value_i): Likewise.
6292 * config/arc/constraints.md (Csc): New constraint.
6294 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6295 John Eric Martin <John.Martin@emmicro-us.com>
6297 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6298 * config/arc/arc.c (_arc_jli_section): New struct.
6299 (arc_jli_section): New type.
6300 (rc_jli_sections): New static variable.
6301 (arc_handle_jli_attribute): New function.
6302 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6303 (arc_file_end): New function.
6304 (TARGET_ASM_FILE_END): Define.
6305 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6306 (arc_add_jli_section): New function.
6307 (jli_call_scan): Likewise.
6308 (arc_reorg): Call jli_call_scan.
6309 (arc_output_addsi): Remove 'S' from printing asm operand.
6310 (arc_is_jli_call_p): New function.
6311 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6313 (movhi_insn): Likewise.
6314 (movsi_insn): Likewise.
6315 (movsi_set_cc_insn): Likewise.
6316 (loadqi_update): Likewise.
6317 (load_zeroextendqisi_update): Likewise.
6318 (load_signextendqisi_update): Likewise.
6319 (loadhi_update): Likewise.
6320 (load_zeroextendhisi_update): Likewise.
6321 (load_signextendhisi_update): Likewise.
6322 (loadsi_update): Likewise.
6323 (loadsf_update): Likewise.
6324 (movsicc_insn): Likewise.
6325 (bset_insn): Likewise.
6326 (bxor_insn): Likewise.
6327 (bclr_insn): Likewise.
6328 (bmsk_insn): Likewise.
6329 (bicsi3_insn): Likewise.
6330 (cmpsi_cc_c_insn): Likewise.
6331 (movsi_ne): Likewise.
6332 (movsi_cond_exec): Likewise.
6333 (clrsbsi2): Likewise.
6344 (call_i): Remove 'S' asm letter, add jli instruction.
6345 (call_value_i): Likewise.
6346 * config/arc/arc.op (mjli-always): New option.
6347 * config/arc/constraints.md (Cji): New constraint.
6348 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6350 (subsf3_fpx): Likewise.
6351 (mulsf3_fpx): Likewise.
6352 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6354 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6356 * doc/invoke.texi (ARC): Document mjli-always option.
6358 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6360 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6361 avoid addition with 0 and use incw and decw where possible.
6363 2018-01-26 Richard Biener <rguenther@suse.de>
6365 PR tree-optimization/81082
6366 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6367 association if it requires casting to unsigned.
6368 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6369 from fold_plusminus_mult_expr to catch important cases late when
6370 range info is available.
6372 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6374 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6375 * configure.ac (hidden_linkonce): New test.
6376 * configure: Regenerate.
6377 * config.in: Regenerate.
6379 2018-01-26 Julia Koval <julia.koval@intel.com>
6381 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6382 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6383 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6384 _mm_mask_bitshuffle_epi64_mask): Fix type.
6385 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6386 USI_FTYPE_V4DI_V4DI_USI): Remove.
6387 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6388 __builtin_ia32_vpshufbitqmb256_mask,
6389 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6390 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6391 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6393 2018-01-26 Alan Modra <amodra@gmail.com>
6396 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6397 UNSPEC_VBPERMQ. Sort other unspecs.
6399 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6401 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6403 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6406 * predict.c (drop_profile): Do not push/pop cfun; update also
6408 (handle_missing_profiles): Fix logic looking for zero profiles.
6410 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6413 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6414 on functions with #pragma omp declare simd or functions with simd
6416 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6417 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6418 Remove trailing \n from warning_at calls.
6420 2018-01-25 Tom de Vries <tom@codesourcery.com>
6423 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6424 for neutered workers.
6426 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6429 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6430 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6432 2018-01-24 Jeff Law <law@redhat.com>
6435 * i386.c (get_probe_interval): Move to earlier point.
6436 (ix86_compute_frame_layout): If -fstack-clash-protection and
6437 the frame is larger than the probe interval, then use pushes
6438 to save registers rather than reg->mem moves.
6439 (ix86_expand_prologue): Remove conditional for int_registers_saved
6442 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6445 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6446 min/max for never referenced object.
6448 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6451 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6453 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6454 attributes from DECL_ATTRIBUTES (decl) without affecting
6455 DECL_ATTRIBUTES (current_function_decl).
6456 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6457 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6459 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6461 PR tree-optimization/83979
6462 * fold-const.c (fold_comparison): Use constant_boolean_node
6463 instead of boolean_{true,false}_node.
6465 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6467 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6470 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6472 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6473 Simplify the clause that sets the length attribute.
6474 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6475 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6476 clause that sets the length attribute.
6477 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6479 2018-01-24 Tom de Vries <tom@codesourcery.com>
6482 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6483 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6484 Add strict parameter.
6485 (prevent_branch_around_nothing): Insert dummy insn between branch to
6486 label and label with no ptx insn inbetween.
6487 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6489 2018-01-24 Tom de Vries <tom@codesourcery.com>
6492 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6493 for neutered threads in warp.
6494 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6496 2018-01-24 Richard Biener <rguenther@suse.de>
6498 PR tree-optimization/83176
6499 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6502 2018-01-24 Richard Biener <rguenther@suse.de>
6504 PR tree-optimization/82819
6505 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6506 code generating pluses that are no-ops in the target precision.
6508 2018-01-24 Richard Biener <rguenther@suse.de>
6511 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6513 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6515 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6516 to merge probabilities.
6517 * predict.c (probably_never_executed): Also mark as cold functions
6518 with global 0 profile and guessed local profile.
6519 * profile-count.c (profile_probability::combine_with_count): New
6521 * profile-count.h (profile_probability::operator*,
6522 profile_probability::operator*=, profile_probability::operator/,
6523 profile_probability::operator/=): Reduce precision to adjusted
6524 and set value to guessed on contradictory divisions.
6525 (profile_probability::combine_with_freq): Remove.
6526 (profile_probability::combine_wiht_count): Declare.
6527 (profile_count::force_nonzero):: Set to adjusted.
6528 (profile_count::probability_in):: Set quality to adjusted.
6529 * tree-ssa-tail-merge.c (replace_block_by): Use
6532 2018-01-23 Andrew Waterman <andrew@sifive.com>
6533 Jim Wilson <jimw@sifive.com>
6535 * config/riscv/riscv.c (riscv_stack_boundary): New.
6536 (riscv_option_override): Set riscv_stack_boundary. Handle
6537 riscv_preferred_stack_boundary_arg.
6538 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6539 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6540 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6541 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6542 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6543 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6545 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6548 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6549 of struct ix86_frame.
6550 (ix86_expand_epilogue): Likewise. Add a local variable for
6551 the reg_save_offset field in struct ix86_frame.
6553 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6555 PR tree-optimization/82604
6556 * tree-loop-distribution.c (enum partition_kind): New enum item
6557 PKIND_PARTIAL_MEMSET.
6558 (partition_builtin_p): Support above new enum item.
6559 (generate_code_for_partition): Ditto.
6560 (compute_access_range): Differentiate cases that equality can be
6561 proven at all loops, the innermost loops or no loops.
6562 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6563 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6564 (finalize_partitions, distribute_loop): Don't fuse partition of
6565 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6566 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6569 2018-01-23 Martin Liska <mliska@suse.cz>
6571 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6572 order to ignore the predictor.
6573 (PRED_POLYMORPHIC_CALL): Likewise.
6574 (PRED_RECURSIVE_CALL): Likewise.
6576 2018-01-23 Martin Liska <mliska@suse.cz>
6578 * tree-profile.c (tree_profiling): Print function header to
6579 aware reader which function we are working on.
6580 * value-prof.c (gimple_find_values_to_profile): Do not print
6581 not interesting value histograms.
6583 2018-01-23 Martin Liska <mliska@suse.cz>
6585 * profile-count.h (enum profile_quality): Add
6586 profile_uninitialized as the first value. Do not number values
6587 as they are zero based.
6588 (profile_count::verify): Update sanity check.
6589 (profile_probability::verify): Likewise.
6591 2018-01-23 Nathan Sidwell <nathan@acm.org>
6593 * doc/invoke.texi (ffor-scope): Deprecate.
6595 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6597 PR tree-optimization/83510
6598 * domwalk.c (set_all_edges_as_executable): New function.
6599 (dom_walker::dom_walker): Convert bool param
6600 "skip_unreachable_blocks" to enum reachability. Move setup of
6601 edge flags to set_all_edges_as_executable and only do it when
6602 reachability is REACHABLE_BLOCKS.
6603 * domwalk.h (enum dom_walker::reachability): New enum.
6604 (dom_walker::dom_walker): Convert bool param
6605 "skip_unreachable_blocks" to enum reachability.
6606 (set_all_edges_as_executable): New decl.
6607 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6608 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6610 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6611 but converting true to REACHABLE_BLOCKS.
6612 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6614 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6615 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6616 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6618 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6619 if check_all_array_refs will be called.
6621 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6623 * tree.c (selftest::test_location_wrappers): Add more test
6626 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6628 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6629 (selftest::test_bit_in_range): Likewise.
6631 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6634 * doc/sourcebuild.texi (vect_float): Say that the selector
6635 only describes the situation when -funsafe-math-optimizations is on.
6636 (vect_float_strict): Document.
6638 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6640 PR tree-optimization/83965
6641 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6642 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6643 instead of checking only for a reduction.
6644 (vect_recog_widen_sum_pattern): Likewise.
6646 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6648 * predict.c (probably_never_executed): Only use precise profile info.
6649 (compute_function_frequency): Skip after inlining hack since we now
6650 have quality checking.
6652 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6654 * profile-count.h (profile_probability::very_unlikely,
6655 profile_probability::unlikely, profile_probability::even): Set
6656 precision to guessed.
6658 2018-01-23 Richard Biener <rguenther@suse.de>
6660 PR tree-optimization/83963
6661 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6662 Properly terminate dominator walk when crossing the exit edge not
6663 when visiting its source block.
6665 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6668 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6669 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6671 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6673 PR tree-optimization/83957
6674 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6675 semicolon after for body surrounded by braces.
6677 PR tree-optimization/83081
6678 * profile-count.h (profile_probability::split): New method.
6679 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6680 Use profile_probability::split.
6681 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6682 when splitting a single conditional jump into 2.
6684 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6686 PR tree-optimization/69452
6687 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6690 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6692 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6693 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6694 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6696 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6698 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6699 * config/rl78/rl78.md (movdi): New define_expand.
6700 * config/rl78/rl78.c (rl78_split_movdi): New function.
6702 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6705 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6707 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6708 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6709 128-bit to produce an UNSPEC move to get the double word with the
6710 signbit and then a shift directly to do signbit.
6711 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6712 implementation with a new version that just does either a direct
6713 move or a regular move. Move memory interface to separate insns.
6714 Move insns so they are next to the expander.
6715 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6716 with signbit move. Split big and little endian case.
6717 (signbit<mode>2_dm_mem_le): Likewise.
6718 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6719 (signbit<mode>2_dm2): Likewise.
6721 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6723 * config/rl78/rl78.md (anddi3): New define_expand.
6725 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6727 * config/rl78/rl78.md (umindi3): New define_expand.
6729 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6731 * config/rl78/rl78.md (smindi3): New define_expand.
6733 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6735 * config/rl78/rl78.md (smaxdi3): New define_expand.
6737 2018-01-22 Carl Love <cel@us.ibm.com>
6739 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6740 LVX_V1TI): Add macro expansion.
6741 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6742 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6743 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6744 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6745 Change check to determine if the instruction is a byte reversing
6746 entry. Fix typo in comment.
6747 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6748 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6749 Add def_builtin calls for new builtins.
6750 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6751 Add define_insn expansion.
6753 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6755 * config/rl78/rl78.md (umaxdi3): New define_expand.
6757 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6759 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6760 for non-QImode registers.
6762 2018-01-22 Richard Biener <rguenther@suse.de>
6764 PR tree-optimization/83963
6765 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6766 including the loop exit block.
6767 (scop_detection::merge_sese): Likewise.
6768 (scop_detection::add_scop): Do it here instead.
6770 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6772 * doc/sourcebuild.texi (arm_softfloat): Document.
6774 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6777 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6778 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6779 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6781 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6782 David Edelsohn <dje.gcc@gmail.com>
6785 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6786 Change "crset eq" to "crset 2".
6787 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6788 (*call_indirect_aix<mode>_nospec): Likewise.
6789 (*call_value_indirect_aix<mode>_nospec): Likewise.
6790 (*call_indirect_elfv2<mode>_nospec): Likewise.
6791 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6792 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6793 change assembly output from . to $.
6794 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6795 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6796 (*tablejump<mode>_internal1_nospec): Likewise.
6798 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6801 * config/sh/sh_optimize_sett_clrt.cc:
6802 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6804 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6806 PR tree-optimization/83940
6807 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6808 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6809 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6810 use it to pass back the definition type.
6811 (vect_check_store_rhs): Likewise rhs_dt_out.
6812 (vect_build_gather_load_calls): Add a mask_dt argument and use
6813 it instead of a call to vect_is_simple_use.
6814 (vectorizable_store): Update calls to vect_check_load_store_mask
6815 and vect_check_store_rhs. Use the dt returned by the latter instead
6816 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6817 instead of calls to vect_is_simple_use. Pass the scalar rather
6818 than the vector operand to vect_is_simple_use when handling
6819 second and subsequent copies of an rhs value.
6820 (vectorizable_load): Update calls to vect_check_load_store_mask
6821 and vect_build_gather_load_calls. Use the cached mask_dt and
6822 gs_info.offset_dt instead of calls to vect_is_simple_use.
6824 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6827 * tree-emutls.c: Include gimplify.h.
6828 (lower_emutls_2): New function.
6829 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6830 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6831 it before further processing.
6834 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6835 UINTVAL (trueop1) instead of INTVAL (op1).
6837 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6841 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6842 INCOMING_FRAME_SP_OFFSET if not defined.
6843 (scan_trace): Add ENTRY argument. If true and
6844 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6845 emit a note to adjust the CFA offset.
6846 (create_cfi_notes): Adjust scan_trace callers.
6847 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6848 INCOMING_FRAME_SP_OFFSET in the CIE.
6849 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6850 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6852 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6853 * doc/tm.texi: Regenerated.
6855 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6857 PR rtl-optimization/83147
6858 * lra-constraints.c (remove_inheritance_pseudos): Use
6859 lra_substitute_pseudo_within_insn.
6861 2018-01-19 Tom de Vries <tom@codesourcery.com>
6862 Cesar Philippidis <cesar@codesourcery.com>
6865 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6867 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6870 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6871 spaces for function labels.
6873 2018-01-19 Martin Liska <mliska@suse.cz>
6875 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6876 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6877 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6878 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6879 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6880 (PRED_CONST_RETURN): Change from 69 to 65.
6881 (PRED_NULL_RETURN): Change from 91 to 71.
6882 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6883 (PRED_LOOP_GUARD): Change from 66 to 73.
6885 2018-01-19 Martin Liska <mliska@suse.cz>
6887 * predict.c (predict_insn_def): Add new assert.
6888 (struct branch_predictor): Change type to signed integer.
6889 (test_prediction_value_range): Amend test to cover
6891 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6892 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6893 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6894 (PRED_LOOP_IV_COMPARE): Likewise.
6895 * predict.h (PROB_UNINITIALIZED): Define new constant.
6897 2018-01-19 Martin Liska <mliska@suse.cz>
6899 * predict.c (dump_prediction): Add new format for
6900 analyze_brprob.py script which is enabled with -details
6902 * profile-count.h (precise_p): New function.
6904 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6906 PR tree-optimization/83922
6907 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6908 there are no statements that need masking.
6909 (vect_active_double_reduction_p): New function.
6910 (vect_analyze_loop_operations): Use it when handling phis that
6911 are not in the loop header.
6913 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6915 PR tree-optimization/83914
6916 * tree-vect-loop.c (vectorizable_induction): Don't convert
6917 init_expr or apply the peeling adjustment for inductions
6918 that are nested within the vectorized loop.
6920 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6922 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6925 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6929 * function.h (gimplify_parameters): Add gimple_seq * argument.
6930 * function.c: Include gimple.h and options.h.
6931 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6932 for the added local temporaries if needed.
6933 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6934 if there are any parameter cleanups, wrap whole body into a
6935 try/finally with the cleanups.
6937 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6940 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6941 Use GET_MODE_CLASS for scalar floating point.
6943 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6947 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6948 Fix call of call_cgraph_insertion_hooks.
6950 2018-01-18 Martin Sebor <msebor@redhat.com>
6952 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6954 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6957 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6960 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6963 * common.opt: (-ffile-prefix-map): New option.
6964 * opts.c (common_handle_option): Defer it.
6965 * opts-global.c (handle_common_deferred_options): Handle it.
6966 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6967 * file-prefix-map.h: New file.
6968 (remap_debug_filename, add_debug_prefix_map): ...here.
6969 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6970 * final.c (debug_prefix_map, add_debug_prefix_map
6971 remap_debug_filename): Move to...
6972 * file-prefix-map.c: New file.
6973 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6974 generalize, get rid of alloca(), use strrchr() instead of strchr().
6975 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6976 Implement in terms of add_prefix_map().
6977 (remap_macro_filename, remap_debug_filename): Implement in term of
6979 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6980 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6981 * dbxout.c: Include file-prefix-map.h.
6982 * varasm.c: Likewise.
6983 * vmsdbgout.c: Likewise.
6984 * xcoffout.c: Likewise.
6985 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6986 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6987 * doc/invoke.texi (-ffile-prefix-map): Document.
6988 (-fdebug-prefix-map): Update description.
6990 2018-01-18 Martin Liska <mliska@suse.cz>
6992 * config/i386/i386.c (indirect_thunk_name): Document that also
6994 (output_indirect_thunk): Document why both instructions
6995 (pause and lfence) are generated.
6997 2018-01-18 Richard Biener <rguenther@suse.de>
6999 PR tree-optimization/83887
7000 * graphite-scop-detection.c
7001 (scop_detection::get_nearest_dom_with_single_entry): Remove.
7002 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
7003 (scop_detection::merge_sese): Re-implement with a flood-fill
7004 algorithm that properly finds a SESE region if it exists.
7006 2018-01-18 Jakub Jelinek <jakub@redhat.com>
7009 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
7010 pointer_diff optimizations use view_convert instead of convert.
7012 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7014 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
7015 Generate different code for -mno-speculate-indirect-jumps.
7016 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
7017 (*call_indirect_aix<mode>): Disable for
7018 -mno-speculate-indirect-jumps.
7019 (*call_indirect_aix<mode>_nospec): New define_insn.
7020 (*call_value_indirect_aix<mode>): Disable for
7021 -mno-speculate-indirect-jumps.
7022 (*call_value_indirect_aix<mode>_nospec): New define_insn.
7023 (*sibcall_nonlocal_sysv<mode>): Generate different code for
7024 -mno-speculate-indirect-jumps.
7025 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
7027 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
7029 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
7030 long double type, set the flags for noting the default long double
7031 type, even if we don't pass or return a long double type.
7033 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
7036 * ipa-inline.c (flatten_function): Do not overwrite final inlining
7039 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
7041 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
7042 support for merge[hl].
7043 (fold_mergehl_helper): New helper function.
7044 (tree-vector-builder.h): New #include for tree_vector_builder usage.
7045 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
7046 (altivec_vmrglw_direct): Add xxmrglw insn.
7048 2018-01-17 Andrew Waterman <andrew@sifive.com>
7050 * config/riscv/riscv.c (riscv_conditional_register_usage): If
7051 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
7053 2018-01-17 David Malcolm <dmalcolm@redhat.com>
7056 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
7057 call the lto_location_cache before reading the
7058 DECL_SOURCE_LOCATION of the types.
7060 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7061 Richard Sandiford <richard.sandiford@linaro.org>
7063 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7064 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7065 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7066 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7067 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7069 * config/aarch64/constraints.md (aarch64_movti_operand):
7071 * config/aarch64/predicates.md (Uti): Add new constraint.
7073 2018-01-17 Carl Love <cel@us.ibm.com>
7075 * config/rs6000/vsx.md (define_expand xl_len_r,
7076 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7077 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7079 (define_expand, define_insn): Move the shift left from the
7080 define_insn to the define_expand for lxvl and stxvl instructions.
7081 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7082 and XL_LEN_R definitions to PURE.
7084 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7086 * config/i386/i386.c (indirect_thunk_name): Declare regno
7087 as unsigned int. Compare regno with INVALID_REGNUM.
7088 (output_indirect_thunk): Ditto.
7089 (output_indirect_thunk_function): Ditto.
7090 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7091 in the call to output_indirect_thunk_function.
7093 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7096 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7097 rather than the size of inner_type to determine the stack slot size
7098 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7100 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7103 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7106 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7108 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7109 endian Linux systems to optionally enable multilibs for selecting
7110 the long double type if the user configured an explicit type.
7111 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7112 have no long double multilibs if not defined.
7113 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7114 warn if the user used -mabi={ieee,ibm}longdouble and we built
7115 multilibs for long double.
7116 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7117 appropriate multilib option.
7118 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7120 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7121 for building long double multilibs.
7122 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7124 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7126 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7129 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7131 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7134 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7137 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7140 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7142 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7143 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7145 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7147 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7148 different rtl trees depending on TARGET_64BIT.
7149 (rs6000_gen_lvx): Likewise.
7151 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7153 * config/visium/visium.md (nop): Tweak comment.
7154 (hazard_nop): Likewise.
7156 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7158 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7159 -mspeculate-indirect-jumps.
7160 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7161 for -mno-speculate-indirect-jumps.
7162 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7163 (*call_value_indirect_elfv2<mode>): Disable for
7164 -mno-speculate-indirect-jumps.
7165 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7166 (indirect_jump): Emit different RTL for
7167 -mno-speculate-indirect-jumps.
7168 (*indirect_jump<mode>): Disable for
7169 -mno-speculate-indirect-jumps.
7170 (*indirect_jump<mode>_nospec): New define_insn.
7171 (tablejump): Emit different RTL for
7172 -mno-speculate-indirect-jumps.
7173 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7174 (tablejumpsi_nospec): New define_expand.
7175 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7176 (tablejumpdi_nospec): New define_expand.
7177 (*tablejump<mode>_internal1): Disable for
7178 -mno-speculate-indirect-jumps.
7179 (*tablejump<mode>_internal1_nospec): New define_insn.
7180 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7183 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7185 * caller-save.c (insert_save): Drop unnecessary parameter. All
7188 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7189 Richard Biener <rguenth@suse.de>
7192 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7193 return early, inline manually is_gimple_sizepos. Make sure if we
7194 call gimplify_expr we don't end up with a gimple constant.
7195 * tree.c (variably_modified_type_p): Don't return true for
7196 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7197 * gimplify.h (is_gimple_sizepos): Remove.
7199 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7201 PR tree-optimization/83857
7202 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7203 vectorizable_live_operation for pure SLP statements.
7204 (vectorizable_live_operation): Handle PHIs.
7206 2018-01-16 Richard Biener <rguenther@suse.de>
7208 PR tree-optimization/83867
7209 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7210 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7212 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7215 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7216 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7217 If off is not INTEGER_CST, issue a may not be aligned warning
7218 rather than isn't aligned. Use isn%'t rather than isn't.
7219 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7221 <case MULT_EXPR>: Improve the case when bottom and one of the
7222 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7223 operand, in that case check if the other operand is multiple of
7224 bottom divided by the INTEGER_CST operand.
7226 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7229 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7230 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7231 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7232 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7233 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7234 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7235 (pa_function_arg_size): New function.
7237 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7239 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7240 in a separate statement.
7242 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7244 PR tree-optimization/83847
7245 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7246 group gathers and scatters.
7248 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7250 PR rtl-optimization/86620
7251 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7253 PR rtl-optimization/83213
7254 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7255 to last if both are JUMP_INSNs.
7257 PR tree-optimization/83843
7258 * gimple-ssa-store-merging.c
7259 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7260 store_immediate_info for bswap/nop orig_stores.
7262 2018-01-15 Andrew Waterman <andrew@sifive.com>
7264 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7266 <UDIV>: Increase cost if !TARGET_DIV.
7268 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7270 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7271 (define_attr "cr_logical_3op"): New.
7272 (cceq_ior_compare): Adjust.
7273 (cceq_ior_compare_complement): Adjust.
7274 (*cceq_rev_compare): Adjust.
7275 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7276 (is_cracked_insn): Adjust.
7277 (insn_must_be_first_in_group): Adjust.
7278 * config/rs6000/40x.md: Adjust.
7279 * config/rs6000/440.md: Adjust.
7280 * config/rs6000/476.md: Adjust.
7281 * config/rs6000/601.md: Adjust.
7282 * config/rs6000/603.md: Adjust.
7283 * config/rs6000/6xx.md: Adjust.
7284 * config/rs6000/7450.md: Adjust.
7285 * config/rs6000/7xx.md: Adjust.
7286 * config/rs6000/8540.md: Adjust.
7287 * config/rs6000/cell.md: Adjust.
7288 * config/rs6000/e300c2c3.md: Adjust.
7289 * config/rs6000/e500mc.md: Adjust.
7290 * config/rs6000/e500mc64.md: Adjust.
7291 * config/rs6000/e5500.md: Adjust.
7292 * config/rs6000/e6500.md: Adjust.
7293 * config/rs6000/mpc.md: Adjust.
7294 * config/rs6000/power4.md: Adjust.
7295 * config/rs6000/power5.md: Adjust.
7296 * config/rs6000/power6.md: Adjust.
7297 * config/rs6000/power7.md: Adjust.
7298 * config/rs6000/power8.md: Adjust.
7299 * config/rs6000/power9.md: Adjust.
7300 * config/rs6000/rs64.md: Adjust.
7301 * config/rs6000/titan.md: Adjust.
7303 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7305 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7306 ix86_indirect_branch_register logic.
7308 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7310 * config/i386/constraints.md (Bs): Update
7311 ix86_indirect_branch_register check. Don't check
7312 ix86_indirect_branch_register with GOT_memory_operand.
7314 * config/i386/predicates.md (GOT_memory_operand): Don't check
7315 ix86_indirect_branch_register here.
7316 (GOT32_symbol_operand): Likewise.
7318 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7320 * config/i386/predicates.md (constant_call_address_operand):
7321 Rewrite ix86_indirect_branch_register logic.
7322 (sibcall_insn_operand): Likewise.
7324 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7326 * config/i386/constraints.md (Bs): Replace
7327 ix86_indirect_branch_thunk_register with
7328 ix86_indirect_branch_register.
7330 * config/i386/i386.md (indirect_jump): Likewise.
7331 (tablejump): Likewise.
7332 (*sibcall_memory): Likewise.
7333 (*sibcall_value_memory): Likewise.
7334 Peepholes of indirect call and jump via memory: Likewise.
7335 * config/i386/i386.opt: Likewise.
7336 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7337 (GOT_memory_operand): Likewise.
7338 (call_insn_operand): Likewise.
7339 (sibcall_insn_operand): Likewise.
7340 (GOT32_symbol_operand): Likewise.
7342 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7345 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7346 type rather than type addr's type points to.
7347 (expand_omp_atomic_mutex): Likewise.
7348 (expand_omp_atomic): Likewise.
7350 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7353 * config/i386/i386.c (output_indirect_thunk_function): Use
7354 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7355 for __x86_return_thunk.
7357 2018-01-15 Richard Biener <rguenther@suse.de>
7360 * expmed.c (extract_bit_field_1): Fix typo.
7362 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7365 * config/arm/iterators.md (VF): New mode iterator.
7366 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7367 Remove integer-related logic from pattern.
7368 (neon_vabd<mode>_3): Likewise.
7370 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7373 * common.opt (fstrict-overflow): No longer an alias.
7374 (fwrapv-pointer): New option.
7375 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7376 also for pointer types based on flag_wrapv_pointer.
7377 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7378 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7379 opts->x_flag_wrapv got set.
7380 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7381 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7382 POINTER_TYPE_OVERFLOW_UNDEFINED.
7383 * match.pd: Likewise in address comparison pattern.
7384 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7386 2018-01-15 Richard Biener <rguenther@suse.de>
7389 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7390 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7391 Reset type names to their identifier if their TYPE_DECL doesn't
7392 have linkage (and thus is used for ODR and devirt).
7393 (save_debug_info_for_decl): Remove.
7394 (save_debug_info_for_type): Likewise.
7395 (add_tree_to_fld_list): Adjust.
7396 * tree-pretty-print.c (dump_generic_node): Make dumping of
7397 type names more robust.
7399 2018-01-15 Richard Biener <rguenther@suse.de>
7401 * BASE-VER: Bump to 8.0.1.
7403 2018-01-14 Martin Sebor <msebor@redhat.com>
7406 * builtins.c (check_access): Avoid warning when the no-warning bit
7409 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7411 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7412 * ira-color (allocno_hard_regs_compare): Likewise.
7414 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7417 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7418 Use .pushsection/.popsection.
7420 2018-01-14 Martin Sebor <msebor@redhat.com>
7423 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7425 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7427 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7428 entry from extra_headers.
7429 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7430 extra_headers, make the list bitwise identical to the i?86-*-* one.
7432 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7434 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7435 -mcmodel=large with -mindirect-branch=thunk,
7436 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7437 -mfunction-return=thunk-extern.
7438 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7439 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7440 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7442 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7444 * config/i386/i386.c (print_reg): Print the name of the full
7445 integer register without '%'.
7446 (ix86_print_operand): Handle 'V'.
7447 * doc/extend.texi: Document 'V' modifier.
7449 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7451 * config/i386/constraints.md (Bs): Disallow memory operand for
7452 -mindirect-branch-register.
7454 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7455 (GOT_memory_operand): Likewise.
7456 (call_insn_operand): Likewise.
7457 (sibcall_insn_operand): Likewise.
7458 (GOT32_symbol_operand): Likewise.
7459 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7460 for -mindirect-branch-register.
7461 (tablejump): Likewise.
7462 (*sibcall_memory): Likewise.
7463 (*sibcall_value_memory): Likewise.
7464 Disallow peepholes of indirect call and jump via memory for
7465 -mindirect-branch-register.
7466 (*call_pop): Replace m with Bw.
7467 (*call_value_pop): Likewise.
7468 (*sibcall_pop_memory): Replace m with Bs.
7469 * config/i386/i386.opt (mindirect-branch-register): New option.
7470 * doc/invoke.texi: Document -mindirect-branch-register option.
7472 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7474 * config/i386/i386-protos.h (ix86_output_function_return): New.
7475 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7476 set function_return_type.
7477 (indirect_thunk_name): Add ret_p to indicate thunk for function
7479 (output_indirect_thunk_function): Pass false to
7480 indirect_thunk_name.
7481 (ix86_output_indirect_branch_via_reg): Likewise.
7482 (ix86_output_indirect_branch_via_push): Likewise.
7483 (output_indirect_thunk_function): Create alias for function
7484 return thunk if regno < 0.
7485 (ix86_output_function_return): New function.
7486 (ix86_handle_fndecl_attribute): Handle function_return.
7487 (ix86_attribute_table): Add function_return.
7488 * config/i386/i386.h (machine_function): Add
7489 function_return_type.
7490 * config/i386/i386.md (simple_return_internal): Use
7491 ix86_output_function_return.
7492 (simple_return_internal_long): Likewise.
7493 * config/i386/i386.opt (mfunction-return=): New option.
7494 (indirect_branch): Mention -mfunction-return=.
7495 * doc/extend.texi: Document function_return function attribute.
7496 * doc/invoke.texi: Document -mfunction-return= option.
7498 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7500 * config/i386/i386-opts.h (indirect_branch): New.
7501 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7502 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7503 with local indirect jump when converting indirect call and jump.
7504 (ix86_set_indirect_branch_type): New.
7505 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7506 (indirectlabelno): New.
7507 (indirect_thunk_needed): Likewise.
7508 (indirect_thunk_bnd_needed): Likewise.
7509 (indirect_thunks_used): Likewise.
7510 (indirect_thunks_bnd_used): Likewise.
7511 (INDIRECT_LABEL): Likewise.
7512 (indirect_thunk_name): Likewise.
7513 (output_indirect_thunk): Likewise.
7514 (output_indirect_thunk_function): Likewise.
7515 (ix86_output_indirect_branch_via_reg): Likewise.
7516 (ix86_output_indirect_branch_via_push): Likewise.
7517 (ix86_output_indirect_branch): Likewise.
7518 (ix86_output_indirect_jmp): Likewise.
7519 (ix86_code_end): Call output_indirect_thunk_function if needed.
7520 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7522 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7523 (ix86_attribute_table): Add indirect_branch.
7524 * config/i386/i386.h (machine_function): Add indirect_branch_type
7525 and has_local_indirect_jump.
7526 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7528 (tablejump): Likewise.
7529 (*indirect_jump): Use ix86_output_indirect_jmp.
7530 (*tablejump_1): Likewise.
7531 (simple_return_indirect_internal): Likewise.
7532 * config/i386/i386.opt (mindirect-branch=): New option.
7533 (indirect_branch): New.
7536 (thunk-inline): Likewise.
7537 (thunk-extern): Likewise.
7538 * doc/extend.texi: Document indirect_branch function attribute.
7539 * doc/invoke.texi: Document -mindirect-branch= option.
7541 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7544 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7546 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7548 * ipa-inline.c (want_inline_small_function_p): Return false if
7549 inlining has already failed with CIF_FINAL_ERROR.
7550 (update_caller_keys): Call want_inline_small_function_p before
7552 (update_callee_keys): Likewise.
7554 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7556 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7558 (rs6000_quadword_masked_address_p): Likewise.
7559 (quad_aligned_load_p): Likewise.
7560 (quad_aligned_store_p): Likewise.
7561 (const_load_sequence_p): Add comment to describe the outer-most loop.
7562 (mimic_memory_attributes_and_flags): New function.
7563 (rs6000_gen_stvx): Likewise.
7564 (replace_swapped_aligned_store): Likewise.
7565 (rs6000_gen_lvx): Likewise.
7566 (replace_swapped_aligned_load): Likewise.
7567 (replace_swapped_load_constant): Capitalize argument name in
7568 comment describing this function.
7569 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7570 and stores that access quad-word aligned addresses and replace
7571 with stvx or lvx instructions when appropriate.
7572 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7573 New function prototype.
7574 (rs6000_quadword_masked_address_p): Likewise.
7575 (rs6000_gen_lvx): Likewise.
7576 (rs6000_gen_stvx): Likewise.
7577 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7578 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7579 when memory address is aligned.
7580 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7581 this split to select lvx instruction when memory address is aligned.
7582 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7583 instruction when memory address is aligned.
7584 (*vsx_le_perm_load_v16qi): Likewise.
7585 (four unnamed splitters): Modify to select the stvx instruction
7586 when memory is aligned.
7588 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7590 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7591 which appears in the queue multiple times.
7593 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7594 Alan Hayward <alan.hayward@arm.com>
7595 David Sherwood <david.sherwood@arm.com>
7597 * tree-vectorizer.h (vec_lower_bound): New structure.
7598 (_loop_vec_info): Add check_nonzero and lower_bounds.
7599 (LOOP_VINFO_CHECK_NONZERO): New macro.
7600 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7601 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7602 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7603 fields. Make seg_len the distance travelled, not including the
7605 (dr_direction_indicator): Declare.
7606 (dr_zero_step_indicator): Likewise.
7607 (dr_known_forward_stride_p): Likewise.
7608 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7610 (runtime_alias_check_p): Allow runtime alias checks with
7612 (operator ==): Compare access_size and align.
7613 (prune_runtime_alias_test_list): Rework for new distinction between
7614 the access_size and seg_len.
7615 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7617 (get_segment_min_max): New function.
7618 (create_intersect_range_checks): Use it.
7619 (dr_step_indicator): New function.
7620 (dr_direction_indicator): Likewise.
7621 (dr_zero_step_indicator): Likewise.
7622 (dr_known_forward_stride_p): Likewise.
7623 * tree-loop-distribution.c (data_ref_segment_size): Return
7624 DR_STEP * (niters - 1).
7625 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7627 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7628 (vect_preserves_scalar_order_p): New function, split out from...
7629 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7630 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7631 (vect_vfa_access_size): New function.
7632 (vect_vfa_align): Likewise.
7633 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7634 (dump_lower_bound): New function.
7635 (vect_check_lower_bound): Likewise.
7636 (vect_small_gap_p): Likewise.
7637 (vectorizable_with_step_bound_p): Likewise.
7638 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7639 depencies if the vectorization factor is 1. Convert the checks
7640 for nonzero steps into checks on the bounds of DR_STEP. Try using
7641 a bunds check for variable steps if the minimum required step is
7642 relatively small. Update calls to the dr_with_seg_len
7643 constructor and to vect_compile_time_alias.
7644 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7646 (vect_loop_versioning): Call it.
7647 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7649 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7651 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7652 Alan Hayward <alan.hayward@arm.com>
7653 David Sherwood <david.sherwood@arm.com>
7655 * doc/sourcebuild.texi (vect_scatter_store): Document.
7656 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7658 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7660 * genopinit.c (main): Add supports_vec_scatter_store and
7661 supports_vec_scatter_store_cached to target_optabs.
7662 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7663 IFN_MASK_SCATTER_STORE.
7664 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7666 * internal-fn.h (internal_store_fn_p): Declare.
7667 (internal_fn_stored_value_index): Likewise.
7668 * internal-fn.c (scatter_store_direct): New macro.
7669 (expand_scatter_store_optab_fn): New function.
7670 (direct_scatter_store_optab_supported_p): New macro.
7671 (internal_store_fn_p): New function.
7672 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7673 IFN_MASK_SCATTER_STORE.
7674 (internal_fn_mask_index): Likewise.
7675 (internal_fn_stored_value_index): New function.
7676 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7678 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7679 * optabs-query.c (supports_vec_scatter_store_p): New function.
7680 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7681 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7682 true for scatter stores.
7683 (vect_gather_scatter_fn_p): Handle scatter stores too.
7684 (vect_check_gather_scatter): Consider using scatter stores if
7685 supports_vec_scatter_store_p.
7686 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7688 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7689 internal_fn_stored_value_index.
7690 (check_load_store_masking): Handle scatter stores too.
7691 (vect_get_store_rhs): Make public.
7692 (vectorizable_call): Use internal_store_fn_p.
7693 (vectorizable_store): Handle scatter store internal functions.
7694 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7695 when deciding whether the end of the group has been reached.
7696 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7697 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7698 (mask_scatter_store<mode>): New insns.
7700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7701 Alan Hayward <alan.hayward@arm.com>
7702 David Sherwood <david.sherwood@arm.com>
7704 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7705 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7706 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7708 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7709 Use vect_truncate_gather_scatter_offset if we can't treat the
7710 operation as a normal gather load or scatter store.
7711 (get_group_load_store_type): Take the gather_scatter_info
7712 as argument. Try using a gather load or scatter store for
7713 single-element groups.
7714 (get_load_store_type): Update calls to get_group_load_store_type
7715 and vect_use_strided_gather_scatters_p.
7717 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7718 Alan Hayward <alan.hayward@arm.com>
7719 David Sherwood <david.sherwood@arm.com>
7721 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7722 optional tree argument.
7723 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7725 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7726 but continue to use the current value as a fallback.
7727 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7728 to compare the updates.
7729 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7730 (get_load_store_type): Use it when handling a strided access.
7731 (vect_get_strided_load_store_ops): New function.
7732 (vect_get_data_ptr_increment): Likewise.
7733 (vectorizable_load): Handle strided gather loads. Always pass
7734 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7737 Alan Hayward <alan.hayward@arm.com>
7738 David Sherwood <david.sherwood@arm.com>
7740 * doc/md.texi (gather_load@var{m}): Document.
7741 (mask_gather_load@var{m}): Likewise.
7742 * genopinit.c (main): Add supports_vec_gather_load and
7743 supports_vec_gather_load_cached to target_optabs.
7744 * optabs-tree.c (init_tree_optimization_optabs): Use
7745 ggc_cleared_alloc to allocate target_optabs.
7746 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7747 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7749 * internal-fn.h (internal_load_fn_p): Declare.
7750 (internal_gather_scatter_fn_p): Likewise.
7751 (internal_fn_mask_index): Likewise.
7752 (internal_gather_scatter_fn_supported_p): Likewise.
7753 * internal-fn.c (gather_load_direct): New macro.
7754 (expand_gather_load_optab_fn): New function.
7755 (direct_gather_load_optab_supported_p): New macro.
7756 (direct_internal_fn_optab): New function.
7757 (internal_load_fn_p): Likewise.
7758 (internal_gather_scatter_fn_p): Likewise.
7759 (internal_fn_mask_index): Likewise.
7760 (internal_gather_scatter_fn_supported_p): Likewise.
7761 * optabs-query.c (supports_at_least_one_mode_p): New function.
7762 (supports_vec_gather_load_p): Likewise.
7763 * optabs-query.h (supports_vec_gather_load_p): Declare.
7764 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7765 and memory_type field.
7766 (NUM_PATTERNS): Bump to 15.
7767 * tree-vect-data-refs.c: Include internal-fn.h.
7768 (vect_gather_scatter_fn_p): New function.
7769 (vect_describe_gather_scatter_call): Likewise.
7770 (vect_check_gather_scatter): Try using internal functions for
7771 gather loads. Recognize existing calls to a gather load function.
7772 (vect_analyze_data_refs): Consider using gather loads if
7773 supports_vec_gather_load_p.
7774 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7775 (vect_get_gather_scatter_offset_type): Likewise.
7776 (vect_convert_mask_for_vectype): Likewise.
7777 (vect_add_conversion_to_patterm): Likewise.
7778 (vect_try_gather_scatter_pattern): Likewise.
7779 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7780 (vect_vect_recog_func_ptrs): Add it.
7781 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7782 internal_fn_mask_index and internal_gather_scatter_fn_p.
7783 (check_load_store_masking): Take the gather_scatter_info as an
7784 argument and handle gather loads.
7785 (vect_get_gather_scatter_ops): New function.
7786 (vectorizable_call): Check internal_load_fn_p.
7787 (vectorizable_load): Likewise. Handle gather load internal
7789 (vectorizable_store): Update call to check_load_store_masking.
7790 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7791 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7792 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7793 (aarch64_gather_scale_operand_d): New predicates.
7794 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7795 (mask_gather_load<mode>): New insns.
7797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7798 Alan Hayward <alan.hayward@arm.com>
7799 David Sherwood <david.sherwood@arm.com>
7801 * optabs.def (fold_left_plus_optab): New optab.
7802 * doc/md.texi (fold_left_plus_@var{m}): Document.
7803 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7804 * internal-fn.c (fold_left_direct): Define.
7805 (expand_fold_left_optab_fn): Likewise.
7806 (direct_fold_left_optab_supported_p): Likewise.
7807 * fold-const-call.c (fold_const_fold_left): New function.
7808 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7809 * tree-parloops.c (valid_reduction_p): New function.
7810 (gather_scalar_reductions): Use it.
7811 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7812 (vect_finish_replace_stmt): Declare.
7813 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7814 (needs_fold_left_reduction_p): New function, split out from...
7815 (vect_is_simple_reduction): ...here. Accept reductions that
7816 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7817 (vect_force_simple_reduction): Also store the reduction type in
7818 the assignment's STMT_VINFO_REDUC_TYPE.
7819 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7820 (merge_with_identity): New function.
7821 (vect_expand_fold_left): Likewise.
7822 (vectorize_fold_left_reduction): Likewise.
7823 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7824 scalar phi in place for it. Check for target support and reject
7825 cases that would reassociate the operation. Defer the transform
7826 phase to vectorize_fold_left_reduction.
7827 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7828 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7829 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7831 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7833 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7834 call to ifc_temp_var.
7836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7837 Alan Hayward <alan.hayward@arm.com>
7838 David Sherwood <david.sherwood@arm.com>
7840 * target.def (legitimize_address_displacement): Take the original
7841 offset as a poly_int.
7842 * targhooks.h (default_legitimize_address_displacement): Update
7844 * targhooks.c (default_legitimize_address_displacement): Likewise.
7845 * doc/tm.texi: Regenerate.
7846 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7847 as an argument, moving assert of ad->disp == ad->disp_term to...
7848 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7849 Try calling targetm.legitimize_address_displacement before expanding
7850 the address rather than afterwards, and adjust for the new interface.
7851 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7852 Match the new hook interface. Handle SVE addresses.
7853 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7858 * Makefile.in (OBJS): Add early-remat.o.
7859 * target.def (select_early_remat_modes): New hook.
7860 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7861 * doc/tm.texi: Regenerate.
7862 * targhooks.h (default_select_early_remat_modes): Declare.
7863 * targhooks.c (default_select_early_remat_modes): New function.
7864 * timevar.def (TV_EARLY_REMAT): New timevar.
7865 * passes.def (pass_early_remat): New pass.
7866 * tree-pass.h (make_pass_early_remat): Declare.
7867 * early-remat.c: New file.
7868 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7870 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7873 Alan Hayward <alan.hayward@arm.com>
7874 David Sherwood <david.sherwood@arm.com>
7876 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7877 vfm1 with a bound_epilog parameter.
7878 (vect_do_peeling): Update calls accordingly, and move the prologue
7879 call earlier in the function. Treat the base bound_epilog as 0 for
7880 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7881 this base when peeling for gaps.
7882 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7883 with fully-masked loops.
7884 (vect_estimate_min_profitable_iters): Handle the single peeled
7885 iteration in that case.
7887 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7888 Alan Hayward <alan.hayward@arm.com>
7889 David Sherwood <david.sherwood@arm.com>
7891 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7892 single-element interleaving even if the size is not a power of 2.
7893 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7894 accesses for single-element interleaving if the group size is
7897 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7898 Alan Hayward <alan.hayward@arm.com>
7899 David Sherwood <david.sherwood@arm.com>
7901 * doc/md.texi (fold_extract_last_@var{m}): Document.
7902 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7903 * optabs.def (fold_extract_last_optab): New optab.
7904 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7905 * internal-fn.c (fold_extract_direct): New macro.
7906 (expand_fold_extract_optab_fn): Likewise.
7907 (direct_fold_extract_optab_supported_p): Likewise.
7908 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7909 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7910 EXTRACT_LAST_REDUCTION.
7911 (get_initial_def_for_reduction): Do not create an initial vector
7912 for EXTRACT_LAST_REDUCTION reductions.
7913 (vectorizable_reduction): Leave the scalar phi in place for
7914 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7915 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7916 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7917 transform phase to vectorizable_condition.
7918 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7920 (vect_finish_stmt_generation): ...here.
7921 (vect_finish_replace_stmt): New function.
7922 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7923 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7925 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7928 Alan Hayward <alan.hayward@arm.com>
7929 David Sherwood <david.sherwood@arm.com>
7931 * doc/md.texi (extract_last_@var{m}): Document.
7932 * optabs.def (extract_last_optab): New optab.
7933 * internal-fn.def (EXTRACT_LAST): New internal function.
7934 * internal-fn.c (cond_unary_direct): New macro.
7935 (expand_cond_unary_optab_fn): Likewise.
7936 (direct_cond_unary_optab_supported_p): Likewise.
7937 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7938 loops using EXTRACT_LAST.
7939 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7940 (extract_last_<mode>): ...this optab.
7941 (vec_extract<mode><Vel>): Update accordingly.
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 * target.def (empty_mask_is_expensive): New hook.
7948 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7949 * doc/tm.texi: Regenerate.
7950 * targhooks.h (default_empty_mask_is_expensive): Declare.
7951 * targhooks.c (default_empty_mask_is_expensive): New function.
7952 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7953 if the target says that empty masks are expensive.
7954 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7956 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7959 Alan Hayward <alan.hayward@arm.com>
7960 David Sherwood <david.sherwood@arm.com>
7962 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7963 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7964 (vect_use_loop_mask_for_alignment_p): New function.
7965 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7966 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7967 niters_skip argument. Make sure that the first niters_skip elements
7968 of the first iteration are inactive.
7969 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7970 Update call to vect_set_loop_masks_directly.
7971 (get_misalign_in_elems): New function, split out from...
7972 (vect_gen_prolog_loop_niters): ...here.
7973 (vect_update_init_of_dr): Take a code argument that specifies whether
7974 the adjustment should be added or subtracted.
7975 (vect_update_init_of_drs): Likewise.
7976 (vect_prepare_for_masked_peels): New function.
7977 (vect_do_peeling): Skip prologue peeling if we're using a mask
7978 instead. Update call to vect_update_inits_of_drs.
7979 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7981 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7982 alignment. Do not include the number of peeled iterations in
7983 the minimum threshold in that case.
7984 (vectorizable_induction): Adjust the start value down by
7985 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7986 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7987 Take the number of skipped iterations into account when calculating
7989 * tree-vect-stmts.c (vect_gen_while_not): New function.
7991 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7992 Alan Hayward <alan.hayward@arm.com>
7993 David Sherwood <david.sherwood@arm.com>
7995 * doc/sourcebuild.texi (vect_fully_masked): Document.
7996 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7998 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
8000 (vect_analyze_loop_2): ...here. Don't check the vectorization
8001 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
8009 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
8010 (dump_groups): Update accordingly.
8011 (iv_use::mem_type): New member variable.
8012 (address_p): New function.
8013 (record_use): Add a mem_type argument and initialize the new
8015 (record_group_use): Add a mem_type argument. Use address_p.
8016 Remove obsolete null checks of base_object. Update call to record_use.
8017 (find_interesting_uses_op): Update call to record_group_use.
8018 (find_interesting_uses_cond): Likewise.
8019 (find_interesting_uses_address): Likewise.
8020 (get_mem_type_for_internal_fn): New function.
8021 (find_address_like_use): Likewise.
8022 (find_interesting_uses_stmt): Try find_address_like_use before
8023 calling find_interesting_uses_op.
8024 (addr_offset_valid_p): Use the iv mem_type field as the type
8025 of the addressed memory.
8026 (add_autoinc_candidates): Likewise.
8027 (get_address_cost): Likewise.
8028 (split_small_address_groups_p): Use address_p.
8029 (split_address_groups): Likewise.
8030 (add_iv_candidate_for_use): Likewise.
8031 (autoinc_possible_for_pair): Likewise.
8032 (rewrite_groups): Likewise.
8033 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
8034 (determine_group_iv_cost): Update after split of USE_ADDRESS.
8035 (get_alias_ptr_type_for_ptr_address): New function.
8036 (rewrite_use_address): Rewrite address uses in calls that were
8037 identified by find_address_like_use.
8039 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8040 Alan Hayward <alan.hayward@arm.com>
8041 David Sherwood <david.sherwood@arm.com>
8043 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
8045 * gimple-expr.h (is_gimple_addressable: Likewise.
8046 * gimple-expr.c (is_gimple_address): Likewise.
8047 * internal-fn.c (expand_call_mem_ref): New function.
8048 (expand_mask_load_optab_fn): Use it.
8049 (expand_mask_store_optab_fn): Likewise.
8051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8052 Alan Hayward <alan.hayward@arm.com>
8053 David Sherwood <david.sherwood@arm.com>
8055 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
8056 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
8057 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
8058 (cond_umax@var{mode}): Document.
8059 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8060 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8061 (cond_umin_optab, cond_umax_optab): New optabs.
8062 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8063 (COND_IOR, COND_XOR): New internal functions.
8064 * internal-fn.h (get_conditional_internal_fn): Declare.
8065 * internal-fn.c (cond_binary_direct): New macro.
8066 (expand_cond_binary_optab_fn): Likewise.
8067 (direct_cond_binary_optab_supported_p): Likewise.
8068 (get_conditional_internal_fn): New function.
8069 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8070 Cope with reduction statements that are vectorized as calls rather
8072 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8073 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8074 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8075 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8076 (UNSPEC_COND_EOR): New unspecs.
8077 (optab): Add mappings for them.
8078 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8079 (sve_int_op, sve_fp_op): New int attributes.
8081 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8082 Alan Hayward <alan.hayward@arm.com>
8083 David Sherwood <david.sherwood@arm.com>
8085 * optabs.def (while_ult_optab): New optab.
8086 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8087 * internal-fn.def (WHILE_ULT): New internal function.
8088 * internal-fn.h (direct_internal_fn_supported_p): New override
8089 that takes two types as argument.
8090 * internal-fn.c (while_direct): New macro.
8091 (expand_while_optab_fn): New function.
8092 (convert_optab_supported_p): Likewise.
8093 (direct_while_optab_supported_p): New macro.
8094 * wide-int.h (wi::udiv_ceil): New function.
8095 * tree-vectorizer.h (rgroup_masks): New structure.
8096 (vec_loop_masks): New typedef.
8097 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8099 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8100 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8101 (vect_max_vf): New function.
8102 (slpeel_make_loop_iterate_ntimes): Delete.
8103 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8104 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8105 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8106 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8107 internal-fn.h, stor-layout.h and optabs-query.h.
8108 (vect_set_loop_mask): New function.
8109 (add_preheader_seq): Likewise.
8110 (add_header_seq): Likewise.
8111 (interleave_supported_p): Likewise.
8112 (vect_maybe_permute_loop_masks): Likewise.
8113 (vect_set_loop_masks_directly): Likewise.
8114 (vect_set_loop_condition_masked): Likewise.
8115 (vect_set_loop_condition_unmasked): New function, split out from
8116 slpeel_make_loop_iterate_ntimes.
8117 (slpeel_make_loop_iterate_ntimes): Rename to..
8118 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8119 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8120 (vect_do_peeling): Update call accordingly.
8121 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8123 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8124 mask_compare_type, can_fully_mask_p and fully_masked_p.
8125 (release_vec_loop_masks): New function.
8126 (_loop_vec_info): Use it to free the loop masks.
8127 (can_produce_all_loop_masks_p): New function.
8128 (vect_get_max_nscalars_per_iter): Likewise.
8129 (vect_verify_full_masking): Likewise.
8130 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8131 retries, and free the mask rgroups before retrying. Check loop-wide
8132 reasons for disallowing fully-masked loops. Make the final decision
8133 about whether use a fully-masked loop or not.
8134 (vect_estimate_min_profitable_iters): Do not assume that peeling
8135 for the number of iterations will be needed for fully-masked loops.
8136 (vectorizable_reduction): Disable fully-masked loops.
8137 (vectorizable_live_operation): Likewise.
8138 (vect_halve_mask_nunits): New function.
8139 (vect_double_mask_nunits): Likewise.
8140 (vect_record_loop_mask): Likewise.
8141 (vect_get_loop_mask): Likewise.
8142 (vect_transform_loop): Handle the case in which the final loop
8143 iteration might handle a partial vector. Call vect_set_loop_condition
8144 instead of slpeel_make_loop_iterate_ntimes.
8145 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8146 (check_load_store_masking): New function.
8147 (prepare_load_store_mask): Likewise.
8148 (vectorizable_store): Handle fully-masked loops.
8149 (vectorizable_load): Likewise.
8150 (supportable_widening_operation): Use vect_halve_mask_nunits for
8152 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8153 (vect_gen_while): New function.
8154 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8155 (aarch64_uqdec<mode>): New insn.
8157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8158 Alan Hayward <alan.hayward@arm.com>
8159 David Sherwood <david.sherwood@arm.com>
8161 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8162 (reduc_xor_scal_optab): New optabs.
8163 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8164 (reduc_xor_scal_@var{m}): Document.
8165 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8166 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8168 * fold-const-call.c (fold_const_call): Handle them.
8169 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8170 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8171 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8172 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8173 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8174 (UNSPEC_XORV): New unspecs.
8175 (optab): Add entries for them.
8176 (BITWISEV): New int iterator.
8177 (bit_reduc_op): New int attributes.
8179 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8180 Alan Hayward <alan.hayward@arm.com>
8181 David Sherwood <david.sherwood@arm.com>
8183 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8184 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8185 * optabs.def (vec_shl_insert_optab): New optab.
8186 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8187 (duplicate_and_interleave): Likewise.
8188 * tree-vect-loop.c: Include internal-fn.h.
8189 (neutral_op_for_slp_reduction): New function, split out from
8190 get_initial_defs_for_reduction.
8191 (get_initial_def_for_reduction): Handle option 2 for variable-length
8192 vectors by loading the neutral value into a vector and then shifting
8193 the initial value into element 0.
8194 (get_initial_defs_for_reduction): Replace the code argument with
8195 the neutral value calculated by neutral_op_for_slp_reduction.
8196 Use gimple_build_vector for constant-length vectors.
8197 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8198 but the first group_size elements have a neutral value.
8199 Use duplicate_and_interleave otherwise.
8200 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8201 Update call to get_initial_defs_for_reduction. Handle SLP
8202 reductions for variable-length vectors by creating one vector
8203 result for each scalar result, with the elements associated
8204 with other scalar results stubbed out with the neutral value.
8205 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8206 Require IFN_VEC_SHL_INSERT for double reductions on
8207 variable-length vectors, or SLP reductions that have
8208 a neutral value. Require can_duplicate_and_interleave_p
8209 support for variable-length unchained SLP reductions if there
8210 is no neutral value, such as for MIN/MAX reductions. Also require
8211 the number of vector elements to be a multiple of the number of
8212 SLP statements when doing variable-length unchained SLP reductions.
8213 Update call to vect_create_epilog_for_reduction.
8214 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8215 and remove initial values.
8216 (duplicate_and_interleave): Make public.
8217 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8218 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8220 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8221 Alan Hayward <alan.hayward@arm.com>
8222 David Sherwood <david.sherwood@arm.com>
8224 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8225 (can_duplicate_and_interleave_p): New function.
8226 (vect_get_and_check_slp_defs): Take the vector of statements
8227 rather than just the current one. Remove excess parentheses.
8228 Restriction rejectinon of vect_constant_def and vect_external_def
8229 for variable-length vectors to boolean types, or types for which
8230 can_duplicate_and_interleave_p is false.
8231 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8232 (duplicate_and_interleave): New function.
8233 (vect_get_constant_vectors): Use gimple_build_vector for
8234 constant-length vectors and suitable variable-length constant
8235 vectors. Use duplicate_and_interleave for other variable-length
8236 vectors. Don't defer the update when inserting new statements.
8238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8239 Alan Hayward <alan.hayward@arm.com>
8240 David Sherwood <david.sherwood@arm.com>
8242 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8243 min_profitable_iters doesn't go negative.
8245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8246 Alan Hayward <alan.hayward@arm.com>
8247 David Sherwood <david.sherwood@arm.com>
8249 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8250 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8251 * optabs.def (vec_mask_load_lanes_optab): New optab.
8252 (vec_mask_store_lanes_optab): Likewise.
8253 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8254 (MASK_STORE_LANES): Likewise.
8255 * internal-fn.c (mask_load_lanes_direct): New macro.
8256 (mask_store_lanes_direct): Likewise.
8257 (expand_mask_load_optab_fn): Handle masked operations.
8258 (expand_mask_load_lanes_optab_fn): New macro.
8259 (expand_mask_store_optab_fn): Handle masked operations.
8260 (expand_mask_store_lanes_optab_fn): New macro.
8261 (direct_mask_load_lanes_optab_supported_p): Likewise.
8262 (direct_mask_store_lanes_optab_supported_p): Likewise.
8263 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8265 (vect_load_lanes_supported): Likewise.
8266 * tree-vect-data-refs.c (strip_conversion): New function.
8267 (can_group_stmts_p): Likewise.
8268 (vect_analyze_data_ref_accesses): Use it instead of checking
8269 for a pair of assignments.
8270 (vect_store_lanes_supported): Take a masked_p parameter.
8271 (vect_load_lanes_supported): Likewise.
8272 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8273 vect_store_lanes_supported and vect_load_lanes_supported.
8274 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8275 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8276 parameter. Don't allow gaps for masked accesses.
8277 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8278 and vect_load_lanes_supported.
8279 (get_load_store_type): Take a masked_p parameter and update
8280 call to get_group_load_store_type.
8281 (vectorizable_store): Update call to get_load_store_type.
8282 Handle IFN_MASK_STORE_LANES.
8283 (vectorizable_load): Update call to get_load_store_type.
8284 Handle IFN_MASK_LOAD_LANES.
8286 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8287 Alan Hayward <alan.hayward@arm.com>
8288 David Sherwood <david.sherwood@arm.com>
8290 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8292 * config/aarch64/aarch64-protos.h
8293 (aarch64_sve_struct_memory_operand_p): Declare.
8294 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8295 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8296 (VPRED, vpred): Handle SVE structure modes.
8297 * config/aarch64/constraints.md (Utx): New constraint.
8298 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8299 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8300 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8301 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8302 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8303 structure modes. Split into pieces after RA.
8304 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8305 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8307 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8308 SVE structure modes.
8309 (aarch64_classify_address): Likewise.
8310 (sizetochar): Move earlier in file.
8311 (aarch64_print_operand): Handle SVE register lists.
8312 (aarch64_array_mode): New function.
8313 (aarch64_sve_struct_memory_operand_p): Likewise.
8314 (TARGET_ARRAY_MODE): Redefine.
8316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8317 Alan Hayward <alan.hayward@arm.com>
8318 David Sherwood <david.sherwood@arm.com>
8320 * target.def (array_mode): New target hook.
8321 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8322 * doc/tm.texi: Regenerate.
8323 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8324 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8325 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8327 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8330 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8331 Alan Hayward <alan.hayward@arm.com>
8332 David Sherwood <david.sherwood@arm.com>
8334 * fold-const.c (fold_binary_loc): Check the argument types
8335 rather than the result type when testing for a vector operation.
8337 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8339 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8340 * doc/tm.texi: Regenerate.
8342 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8343 Alan Hayward <alan.hayward@arm.com>
8344 David Sherwood <david.sherwood@arm.com>
8346 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8347 (sve): Document new AArch64 extension.
8348 * doc/md.texi (w): Extend the description of the AArch64
8349 constraint to include SVE vectors.
8350 (Upl, Upa): Document new AArch64 predicate constraints.
8351 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8353 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8354 (msve-vector-bits=): New option.
8355 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8356 SVE when these are disabled.
8357 (sve): New extension.
8358 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8359 modes. Adjust their number of units based on aarch64_sve_vg.
8360 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8361 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8362 aarch64_addr_query_type.
8363 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8364 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8365 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8366 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8367 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8368 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8369 (aarch64_simd_imm_zero_p): Delete.
8370 (aarch64_check_zero_based_sve_index_immediate): Declare.
8371 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8372 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8373 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8374 (aarch64_sve_float_mul_immediate_p): Likewise.
8375 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8377 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8378 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8379 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8380 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8381 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8382 (aarch64_regmode_natural_size): Likewise.
8383 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8384 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8386 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8387 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8388 for VG and the SVE predicate registers.
8389 (V_ALIASES): Add a "z"-prefixed alias.
8390 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8391 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8392 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8393 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8394 (REG_CLASS_NAMES): Add entries for them.
8395 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8396 and the predicate registers.
8397 (aarch64_sve_vg): Declare.
8398 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8399 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8400 (REGMODE_NATURAL_SIZE): Define.
8401 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8403 * config/aarch64/aarch64.c: Include cfgrtl.h.
8404 (simd_immediate_info): Add a constructor for series vectors,
8405 and an associated step field.
8406 (aarch64_sve_vg): New variable.
8407 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8408 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8409 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8410 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8411 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8412 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8413 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8414 (aarch64_get_mask_mode): New functions.
8415 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8416 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8417 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8418 predicate modes and predicate registers. Explicitly restrict
8419 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8420 to store a vector mode if it is recognized by
8421 aarch64_classify_vector_mode.
8422 (aarch64_regmode_natural_size): New function.
8423 (aarch64_hard_regno_caller_save_mode): Return the original mode
8425 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8426 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8427 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8428 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8430 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8431 does not overlap dest if the function is frame-related. Handle
8433 (aarch64_split_add_offset): New function.
8434 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8435 them aarch64_add_offset.
8436 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8437 and update call to aarch64_sub_sp.
8438 (aarch64_add_cfa_expression): New function.
8439 (aarch64_expand_prologue): Pass extra temporary registers to the
8440 functions above. Handle the case in which we need to emit new
8441 DW_CFA_expressions for registers that were originally saved
8442 relative to the stack pointer, but now have to be expressed
8443 relative to the frame pointer.
8444 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8446 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8447 IP0 and IP1 values for SVE frames.
8448 (aarch64_expand_vec_series): New function.
8449 (aarch64_expand_sve_widened_duplicate): Likewise.
8450 (aarch64_expand_sve_const_vector): Likewise.
8451 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8452 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8453 into the register, rather than emitting a SET directly.
8454 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8455 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8456 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8457 (offset_9bit_signed_scaled_p): New functions.
8458 (aarch64_replicate_bitmask_imm): New function.
8459 (aarch64_bitmask_imm): Use it.
8460 (aarch64_cannot_force_const_mem): Reject expressions involving
8461 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8462 (aarch64_classify_index): Handle SVE indices, by requiring
8463 a plain register index with a scale that matches the element size.
8464 (aarch64_classify_address): Handle SVE addresses. Assert that
8465 the mode of the address is VOIDmode or an integer mode.
8466 Update call to aarch64_classify_symbol.
8467 (aarch64_classify_symbolic_expression): Update call to
8468 aarch64_classify_symbol.
8469 (aarch64_const_vec_all_in_range_p): New function.
8470 (aarch64_print_vector_float_operand): Likewise.
8471 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8472 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8473 and the FP immediates 1.0 and 0.5.
8474 (aarch64_print_address_internal): Handle SVE addresses.
8475 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8476 (aarch64_regno_regclass): Handle predicate registers.
8477 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8479 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8480 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8481 (aarch64_convert_sve_vector_bits): New function.
8482 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8483 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8485 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8486 Handle SVE vector and predicate modes. Accept VL-based constants
8487 that need only one temporary register, and VL offsets that require
8488 no temporary registers.
8489 (aarch64_conditional_register_usage): Mark the predicate registers
8490 as fixed if SVE isn't available.
8491 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8492 Return true for SVE vector and predicate modes.
8493 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8494 rather than an unsigned int. Handle SVE modes.
8495 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8497 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8499 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8500 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8501 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8502 (aarch64_sve_float_mul_immediate_p): New functions.
8503 (aarch64_sve_valid_immediate): New function.
8504 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8505 Explicitly reject structure modes. Check for INDEX constants.
8506 Handle PTRUE and PFALSE constants.
8507 (aarch64_check_zero_based_sve_index_immediate): New function.
8508 (aarch64_simd_imm_zero_p): Delete.
8509 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8510 vector modes. Accept constants in the range of CNT[BHWD].
8511 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8512 ask for an Advanced SIMD mode.
8513 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8514 (aarch64_simd_vector_alignment): Handle SVE predicates.
8515 (aarch64_vectorize_preferred_vector_alignment): New function.
8516 (aarch64_simd_vector_alignment_reachable): Use it instead of
8518 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8519 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8521 (MAX_VECT_LEN): Delete.
8522 (expand_vec_perm_d): Add a vec_flags field.
8523 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8524 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8525 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8527 (aarch64_evpc_rev): Rename to...
8528 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8529 (aarch64_evpc_rev_global): New function.
8530 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8531 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8533 (aarch64_evpc_sve_tbl): New function.
8534 (aarch64_expand_vec_perm_const_1): Update after rename of
8535 aarch64_evpc_rev. Handle SVE permutes too, trying
8536 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8537 than aarch64_evpc_tbl.
8538 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8539 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8540 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8541 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8542 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8543 (aarch64_expand_sve_vcond): New functions.
8544 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8545 of aarch64_vector_mode_p.
8546 (aarch64_dwarf_poly_indeterminate_value): New function.
8547 (aarch64_compute_pressure_classes): Likewise.
8548 (aarch64_can_change_mode_class): Likewise.
8549 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8550 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8551 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8552 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8553 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8554 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8555 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8556 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8558 (Dn, Dl, Dr): Accept const as well as const_vector.
8559 (Dz): Likewise. Compare against CONST0_RTX.
8560 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8561 of "vector" where appropriate.
8562 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8563 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8564 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8565 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8566 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8567 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8568 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8569 (v_int_equiv): Extend to SVE modes.
8570 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8572 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8573 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8574 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8575 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8576 (SVE_COND_FP_CMP): New int iterators.
8577 (perm_hilo): Handle the new unpack unspecs.
8578 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8580 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8581 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8582 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8583 (aarch64_equality_operator, aarch64_constant_vector_operand)
8584 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8585 (aarch64_sve_nonimmediate_operand): Likewise.
8586 (aarch64_sve_general_operand): Likewise.
8587 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8588 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8589 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8590 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8591 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8592 (aarch64_sve_float_arith_immediate): Likewise.
8593 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8594 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8595 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8596 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8597 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8598 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8599 (aarch64_sve_float_arith_operand): Likewise.
8600 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8601 (aarch64_sve_float_mul_operand): Likewise.
8602 (aarch64_sve_vec_perm_operand): Likewise.
8603 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8604 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8605 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8606 as well as const_vector.
8607 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8608 in file. Use CONST0_RTX and CONSTM1_RTX.
8609 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8610 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8611 Use aarch64_simd_imm_zero.
8612 * config/aarch64/aarch64-sve.md: New file.
8613 * config/aarch64/aarch64.md: Include it.
8614 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8615 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8616 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8617 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8618 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8619 (sve): New attribute.
8620 (enabled): Disable instructions with the sve attribute unless
8622 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8623 aarch64_expand_mov_immediate.
8624 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8625 CNT[BHSD] immediates.
8626 (movti): Split CONST_POLY_INT moves into two halves.
8627 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8628 Split additions that need a temporary here if the destination
8629 is the stack pointer.
8630 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8631 (*add<mode>3_poly_1): New instruction.
8632 (set_clobber_cc): New expander.
8634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8636 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8637 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8638 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8639 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8640 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8641 Change innermode from fixed_mode_size to machine_mode.
8642 (simplify_subreg): Update call accordingly. Handle a constant-sized
8643 subreg of a variable-length CONST_VECTOR.
8645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8646 Alan Hayward <alan.hayward@arm.com>
8647 David Sherwood <david.sherwood@arm.com>
8649 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8650 (add_offset_to_base): New function, split out from...
8651 (create_mem_ref): ...here. When handling a scale other than 1,
8652 check first whether the address is valid without the offset.
8653 Add it into the base if so, leaving the index and scale as-is.
8655 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8658 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8659 fold_for_warn before checking if arg2 is INTEGER_CST.
8661 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8663 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8664 (store_multiple_operation): Delete.
8665 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8666 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8667 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8668 guarded by TARGET_STRING.
8669 (rs6000_output_load_multiple): Delete.
8670 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8671 OPTION_MASK_STRING / TARGET_STRING handling.
8672 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8673 (const rs6000_opt_masks) <"string">: Change mask to 0.
8674 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8675 (MASK_STRING): Delete.
8676 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8678 (load_multiple): Delete.
8685 (store_multiple): Delete.
8692 (movmemsi_8reg): Delete.
8693 (corresponding unnamed define_insn): Delete.
8694 (movmemsi_6reg): Delete.
8695 (corresponding unnamed define_insn): Delete.
8696 (movmemsi_4reg): Delete.
8697 (corresponding unnamed define_insn): Delete.
8698 (movmemsi_2reg): Delete.
8699 (corresponding unnamed define_insn): Delete.
8700 (movmemsi_1reg): Delete.
8701 (corresponding unnamed define_insn): Delete.
8702 * config/rs6000/rs6000.opt (mno-string): New.
8703 (mstring): Replace by deprecation warning stub.
8704 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8706 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8708 * regrename.c (regrename_do_replace): If replacing the same
8709 reg multiple times, try to reuse last created gen_raw_REG.
8712 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8713 main to workaround a bug in GDB.
8715 2018-01-12 Tom de Vries <tom@codesourcery.com>
8718 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8720 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8722 PR rtl-optimization/80481
8723 * ira-color.c (get_cap_member): New function.
8724 (allocnos_conflict_by_live_ranges_p): Use it.
8725 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8726 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8728 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8731 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8732 (*saddl_se_1): Ditto.
8734 (*ssubl_se_1): Ditto.
8736 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8738 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8739 rather than wi::to_widest for DR_INITs.
8740 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8741 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8742 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8744 (vect_analyze_group_access_1): Note that here.
8746 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8748 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8749 polynomial type sizes.
8751 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8753 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8754 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8755 (gimple_add_tmp_var): Likewise.
8757 2018-01-12 Martin Liska <mliska@suse.cz>
8759 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8760 (gimple_alloc_sizes): Likewise.
8761 (dump_gimple_statistics): Use PRIu64 in printf format.
8762 * gimple.h: Change uint64_t to int.
8764 2018-01-12 Martin Liska <mliska@suse.cz>
8766 * tree-core.h: Use uint64_t instead of int.
8767 * tree.c (tree_node_counts): Likewise.
8768 (tree_node_sizes): Likewise.
8769 (dump_tree_statistics): Use PRIu64 in printf format.
8771 2018-01-12 Martin Liska <mliska@suse.cz>
8773 * Makefile.in: As qsort_chk is implemented in vec.c, add
8774 vec.o to linkage of gencfn-macros.
8775 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8776 passing the info to record_node_allocation_statistics.
8777 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8779 * ggc-common.c (struct ggc_usage): Add operator== and use
8780 it in operator< and compare function.
8781 * mem-stats.h (struct mem_usage): Likewise.
8782 * vec.c (struct vec_usage): Remove operator< and compare
8783 function. Can be simply inherited.
8785 2018-01-12 Martin Jambor <mjambor@suse.cz>
8788 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8789 * tree-ssa-math-opts.c: Include domwalk.h.
8790 (convert_mult_to_fma_1): New function.
8791 (fma_transformation_info): New type.
8792 (fma_deferring_state): Likewise.
8793 (cancel_fma_deferring): New function.
8794 (result_of_phi): Likewise.
8795 (last_fma_candidate_feeds_initial_phi): Likewise.
8796 (convert_mult_to_fma): Added deferring logic, split actual
8797 transformation to convert_mult_to_fma_1.
8798 (math_opts_dom_walker): New type.
8799 (math_opts_dom_walker::after_dom_children): New method, body moved
8800 here from pass_optimize_widening_mul::execute, added deferring logic
8802 (pass_optimize_widening_mul::execute): Moved most of code to
8803 math_opts_dom_walker::after_dom_children.
8804 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8805 * config/i386/i386.c (ix86_option_override_internal): Added
8806 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8808 2018-01-12 Richard Biener <rguenther@suse.de>
8811 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8812 inline instance vars.
8814 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8817 * config/rx/rx.c (rx_is_restricted_memory_address):
8820 2018-01-12 Richard Biener <rguenther@suse.de>
8822 PR tree-optimization/80846
8823 * target.def (split_reduction): New target hook.
8824 * targhooks.c (default_split_reduction): New function.
8825 * targhooks.h (default_split_reduction): Declare.
8826 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8827 target requests first reduce vectors by combining low and high
8829 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8830 (get_vectype_for_scalar_type_and_size): Export.
8831 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8832 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8833 * doc/tm.texi: Regenerate.
8834 * config/i386/i386.c (ix86_split_reduction): Implement
8835 TARGET_VECTORIZE_SPLIT_REDUCTION.
8837 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8840 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8841 in PIC mode except for TARGET_VXWORKS_RTP.
8842 * config/sparc/sparc.c: Include cfgrtl.h.
8843 (TARGET_INIT_PIC_REG): Define.
8844 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8845 (sparc_pic_register_p): New predicate.
8846 (sparc_legitimate_address_p): Use it.
8847 (sparc_legitimize_pic_address): Likewise.
8848 (sparc_delegitimize_address): Likewise.
8849 (sparc_mode_dependent_address_p): Likewise.
8850 (gen_load_pcrel_sym): Remove 4th parameter.
8851 (load_got_register): Adjust call to above. Remove obsolete stuff.
8852 (sparc_expand_prologue): Do not call load_got_register here.
8853 (sparc_flat_expand_prologue): Likewise.
8854 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8855 (sparc_use_pseudo_pic_reg): New function.
8856 (sparc_init_pic_reg): Likewise.
8857 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8858 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8860 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8862 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8863 Add item for branch_cost.
8865 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8867 PR rtl-optimization/83565
8868 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8869 not extend the result to a larger mode for rotate operations.
8870 (num_sign_bit_copies1): Likewise.
8872 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8875 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8877 Use values-Xc.o for -pedantic.
8878 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8880 2018-01-12 Martin Liska <mliska@suse.cz>
8883 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8885 (possible_polymorphic_call_targets): Use it.
8886 (ipa_devirt): Likewise.
8888 2018-01-12 Martin Liska <mliska@suse.cz>
8890 * profile-count.h (enum profile_quality): Use 0 as invalid
8891 enum value of profile_quality.
8893 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8895 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8896 -mext-string options.
8898 2018-01-12 Richard Biener <rguenther@suse.de>
8900 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8901 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8902 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8904 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8906 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8908 * configure.ac (--with-long-double-format): Add support for the
8909 configuration option to change the default long double format on
8911 * config.gcc (powerpc*-linux*-*): Likewise.
8912 * configure: Regenerate.
8913 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8914 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8915 used without modification.
8917 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8919 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8920 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8921 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8922 MISC_BUILTIN_SPEC_BARRIER.
8923 (rs6000_init_builtins): Likewise.
8924 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8926 (speculation_barrier): New define_insn.
8927 * doc/extend.texi: Document __builtin_speculation_barrier.
8929 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8932 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8933 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8934 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8936 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8937 integral modes instead of "ss" and "sd".
8938 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8939 vectors with 32-bit and 64-bit elements.
8940 (vecdupssescalarmodesuffix): New mode attribute.
8941 (vec_dup<mode>): Use it.
8943 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8946 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8947 frame if argument is passed on stack.
8949 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8952 * ree.c (combine_reaching_defs): Optimize also
8953 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8954 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8956 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8959 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8961 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8964 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8965 after they are computed.
8967 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8969 PR tree-optimization/83695
8970 * gimple-loop-linterchange.cc
8971 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8972 reset cached scev information after interchange.
8973 (pass_linterchange::execute): Remove call to scev_reset_htab.
8975 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8977 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8978 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8979 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8980 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8981 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8982 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8983 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8984 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8985 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8986 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8987 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8988 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8989 (V_lane_reg): Likewise.
8990 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8992 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8993 (vfmal_lane_low<mode>_intrinsic,
8994 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8995 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8996 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8997 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8998 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8999 vfmsl_lane_high<mode>_intrinsic): New define_insns.
9001 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9003 * config/arm/arm-cpus.in (fp16fml): New feature.
9004 (ALL_SIMD): Add fp16fml.
9005 (armv8.2-a): Add fp16fml as an option.
9006 (armv8.3-a): Likewise.
9007 (armv8.4-a): Add fp16fml as part of fp16.
9008 * config/arm/arm.h (TARGET_FP16FML): Define.
9009 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
9011 * config/arm/arm-modes.def (V2HF): Define.
9012 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
9013 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
9014 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
9015 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
9016 vfmsl_low, vfmsl_high): New set of builtins.
9017 * config/arm/iterators.md (PLUSMINUS): New code iterator.
9018 (vfml_op): New code attribute.
9019 (VFMLHALVES): New int iterator.
9020 (VFML, VFMLSEL): New mode attributes.
9021 (V_reg): Define mapping for V2HF.
9022 (V_hi, V_lo): New mode attributes.
9023 (VF_constraint): Likewise.
9024 (vfml_half, vfml_half_selector): New int attributes.
9025 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
9027 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
9028 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
9030 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
9031 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
9032 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
9033 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
9035 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
9036 Document new effective target and option set.
9038 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9040 * config/arm/arm-cpus.in (armv8_4): New feature.
9041 (ARMv8_4a): New fgroup.
9042 (armv8.4-a): New arch.
9043 * config/arm/arm-tables.opt: Regenerate.
9044 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
9045 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
9046 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
9047 Add matching rules for -march=armv8.4-a and extensions.
9048 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
9050 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
9053 * config/rx/rx.md (BW): New mode attribute.
9054 (sync_lock_test_and_setsi): Add mode suffix to insn output.
9056 2018-01-11 Richard Biener <rguenther@suse.de>
9058 PR tree-optimization/83435
9059 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9060 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9061 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9063 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9064 Alan Hayward <alan.hayward@arm.com>
9065 David Sherwood <david.sherwood@arm.com>
9067 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9069 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9070 (aarch64_print_address_internal): Use it to check for a zero offset.
9072 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9073 Alan Hayward <alan.hayward@arm.com>
9074 David Sherwood <david.sherwood@arm.com>
9076 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9077 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9078 Return a poly_int64 rather than a HOST_WIDE_INT.
9079 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9080 rather than a HOST_WIDE_INT.
9081 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9082 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9083 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9084 final_offset from HOST_WIDE_INT to poly_int64.
9085 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9086 to_constant when getting the number of units in an Advanced SIMD
9088 (aarch64_builtin_vectorized_function): Check for a constant number
9090 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9092 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9093 attribute instead of GET_MODE_NUNITS.
9094 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9095 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9096 GET_MODE_SIZE for fixed-size registers.
9097 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9098 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9099 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9100 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9101 (aarch64_print_operand, aarch64_print_address_internal)
9102 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9103 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9104 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9105 Handle polynomial GET_MODE_SIZE.
9106 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9107 wider than SImode without modification.
9108 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9109 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9110 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9111 passing and returning SVE modes.
9112 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9113 rather than GEN_INT.
9114 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9115 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9116 (aarch64_allocate_and_probe_stack_space): Likewise.
9117 (aarch64_layout_frame): Cope with polynomial offsets.
9118 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9119 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9121 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9122 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9123 poly_int64 rather than a HOST_WIDE_INT.
9124 (aarch64_get_separate_components, aarch64_process_components)
9125 (aarch64_expand_prologue, aarch64_expand_epilogue)
9126 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9127 (aarch64_anchor_offset): New function, split out from...
9128 (aarch64_legitimize_address): ...here.
9129 (aarch64_builtin_vectorization_cost): Handle polynomial
9130 TYPE_VECTOR_SUBPARTS.
9131 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9133 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9134 number of elements from the PARALLEL rather than the mode.
9135 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9136 rather than GET_MODE_BITSIZE.
9137 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9138 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9139 (aarch64_expand_vec_perm_const_1): Handle polynomial
9140 d->perm.length () and d->perm elements.
9141 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9142 Apply to_constant to d->perm elements.
9143 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9144 polynomial CONST_VECTOR_NUNITS.
9145 (aarch64_move_pointer): Take amount as a poly_int64 rather
9147 (aarch64_progress_pointer): Avoid temporary variable.
9148 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9149 the mode attribute instead of GET_MODE.
9151 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9152 Alan Hayward <alan.hayward@arm.com>
9153 David Sherwood <david.sherwood@arm.com>
9155 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9156 x exists before using it.
9157 (aarch64_add_constant_internal): Rename to...
9158 (aarch64_add_offset_1): ...this. Replace regnum with separate
9159 src and dest rtxes. Handle the case in which they're different,
9160 including when the offset is zero. Replace scratchreg with an rtx.
9161 Use 2 additions if there is no spare register into which we can
9162 move a 16-bit constant.
9163 (aarch64_add_constant): Delete.
9164 (aarch64_add_offset): Replace reg with separate src and dest
9165 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9166 Use aarch64_add_offset_1.
9167 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9168 an rtx rather than an int. Take the delta as a poly_int64
9169 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9170 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9171 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9172 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9173 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9175 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9176 aarch64_add_constant.
9178 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9180 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9181 Use scalar_float_mode.
9183 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9185 * config/aarch64/aarch64-simd.md
9186 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9187 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9188 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9189 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9190 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9191 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9192 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9193 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9194 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9195 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9197 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9200 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9201 targ_options->x_arm_arch_string is non NULL.
9203 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9205 * config/aarch64/aarch64.h
9206 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9208 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9211 * expmed.c (emit_store_flag_force): Swap if const op0
9212 and change VOIDmode to mode of op0.
9214 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9216 PR rtl-optimization/83761
9217 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9218 than bytes to mode_for_size.
9220 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9223 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9224 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9227 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9230 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9231 when in layout mode.
9232 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9233 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9236 2018-01-10 Michael Collison <michael.collison@arm.com>
9238 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9239 * config/aarch64/aarch64-option-extension.def: Add
9240 AARCH64_OPT_EXTENSION of 'fp16fml'.
9241 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9242 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9243 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9244 * config/aarch64/constraints.md (Ui7): New constraint.
9245 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9246 (VFMLA_SEL_W): Ditto.
9249 (VFMLA16_LOW): New int iterator.
9250 (VFMLA16_HIGH): Ditto.
9251 (UNSPEC_FMLAL): New unspec.
9252 (UNSPEC_FMLSL): Ditto.
9253 (UNSPEC_FMLAL2): Ditto.
9254 (UNSPEC_FMLSL2): Ditto.
9255 (f16mac): New code attribute.
9256 * config/aarch64/aarch64-simd-builtins.def
9257 (aarch64_fmlal_lowv2sf): Ditto.
9258 (aarch64_fmlsl_lowv2sf): Ditto.
9259 (aarch64_fmlalq_lowv4sf): Ditto.
9260 (aarch64_fmlslq_lowv4sf): Ditto.
9261 (aarch64_fmlal_highv2sf): Ditto.
9262 (aarch64_fmlsl_highv2sf): Ditto.
9263 (aarch64_fmlalq_highv4sf): Ditto.
9264 (aarch64_fmlslq_highv4sf): Ditto.
9265 (aarch64_fmlal_lane_lowv2sf): Ditto.
9266 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9267 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9268 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9269 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9270 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9271 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9272 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9273 (aarch64_fmlal_lane_highv2sf): Ditto.
9274 (aarch64_fmlsl_lane_highv2sf): Ditto.
9275 (aarch64_fmlal_laneq_highv2sf): Ditto.
9276 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9277 (aarch64_fmlalq_lane_highv4sf): Ditto.
9278 (aarch64_fmlsl_lane_highv4sf): Ditto.
9279 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9280 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9281 * config/aarch64/aarch64-simd.md:
9282 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9283 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9284 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9285 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9286 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9287 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9288 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9289 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9290 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9291 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9292 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9293 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9294 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9295 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9296 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9297 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9298 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9299 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9300 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9301 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9302 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9303 (vfmlsl_low_u32): Ditto.
9304 (vfmlalq_low_u32): Ditto.
9305 (vfmlslq_low_u32): Ditto.
9306 (vfmlal_high_u32): Ditto.
9307 (vfmlsl_high_u32): Ditto.
9308 (vfmlalq_high_u32): Ditto.
9309 (vfmlslq_high_u32): Ditto.
9310 (vfmlal_lane_low_u32): Ditto.
9311 (vfmlsl_lane_low_u32): Ditto.
9312 (vfmlal_laneq_low_u32): Ditto.
9313 (vfmlsl_laneq_low_u32): Ditto.
9314 (vfmlalq_lane_low_u32): Ditto.
9315 (vfmlslq_lane_low_u32): Ditto.
9316 (vfmlalq_laneq_low_u32): Ditto.
9317 (vfmlslq_laneq_low_u32): Ditto.
9318 (vfmlal_lane_high_u32): Ditto.
9319 (vfmlsl_lane_high_u32): Ditto.
9320 (vfmlal_laneq_high_u32): Ditto.
9321 (vfmlsl_laneq_high_u32): Ditto.
9322 (vfmlalq_lane_high_u32): Ditto.
9323 (vfmlslq_lane_high_u32): Ditto.
9324 (vfmlalq_laneq_high_u32): Ditto.
9325 (vfmlslq_laneq_high_u32): Ditto.
9326 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9327 (AARCH64_FL_FOR_ARCH8_4): New.
9328 (AARCH64_ISA_F16FML): New ISA flag.
9329 (TARGET_F16FML): New feature flag for fp16fml.
9330 (doc/invoke.texi): Document new fp16fml option.
9332 2018-01-10 Michael Collison <michael.collison@arm.com>
9334 * config/aarch64/aarch64-builtins.c:
9335 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9336 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9337 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9338 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9339 (AARCH64_ISA_SHA3): New ISA flag.
9340 (TARGET_SHA3): New feature flag for sha3.
9341 * config/aarch64/iterators.md (sha512_op): New int attribute.
9342 (CRYPTO_SHA512): New int iterator.
9343 (UNSPEC_SHA512H): New unspec.
9344 (UNSPEC_SHA512H2): Ditto.
9345 (UNSPEC_SHA512SU0): Ditto.
9346 (UNSPEC_SHA512SU1): Ditto.
9347 * config/aarch64/aarch64-simd-builtins.def
9348 (aarch64_crypto_sha512hqv2di): New builtin.
9349 (aarch64_crypto_sha512h2qv2di): Ditto.
9350 (aarch64_crypto_sha512su0qv2di): Ditto.
9351 (aarch64_crypto_sha512su1qv2di): Ditto.
9352 (aarch64_eor3qv8hi): Ditto.
9353 (aarch64_rax1qv2di): Ditto.
9354 (aarch64_xarqv2di): Ditto.
9355 (aarch64_bcaxqv8hi): Ditto.
9356 * config/aarch64/aarch64-simd.md:
9357 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9358 (aarch64_crypto_sha512su0qv2di): Ditto.
9359 (aarch64_crypto_sha512su1qv2di): Ditto.
9360 (aarch64_eor3qv8hi): Ditto.
9361 (aarch64_rax1qv2di): Ditto.
9362 (aarch64_xarqv2di): Ditto.
9363 (aarch64_bcaxqv8hi): Ditto.
9364 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9365 (vsha512h2q_u64): Ditto.
9366 (vsha512su0q_u64): Ditto.
9367 (vsha512su1q_u64): Ditto.
9368 (veor3q_u16): Ditto.
9369 (vrax1q_u64): Ditto.
9371 (vbcaxq_u16): Ditto.
9372 * config/arm/types.md (crypto_sha512): New type attribute.
9373 (crypto_sha3): Ditto.
9374 (doc/invoke.texi): Document new sha3 option.
9376 2018-01-10 Michael Collison <michael.collison@arm.com>
9378 * config/aarch64/aarch64-builtins.c:
9379 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9380 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9381 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9382 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9383 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9384 (AARCH64_ISA_SM4): New ISA flag.
9385 (TARGET_SM4): New feature flag for sm4.
9386 * config/aarch64/aarch64-simd-builtins.def
9387 (aarch64_sm3ss1qv4si): Ditto.
9388 (aarch64_sm3tt1aq4si): Ditto.
9389 (aarch64_sm3tt1bq4si): Ditto.
9390 (aarch64_sm3tt2aq4si): Ditto.
9391 (aarch64_sm3tt2bq4si): Ditto.
9392 (aarch64_sm3partw1qv4si): Ditto.
9393 (aarch64_sm3partw2qv4si): Ditto.
9394 (aarch64_sm4eqv4si): Ditto.
9395 (aarch64_sm4ekeyqv4si): Ditto.
9396 * config/aarch64/aarch64-simd.md:
9397 (aarch64_sm3ss1qv4si): Ditto.
9398 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9399 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9400 (aarch64_sm4eqv4si): Ditto.
9401 (aarch64_sm4ekeyqv4si): Ditto.
9402 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9403 (sm3part_op): Ditto.
9404 (CRYPTO_SM3TT): Ditto.
9405 (CRYPTO_SM3PART): Ditto.
9406 (UNSPEC_SM3SS1): New unspec.
9407 (UNSPEC_SM3TT1A): Ditto.
9408 (UNSPEC_SM3TT1B): Ditto.
9409 (UNSPEC_SM3TT2A): Ditto.
9410 (UNSPEC_SM3TT2B): Ditto.
9411 (UNSPEC_SM3PARTW1): Ditto.
9412 (UNSPEC_SM3PARTW2): Ditto.
9413 (UNSPEC_SM4E): Ditto.
9414 (UNSPEC_SM4EKEY): Ditto.
9415 * config/aarch64/constraints.md (Ui2): New constraint.
9416 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9417 * config/arm/types.md (crypto_sm3): New type attribute.
9418 (crypto_sm4): Ditto.
9419 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9420 (vsm3tt1aq_u32): Ditto.
9421 (vsm3tt1bq_u32): Ditto.
9422 (vsm3tt2aq_u32): Ditto.
9423 (vsm3tt2bq_u32): Ditto.
9424 (vsm3partw1q_u32): Ditto.
9425 (vsm3partw2q_u32): Ditto.
9426 (vsm4eq_u32): Ditto.
9427 (vsm4ekeyq_u32): Ditto.
9428 (doc/invoke.texi): Document new sm4 option.
9430 2018-01-10 Michael Collison <michael.collison@arm.com>
9432 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9433 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9434 (AARCH64_FL_FOR_ARCH8_4): New.
9435 (AARCH64_FL_V8_4): New flag.
9436 (doc/invoke.texi): Document new armv8.4-a option.
9438 2018-01-10 Michael Collison <michael.collison@arm.com>
9440 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9441 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9442 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9443 * config/aarch64/aarch64-option-extension.def: Add
9444 AARCH64_OPT_EXTENSION of 'sha2'.
9445 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9446 (crypto): Disable sha2 and aes if crypto disabled.
9447 (crypto): Enable aes and sha2 if enabled.
9448 (simd): Disable sha2 and aes if simd disabled.
9449 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9451 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9452 (TARGET_SHA2): New feature flag for sha2.
9453 (TARGET_AES): New feature flag for aes.
9454 * config/aarch64/aarch64-simd.md:
9455 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9456 conditional on TARGET_AES.
9457 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9458 (aarch64_crypto_sha1hsi): Make pattern conditional
9460 (aarch64_crypto_sha1hv4si): Ditto.
9461 (aarch64_be_crypto_sha1hv4si): Ditto.
9462 (aarch64_crypto_sha1su1v4si): Ditto.
9463 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9464 (aarch64_crypto_sha1su0v4si): Ditto.
9465 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9466 (aarch64_crypto_sha256su0v4si): Ditto.
9467 (aarch64_crypto_sha256su1v4si): Ditto.
9468 (doc/invoke.texi): Document new aes and sha2 options.
9470 2018-01-10 Martin Sebor <msebor@redhat.com>
9472 PR tree-optimization/83781
9473 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9476 2018-01-11 Martin Sebor <msebor@gmail.com>
9477 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9479 PR tree-optimization/83501
9480 PR tree-optimization/81703
9482 * tree-ssa-strlen.c (get_string_cst): Rename...
9483 (get_string_len): ...to this. Handle global constants.
9484 (handle_char_store): Adjust.
9486 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9487 Jim Wilson <jimw@sifive.com>
9489 * config/riscv/riscv-protos.h (riscv_output_return): New.
9490 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9491 (riscv_attribute_table, riscv_output_return),
9492 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9493 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9494 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9496 (riscv_expand_prologue): Add early return for naked function.
9497 (riscv_expand_epilogue): Likewise.
9498 (riscv_function_ok_for_sibcall): Return false for naked function.
9499 (riscv_set_current_function): New.
9500 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9501 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9502 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9503 * doc/extend.texi (RISC-V Function Attributes): New.
9505 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9507 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9508 check for 128-bit long double before checking TCmode.
9509 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9510 128-bit long doubles before checking TFmode or TCmode.
9511 (FLOAT128_IBM_P): Likewise.
9513 2018-01-10 Martin Sebor <msebor@redhat.com>
9515 PR tree-optimization/83671
9516 * builtins.c (c_strlen): Unconditionally return zero for the empty
9518 Use -Warray-bounds for warnings.
9519 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9520 for non-constant array indices with COMPONENT_REF, arrays of
9521 arrays, and pointers to arrays.
9522 (gimple_fold_builtin_strlen): Determine and set length range for
9523 non-constant character arrays.
9525 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9528 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9531 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9533 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9535 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9538 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9539 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9540 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9541 indexed_or_indirect_operand predicate.
9542 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9543 (*vsx_le_perm_load_v8hi): Likewise.
9544 (*vsx_le_perm_load_v16qi): Likewise.
9545 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9546 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9547 (*vsx_le_perm_store_v8hi): Likewise.
9548 (*vsx_le_perm_store_v16qi): Likewise.
9549 (eight unnamed splitters): Likewise.
9551 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9553 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9554 * config/rs6000/emmintrin.h: Likewise.
9555 * config/rs6000/mmintrin.h: Likewise.
9556 * config/rs6000/xmmintrin.h: Likewise.
9558 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9561 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9563 * tree.c (tree_nop_conversion): Return true for location wrapper
9565 (maybe_wrap_with_location): New function.
9566 (selftest::check_strip_nops): New function.
9567 (selftest::test_location_wrappers): New function.
9568 (selftest::tree_c_tests): Call it.
9569 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9570 (maybe_wrap_with_location): New decl.
9571 (EXPR_LOCATION_WRAPPER_P): New macro.
9572 (location_wrapper_p): New inline function.
9573 (tree_strip_any_location_wrapper): New inline function.
9575 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9578 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9579 stack_realign_offset for the largest alignment of stack slot
9581 (ix86_find_max_used_stack_alignment): New function.
9582 (ix86_finalize_stack_frame_flags): Use it. Set
9583 max_used_stack_alignment if we don't realign stack.
9584 * config/i386/i386.h (machine_function): Add
9585 max_used_stack_alignment.
9587 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9589 * config/arm/arm.opt (-mbranch-cost): New option.
9590 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9593 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9596 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9597 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9599 2018-01-10 Richard Biener <rguenther@suse.de>
9602 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9603 early out so it also covers the case where we have a non-NULL
9606 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9608 PR tree-optimization/83753
9609 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9610 for non-strided grouped accesses if the number of elements is 1.
9612 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9615 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9616 * i386.h (TARGET_USE_GATHER): Define.
9617 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9619 2018-01-10 Martin Liska <mliska@suse.cz>
9622 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9623 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9625 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9626 CLEANUP_NO_PARTITIONING is not set.
9628 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9630 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9631 for vectors, as a partial revert of r254296.
9632 * rtl.h (const_vec_p): Delete.
9633 (const_vec_duplicate_p): Don't test for vector CONSTs.
9634 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9635 * expmed.c (make_tree): Likewise.
9638 * common.md (E, F): Use CONSTANT_P instead of checking for
9640 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9641 checking for CONST_VECTOR.
9643 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9646 * predict.c (force_edge_cold): Handle in more sane way edges
9649 2018-01-09 Carl Love <cel@us.ibm.com>
9651 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9653 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9654 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9655 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9656 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9657 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9658 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9659 * config/rs6000/rs6000-protos.h: Add extern defition for
9660 rs6000_generate_float2_double_code.
9661 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9663 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9664 (float2_v2df): Add define_expand.
9666 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9669 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9670 op_mode in the force_to_mode call.
9672 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9674 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9675 instead of checking each element individually.
9676 (aarch64_evpc_uzp): Likewise.
9677 (aarch64_evpc_zip): Likewise.
9678 (aarch64_evpc_ext): Likewise.
9679 (aarch64_evpc_rev): Likewise.
9680 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9681 instead of checking each element individually. Return true without
9683 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9684 whether all selected elements come from the same input, instead of
9685 checking each element individually. Remove calls to gen_rtx_REG,
9686 start_sequence and end_sequence and instead assert that no rtl is
9689 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9691 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9692 order of HIGH and CONST checks.
9694 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9696 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9697 if the destination isn't an SSA_NAME.
9699 2018-01-09 Richard Biener <rguenther@suse.de>
9701 PR tree-optimization/83668
9702 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9704 (canonicalize_loop_form): ... here, renamed from ...
9705 (canonicalize_loop_closed_ssa_form): ... this and amended to
9706 swap successor edges for loop exit blocks to make us use
9707 the RPO order we need for initial schedule generation.
9709 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9711 PR tree-optimization/64811
9712 * match.pd: When optimizing comparisons with Inf, avoid
9713 introducing or losing exceptions from comparisons with NaN.
9715 2018-01-09 Martin Liska <mliska@suse.cz>
9718 * asan.c (shadow_mem_size): Add gcc_assert.
9720 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9722 Don't save registers in main().
9725 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9726 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9727 * config/avr/avr.c (avr_set_current_function): Don't error if
9728 naked, OS_task or OS_main are specified at the same time.
9729 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9731 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9733 * common/config/avr/avr-common.c (avr_option_optimization_table):
9734 Switch on -mmain-is-OS_task for optimizing compilations.
9736 2018-01-09 Richard Biener <rguenther@suse.de>
9738 PR tree-optimization/83572
9739 * graphite.c: Include cfganal.h.
9740 (graphite_transform_loops): Connect infinite loops to exit
9741 and remove fake edges at the end.
9743 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9745 * ipa-inline.c (edge_badness): Revert accidental checkin.
9747 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9750 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9751 symbols; not inline clones.
9753 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9756 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9757 hard registers. Formatting fixes.
9759 PR preprocessor/83722
9760 * gcc.c (try_generate_repro): Pass
9761 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9762 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9765 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9766 Kito Cheng <kito.cheng@gmail.com>
9768 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9769 (riscv_leaf_function_p): Delete.
9770 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9772 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9774 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9776 (do_ifelse): New function.
9777 (do_isel): New function.
9778 (do_sub3): New function.
9779 (do_add3): New function.
9780 (do_load_mask_compare): New function.
9781 (do_overlap_load_compare): New function.
9782 (expand_compare_loop): New function.
9783 (expand_block_compare): Call expand_compare_loop() when appropriate.
9784 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9786 (-mblock-compare-inline-loop-limit): New option.
9788 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9791 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9792 Reverse order of second and third operands in first alternative.
9793 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9794 of first and second elements in UNSPEC_VPERMR vector.
9795 (altivec_expand_vec_perm_le): Likewise.
9797 2018-01-08 Jeff Law <law@redhat.com>
9799 PR rtl-optimizatin/81308
9800 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9801 (process_switch): If group_case_labels makes a change, then set
9803 (pass_convert_switch::execute): If a switch is converted, then
9804 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9806 PR rtl-optimization/81308
9807 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9810 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9812 PR target/83663 - Revert r255946
9813 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9814 generation for cases where splatting a value is not useful.
9815 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9816 across a vec_duplicate and a paradoxical subreg forming a vector
9817 mode to a vec_concat.
9819 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9821 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9822 -march=armv8.3-a variants.
9823 * config/arm/t-multilib: Likewise.
9824 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9826 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9828 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9830 (cceq_ior_compare_complement): Give it a name so I can use it, and
9831 change boolean_or_operator predicate to boolean_operator so it can
9832 be used to generate a crand.
9833 (eqne): New code iterator.
9834 (bd/bd_neg): New code_attrs.
9835 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9836 a single define_insn.
9837 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9838 decrement (bdnzt/bdnzf/bdzt/bdzf).
9839 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9840 with the new names of the branch decrement patterns, and added the
9841 names of the branch decrement conditional patterns.
9843 2018-01-08 Richard Biener <rguenther@suse.de>
9845 PR tree-optimization/83563
9846 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9849 2018-01-08 Richard Biener <rguenther@suse.de>
9852 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9854 2018-01-08 Richard Biener <rguenther@suse.de>
9856 PR tree-optimization/83685
9857 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9858 references to abnormals.
9860 2018-01-08 Richard Biener <rguenther@suse.de>
9863 * dwarf2out.c (output_indirect_strings): Handle empty
9864 skeleton_debug_str_hash.
9865 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9867 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9869 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9870 (emit_store_direct): Likewise.
9871 (arc_trampoline_adjust_address): Likewise.
9872 (arc_asm_trampoline_template): New function.
9873 (arc_initialize_trampoline): Use asm_trampoline_template.
9874 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9875 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9876 * config/arc/arc.md (flush_icache): Delete pattern.
9878 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9880 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9881 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9884 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9887 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9888 by not USED_FOR_TARGET.
9889 (make_pass_resolve_sw_modes): Likewise.
9891 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9893 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9896 2018-01-08 Richard Biener <rguenther@suse.de>
9899 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9901 2018-01-08 Richard Biener <rguenther@suse.de>
9904 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9906 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9909 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9910 basic blocks with a small number of successors.
9911 (convert_control_dep_chain_into_preds): Improve handling of
9913 (dump_predicates): Split apart into...
9914 (dump_pred_chain): ...here...
9915 (dump_pred_info): ...and here.
9916 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9917 (can_chain_union_be_invalidated_p): Improve check for invalidation
9919 (uninit_uses_cannot_happen): Avoid unnecessary if
9920 convert_control_dep_chain_into_preds yielded nothing.
9922 2018-01-06 Martin Sebor <msebor@redhat.com>
9924 PR tree-optimization/83640
9925 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9926 subtracting negative offset from size.
9927 (builtin_access::overlap): Adjust offset bounds of the access to fall
9928 within the size of the object if possible.
9930 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9932 PR rtl-optimization/83699
9933 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9934 extract_bit_field_as_subreg to cases in which the extracted
9935 value is also a vector.
9937 * lra-constraints.c (process_alt_operands): Test for the equivalence
9938 substitutions when detecting a possible reload cycle.
9940 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9943 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9944 by default if flag_selective_schedling{,2}. Formatting fixes.
9946 PR rtl-optimization/83682
9947 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9948 if it has non-VECTOR_MODE element mode.
9949 (vec_duplicate_p): Likewise.
9952 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9953 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9955 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9958 * config/i386/i386-builtin.def
9959 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9960 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9961 Require also OPTION_MASK_ISA_AVX512F in addition to
9962 OPTION_MASK_ISA_GFNI.
9963 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9964 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9965 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9966 to OPTION_MASK_ISA_GFNI.
9967 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9968 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9969 OPTION_MASK_ISA_AVX512BW.
9970 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9971 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9972 addition to OPTION_MASK_ISA_GFNI.
9973 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9974 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9975 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9976 to OPTION_MASK_ISA_GFNI.
9977 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9978 a requirement for all ISAs rather than any of them with a few
9980 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9982 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9983 bitmasks to be enabled with 3 exceptions, instead of requiring any
9984 enabled ISA with lots of exceptions.
9985 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9986 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9987 Change avx512bw in isa attribute to avx512f.
9988 * config/i386/sgxintrin.h: Add license boilerplate.
9989 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9990 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9991 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9992 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9994 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9995 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9996 temporarily sse2 rather than sse if not enabled already.
9999 * config/i386/sse.md (VI248_VLBW): Rename to ...
10000 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
10001 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
10002 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
10003 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
10004 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
10005 mode iterator instead of VI248_VLBW.
10007 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
10009 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
10010 (record_modified): Skip clobbers; add debug output.
10011 (param_change_prob): Use sreal frequencies.
10013 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10015 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
10016 punt for user-aligned variables.
10018 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
10020 * tree-chrec.c (chrec_contains_symbols): Return true for
10023 2018-01-05 Sudakshina Das <sudi.das@arm.com>
10026 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
10027 of (x|y) == x for BICS pattern.
10029 2018-01-05 Jakub Jelinek <jakub@redhat.com>
10031 PR tree-optimization/83605
10032 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
10033 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
10036 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
10038 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
10039 * config/epiphany/rtems.h: New file.
10041 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10042 Uros Bizjak <ubizjak@gmail.com>
10045 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
10046 QIreg_operand instead of register_operand predicate.
10047 * config/i386/i386.c (ix86_rop_should_change_byte_p,
10048 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
10049 comments instead of -fmitigate[-_]rop.
10051 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10054 * cgraphunit.c (symbol_table::compile): Switch to text_section
10055 before calling assembly_start debug hook.
10056 * run-rtl-passes.c (run_rtl_passes): Likewise.
10059 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10061 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10062 range_int_cst_p rather than !symbolic_range_p before calling
10063 extract_range_from_multiplicative_op_1.
10065 2018-01-04 Jeff Law <law@redhat.com>
10067 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10068 redundant test in assertion.
10070 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10072 * doc/rtl.texi: Document machine_mode wrapper classes.
10074 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10076 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10077 using tree_to_uhwi.
10079 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10081 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10082 the VEC_PERM_EXPR fold to fail.
10084 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10087 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10088 to switched_sections.
10090 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10093 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10094 test for d.testing.
10096 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10099 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10100 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10102 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10105 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10106 is BLKmode and bitpos not zero or mode change is needed.
10108 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10111 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10114 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10117 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10118 instead of MULT rtx. Update all corresponding splitters.
10119 (*saddl_se): Ditto.
10120 (*ssub<modesuffix>): Ditto.
10121 (*ssubl_se): Ditto.
10122 (*cmp_sadd_di): Update split patterns.
10123 (*cmp_sadd_si): Ditto.
10124 (*cmp_sadd_sidi): Ditto.
10125 (*cmp_ssub_di): Ditto.
10126 (*cmp_ssub_si): Ditto.
10127 (*cmp_ssub_sidi): Ditto.
10128 * config/alpha/predicates.md (const23_operand): New predicate.
10129 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10130 Look for ASHIFT, not MULT inner operand.
10131 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10133 2018-01-04 Martin Liska <mliska@suse.cz>
10135 PR gcov-profile/83669
10136 * gcov.c (output_intermediate_file): Add version to intermediate
10138 * doc/gcov.texi: Document new field 'version' in intermediate
10139 file format. Fix location of '-k' option of gcov command.
10141 2018-01-04 Martin Liska <mliska@suse.cz>
10144 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10146 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10148 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10150 2018-01-03 Martin Sebor <msebor@redhat.com>
10152 PR tree-optimization/83655
10153 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10154 checking calls with invalid arguments.
10156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10158 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10159 (vectorizable_mask_load_store): Delete.
10160 (vectorizable_call): Return false for masked loads and stores.
10161 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10162 instead of gimple_assign_rhs1.
10163 (vectorizable_load): Handle IFN_MASK_LOAD.
10164 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10166 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10168 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10170 (vectorizable_mask_load_store): ...here.
10171 (vectorizable_load): ...and here.
10173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10175 * tree-vect-stmts.c (vect_build_all_ones_mask)
10176 (vect_build_zero_merge_argument): New functions, split out from...
10177 (vectorizable_load): ...here.
10179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10181 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10183 (vectorizable_mask_load_store): ...here.
10184 (vectorizable_store): ...and here.
10186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10188 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10190 (vectorizable_mask_load_store): ...here.
10192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10194 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10195 (vect_model_store_cost): Take a vec_load_store_type instead of a
10197 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10198 (vect_model_store_cost): Take a vec_load_store_type instead of a
10200 (vectorizable_mask_load_store): Update accordingly.
10201 (vectorizable_store): Likewise.
10202 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10206 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10207 IFN_MASK_LOAD calls here rather than...
10208 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10211 Alan Hayward <alan.hayward@arm.com>
10212 David Sherwood <david.sherwood@arm.com>
10214 * expmed.c (extract_bit_field_1): For vector extracts,
10215 fall back to extract_bit_field_as_subreg if vec_extract
10218 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10219 Alan Hayward <alan.hayward@arm.com>
10220 David Sherwood <david.sherwood@arm.com>
10222 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10223 they are variable or constant sized.
10224 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10225 slots for constant-sized data.
10227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10228 Alan Hayward <alan.hayward@arm.com>
10229 David Sherwood <david.sherwood@arm.com>
10231 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10232 handling COND_EXPRs with boolean comparisons, try to find a better
10233 basis for the mask type than the boolean itself.
10235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10237 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10238 is calculated and how it can be overridden.
10239 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10240 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10242 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10245 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10246 Alan Hayward <alan.hayward@arm.com>
10247 David Sherwood <david.sherwood@arm.com>
10249 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10250 Remove the mode argument.
10251 (aarch64_simd_valid_immediate): Remove the mode and inverse
10253 * config/aarch64/iterators.md (bitsize): New iterator.
10254 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10255 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10256 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10257 aarch64_simd_valid_immediate.
10258 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10259 (aarch64_reg_or_bic_imm): Likewise.
10260 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10261 with an insn_type enum and msl with a modifier_type enum.
10262 Replace element_width with a scalar_mode. Change the shift
10263 to unsigned int. Add constructors for scalar_float_mode and
10264 scalar_int_mode elements.
10265 (aarch64_vect_float_const_representable_p): Delete.
10266 (aarch64_can_const_movi_rtx_p)
10267 (aarch64_simd_scalar_immediate_valid_for_move)
10268 (aarch64_simd_make_constant): Update call to
10269 aarch64_simd_valid_immediate.
10270 (aarch64_advsimd_valid_immediate_hs): New function.
10271 (aarch64_advsimd_valid_immediate): Likewise.
10272 (aarch64_simd_valid_immediate): Remove mode and inverse
10273 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10274 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10275 and aarch64_float_const_representable_p on the result.
10276 (aarch64_output_simd_mov_immediate): Remove mode argument.
10277 Update call to aarch64_simd_valid_immediate and use of
10278 simd_immediate_info.
10279 (aarch64_output_scalar_simd_mov_immediate): Update call
10282 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10283 Alan Hayward <alan.hayward@arm.com>
10284 David Sherwood <david.sherwood@arm.com>
10286 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10287 (mode_nunits): Likewise CONST_MODE_NUNITS.
10288 * machmode.def (ADJUST_NUNITS): Document.
10289 * genmodes.c (mode_data::need_nunits_adj): New field.
10290 (blank_mode): Update accordingly.
10291 (adj_nunits): New variable.
10292 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10294 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10295 listed in adj_nunits.
10296 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10297 listed in adj_nunits. Don't emit case statements for such modes.
10298 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10299 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10300 nothing if adj_nunits is nonnull.
10301 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10302 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10303 (emit_mode_fbit): Update use of print_maybe_const_decl.
10304 (emit_move_size): Likewise. Treat the array as non-const
10306 (emit_mode_adjustments): Handle adj_nunits.
10308 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10310 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10311 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10312 (VECTOR_MODES): Use it.
10313 (make_vector_modes): Take the prefix as an argument.
10315 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10316 Alan Hayward <alan.hayward@arm.com>
10317 David Sherwood <david.sherwood@arm.com>
10319 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10320 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10321 for MODE_VECTOR_BOOL.
10322 * machmode.def (VECTOR_BOOL_MODE): Document.
10323 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10324 (make_vector_bool_mode): New function.
10325 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10327 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10328 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10330 * stor-layout.c (int_mode_for_mode): Likewise.
10331 * tree.c (build_vector_type_for_mode): Likewise.
10332 * varasm.c (output_constant_pool_2): Likewise.
10333 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10334 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10335 for MODE_VECTOR_BOOL.
10336 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10337 of mode class checks.
10338 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10339 instead of a list of mode class checks.
10340 (expand_vector_scalar_condition): Likewise.
10341 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10343 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10344 Alan Hayward <alan.hayward@arm.com>
10345 David Sherwood <david.sherwood@arm.com>
10347 * machmode.h (mode_size): Change from unsigned short to
10349 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10350 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10351 or if measurement_type is not polynomial.
10352 (fixed_size_mode::includes_p): Check for constant-sized modes.
10353 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10354 return a poly_uint16 rather than an unsigned short.
10355 (emit_mode_size): Change the type of mode_size from unsigned short
10356 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10357 (emit_mode_adjustments): Cope with polynomial vector sizes.
10358 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10360 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10362 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10363 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10364 * caller-save.c (setup_save_areas): Likewise.
10365 (replace_reg_with_saved_mem): Likewise.
10366 * calls.c (emit_library_call_value_1): Likewise.
10367 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10368 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10369 (gen_lowpart_for_combine): Likewise.
10370 * convert.c (convert_to_integer_1): Likewise.
10371 * cse.c (equiv_constant, cse_insn): Likewise.
10372 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10373 (cselib_subst_to_values): Likewise.
10374 * dce.c (word_dce_process_block): Likewise.
10375 * df-problems.c (df_word_lr_mark_ref): Likewise.
10376 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10377 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10378 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10379 (rtl_for_decl_location): Likewise.
10380 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10381 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10382 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10383 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10384 (expand_expr_real_1): Likewise.
10385 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10386 (pad_below): Likewise.
10387 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10388 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10389 * ira.c (get_subreg_tracking_sizes): Likewise.
10390 * ira-build.c (ira_create_allocno_objects): Likewise.
10391 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10392 (ira_sort_regnos_for_alter_reg): Likewise.
10393 * ira-costs.c (record_operand_costs): Likewise.
10394 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10395 (resolve_simple_move): Likewise.
10396 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10397 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10398 (lra_constraints): Likewise.
10399 (CONST_POOL_OK_P): Reject variable-sized modes.
10400 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10401 (add_pseudo_to_slot, lra_spill): Likewise.
10402 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10403 * optabs-query.c (get_best_extraction_insn): Likewise.
10404 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10405 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10406 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10407 * recog.c (offsettable_address_addr_space_p): Likewise.
10408 * regcprop.c (maybe_mode_change): Likewise.
10409 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10410 * regrename.c (build_def_use): Likewise.
10411 * regstat.c (dump_reg_info): Likewise.
10412 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10413 (find_reloads, find_reloads_subreg_address): Likewise.
10414 * reload1.c (eliminate_regs_1): Likewise.
10415 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10416 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10417 (simplify_binary_operation_1, simplify_subreg): Likewise.
10418 * targhooks.c (default_function_arg_padding): Likewise.
10419 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10420 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10421 (verify_gimple_assign_ternary): Likewise.
10422 * tree-inline.c (estimate_move_cost): Likewise.
10423 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10424 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10425 (get_address_cost_ainc): Likewise.
10426 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10427 (vect_supportable_dr_alignment): Likewise.
10428 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10429 (vectorizable_reduction): Likewise.
10430 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10431 (vectorizable_operation, vectorizable_load): Likewise.
10432 * tree.c (build_same_sized_truth_vector_type): Likewise.
10433 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10434 * var-tracking.c (emit_note_insn_var_location): Likewise.
10435 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10436 (ADDR_VEC_ALIGN): Likewise.
10438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10439 Alan Hayward <alan.hayward@arm.com>
10440 David Sherwood <david.sherwood@arm.com>
10442 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10444 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10445 or if measurement_type is polynomial.
10446 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10447 * combine.c (make_extraction): Likewise.
10448 * dse.c (find_shift_sequence): Likewise.
10449 * dwarf2out.c (mem_loc_descriptor): Likewise.
10450 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10451 (extract_bit_field, extract_low_bits): Likewise.
10452 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10453 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10454 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10455 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10456 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10457 * reload.c (find_reloads): Likewise.
10458 * reload1.c (alter_reg): Likewise.
10459 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10460 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10461 * tree-if-conv.c (predicate_mem_writes): Likewise.
10462 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10463 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10464 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10465 * valtrack.c (dead_debug_insert_temp): Likewise.
10466 * varasm.c (mergeable_constant_section): Likewise.
10467 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10470 Alan Hayward <alan.hayward@arm.com>
10471 David Sherwood <david.sherwood@arm.com>
10473 * expr.c (expand_assignment): Cope with polynomial mode sizes
10474 when assigning to a CONCAT.
10476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10477 Alan Hayward <alan.hayward@arm.com>
10478 David Sherwood <david.sherwood@arm.com>
10480 * machmode.h (mode_precision): Change from unsigned short to
10482 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10484 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10485 or if measurement_type is not polynomial.
10486 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10487 in which the mode is already known to be a scalar_int_mode.
10488 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10489 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10491 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10492 for GET_MODE_PRECISION.
10493 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10494 for GET_MODE_PRECISION.
10495 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10497 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10498 (expand_field_assignment, make_extraction): Likewise.
10499 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10500 (get_last_value): Likewise.
10501 * convert.c (convert_to_integer_1): Likewise.
10502 * cse.c (cse_insn): Likewise.
10503 * expr.c (expand_expr_real_1): Likewise.
10504 * lra-constraints.c (simplify_operand_subreg): Likewise.
10505 * optabs-query.c (can_atomic_load_p): Likewise.
10506 * optabs.c (expand_atomic_load): Likewise.
10507 (expand_atomic_store): Likewise.
10508 * ree.c (combine_reaching_defs): Likewise.
10509 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10510 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10511 * tree.h (type_has_mode_precision_p): Likewise.
10512 * ubsan.c (instrument_si_overflow): Likewise.
10514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10515 Alan Hayward <alan.hayward@arm.com>
10516 David Sherwood <david.sherwood@arm.com>
10518 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10519 polynomial numbers of units.
10520 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10521 (valid_vector_subparts_p): New function.
10522 (build_vector_type): Remove temporary shim and take the number
10523 of units as a poly_uint64 rather than an int.
10524 (build_opaque_vector_type): Take the number of units as a
10525 poly_uint64 rather than an int.
10526 * tree.c (build_vector_from_ctor): Handle polynomial
10527 TYPE_VECTOR_SUBPARTS.
10528 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10529 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10530 (build_vector_from_val): If the number of units is variable,
10531 use build_vec_duplicate_cst for constant operands and
10532 VEC_DUPLICATE_EXPR otherwise.
10533 (make_vector_type): Remove temporary is_constant ().
10534 (build_vector_type, build_opaque_vector_type): Take the number of
10535 units as a poly_uint64 rather than an int.
10536 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10538 * cfgexpand.c (expand_debug_expr): Likewise.
10539 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10540 (store_constructor, expand_expr_real_1): Likewise.
10541 (const_scalar_mask_from_tree): Likewise.
10542 * fold-const-call.c (fold_const_reduction): Likewise.
10543 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10544 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10545 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10546 (fold_relational_const): Likewise.
10547 (native_interpret_vector): Likewise. Change the size from an
10548 int to an unsigned int.
10549 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10550 TYPE_VECTOR_SUBPARTS.
10551 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10552 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10553 duplicating a non-constant operand into a variable-length vector.
10554 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10555 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10556 * ipa-icf.c (sem_variable::equals): Likewise.
10557 * match.pd: Likewise.
10558 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10559 * print-tree.c (print_node): Likewise.
10560 * stor-layout.c (layout_type): Likewise.
10561 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10562 * tree-cfg.c (verify_gimple_comparison): Likewise.
10563 (verify_gimple_assign_binary): Likewise.
10564 (verify_gimple_assign_ternary): Likewise.
10565 (verify_gimple_assign_single): Likewise.
10566 * tree-pretty-print.c (dump_generic_node): Likewise.
10567 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10568 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10569 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10570 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10571 (vect_shift_permute_load_chain): Likewise.
10572 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10573 (expand_vector_condition, optimize_vector_constructor): Likewise.
10574 (lower_vec_perm, get_compute_type): Likewise.
10575 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10576 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10577 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10578 (vect_recog_mask_conversion_pattern): Likewise.
10579 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10580 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10581 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10582 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10583 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10584 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10585 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10586 (supportable_widening_operation): Likewise.
10587 (supportable_narrowing_operation): Likewise.
10588 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10590 * varasm.c (output_constant): Likewise.
10592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10593 Alan Hayward <alan.hayward@arm.com>
10594 David Sherwood <david.sherwood@arm.com>
10596 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10597 so that both the length == 3 and length != 3 cases set up their
10598 own permute vectors. Add comments explaining why we know the
10599 number of elements is constant.
10600 (vect_permute_load_chain): Likewise.
10602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10603 Alan Hayward <alan.hayward@arm.com>
10604 David Sherwood <david.sherwood@arm.com>
10606 * machmode.h (mode_nunits): Change from unsigned char to
10608 (ONLY_FIXED_SIZE_MODES): New macro.
10609 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10610 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10611 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10613 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10614 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10615 or if measurement_type is not polynomial.
10616 * genmodes.c (ZERO_COEFFS): New macro.
10617 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10619 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10620 Use ZERO_COEFFS when emitting initializers.
10621 * data-streamer.h (bp_pack_poly_value): New function.
10622 (bp_unpack_poly_value): Likewise.
10623 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10624 for GET_MODE_NUNITS.
10625 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10626 for GET_MODE_NUNITS.
10627 * tree.c (make_vector_type): Remove temporary shim and make
10628 the real function take the number of units as a poly_uint64
10629 rather than an int.
10630 (build_vector_type_for_mode): Handle polynomial nunits.
10631 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10632 * emit-rtl.c (const_vec_series_p_1): Likewise.
10633 (gen_rtx_CONST_VECTOR): Likewise.
10634 * fold-const.c (test_vec_duplicate_folding): Likewise.
10635 * genrecog.c (validate_pattern): Likewise.
10636 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10637 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10638 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10639 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10640 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10641 * rtlanal.c (subreg_get_info): Likewise.
10642 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10643 (vect_grouped_load_supported): Likewise.
10644 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10645 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10646 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10647 (simplify_const_unary_operation, simplify_binary_operation_1)
10648 (simplify_const_binary_operation, simplify_ternary_operation)
10649 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10650 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10651 instead of CONST_VECTOR_NUNITS.
10652 * varasm.c (output_constant_pool_2): Likewise.
10653 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10654 explicit-encoded elements in the XVEC for variable-length vectors.
10656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10658 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10660 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10661 Alan Hayward <alan.hayward@arm.com>
10662 David Sherwood <david.sherwood@arm.com>
10664 * coretypes.h (fixed_size_mode): Declare.
10665 (fixed_size_mode_pod): New typedef.
10666 * builtins.h (target_builtins::x_apply_args_mode)
10667 (target_builtins::x_apply_result_mode): Change type to
10668 fixed_size_mode_pod.
10669 * builtins.c (apply_args_size, apply_result_size, result_vector)
10670 (expand_builtin_apply_args_1, expand_builtin_apply)
10671 (expand_builtin_return): Update accordingly.
10673 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10675 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10676 * cselib.c (cselib_hash_rtx): Likewise.
10677 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10678 CONST_VECTOR encoding.
10680 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10681 Jeff Law <law@redhat.com>
10684 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10685 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10686 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10687 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10690 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10691 explicitly probe *sp in a noreturn function if there were any callee
10692 register saves or frame pointer is needed.
10694 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10697 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10698 BLKmode for ternary, binary or unary expressions.
10701 * var-tracking.c (delete_vta_debug_insn): New inline function.
10702 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10703 insns from get_insns () to NULL instead of each bb separately.
10704 Use delete_vta_debug_insn. No longer static.
10705 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10706 delete_vta_debug_insns callers.
10707 * rtl.h (delete_vta_debug_insns): Declare.
10708 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10709 instead of variable_tracking_main.
10711 2018-01-03 Martin Sebor <msebor@redhat.com>
10713 PR tree-optimization/83603
10714 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10715 arguments past the endof the argument list in functions declared
10716 without a prototype.
10717 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10718 Avoid checking when arguments are null.
10720 2018-01-03 Martin Sebor <msebor@redhat.com>
10723 * doc/extend.texi (attribute const): Fix a typo.
10724 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10725 issuing -Wsuggest-attribute for void functions.
10727 2018-01-03 Martin Sebor <msebor@redhat.com>
10729 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10730 offset_int::from instead of wide_int::to_shwi.
10731 (maybe_diag_overlap): Remove assertion.
10732 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10733 * gimple-ssa-sprintf.c (format_directive): Same.
10734 (parse_directive): Same.
10735 (sprintf_dom_walker::compute_format_length): Same.
10736 (try_substitute_return_value): Same.
10738 2018-01-03 Jeff Law <law@redhat.com>
10740 PR middle-end/83654
10741 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10742 non-constant residual for zero at runtime and avoid probing in
10743 that case. Reorganize code for trailing problem to mirror handling
10746 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10748 PR tree-optimization/83501
10749 * tree-ssa-strlen.c (get_string_cst): New.
10750 (handle_char_store): Call get_string_cst.
10752 2018-01-03 Martin Liska <mliska@suse.cz>
10754 PR tree-optimization/83593
10755 * tree-ssa-strlen.c: Include tree-cfg.h.
10756 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10757 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10758 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10760 (strlen_dom_walker::before_dom_children): Call
10761 gimple_purge_dead_eh_edges. Dump tranformation with details
10763 (strlen_dom_walker::before_dom_children): Update call by adding
10764 new argument cleanup_eh.
10765 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10767 2018-01-03 Martin Liska <mliska@suse.cz>
10770 * cif-code.def (VARIADIC_THUNK): New enum value.
10771 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10774 2018-01-03 Jan Beulich <jbeulich@suse.com>
10776 * sse.md (mov<mode>_internal): Tighten condition for when to use
10777 vmovdqu<ssescalarsize> for TI and OI modes.
10779 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10781 Update copyright years.
10783 2018-01-03 Martin Liska <mliska@suse.cz>
10786 * ipa-visibility.c (function_and_variable_visibility): Skip
10787 functions with noipa attribure.
10789 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10791 * gcc.c (process_command): Update copyright notice dates.
10792 * gcov-dump.c (print_version): Ditto.
10793 * gcov.c (print_version): Ditto.
10794 * gcov-tool.c (print_version): Ditto.
10795 * gengtype.c (create_file): Ditto.
10796 * doc/cpp.texi: Bump @copying's copyright year.
10797 * doc/cppinternals.texi: Ditto.
10798 * doc/gcc.texi: Ditto.
10799 * doc/gccint.texi: Ditto.
10800 * doc/gcov.texi: Ditto.
10801 * doc/install.texi: Ditto.
10802 * doc/invoke.texi: Ditto.
10804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10806 * vector-builder.h (vector_builder::m_full_nelts): Change from
10807 unsigned int to poly_uint64.
10808 (vector_builder::full_nelts): Update prototype accordingly.
10809 (vector_builder::new_vector): Likewise.
10810 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10811 (vector_builder::operator ==): Likewise.
10812 (vector_builder::finalize): Likewise.
10813 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10814 Take the number of elements as a poly_uint64 rather than an
10816 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10817 from unsigned int to poly_uint64.
10818 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10819 (vec_perm_indices::new_vector): Likewise.
10820 (vec_perm_indices::length): Likewise.
10821 (vec_perm_indices::nelts_per_input): Likewise.
10822 (vec_perm_indices::input_nelts): Likewise.
10823 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10824 number of elements per input as a poly_uint64 rather than an
10825 unsigned int. Use the original encoding for variable-length
10826 vectors, rather than clamping each individual element.
10827 For the second and subsequent elements in each pattern,
10828 clamp the step and base before clamping their sum.
10829 (vec_perm_indices::series_p): Handle polynomial element counts.
10830 (vec_perm_indices::all_in_range_p): Likewise.
10831 (vec_perm_indices_to_tree): Likewise.
10832 (vec_perm_indices_to_rtx): Likewise.
10833 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10834 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10835 (tree_vector_builder::new_binary_operation): Handle polynomial
10836 element counts. Return false if we need to know the number
10837 of elements at compile time.
10838 * fold-const.c (fold_vec_perm): Punt if the number of elements
10839 isn't known at compile time.
10841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10843 * vec-perm-indices.h (vec_perm_builder): Change element type
10844 from HOST_WIDE_INT to poly_int64.
10845 (vec_perm_indices::element_type): Update accordingly.
10846 (vec_perm_indices::clamp): Handle polynomial element_types.
10847 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10848 (vec_perm_indices::all_in_range_p): Likewise.
10849 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10851 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10852 polynomial vec_perm_indices element types.
10853 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10854 * fold-const.c (fold_vec_perm): Likewise.
10855 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10856 * tree-vect-generic.c (lower_vec_perm): Likewise.
10857 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10858 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10859 element type to HOST_WIDE_INT.
10861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10862 Alan Hayward <alan.hayward@arm.com>
10863 David Sherwood <david.sherwood@arm.com>
10865 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10866 rather than an int. Use plus_constant.
10867 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10868 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10870 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10871 Alan Hayward <alan.hayward@arm.com>
10872 David Sherwood <david.sherwood@arm.com>
10874 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10875 a HOST_WIDE_INT to a poly_int64.
10877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10878 Alan Hayward <alan.hayward@arm.com>
10879 David Sherwood <david.sherwood@arm.com>
10881 * calls.c (load_register_parameters): Cope with polynomial
10882 mode sizes. Require a constant size for BLKmode parameters
10883 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10884 forces a parameter to be padded at the lsb end in order to
10885 fill a complete number of words, require the parameter size
10886 to be ordered wrt UNITS_PER_WORD.
10888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10889 Alan Hayward <alan.hayward@arm.com>
10890 David Sherwood <david.sherwood@arm.com>
10892 * reload1.c (spill_stack_slot_width): Change element type
10893 from unsigned int to poly_uint64_pod.
10894 (alter_reg): Treat mode sizes as polynomial.
10896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10897 Alan Hayward <alan.hayward@arm.com>
10898 David Sherwood <david.sherwood@arm.com>
10900 * reload.c (complex_word_subreg_p): New function.
10901 (reload_inner_reg_of_subreg, push_reload): Use it.
10903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10904 Alan Hayward <alan.hayward@arm.com>
10905 David Sherwood <david.sherwood@arm.com>
10907 * lra-constraints.c (process_alt_operands): Reject matched
10908 operands whose sizes aren't ordered.
10909 (match_reload): Refer to this check here.
10911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10912 Alan Hayward <alan.hayward@arm.com>
10913 David Sherwood <david.sherwood@arm.com>
10915 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10916 that the mode size is in the set {1, 2, 4, 8, 16}.
10918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10919 Alan Hayward <alan.hayward@arm.com>
10920 David Sherwood <david.sherwood@arm.com>
10922 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10923 Use plus_constant instead of gen_rtx_PLUS.
10925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10926 Alan Hayward <alan.hayward@arm.com>
10927 David Sherwood <david.sherwood@arm.com>
10929 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10930 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10931 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10932 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10933 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10934 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10935 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10936 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10937 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10938 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10940 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10941 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10942 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10943 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10944 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10945 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10946 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10947 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10948 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10949 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10951 * expr.c (emit_move_resolve_push): Treat the input and result
10952 of PUSH_ROUNDING as a poly_int64.
10953 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10954 (emit_push_insn): Likewise.
10955 * lra-eliminations.c (mark_not_eliminable): Likewise.
10956 * recog.c (push_operand): Likewise.
10957 * reload1.c (elimination_effects): Likewise.
10958 * rtlanal.c (nonzero_bits1): Likewise.
10959 * calls.c (store_one_arg): Likewise. Require the padding to be
10960 known at compile time.
10962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10963 Alan Hayward <alan.hayward@arm.com>
10964 David Sherwood <david.sherwood@arm.com>
10966 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10967 Use plus_constant instead of gen_rtx_PLUS.
10969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10970 Alan Hayward <alan.hayward@arm.com>
10971 David Sherwood <david.sherwood@arm.com>
10973 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10974 rather than an int.
10976 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10977 Alan Hayward <alan.hayward@arm.com>
10978 David Sherwood <david.sherwood@arm.com>
10980 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10981 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10982 via stack temporaries. Treat the mode size as polynomial too.
10984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10985 Alan Hayward <alan.hayward@arm.com>
10986 David Sherwood <david.sherwood@arm.com>
10988 * expr.c (expand_expr_real_2): When handling conversions involving
10989 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10990 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10991 as a poly_uint64 too.
10993 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10994 Alan Hayward <alan.hayward@arm.com>
10995 David Sherwood <david.sherwood@arm.com>
10997 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10999 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11000 Alan Hayward <alan.hayward@arm.com>
11001 David Sherwood <david.sherwood@arm.com>
11003 * combine.c (can_change_dest_mode): Handle polynomial
11004 REGMODE_NATURAL_SIZE.
11005 * expmed.c (store_bit_field_1): Likewise.
11006 * expr.c (store_constructor): Likewise.
11007 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
11008 and polynomial REGMODE_NATURAL_SIZE.
11009 (gen_lowpart_common): Likewise.
11010 * reginfo.c (record_subregs_of_mode): Likewise.
11011 * rtlanal.c (read_modify_subreg_p): Likewise.
11013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11014 Alan Hayward <alan.hayward@arm.com>
11015 David Sherwood <david.sherwood@arm.com>
11017 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
11018 numbers of elements.
11020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11021 Alan Hayward <alan.hayward@arm.com>
11022 David Sherwood <david.sherwood@arm.com>
11024 * match.pd: Cope with polynomial numbers of vector elements.
11026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11027 Alan Hayward <alan.hayward@arm.com>
11028 David Sherwood <david.sherwood@arm.com>
11030 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
11031 in a POINTER_PLUS_EXPR.
11033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11034 Alan Hayward <alan.hayward@arm.com>
11035 David Sherwood <david.sherwood@arm.com>
11037 * omp-simd-clone.c (simd_clone_subparts): New function.
11038 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
11039 (ipa_simd_modify_function_body): Likewise.
11041 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11042 Alan Hayward <alan.hayward@arm.com>
11043 David Sherwood <david.sherwood@arm.com>
11045 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
11046 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
11047 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
11048 (expand_vector_condition, vector_element): Likewise.
11049 (subparts_gt): New function.
11050 (get_compute_type): Use subparts_gt.
11051 (count_type_subparts): Delete.
11052 (expand_vector_operations_1): Use subparts_gt instead of
11053 count_type_subparts.
11055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11056 Alan Hayward <alan.hayward@arm.com>
11057 David Sherwood <david.sherwood@arm.com>
11059 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11060 (vect_compile_time_alias): ...this new function. Do the calculation
11061 on poly_ints rather than trees.
11062 (vect_prune_runtime_alias_test_list): Update call accordingly.
11064 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11065 Alan Hayward <alan.hayward@arm.com>
11066 David Sherwood <david.sherwood@arm.com>
11068 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11070 (vect_schedule_slp_instance): Likewise.
11072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11073 Alan Hayward <alan.hayward@arm.com>
11074 David Sherwood <david.sherwood@arm.com>
11076 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11077 constant and extern definitions for variable-length vectors.
11078 (vect_get_constant_vectors): Note that the number of units
11079 is known to be constant.
11081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11082 Alan Hayward <alan.hayward@arm.com>
11083 David Sherwood <david.sherwood@arm.com>
11085 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11086 of units as polynomial. Choose between WIDE and NARROW based
11089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11090 Alan Hayward <alan.hayward@arm.com>
11091 David Sherwood <david.sherwood@arm.com>
11093 * tree-vect-stmts.c (simd_clone_subparts): New function.
11094 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11097 Alan Hayward <alan.hayward@arm.com>
11098 David Sherwood <david.sherwood@arm.com>
11100 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11101 vectors as polynomial. Use build_index_vector for
11102 IFN_GOMP_SIMD_LANE.
11104 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11105 Alan Hayward <alan.hayward@arm.com>
11106 David Sherwood <david.sherwood@arm.com>
11108 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11109 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11110 for variable-length vectors.
11111 (vectorizable_mask_load_store): Treat the number of units as
11112 polynomial, asserting that it is constant if the condition has
11113 already been enforced.
11114 (vectorizable_store, vectorizable_load): Likewise.
11116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11117 Alan Hayward <alan.hayward@arm.com>
11118 David Sherwood <david.sherwood@arm.com>
11120 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11121 of units as polynomial. Punt if we can't tell at compile time
11122 which vector contains the final result.
11124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11125 Alan Hayward <alan.hayward@arm.com>
11126 David Sherwood <david.sherwood@arm.com>
11128 * tree-vect-loop.c (vectorizable_induction): Treat the number
11129 of units as polynomial. Punt on SLP inductions. Use an integer
11130 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11131 cast of such a series for variable-length floating-point
11134 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11135 Alan Hayward <alan.hayward@arm.com>
11136 David Sherwood <david.sherwood@arm.com>
11138 * tree.h (build_index_vector): Declare.
11139 * tree.c (build_index_vector): New function.
11140 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11141 of units as polynomial, forcibly converting it to a constant if
11142 vectorizable_reduction has already enforced the condition.
11143 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11144 to create a {1,2,3,...} vector.
11145 (vectorizable_reduction): Treat the number of units as polynomial.
11146 Choose vectype_in based on the largest scalar element size rather
11147 than the smallest number of units. Enforce the restrictions
11150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11151 Alan Hayward <alan.hayward@arm.com>
11152 David Sherwood <david.sherwood@arm.com>
11154 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11155 number of units as polynomial.
11157 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11158 Alan Hayward <alan.hayward@arm.com>
11159 David Sherwood <david.sherwood@arm.com>
11161 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11162 * target.def (autovectorize_vector_sizes): Return the vector sizes
11163 by pointer, using vector_sizes rather than a bitmask.
11164 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11165 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11166 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11168 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11169 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11170 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11171 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11172 * omp-general.c (omp_max_vf): Likewise.
11173 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11174 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11175 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11176 * tree-vect-slp.c (vect_slp_bb): Likewise.
11177 * doc/tm.texi: Regenerate.
11178 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11180 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11181 the vector size as a poly_uint64 rather than an unsigned int.
11182 (current_vector_size): Change from an unsigned int to a poly_uint64.
11183 (get_vectype_for_scalar_type): Update accordingly.
11184 * tree.h (build_truth_vector_type): Take the size and number of
11185 units as a poly_uint64 rather than an unsigned int.
11186 (build_vector_type): Add a temporary overload that takes
11187 the number of units as a poly_uint64 rather than an unsigned int.
11188 * tree.c (make_vector_type): Likewise.
11189 (build_truth_vector_type): Take the number of units as a poly_uint64
11190 rather than an unsigned int.
11192 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11193 Alan Hayward <alan.hayward@arm.com>
11194 David Sherwood <david.sherwood@arm.com>
11196 * target.def (get_mask_mode): Take the number of units and length
11197 as poly_uint64s rather than unsigned ints.
11198 * targhooks.h (default_get_mask_mode): Update accordingly.
11199 * targhooks.c (default_get_mask_mode): Likewise.
11200 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11201 * doc/tm.texi: Regenerate.
11203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11204 Alan Hayward <alan.hayward@arm.com>
11205 David Sherwood <david.sherwood@arm.com>
11207 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11208 * omp-general.c (omp_max_vf): Likewise.
11209 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11210 (expand_omp_simd): Handle polynomial safelen.
11211 * omp-low.c (omplow_simd_context): Add a default constructor.
11212 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11213 (lower_rec_simd_input_clauses): Update accordingly.
11214 (lower_rec_input_clauses): Likewise.
11216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11217 Alan Hayward <alan.hayward@arm.com>
11218 David Sherwood <david.sherwood@arm.com>
11220 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11221 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11222 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11223 (vect_analyze_slp_cost): Likewise.
11224 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11225 (vect_model_load_cost): Likewise.
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 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11232 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11233 from an unsigned int * to a poly_uint64_pod *.
11234 (calculate_unrolling_factor): New function.
11235 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11238 Alan Hayward <alan.hayward@arm.com>
11239 David Sherwood <david.sherwood@arm.com>
11241 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11242 from an unsigned int to a poly_uint64.
11243 (_loop_vec_info::slp_unrolling_factor): Likewise.
11244 (_loop_vec_info::vectorization_factor): Change from an int
11246 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11247 (vect_get_num_vectors): New function.
11248 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11249 (vect_get_num_copies): Use vect_get_num_vectors.
11250 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11251 to an unsigned int *.
11252 (vect_analyze_data_refs): Change min_vf from an int * to a
11254 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11255 than an unsigned HOST_WIDE_INT.
11256 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11257 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11258 to an unsigned int *.
11259 (vect_analyze_data_ref_dependences): Likewise.
11260 (vect_compute_data_ref_alignment): Handle polynomial vf.
11261 (vect_enhance_data_refs_alignment): Likewise.
11262 (vect_prune_runtime_alias_test_list): Likewise.
11263 (vect_shift_permute_load_chain): Likewise.
11264 (vect_supportable_dr_alignment): Likewise.
11265 (dependence_distance_ge_vf): Take the vectorization factor as a
11266 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11267 (vect_analyze_data_refs): Change min_vf from an int * to a
11269 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11270 vfm1 as a poly_uint64 rather than an int. Make the same change
11271 for the returned bound_scalar.
11272 (vect_gen_vector_loop_niters): Handle polynomial vf.
11273 (vect_do_peeling): Likewise. Update call to
11274 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11275 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11277 * tree-vect-loop.c (vect_determine_vectorization_factor)
11278 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11279 (vect_get_known_peeling_cost): Likewise.
11280 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11281 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11282 (vect_transform_loop): Likewise. Use the lowest possible VF when
11283 updating the upper bounds of the loop.
11284 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11285 rather than an int.
11286 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11287 polynomial unroll factors.
11288 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11289 (vect_make_slp_decision): Likewise.
11290 (vect_supported_load_permutation_p): Likewise, and polynomial
11292 (vect_analyze_slp_cost): Handle polynomial vf.
11293 (vect_slp_analyze_node_operations): Likewise.
11294 (vect_slp_analyze_bb_1): Likewise.
11295 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11296 than an unsigned HOST_WIDE_INT.
11297 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11298 (vectorizable_load): Handle polynomial vf.
11299 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11301 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
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 * match.pd: Handle bit operations involving three constants
11308 and try to fold one pair.
11310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11312 * tree-vect-loop-manip.c: Include gimple-fold.h.
11313 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11314 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11315 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11316 Add a path that uses a step of VF instead of 1, but disable it
11318 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11319 and niters_no_overflow parameters. Update calls to
11320 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11321 Create a new SSA name if the latter choses to use a ste other
11322 than zero, and return it via niters_vector_mult_vf_var.
11323 * tree-vect-loop.c (vect_transform_loop): Update calls to
11324 vect_do_peeling, vect_gen_vector_loop_niters and
11325 slpeel_make_loop_iterate_ntimes.
11326 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11327 (vect_gen_vector_loop_niters): Update declarations after above changes.
11329 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11331 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11332 128-bit round to integer instructions.
11333 (ceil<mode>2): Likewise.
11334 (btrunc<mode>2): Likewise.
11335 (round<mode>2): Likewise.
11337 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11339 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11340 unaligned VSX load/store on P8/P9.
11341 (expand_block_clear): Allow the use of unaligned VSX
11342 load/store on P8/P9.
11344 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11346 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11348 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11349 swap associated with both a load and a store.
11351 2018-01-02 Andrew Waterman <andrew@sifive.com>
11353 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11354 * config/riscv/riscv.md (clear_cache): Use it.
11356 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11358 * web.c: Remove out-of-date comment.
11360 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11362 * expr.c (fixup_args_size_notes): Check that any existing
11363 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11364 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11365 (emit_single_push_insn): ...here.
11367 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11369 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11370 (const_vector_encoded_nelts): New function.
11371 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11372 (const_vector_int_elt, const_vector_elt): Declare.
11373 * emit-rtl.c (const_vector_int_elt_1): New function.
11374 (const_vector_elt): Likewise.
11375 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11376 of CONST_VECTOR_ELT.
11378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11380 * expr.c: Include rtx-vector-builder.h.
11381 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11382 directly on the tree encoding.
11383 (const_vector_from_tree): Likewise.
11384 * optabs.c: Include rtx-vector-builder.h.
11385 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11386 sequence of "u" values.
11387 * vec-perm-indices.c: Include rtx-vector-builder.h.
11388 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11389 directly on the vec_perm_indices encoding.
11391 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11393 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11394 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11395 * rtx-vector-builder.h: New file.
11396 * rtx-vector-builder.c: Likewise.
11397 * rtl.h (rtx_def::u2): Add a const_vector field.
11398 (CONST_VECTOR_NPATTERNS): New macro.
11399 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11400 (CONST_VECTOR_DUPLICATE_P): Likewise.
11401 (CONST_VECTOR_STEPPED_P): Likewise.
11402 (CONST_VECTOR_ENCODED_ELT): Likewise.
11403 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11404 (unwrap_const_vec_duplicate): Likewise.
11405 (const_vec_series_p): Check for a non-duplicated vector encoding.
11406 Say that the function only returns true for integer vectors.
11407 * emit-rtl.c: Include rtx-vector-builder.h.
11408 (gen_const_vec_duplicate_1): Delete.
11409 (gen_const_vector): Call gen_const_vec_duplicate instead of
11410 gen_const_vec_duplicate_1.
11411 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11412 (gen_const_vec_duplicate): Use rtx_vector_builder.
11413 (gen_const_vec_series): Likewise.
11414 (gen_rtx_CONST_VECTOR): Likewise.
11415 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11416 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11417 Build a new vector rather than modifying a CONST_VECTOR in-place.
11418 (handle_special_swappables): Update call accordingly.
11419 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11420 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11421 Build a new vector rather than modifying a CONST_VECTOR in-place.
11422 (handle_special_swappables): Update call accordingly.
11424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11426 * simplify-rtx.c (simplify_const_binary_operation): Use
11427 CONST_VECTOR_ELT instead of XVECEXP.
11429 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11431 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11432 the selector elements to be different from the data elements
11433 if the selector is a VECTOR_CST.
11434 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11435 ssizetype for the selector.
11437 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11439 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11440 before testing each element individually.
11441 * tree-vect-generic.c (lower_vec_perm): Likewise.
11443 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11445 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11446 * selftest-run-tests.c (selftest::run_tests): Call it.
11447 * vector-builder.h (vector_builder::operator ==): New function.
11448 (vector_builder::operator !=): Likewise.
11449 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11450 (vec_perm_indices::all_from_input_p): New function.
11451 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11452 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11453 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11454 instead of reading the VECTOR_CST directly. Detect whether both
11455 vector inputs are the same before constructing the vec_perm_indices,
11456 and update the number of inputs argument accordingly. Use the
11457 utility functions added above. Only construct sel2 if we need to.
11459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11461 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11462 the broadcast of the low byte.
11463 (expand_mult_highpart): Use an explicit encoding for the permutes.
11464 * optabs-query.c (can_mult_highpart_p): Likewise.
11465 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11466 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11467 (vectorizable_bswap): Likewise.
11468 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11469 explicit encoding for the power-of-2 permutes.
11470 (vect_permute_store_chain): Likewise.
11471 (vect_grouped_load_supported): Likewise.
11472 (vect_permute_load_chain): Likewise.
11474 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11476 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11477 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11478 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11479 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11480 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11481 (vect_gen_perm_mask_any): Likewise.
11483 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11485 * int-vector-builder.h: New file.
11486 * vec-perm-indices.h: Include int-vector-builder.h.
11487 (vec_perm_indices): Redefine as an int_vector_builder.
11488 (auto_vec_perm_indices): Delete.
11489 (vec_perm_builder): Redefine as a stand-alone class.
11490 (vec_perm_indices::vec_perm_indices): New function.
11491 (vec_perm_indices::clamp): Likewise.
11492 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11493 (vec_perm_indices::new_vector): New function.
11494 (vec_perm_indices::new_expanded_vector): Update for new
11495 vec_perm_indices class.
11496 (vec_perm_indices::rotate_inputs): New function.
11497 (vec_perm_indices::all_in_range_p): Operate directly on the
11498 encoded form, without computing elided elements.
11499 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11500 encoding. Update for new vec_perm_indices class.
11501 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11502 the given vec_perm_builder.
11503 (expand_vec_perm_var): Update vec_perm_builder constructor.
11504 (expand_mult_highpart): Use vec_perm_builder instead of
11505 auto_vec_perm_indices.
11506 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11507 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11508 or double series encoding as appropriate.
11509 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11510 vec_perm_indices instead of auto_vec_perm_indices.
11511 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11512 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11513 (vect_permute_store_chain): Likewise.
11514 (vect_grouped_load_supported): Likewise.
11515 (vect_permute_load_chain): Likewise.
11516 (vect_shift_permute_load_chain): Likewise.
11517 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11518 (vect_transform_slp_perm_load): Likewise.
11519 (vect_schedule_slp_instance): Likewise.
11520 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11521 (vectorizable_mask_load_store): Likewise.
11522 (vectorizable_bswap): Likewise.
11523 (vectorizable_store): Likewise.
11524 (vectorizable_load): Likewise.
11525 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11526 vec_perm_indices instead of auto_vec_perm_indices. Use
11527 tree_to_vec_perm_builder to read the vector from a tree.
11528 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11529 vec_perm_builder instead of a vec_perm_indices.
11530 (have_whole_vector_shift): Use vec_perm_builder and
11531 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11532 truncation to calc_vec_perm_mask_for_shift.
11533 (vect_create_epilog_for_reduction): Likewise.
11534 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11535 from auto_vec_perm_indices to vec_perm_indices.
11536 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11537 instead of changing individual elements.
11538 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11539 the vector in d.perm.
11540 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11541 from auto_vec_perm_indices to vec_perm_indices.
11542 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11543 instead of changing individual elements.
11544 (arm_vectorize_vec_perm_const): Use new_vector to install
11545 the vector in d.perm.
11546 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11547 Update vec_perm_builder constructor.
11548 (rs6000_expand_interleave): Likewise.
11549 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11550 (rs6000_expand_interleave): Likewise.
11552 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11554 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11555 to qimode could truncate the indices.
11556 * optabs.c (expand_vec_perm_var): Likewise.
11558 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11560 * Makefile.in (OBJS): Add vec-perm-indices.o.
11561 * vec-perm-indices.h: New file.
11562 * vec-perm-indices.c: Likewise.
11563 * target.h (vec_perm_indices): Replace with a forward class
11565 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11566 * optabs.h: Include vec-perm-indices.h.
11567 (expand_vec_perm): Delete.
11568 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11569 (expand_vec_perm_const): Declare.
11570 * target.def (vec_perm_const_ok): Replace with...
11571 (vec_perm_const): ...this new hook.
11572 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11573 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11574 * doc/tm.texi: Regenerate.
11575 * optabs.def (vec_perm_const): Delete.
11576 * doc/md.texi (vec_perm_const): Likewise.
11577 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11578 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11579 expand_vec_perm for constant permutation vectors. Assert that
11580 the mode of variable permutation vectors is the integer equivalent
11581 of the mode that is being permuted.
11582 * optabs-query.h (selector_fits_mode_p): Declare.
11583 * optabs-query.c: Include vec-perm-indices.h.
11584 (selector_fits_mode_p): New function.
11585 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11586 is defined, instead of checking whether the vec_perm_const_optab
11587 exists. Use targetm.vectorize.vec_perm_const instead of
11588 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11589 fit in the vector mode before using a variable permute.
11590 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11591 vec_perm_indices instead of an rtx.
11592 (expand_vec_perm): Replace with...
11593 (expand_vec_perm_const): ...this new function. Take the selector
11594 as a vec_perm_indices rather than an rtx. Also take the mode of
11595 the selector. Update call to shift_amt_for_vec_perm_mask.
11596 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11597 Use vec_perm_indices::new_expanded_vector to expand the original
11598 selector into bytes. Check whether the indices fit in the vector
11599 mode before using a variable permute.
11600 (expand_vec_perm_var): Make global.
11601 (expand_mult_highpart): Use expand_vec_perm_const.
11602 * fold-const.c: Includes vec-perm-indices.h.
11603 * tree-ssa-forwprop.c: Likewise.
11604 * tree-vect-data-refs.c: Likewise.
11605 * tree-vect-generic.c: Likewise.
11606 * tree-vect-loop.c: Likewise.
11607 * tree-vect-slp.c: Likewise.
11608 * tree-vect-stmts.c: Likewise.
11609 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11611 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11612 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11613 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11614 (aarch64_vectorize_vec_perm_const): ...this new function.
11615 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11616 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11617 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11618 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11619 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11620 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11621 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11623 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11624 check for NEON modes.
11625 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11626 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11627 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11628 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11630 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11631 the old VEC_PERM_CONST conditions.
11632 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11633 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11634 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11635 (ia64_vectorize_vec_perm_const_ok): Merge into...
11636 (ia64_vectorize_vec_perm_const): ...this new function.
11637 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11638 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11639 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11640 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11641 * config/mips/mips.c (mips_expand_vec_perm_const)
11642 (mips_vectorize_vec_perm_const_ok): Merge into...
11643 (mips_vectorize_vec_perm_const): ...this new function.
11644 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11645 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11646 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11647 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11648 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11649 (rs6000_expand_vec_perm_const): Delete.
11650 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11652 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11653 (altivec_expand_vec_perm_const_le): Take each operand individually.
11654 Operate on constant selectors rather than rtxes.
11655 (altivec_expand_vec_perm_const): Likewise. Update call to
11656 altivec_expand_vec_perm_const_le.
11657 (rs6000_expand_vec_perm_const): Delete.
11658 (rs6000_vectorize_vec_perm_const_ok): Delete.
11659 (rs6000_vectorize_vec_perm_const): New function.
11660 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11661 an element count and rtx array.
11662 (rs6000_expand_extract_even): Update call accordingly.
11663 (rs6000_expand_interleave): Likewise.
11664 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11665 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11666 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11667 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11668 (rs6000_expand_vec_perm_const): Delete.
11669 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11670 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11671 (altivec_expand_vec_perm_const_le): Take each operand individually.
11672 Operate on constant selectors rather than rtxes.
11673 (altivec_expand_vec_perm_const): Likewise. Update call to
11674 altivec_expand_vec_perm_const_le.
11675 (rs6000_expand_vec_perm_const): Delete.
11676 (rs6000_vectorize_vec_perm_const_ok): Delete.
11677 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11678 reference to the SPE evmerge intructions.
11679 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11680 an element count and rtx array.
11681 (rs6000_expand_extract_even): Update call accordingly.
11682 (rs6000_expand_interleave): Likewise.
11683 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11684 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11686 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11688 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11690 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11691 vector mode and that that mode matches the mode of the data
11693 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11694 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11695 directly using expand_vec_perm_1 when forcing selectors into
11697 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11699 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11701 * optabs-query.h (can_vec_perm_p): Delete.
11702 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11703 * optabs-query.c (can_vec_perm_p): Split into...
11704 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11705 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11706 particular selector is valid.
11707 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11708 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11709 (vect_grouped_load_supported): Likewise.
11710 (vect_shift_permute_load_chain): Likewise.
11711 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11712 (vect_transform_slp_perm_load): Likewise.
11713 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11714 (vectorizable_bswap): Likewise.
11715 (vect_gen_perm_mask_checked): Likewise.
11716 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11717 implementations of variable permutation vectors into account
11718 when deciding which selector to use.
11719 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11720 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11721 with a false third argument.
11722 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11723 to test whether the constant selector is valid and can_vec_perm_var_p
11724 to test whether a variable selector is valid.
11726 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11728 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11729 * optabs-query.c (can_vec_perm_p): Likewise.
11730 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11731 instead of vec_perm_indices.
11732 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11733 (vect_gen_perm_mask_checked): Likewise,
11734 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11735 (vect_gen_perm_mask_checked): Likewise,
11737 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11739 * optabs-query.h (qimode_for_vec_perm): Declare.
11740 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11741 (qimode_for_vec_perm): ...this new function.
11742 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11744 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11746 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11747 does not have a conditional at the top.
11749 2018-01-02 Richard Biener <rguenther@suse.de>
11751 * ipa-inline.c (big_speedup_p): Fix expression.
11753 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11756 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11759 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11763 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11764 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11765 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11766 cond_taken_branch_cost 3->4.
11768 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11770 PR tree-optimization/83581
11771 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11772 TODO_cleanup_cfg if any changes have been made.
11774 PR middle-end/83608
11775 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11776 convert_modes if target mode has the right side, but different mode
11779 PR middle-end/83609
11780 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11781 last argument when extracting from CONCAT. If either from_real or
11782 from_imag is NULL, use expansion through memory. If result is not
11783 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11784 the parts directly to inner mode, if even that fails, use expansion
11787 PR middle-end/83623
11788 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11789 check for bswap in mode rather than HImode and use that in expand_unop
11792 Copyright (C) 2018 Free Software Foundation, Inc.
11794 Copying and distribution of this file, with or without modification,
11795 are permitted in any medium without royalty provided the copyright
11796 notice and this notice are preserved.