| blob | af84646286026e46bdbe227c9ee4db4822719a7b |
1 2018-04-30 Richard Biener <rguenther@suse.de>
3 * tree-cfg.c (verify_address): Remove base argument, add
4 flag whether to check TREE_ADDRESSABLE and do that.
5 (verify_expr): Remove.
6 (verify_types_in_gimple_reference): Add pieces from verify_expr.
7 (verify_gimple_assign_single): Likewise.
8 (verify_gimple_switch): Likewise.
9 (verify_expr_location_1): Dereference tp once. Add (disabled)
10 piece from verify_expr.
11 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
13 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
17 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
19 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20 (small_data_pattern): Likewise.
21 (arc_rewrite_small_data): Likewise.
22 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
23 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
24 (get_symbol_alignment): New function.
25 (legitimate_small_data_address_p): Likewise.
26 (legitimate_scaled_address): Update, call
27 legitimate_small_data_address_p.
28 (output_sdata): New static variable.
29 (arc_print_operand): Update how we handle small data operands.
30 (arc_print_operand_address): Likewise.
31 (arc_legitimate_address_p): Update, use
32 legitimate_small_data_address_p.
33 (arc_rewrite_small_data_p): Remove.
34 (arc_rewrite_small_data_1): Likewise.
35 (arc_rewrite_small_data): Likewise.
36 (small_data_pattern): Likewise.
37 (compact_sda_memory_operand): Update to use
38 legitimate_small_data_address_p and get_symbol_alignment.
39 (prepare_move_operands): Don't rewite sdata pattern.
40 (prepare_extend_operands): Remove.
41 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
42 pattern.
43 (zero_extendqisi2): Likewise.
44 (zero_extendhisi2): Likewise.
45 (extendqihi2): Likewise.
46 (extendqisi2): Likewise.
47 (extendhisi2): Likewise.
48 (addsi3): Likewise.
49 (subsi3): Likewise.
50 (andsi3): Likewise.
51 * config/arc/constraints.md (Usd): Change it to memory constraint.
53 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
55 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
56 as source of std instructions.
57 * config/arc/arc.md (movsi_insn): Update pattern predicate to
58 allow 6-bit constants as source for store instructions.
59 (movdi_insn): Update instruction pattern to allow 6-bit constants
60 as source for store instructions.
62 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
64 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
66 2018-04-30 Nathan Sidwell <nathan@acm.org>
67 Sandra Loosemore <sandra@codesourcery.com>
69 * dumpfile.c (dump_open): Allow '-' for stdout.
70 * doc/invoke.texi (Developer Options): Document dump filename
71 determination early. Document stdin/stdout selection.
73 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
75 Microblaze Target: PIC data text relative
77 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
78 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
79 Add declaration.
80 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
81 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
82 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
83 New addressing mode for data-text relative position indepenedent code.
84 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
85 'ADDRESS_SYMBOLIC_TXT_REL'.
86 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
87 (microblaze_legitimate_pic_operand): Exclude function calls from
88 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
89 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
90 addresses cases.
91 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
92 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
93 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
94 for 'address + offset'.
95 (microblaze_expand_prologue): Add new function prologue call for
96 'r20' assignation.
97 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
98 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
99 table in case of TARGET_PIC_DATA_TEXT_REL.
100 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
101 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
102 Add new macros 'UNSPEC_TEXT',
103 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
104 + exclude function calls from 'UNSPEC_PLT' in case of data text
105 relative mode.
106 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
107 new target hook for generating address diff vector tables in case of
108 flag_pic.
109 * doc/tm.texi : Regenerate.
110 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
111 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
112 of addr diff vector generation.
113 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
114 target hook definition.
115 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
116 Add default function for generate_pic_addr_diff_vec -> flag_pic.
117 * doc/invoke.texi (Add new pic option): Add new microblaze pic
118 option for data text relative.
120 2018-04-30 Richard Biener <rguenther@suse.de>
122 * tree-chrec.h (evolution_function_is_constant_p): Remove
123 redundant check.
124 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
126 2018-04-30 Richard Biener <rguenther@suse.de>
128 PR bootstrap/85571
129 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
131 2018-04-30 Richard Biener <rguenther@suse.de>
133 PR tree-optimization/28364
134 PR tree-optimization/85275
135 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
136 copying first exit test.
138 2018-04-28 Mark Wielaard <mark@klomp.org>
140 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
141 dwarf_version >= 5.
142 (dwarf_AT): Handle DW_AT_addr_base.
143 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
145 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
147 PR target/84431
148 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
149 (*ashl<dwi>3_doubleword_mask_1): Ditto.
150 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
151 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
153 2018-04-28 Richard Biener <rguenther@suse.de>
155 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
156 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
157 to reflect use. Only add interesting stmts.
159 2018-04-27 Martin Jambor <mjambor@suse.cz>
161 PR ipa/85549
162 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
163 the jump function allows for passing through aggregate values.
165 2018-04-27 David Malcolm <dmalcolm@redhat.com>
167 * input.h (in_system_header_at): Convert from macro to inline
168 function.
169 (from_macro_expansion_at): Likewise.
170 (from_macro_definition_at): Likewise.
172 2018-04-27 Jeff Law <law@redhat.com>
174 * config.gcc: Mark tile* targets as deprecated/obsolete.
176 2018-04-27 Richard Biener <rguenther@suse.de>
178 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
179 fix for ILP32.
181 2018-04-27 Richard Biener <rguenther@suse.de>
183 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
185 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
187 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
188 with Yd constraint. Set "preferred_for_speed" attribute from
189 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
190 with Yd constraint.
191 (*movdi_internal): Ditto.
192 (movti_interunit splitters): Remove
193 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
194 (movdi_interunit splitters): Ditto.
195 * config/i386/constraints.md (Ye): Remove.
196 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
198 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
200 PR target/85512
201 * config/aarch64/constraints.md (Usg): Limit to 31.
202 (Usj): Limit to 63.
204 2018-04-27 Jakub Jelinek <jakub@redhat.com>
206 PR tree-optimization/85529
207 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
208 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
209 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
210 zero extension or masking of the MSB bit.
211 (optimize_range_tests): Add FIRST_BB argument, pass it through
212 to optimize_range_tests_var_bound.
213 (maybe_optimize_range_tests, reassociate_bb): Adjust
214 optimize_range_tests callers.
216 2018-04-26 Richard Biener <rguenther@suse.de>
217 Jakub Jelinek <jakub@redhat.com>
219 * cgraph.h (symbol_table): Just declare debug method here.
220 * symtab.c (symbol_table::debug): Define.
222 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
224 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
226 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
228 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
229 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
230 (*movdi_internal): Substitute Yi and Yj constraint with x
231 and Ym and Yn constraint with y constraint. Update "isa"
232 attribute and set "preferred_for_speed" attribute from
233 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
234 (*movsi_internal): Ditto.
235 (*movdf_internal): Ditto.
236 (*movsf_internal): Ditto.
237 (*zero_extendsidi2): Ditto.
238 * config/i386/sse.md (vec_set<mode>_0): Ditto.
239 (sse2_loadld): Ditto.
240 (*vec_extract<ssevecmodelower>_0): Ditto.
241 (*vec_extractv4si_0_zext_sse4): Ditto.
242 (vec_concatv2di): Ditto.
243 (*vec_dup<mode>): Ditto.
244 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
245 * config/i386/constraints.md (Yi): Remove.
246 (Yj): Remove.
247 (Ym): Remove.
248 (Yn): Remove.
250 2018-04-26 Nathan Sidwell <nathan@acm.org>
252 * dumpfile.c (dump_open): New.
253 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
254 (dump_finish): Detect stdio/stderr by value not name.
256 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
258 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
260 2018-04-26 Tom de Vries <tom@codesourcery.com>
262 PR target/84952
263 * config/nvptx/nvptx.c (verify_neutering_jumps)
264 (verify_neutering_labels): New function
265 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
267 2018-04-26 Tom de Vries <tom@codesourcery.com>
269 PR target/84025
270 * config/nvptx/nvptx.c (needs_neutering_p): New function.
271 (nvptx_single): Use needs_neutering_p to skip over insns that do not
272 need neutering.
274 2018-04-26 Richard Biener <rguenther@suse.de>
275 Tom de Vries <tom@codesourcery.com>
277 PR lto/85422
278 * lto-streamer-out.c (output_function): Fixup loops if required to match
279 discovery done in the reader.
281 2018-04-26 Richard Biener <rguenther@suse.de>
283 PR tree-optimization/85116
284 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
285 have a loop exit from the single latch predecessor. Remove
286 case of header with just condition.
287 (ch_base::copy_headers): Exclude infinite loops from any
288 processing.
289 (pass_ch::execute): Record exits.
291 2018-04-26 Richard Biener <rguenther@suse.de>
293 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
294 prologue cost vector and pass it to vect_get_load_cost.
295 (vect_get_peeling_costs_all_drs): Likewise.
296 (vect_peeling_hash_get_lowest_cost): Likewise.
297 (vect_enhance_data_refs_alignment): Likewise.
299 2018-04-26 Richard Biener <rguenther@suse.de>
301 PR middle-end/85450
302 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
303 checking of integer<->pointer conversions.
304 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
305 sign-/zero-extending pointer types.
306 (expand_omp_for_static_chunk): Likewise.
308 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
309 Jean Lee <xiaoyur347@gmail.com>
311 * config/mips/mips.c (mips_asan_shadow_offset): New function.
312 (TARGET_ASAN_SHADOW_OFFSET): Define.
313 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
314 true for -fsanitize=address.
316 2018-04-25 Mark Wielaard <mark@klomp.org>
318 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
319 shorter ones.
321 2018-04-25 Jakub Jelinek <jakub@redhat.com>
323 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
324 than "alu", remove explicit "memory" and "imm_disp" attributes.
325 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
327 PR middle-end/85414
328 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
329 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
330 gen_lowpart_no_emit.
332 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
334 PR target/85473
335 * config/i386/i386.c (ix86_expand_builtin): Change memory
336 operand to XI, extend p0 to Pmode.
337 * config/i386/i386.md: Change unspec volatile and operand
338 1 mode to XI, change operand 0 mode to P.
340 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
342 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
343 GET_MODE_MASK before any checking.
344 (nds32_can_use_bset_p): Likewise.
345 (nds32_can_use_btgl_p): Likewise.
347 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
349 * config/nds32/nds32-doubleword.md: New define_split pattern for
350 illegal register number.
352 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
354 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
356 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
358 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
360 2018-04-25 Richard Biener <rguenther@suse.de>
362 * lto-streamer.h (LTO_major_version): Bump to 8.
364 2018-04-25 Jakub Jelinek <jakub@redhat.com>
366 * BASE-VER: Set to 9.0.0.
368 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
370 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
371 in __abskf2 and __powikf2.
373 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
375 PR target/85512
376 * config/aarch64/constraints.md (Usg, Usj): New constraints.
377 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
378 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
379 Use the above on operand 2. Reindent.
380 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
382 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
384 PR target/85485
385 * common/config/i386/i386-common.c (ix86_handle_option): Don't
386 handle OPT_mcet.
387 * config/i386/i386.opt (mcet): Removed.
388 * doc/install.texi: Remove -mcet documentation.
389 * doc/invoke.texi: Likewise.
391 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
393 PR target/85485
394 * doc/install.texi: Remove -mcet from bootstrap-cet.
396 2018-04-24 Jakub Jelinek <jakub@redhat.com>
398 PR target/85511
399 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
400 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
401 if TARGET_64BIT.
403 PR target/85503
404 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
405 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
406 containing a CONST_VECTOR.
408 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
410 * doc/install.texi: Update newlib dependency for nvptx.
412 2018-04-24 Jakub Jelinek <jakub@redhat.com>
414 PR target/85508
415 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
416 instead of INTVAL when shifting x left.
418 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
420 PR tree-optimization/85478
421 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
422 vect_grouped_store_supported for single element vectors.
424 2018-04-24 Richard Biener <rguenther@suse.de>
426 PR target/85491
427 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
428 load cost increase to the case of non-constant step.
430 2018-04-24 Jakub Jelinek <jakub@redhat.com>
432 PR target/84828
433 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
434 destination if any_malformed_asm.
436 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
438 PR middle-end/85496
439 * expr.c (store_field): In the bitfield case, if the value comes from
440 a function call and is returned in registers by means of a PARALLEL,
441 do not change the mode of the temporary unless BLKmode and VOIDmode.
443 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
445 PR rtl-optimization/85423
446 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
447 dependencies to debug insns when the previous insn is non-debug.
449 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
451 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
452 enums into a single definition.
453 (fls): Fix predicates and printing.
454 (seti): Likewise.
456 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
458 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
459 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
460 and short u6 immediate.
461 (check_if_valid_sleep_operand): Remove.
462 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
464 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
466 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
467 flag_always_save_lp condition.
468 * config/nds32/nds32.opt (malways-save-lp): New option.
470 2018-04-22 Shiva Chen <shiva0217@gmail.com>
472 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
473 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
474 * config/nds32/nds32.h
475 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
476 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
478 2018-04-22 Shiva Chen <shiva0217@gmail.com>
480 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
481 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
483 2018-04-22 Shiva Chen <shiva0217@gmail.com>
484 Chung-Ju Wu <jasonwucj@gmail.com>
486 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
487 Declare.
488 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
489 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
491 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
493 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
495 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
497 * config/nds32/nds32-protos.h (nds32_data_alignment,
498 nds32_local_alignment): Declare.
499 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
500 nds32_local_alignment): New functions.
501 (TARGET_CONSTANT_ALIGNMENT): Define.
502 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
504 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
506 * config/nds32/nds32.c
507 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
508 (TARGET_MODES_TIEABLE_P): Likewise.
510 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
512 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
513 level Ofast and Og.
515 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
516 Chung-Ju Wu <jasonwucj@gmail.com>
518 * config/nds32/constants.md (unspec_volatile_element): Add enum values
519 for unaligned access.
520 * config/nds32/nds32-intrinsic.c: Implementation of expanding
521 unaligned access.
522 * config/nds32/nds32-intrinsic.md: Likewise.
523 * config/nds32/nds32_intrinsic.h: Likewise.
524 * config/nds32/nds32.h (nds32_builtins): Likewise.
525 * config/nds32/nds32.opt (munaligned-access): New option.
526 * config/nds32/nds32.c (nds32_asm_file_start): Display
527 flag_unaligned_access status.
529 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
531 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
532 -mno-relax is present.
533 * config/riscv/linux.h (LINK_SPEC): Ditto.
535 2018-04-20 Martin Sebor <msebor@redhat.com>
537 PR c/85365
538 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
539 for null pointers.
540 (gimple_fold_builtin_stxcpy_chk): Same.
541 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
543 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
545 PR target/85456
546 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
547 __powikf2 when long double is IEEE 128-bit.
549 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
551 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
552 step to make sure stack always aligned.
554 2018-04-20 Carl Love <cel@us.ibm.com>
556 PR target/83402
557 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
558 size check for arg0.
560 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
561 Tom de Vries <tom@codesourcery.com>
563 PR target/85445
564 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
565 Emit insns for calls too.
566 (nvptx_find_par): Always look for worker-level predecessor insn.
567 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
568 calls.
569 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
570 (nvptx_process_pars): Propagate frames for calls.
572 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
574 PR target/85469
575 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
576 Removed.
577 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
578 (ix86_handle_option): Don't handle OPT_mibt.
579 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
580 __SHSTK__.
581 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
582 has_ibt and ibt.
583 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
584 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
585 (ix86_target_macros): Define __CET__ with flag_cf_protection
586 for -fcf-protection.
587 * config/i386/i386.c (isa2_opts): Remove -mibt.
588 * config/i386/i386.h (TARGET_IBT): Removed.
589 (TARGET_IBT_P): Likewise.
590 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
591 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
592 * config/i386/i386.opt (mcet): Update help message.
593 (mshstk): Likewise.
594 (mibt): Removed.
595 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
596 -mcet as an alias for -mshstk.
598 2018-04-20 Richard Biener <rguenther@suse.de>
600 PR middle-end/85475
601 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
602 complexity by forcing a single use of the multiply operand.
604 2018-04-20 Martin Jambor <mjambor@suse.cz>
606 ipa/85449
607 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
608 recursion dependency to only apply to non-clones.
610 2018-04-20 Martin Jambor <mjambor@suse.cz>
612 ipa/85447
613 * ipa-cp.c (create_specialized_node): Check that clones of
614 self-recursive edges exist during IPA-CP.
616 2018-04-19 Toon Moene <toon@moene.org>
618 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
619 by -O3.
621 2018-04-19 Jakub Jelinek <jakub@redhat.com>
623 PR tree-optimization/85467
624 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
625 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
626 VECTOR_CST element to type.
628 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
630 PR target/85397
631 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
632 * config/i386/i386.md (builtin_setjmp_setup): Removed.
633 (builtin_longjmp): Likewise.
634 (save_stack_nonlocal): New pattern.
635 (restore_stack_nonlocal): Likewise.
637 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
639 PR target/85404
640 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
641 Replace ASM_OUTPUT_LABEL with fprintf.
643 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
645 PR target/85417
646 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
647 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
648 * config/i386/i386-c.c (ix86_target_macros_internal): Also
649 define __IBT__ and __SHSTK__ for -fcf-protection.
650 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
651 TARGET_IBT.
652 (ix86_trampoline_init): Likewise.
653 (x86_output_mi_thunk): Likewise.
654 (ix86_notrack_prefixed_insn_p): Likewise.
655 (ix86_option_override_internal): Don't disallow -fcf-protection.
656 * config/i386/i386.md (rdssp<mode>): Also enable for
657 -fcf-protection.
658 (incssp<mode>): Likewise.
659 (nop_endbr): Likewise.
660 * config/i386/i386.opt (mcet): Change help message to built-in
661 functions only.
662 (mibt): Likewise.
663 (mshstk): Likewise.
664 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
665 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
666 enable CET built-in functions.
668 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
670 * common/config/i386/i386-common.c
671 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
672 OPTION_MASK_ISA_MOVDIRI_UNSET,
673 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
674 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
675 * config.gcc (movdirintrin.h): New header.
676 * config/i386/cpuid.h (bit_MOVDIRI,
677 bit_MOVDIR64B): New bits.
678 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
679 and -mmvodir64b.
680 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
681 (VOID, PVOID, PCVOID)): New function types.
682 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
683 __builtin_ia32_directstoreu_u64,
684 __builtin_ia32_movdir64b): New builtins.
685 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
686 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
687 and -mmovdiri.
688 (ix86_valid_target_attribute_inner_p): Ditto.
689 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
690 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
691 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
692 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
693 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
694 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
695 (movdiri<mode>, movdir64b_<mode>): New.
696 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
697 * config/i386/immintrin.h: Include movdirintrin.h.
698 * config/i386/movdirintrin.h: New file.
699 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
701 2018-04-19 Richard Biener <rguenther@suse.de>
703 PR middle-end/85455
704 * cfg.c (clear_bb_flags): When loop state says we have
705 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
707 2018-04-19 Richard Biener <rguenther@suse.de>
709 PR tree-optimization/84737
710 * tree-vect-data-refs.c (vect_copy_ref_info): New function
711 copying restrict info.
712 (vect_setup_realignment): Use it.
713 * tree-vectorizer.h (vect_copy_ref_info): Declare.
714 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
715 the first DR to all generated stores.
716 (vectorizable_load): Likewise for loads.
718 2018-04-19 Jakub Jelinek <jakub@redhat.com>
720 PR tree-optimization/85446
721 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
722 the integral and pointer types to have the same precision.
724 * doc/install.texi: Document --disable-cet being the default and
725 --enable-cet=auto.
727 2018-04-18 Martin Liska <mliska@suse.cz>
729 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
730 style.
732 2018-04-18 Martin Liska <mliska@suse.cz>
734 Revert
735 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
737 PR ipa/83983
738 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
739 arguments if they are comparable.
741 2018-04-18 Martin Liska <mliska@suse.cz>
743 Revert
744 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
746 PR lto/84805
747 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
748 incomplete types.
750 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
752 PR target/85388
753 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
754 ENDBR after calling __morestack.
756 2018-04-18 David Malcolm <dmalcolm@redhat.com>
758 PR jit/85384
759 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
760 by using gcc_base_ver to generate a gcc_driver_version, and use
761 it when generating GCC_DRIVER_NAME.
762 * configure: Regenerate.
764 2018-04-18 Jakub Jelinek <jakub@redhat.com>
766 PR target/81084
767 * config.gcc: Obsolete powerpc*-*-*spe*.
769 2018-04-17 Jakub Jelinek <jakub@redhat.com>
771 PR debug/84637
772 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
773 (stabstr_D): Change type of unum from unsigned int to
774 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
775 type.
777 2018-04-17 Jim Wilson <jimw@sifive.com>
779 PR 84856
780 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
781 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
782 Set arg_pointer_offset after using pretend_args_size.
784 2018-04-17 Jakub Jelinek <jakub@redhat.com>
786 PR rtl-optimization/85431
787 * dse.c (record_store): Ignore zero width stores.
789 PR sanitizer/85230
790 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
791 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
792 __builtin_stack_restore rather than after it.
793 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
794 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
795 argument instead of virtual_dynamic_stack_rtx.
797 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
799 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
800 New prototype.
801 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
802 Add note to error message to explain internal mapping of overloaded
803 built-in function name to non-overloaded built-in function name.
804 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
805 function.
807 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
809 PR target/85424
810 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
811 where the inputs overlap with the output.
813 2018-04-17 Jakub Jelinek <jakub@redhat.com>
815 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
816 (=v, v) alternative and explicit "memory" attribute.
817 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
818 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
819 attributes.
820 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
821 "sselog1" type instead of "sselog".
822 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
823 "sselog". Remove explicit "memory" attribute.
824 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
825 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
826 attributes.
827 (vec_extract_hi_v32hi): Merge all alternatives into one, use
828 "sselog1" type instead of "sselog". Remove explicit "memory"
829 attribute.
830 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
831 use "sselog1" type instead of "sselog". Remove explicit "memory"
832 attribute.
833 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
834 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
835 attributes.
836 (vec_extract_hi_v64qi): Merge all alternatives into one, use
837 "sselog1" type instead of "sselog". Remove explicit "memory"
838 attribute.
839 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
840 use "sselog1" type instead of "sselog". Remove explicit "memory"
841 attribute.
843 PR target/85430
844 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
846 PR middle-end/85414
847 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
848 on a SUBREG.
850 2018-04-17 Martin Jambor <mjambor@suse.cz>
852 PR ipa/85421
853 * ipa-cp.c (create_specialized_node): Call
854 expand_all_artificial_thunks if necessary.
856 2018-04-17 Martin Liska <mliska@suse.cz>
858 PR lto/85405
859 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
860 in message, remote space in between '_G' and '('.
862 2018-04-17 Jakub Jelinek <jakub@redhat.com>
864 PR target/85281
865 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
866 avx512f_vmcmp<mode>3<round_saeonly_name>,
867 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
868 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
869 avx512f_rndscale<mode><round_saeonly_name>,
870 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
871 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
872 Use %<iptr>2 instead of %2 for -masm=intel.
873 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
874 avx512f_vcvttss2usi<round_saeonly_name>,
875 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
876 -masm=intel.
877 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
878 avx512f_vcvttsd2usi<round_saeonly_name>,
879 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
880 Use %q1 instead of %1 for -masm=intel.
881 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
882 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
883 of %3 for -masm=intel.
884 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
885 -masm=intel.
886 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
887 -masm=intel.
888 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
889 -masm=intel.
890 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
891 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
892 %g1.
893 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
894 -masm=intel.
895 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
896 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
897 %g1 and one with %0 and %1.
898 (avx512er_vmrcp28<mode><round_saeonly_name>,
899 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
900 %1 for -masm=intel.
901 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
902 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
903 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
904 of %0 and %{%4%} for -masm=intel.
905 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
906 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
907 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
908 order of %0 and %{%5%}%{z%} for -masm=intel.
910 2018-04-17 Jan Hubicka <jh@suse.cz>
912 PR lto/85405
913 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
915 2018-04-17 Martin Liska <mliska@suse.cz>
917 PR ipa/85329
918 * multiple_target.c (create_dispatcher_calls): Set apostrophes
919 for target_clone error message. Make default implementation
920 clone to be a local declaration.
921 (separate_attrs): Add new argument and check for an empty
922 string.
923 (expand_target_clones): Handle it.
924 (ipa_target_clone): Make redirection just for target_clones
925 functions.
927 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
928 Tom de Vries <tom@codesourcery.com>
930 PR middle-end/84955
931 * omp-expand.c (expand_oacc_for): Add dummy false branch for
932 tiled basic blocks without omp continue statements.
934 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
936 PR target/83660
937 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
938 vec_extract expression as having side effects to make sure it gets
939 a cleanup point.
941 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
943 PR target/85403
944 * config/i386/i386.c (get_builtin_code_for_version): Check
945 error_mark_node.
947 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
949 PR target/84331
950 * gcc/config.gcc: Support "skylake".
951 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
952 PROCESSOR_SKYLAKE.
953 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
954 (processor_target_table): Add "skylake".
955 (ix86_option_override_internal): Add "skylake".
956 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
957 PROCESSOR_CANNONLAKE.
958 (get_builtin_code_for_version): Fix priority for
959 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
960 PROCESSOR_SKYLAKE-AVX512.
961 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
962 (processor_type): Add PROCESSOR_SKYLAKE.
964 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
965 Jason Merrill <jason@redhat.com>
967 PR c++/85112
968 * convert.c (convert_to_integer_1): Use direct recursion for
969 enumeral types and types with a precision less than the number
970 of bits in their mode.
972 2018-04-16 Julia Koval <julia.koval@intel.com>
974 PR target/84413
975 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
976 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
978 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
980 PR target/85293
981 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
982 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
983 and -mno-direct-move.
985 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
987 PR target/83402
988 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
989 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
990 Ensure negative shifts result in {0}.
992 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
994 PR rtl-optimization/79916
995 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
996 regs (if any) to define how to gnerate SD moves when LRA is in
997 progress.
999 2018-04-13 Jakub Jelinek <jakub@redhat.com>
1001 PR rtl-optimization/85393
1002 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
1003 * except.c (expand_dw2_landing_pad_for_region): Make static.
1004 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
1005 a label and unconditional jump to old_bb, rather than
1006 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
1007 basic block.
1009 PR rtl-optimization/85376
1010 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
1011 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
1012 instead of a specific value.
1014 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1015 Bin Cheng <bin.cheng@arm.com>
1017 PR tree-optimization/82965
1018 PR tree-optimization/83991
1019 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
1020 by_profile_only parameter.
1021 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
1022 information if the loop was predicted to iterate too many times.
1023 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
1025 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
1027 PR lto/71991
1028 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
1029 always inline.
1031 2018-04-13 Martin Liska <mliska@suse.cz>
1032 Jakub Jelinek <jakub@redhat.com>
1034 PR middle-end/81657
1035 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
1036 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
1037 * builtins.c (expand_builtin_memory_copy_args): Use
1038 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
1039 handle dest_addr == pc_rtx.
1041 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
1043 PR target/85291
1044 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
1045 asked to not generate direct moves.
1046 (fix_trunc<mode>si2_stfiwx): Similar.
1047 (fix_trunc<mode>si2_internal): Similar.
1049 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1051 PR debug/83157
1052 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
1053 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
1054 lookup if dest in some wider mode is known to be const0_rtx and
1055 if so, record permanent equivalence for it to be ZERO_EXTEND of
1056 the narrower mode destination.
1058 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1060 * lto-streamer-out.c (output_function): Revert 259346.
1061 * omp-expand.c (expand_oacc_for): Likewise.
1063 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
1065 PR rtl-optimization/85354
1066 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
1067 * sel-sched.c (sel_global_init): ... here.
1069 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
1071 PR target/85238
1072 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
1073 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
1074 mode for PE-COFF targets.
1075 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
1076 (i386_pe_asm_lto_end): Likewise.
1077 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
1078 (TARGET_ASM_LTO_END): Likewise.
1079 * config/i386/winnt.c (saved_debug_info_level): New static variable.
1080 (i386_pe_asm_lto_start): New function.
1081 (i386_pe_asm_lto_end): Likewise.
1083 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
1084 Richard Biener <rguenther@suse.de>
1086 PR middle-end/84955
1087 * lto-streamer-out.c (output_function): Fix CFG loop state before
1088 streaming out.
1089 * omp-expand.c (expand_oacc_for): Handle calls to internal
1090 functions like regular functions.
1092 2018-04-12 Richard Biener <rguenther@suse.de>
1094 PR lto/85371
1095 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
1096 for the early LTO debug to properly generate references to it
1097 during DIE emission. Do not re-use that for the skeleton for
1098 split-dwarf.
1099 (dwarf2out_early_finish): Likewise.
1101 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1103 PR target/85328
1104 * config/i386/sse.md
1105 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
1106 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
1107 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
1108 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
1109 and output is a reg, avoid creating invalid lowpart subreg, but
1110 instead split into a 512-bit move. Don't split if not AVX512VL,
1111 input is xmm16+ reg and output is a mem.
1112 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
1113 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
1114 xmm16+ reg and output is a mem.
1116 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1118 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
1119 also for flag_dwarf2_cfi_asm.
1121 2018-04-12 Jakub Jelinek <jakub@redhat.com>
1123 PR rtl-optimization/85342
1124 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
1125 a bool scalar var inside of the loop instead. Don't try to update
1126 recog_data.operand after failed apply_change_group.
1128 2018-04-12 Tom de Vries <tom@codesourcery.com>
1130 PR target/85296
1131 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
1132 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
1133 array with flexible array member as array without given dimension.
1134 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
1135 argument for undefined param to true.
1137 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
1139 PR target/85321
1140 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1141 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
1142 from PowerPC section.
1143 * config/rs6000/sysv4.opt (mcall-): Improve help text.
1144 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
1145 help text that is too long.
1146 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
1147 help text that is too long.
1148 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
1149 help text that is too long.
1151 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
1153 * config/alpha/alpha.md (stack_probe_internal): Rename
1154 from "probe_stack". Update all callers.
1156 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1158 PR rtl-optimization/84566
1159 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
1160 sched_macro_fuse_insns.
1162 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1164 PR target/84301
1165 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
1166 (compute_block_dependences): ... from here.
1168 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1170 PR tree-optimization/85331
1171 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
1172 from int to HOST_WIDE_INT.
1174 2018-04-11 Martin Jambor <mjambor@suse.cz>
1176 PR ipa/84149
1177 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
1178 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
1179 not the same as the source val.
1180 (cgraph_edge_brings_value_p): New parameter.
1181 (gather_edges_for_value): Pass destination value to
1182 cgraph_edge_brings_value_p.
1183 (perhaps_add_new_callers): Likewise.
1184 (get_info_about_necessary_edges): Likewise and exclude values brought
1185 only by self-recursive edges.
1186 (create_specialized_node): Redirect only clones of self-calling edges.
1187 (+self_recursive_pass_through_p): New function.
1188 (find_more_scalar_values_for_callers_subset): Use it.
1189 (find_aggregate_values_for_callers_subset): Likewise.
1190 (known_aggs_to_agg_replacement_list): Removed.
1191 (decide_whether_version_node): Re-calculate known constants for all
1192 remaining context clones.
1194 2018-04-11 Richard Biener <rguenther@suse.de>
1196 PR lto/85339
1197 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
1198 from early DWARF output.
1199 (dwarf2out_early_finish): Output line info unconditionally into
1200 early DWARF and add reference to it.
1202 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1204 PR target/85281
1205 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
1206 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
1207 other than V2DFmode using iptr mode attribute.
1208 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
1210 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
1212 PR rtl-optimization/84659
1213 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
1215 2018-04-11 Jakub Jelinek <jakub@redhat.com>
1217 PR debug/85302
1218 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
1219 SIZEP is NULL.
1220 (output_loc_list): Pass address of a dummy size variable even in the
1221 locview handling loop.
1222 (index_location_lists): Add comment on why skip_loc_list_entry can't
1223 call size_of_locs.
1225 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
1227 PR target/85261
1228 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
1229 into register.
1231 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
1233 PR target/85321
1234 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
1235 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
1236 and -mstring-compare-inline-limit.
1238 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1240 PR target/85287
1241 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
1242 for stack clash protection in a register whenever we need it to be in
1243 a register.
1245 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1247 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
1248 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
1250 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
1252 PR target/85321
1253 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
1254 the help text.
1255 (mlong-double-): Ditto.
1256 * config/rs6000/sysv4.opt (msdata=): Ditto.
1257 (mtls-size=): Ditto.
1259 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
1261 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1262 erroneous entries for
1263 "vector int vec_ldl (int, long int *)", and
1264 "vector unsigned int vec_ldl (int, unsigned long int *)".
1265 Add comments and entries for
1266 "vector bool char vec_ldl (int, bool char *)",
1267 "vector bool short vec_ldl (int, bool short *)",
1268 "vector bool int vec_ldl (int, bool int *)",
1269 "vector bool long long vec_ldl (int, bool long long *)",
1270 "vector pixel vec_ldl (int, pixel *)",
1271 "vector long long vec_ldl (int, long long *)",
1272 "vector unsigned long long vec_ldl (int, unsigned long long *)".
1273 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
1274 type tree bool_long_long_type_node and correct definition of
1275 bool_V2DI_type_node to make reference to this new type tree.
1276 (rs6000_mangle_type): Replace erroneous reference to
1277 bool_long_type_node with bool_long_long_type_node.
1278 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
1279 comments to emphasize sign distinctions for char and int types and
1280 replace RS6000_BTI_bool_long constant with
1281 RS6000_BTI_bool_long_long constant. Also add comment to restrict
1282 use of RS6000_BTI_pixel.
1283 (bool_long_type_node): Remove this macro definition.
1284 (bool_long_long_type_node): New macro definition
1286 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1288 PR rtl-optimization/85300
1289 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
1290 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
1291 simplify_unary_operation fails.
1293 2018-04-10 Martin Liska <mliska@suse.cz>
1295 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
1296 cgraph_edge and ipa_ref.
1298 2018-04-10 Jakub Jelinek <jakub@redhat.com>
1300 PR target/85177
1301 PR target/85255
1302 * config/i386/sse.md
1303 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
1304 computation of the VEC_MERGE selector from mask.
1305 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
1306 Fix decoding of the VEC_MERGE selector into mask.
1308 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1310 PR tree-optimization/85286
1311 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
1313 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
1315 * final.c (final_1): Set insn_last_address as well as
1316 insn_current_address.
1318 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1320 PR target/85173
1321 * explow.c (emit_stack_probe): Call validize_mem on memory location
1322 before passing it to gen_probe_stack. Create address operand and
1323 legitimize it for the probe_stack_address case.
1325 2018-04-09 Jan Hubicka <jh@suse.cz>
1327 PR lto/85078
1328 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
1329 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
1330 * tree.c (free_lang_data_in_type): Fix handling of binfos;
1331 walk basetypes.
1332 (free_lang_data): Rebuild type inheritance graph.
1334 2018-04-09 Martin Sebor <msebor@redhat.com>
1336 * invoke.texi (-finline-small-functions): Mention other optimization
1337 options.
1338 (-findirect-inlining, -fpartial-inlining): Same.
1339 (-finline-functions-called-once): Same.
1340 (-freorder-blocks-and-partition): Same.
1342 2018-04-09 Jan Hubicka <jh@suse.cz>
1344 PR rtl/84058
1345 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
1346 jumps; choose last target that matches the criteria (i.e.
1347 no partition changes for non-crossing jumps).
1348 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
1349 support for redirecting crossing jumps to non-crossing.
1351 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
1353 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
1354 also for naked functions.
1356 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
1358 * config/arc/arc.md (add_shift): New pattern.
1359 (add_shift2): Likewise.
1360 (sub_shift): Likewise.
1361 (sub_shift_cmp0_noout): Likewise.
1362 (compare_si_ashiftsi): Likewise.
1363 (xbfu_cmp0_noout): New combine pattern.
1364 (xbfu_cmp0"): Likewise.
1365 (movsi_set_cc_insn): Place the predicable variant first.
1366 (commutative_binary_cmp0_noout): Remove clobber.
1367 (commutative_binary_cmp0): New pattern.
1368 (noncommutative_binary_cmp0): Likewise.
1369 (noncommutative_binary_cmp0_noout): Likewise.
1370 (noncommutative_binary_comparison_result_used): Removed.
1371 (rsub_cmp0): New pattern.
1372 (rsub_cmp0_noout): Likewise.
1373 (extzvsi): Changed, keep only meaningful variants.
1374 (SQH, SEZ): New iterators.
1375 (SQH_postfix): New mode attribute.
1376 (SEZ_prefix): New code attribute.
1377 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
1378 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
1379 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
1380 of numerical value.
1381 (noncommutative_operator): Check the availability of barrel
1382 shifter option.
1384 2018-04-09 Richard Biener <rguenther@suse.de>
1386 PR tree-optimization/85284
1387 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
1388 Only use the niter constraining form of simple_iv when the exit
1389 is always executed.
1391 2018-04-09 Tom de Vries <tom@codesourcery.com>
1393 PR target/84041
1394 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
1395 (define_expand "*memory_barrier"): New define_expand.
1396 (define_insn "memory_barrier"): New insn.
1398 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1400 PR rtl-optimization/80463
1401 PR rtl-optimization/83972
1402 PR rtl-optimization/83480
1404 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
1405 correct producer for the insn.
1406 (tidy_control_flow): Fixup seqnos in case of debug insns.
1408 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1410 PR rtl-optimization/83913
1412 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
1413 different sched-times when merging exprs.
1415 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1417 PR rtl-optimization/83962
1419 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
1420 tidy_fallthru_edge and tidy_control_flow.
1422 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
1424 PR rtl-optimization/83530
1426 * sel-sched.c (force_next_insn): New global variable.
1427 (remove_insn_for_debug): When force_next_insn is true, also leave only
1428 next insn in the ready list.
1429 (sel_sched_region): When the region wasn't scheduled, make another pass
1430 over it with force_next_insn set to 1.
1432 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1434 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
1435 into tm_file.
1436 * config/nds32/constants.md (unspec_volatile_element): Add enum values
1437 for interrupt control.
1438 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1439 functions for interrupt control.
1440 * config/nds32/nds32-intrinsic.md: Likewise.
1441 * config/nds32/nds32_intrinsic.h: Likewise.
1442 * config/nds32/nds32.h (nds32_builtins): Likewise.
1444 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1446 * config/nds32/nds32.c (nds32_init_machine_status,
1447 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
1448 strict_aligned_p field.
1449 (nds32_expand_to_rtl_hook): New function.
1450 (TARGET_EXPAND_TO_RTL_HOOK): Define.
1451 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
1453 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1454 Chung-Ju Wu <jasonwucj@gmail.com>
1456 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
1457 * config/nds32/nds32-n7.md: New file.
1458 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
1459 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
1460 pipeline.
1461 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
1462 * config/nds32/nds32.md (pipeline_model): Add n7.
1463 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
1464 * config/nds32/pipelines.md: Include n7 settings.
1466 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1467 Chung-Ju Wu <jasonwucj@gmail.com>
1469 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
1470 * config/nds32/nds32-e8.md: New file.
1471 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
1472 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
1473 pipeline.
1474 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
1475 * config/nds32/nds32.md (pipeline_model): Add e8.
1476 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
1477 * config/nds32/pipelines.md: Include e8 settings.
1479 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1480 Chung-Ju Wu <jasonwucj@gmail.com>
1482 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
1483 * config/nds32/nds32-n8.md: New file.
1484 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
1485 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
1486 pipeline.
1487 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
1488 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
1489 * config/nds32/nds32.md (pipeline_model): Add n8.
1490 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
1491 * config/nds32/pipelines.md: Include n8 settings.
1493 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
1494 Chung-Ju Wu <jasonwucj@gmail.com>
1496 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
1497 * config/nds32/nds32-n9-2r1w.md: New file.
1498 * config/nds32/nds32-n9-3r2w.md: New file.
1499 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
1500 nds32_register_ports): New or modify for cpu n9.
1501 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
1502 pipeline.
1503 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
1504 * config/nds32/nds32-utils.c: New file.
1505 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
1506 TARGET_MUL_SLOW): Define.
1507 * config/nds32/nds32.md (pipeline_model): New attribute.
1508 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
1509 New options that support cpu n9.
1510 * config/nds32/pipelines.md: Include n9 settings.
1511 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
1513 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
1515 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
1516 information if necessary.
1517 (output_cond_branch_compare_zero): Likewise.
1518 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
1519 (nds32_target_alignment): Refine for alignment.
1520 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
1521 (FUNCTION_BOUNDARY): Modify.
1522 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
1523 align case.
1524 * config/nds32/nds32.opt (malways-align, malign-functions): New.
1526 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
1528 * config/nds32/constants.md (unspec_volatile_element): Add values for
1529 TLB operation and data prefetch.
1530 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
1531 functions for TLB operation and data prefetch.
1532 * config/nds32/nds32-intrinsic.md: Likewise.
1533 * config/nds32/nds32_intrinsic.h: Likewise.
1534 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
1535 (nds32_print_operand): Likewise.
1536 * config/nds32/nds32.h (nds32_builtins): Likewise.
1538 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
1539 Andrew Pinski <pinsika@gcc.gnu.org>
1541 PR middle-end/82976
1542 * match.pd: Use constant_boolean_node of correct type instead of
1543 boolean_true_node or boolean_false_node for simplifying
1544 pointer comparisons to zero.
1546 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1548 PR tree-optimization/80021
1549 * tree.c (verify_type_variant): Make error call in verify_variant_match
1550 translatable and remove final full stop.
1552 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1554 * config/nds32/constants.md (unspec_volatile_element): Add
1555 UNSPEC_VOLATILE_EH_RETURN.
1556 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
1557 nds32_output_stack_pop): Support dwarf exception handling process.
1558 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
1559 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
1560 exception handling process.
1561 (nds32_compute_stack_frame): Likewise.
1562 (nds32_return_addr_rtx): Likewise.
1563 (nds32_initial_elimination_offset): Likewise.
1564 (nds32_expand_prologue): Likewise.
1565 (nds32_expand_epilogue): Likewise.
1566 (nds32_dynamic_chain_address): New function.
1567 * config/nds32/nds32.h (machine_function): Add fields for dwarf
1568 exception handling.
1569 (DYNAMIC_CHAIN_ADDRESS): Define.
1570 (EH_RETURN_DATA_REGNO): Define.
1571 (EH_RETURN_STACKADJ_RTX): Define.
1572 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
1573 patterns for dwarf exception handling.
1575 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1577 * config/nds32/nds32.h: Clean up obsolete macros.
1579 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1581 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1582 Add enum values for particular instructions.
1583 * config/nds32/nds32-intrinsic.c: Implementation of expanding
1584 particular intrinsic functions.
1585 * config/nds32/nds32-intrinsic.md: Likewise.
1586 * config/nds32/nds32_intrinsic.h: Likewise.
1587 * config/nds32/nds32.h (nds32_builtins): Likewise.
1588 * config/nds32/nds32.md (type): Add pbsad and pbsada.
1589 (btst, ave): New patterns for particular instructions.
1591 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1593 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
1594 Add enum values for atomic load/store and memory sync.
1595 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
1596 and memory sync.
1597 * config/nds32/nds32-intrinsic.md: Likewise.
1598 * config/nds32/nds32_intrinsic.h: Likewise.
1599 * config/nds32/nds32.h (nds32_builtins): Likewise.
1601 2018-04-07 Jakub Jelinek <jakub@redhat.com>
1603 PR tree-optimization/85257
1604 * fold-const.c (native_encode_vector): If not all elts could fit
1605 and off is -1, return 0 rather than offset.
1606 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
1607 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
1608 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
1609 adjust buffer in native_interpret_expr call.
1611 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1613 * config/nds32/constants.md (unspec_volatile_element): Add cache
1614 control enum values.
1615 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
1616 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
1617 * config/nds32/nds32.c (nds32_cctl_names): New.
1618 (nds32_print_operand): Handle cache control register names.
1619 * config/nds32/nds32.h (nds32_builtins): New enum values.
1620 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
1621 macros.
1622 * config/nds32/nds32.md (type): Add mmu.
1623 * config/nds32/pipelines.md (simple_insn): Add mmu.
1625 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1627 * config/nds32/nds32.md (type): Remove call.
1628 * config/nds32/pipelines.md (simple_insn): Likewise.
1630 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1632 * config/nds32/constants.md (unspec_volatile_element): Add
1633 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
1634 UNSPEC_VOLATILE_FMFCFG.
1635 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
1636 description for fmfcfg and fmfcsr.
1637 (bdesc_1arg): Add fmtcsr.
1638 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
1639 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
1640 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
1641 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
1642 unspec_fmfcfg): New patterns.
1643 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
1644 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
1645 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
1646 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
1647 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
1648 __nds32__fmfcfg): Define.
1650 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
1652 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
1653 intrinsic register names.
1654 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
1655 intrinsic register enum values and macros.
1657 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
1659 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
1660 for load/store addressing form.
1661 (nds32_print_operand_address): Likewise.
1663 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
1665 PR target/85196
1666 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
1667 based on LABEL_REF. Remove useless assertion.
1668 (pic_address_needs_scratch): Fix formatting.
1669 (sparc_legitimize_pic_address): Minor tweaks.
1670 (sparc_delegitimize_address): Adjust assertion accordingly.
1671 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
1672 into symbolic_operand.
1673 (movsi_high_pic_label_ref): Likewise.
1674 (movsi_lo_sum_pic_label_ref): Likewise.
1675 (movdi_pic_label_ref): Likewise.
1676 (movdi_high_pic_label_ref): Likewise.
1677 (movdi_lo_sum_pic_label_ref): Likewise.
1679 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
1681 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
1682 custom LIB_SPEC setup.
1684 2018-04-06 Ruslan Bukin <br@bsdpad.com>
1685 Kito Cheng <kito.cheng@gmail.com>
1687 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
1688 * config/riscv/freebsd.h: New.
1690 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1692 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
1693 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
1694 file.
1696 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1697 Kito Cheng <kito.cheng@gmail.com>
1699 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
1700 nds32_output_call, nds32_symbol_binds_local_p): New functions.
1701 * config/nds32/nds32-protos.h (nds32_output_call,
1702 nds32_output_return): Declare.
1703 * config/nds32/nds32.md: Refine all the call and return patterns.
1705 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1707 PR debug/85252
1708 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
1709 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
1711 PR rtl-optimization/84872
1712 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
1713 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
1714 EDGE_CROSSING edge.
1716 2018-04-06 Tamar Christina <tamar.christina@arm.com>
1718 * expr.c (copy_blkmode_to_reg): Revert 254862.
1719 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
1721 2018-04-06 Richard Biener <rguenther@suse.de>
1723 PR middle-end/85244
1724 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
1725 after seeing a component reference with an adjacent field. Treat
1726 refs to arrays at struct end of external decls similar to
1727 refs to unconstrained commons.
1729 2018-04-06 Jakub Jelinek <jakub@redhat.com>
1731 PR sanitizer/85213
1732 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
1733 look through SAVE_EXPRs with non-side-effects argument. Adjust
1734 recursive calls.
1735 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
1736 save_p here.
1738 2018-04-06 Richard Biener <rguenther@suse.de>
1740 PR middle-end/85180
1741 * alias.c (find_base_term): New wrapper around find_base_term
1742 unwinding CSELIB_VAL_PTR changes.
1743 (find_base_term): Do not restore CSELIB_VAL_PTR during the
1744 recursion.
1746 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1748 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
1749 instructions.
1750 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
1751 constant definitions.
1752 ("nop"): lr 0,0 -> nopr r0
1753 ("nop_lr0", "nop_lr1"): New insn definitions.
1755 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
1757 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
1758 NDS32_V3PUSH_AVAILABLE_P macro.
1760 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
1761 Chung-Ju Wu <jasonwucj@gmail.com>
1763 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
1764 (nds32*-*-*): Add float and fpu_config into supported_defaults.
1765 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
1766 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
1767 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
1768 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
1769 * config/nds32/constraints.md: New constraints and checking for hard
1770 float configuration.
1771 * config/nds32/iterators.md: New mode iterator and attribute for hard
1772 float configuration.
1773 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
1774 patterns.
1775 * config/nds32/nds32-fpu.md: New file.
1776 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
1777 deal with hard float code generation.
1778 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
1779 ARCH_V3S.
1780 (abi_type, float_reg_number): New enum type.
1781 * config/nds32/nds32-predicates.c: New predicates for hard float.
1782 * config/nds32/nds32-protos.h: Declare functions for hard float.
1783 * config/nds32/nds32.c: Implementation for hard float configuration.
1784 * config/nds32/nds32.h: Definitions for hard float configuration.
1785 * config/nds32/nds32.md: Include hard float machine description and
1786 modify patterns for hard float configuration.
1787 * config/nds32/nds32.opt: New options for hard float configuration.
1788 * config/nds32/predicates.md: New predicates for hard float
1789 configuration.
1791 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1793 * common/config/nds32/nds32-common.c
1794 (nds32_option_optimization_table): Enable -mreleax-hint by default.
1796 2018-04-05 Jakub Jelinek <jakub@redhat.com>
1798 PR middle-end/85195
1799 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
1800 CONSTRUCTOR_ELT (ctor, ...)->value.
1802 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
1804 PR target/85193
1805 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
1807 2018-04-05 Tom de Vries <tom@codesourcery.com>
1809 PR target/85204
1810 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
1811 cond jump.
1813 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1814 Kito Cheng <kito.cheng@gmail.com>
1816 * config/nds32/constraints.md (U33): Fine-tune checking condition.
1817 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
1818 * config/nds32/nds32.h (nds32_16bit_address_type): Add
1819 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
1821 2018-04-05 Shiva Chen <shiva0217@gmail.com>
1822 Kito Cheng <kito.cheng@gmail.com>
1824 * config/nds32/constraints.md (Ufe): New memory constraint.
1825 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
1826 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
1827 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
1828 operands.
1829 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
1830 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
1832 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1834 * config/nds32/nds32.md: Use optimize_size in the condition for
1835 alu-shift instructions.
1837 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1839 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
1841 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32.md (negsi2): Refine pattern.
1845 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1846 Chung-Ju Wu <jasonwucj@gmail.com>
1848 * config/nds32/iterators.md (shift_rotate): New code iterator.
1849 (shift): New code attribute.
1850 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
1851 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
1852 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
1853 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1854 bit-wise operations.
1855 (andsi3, *andsi3): Ditto.
1856 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1857 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1858 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1859 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1860 nds32_ior_operand, nds32_xor_operand): New predicates.
1862 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1864 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1865 (addsi3, subsi3): ... this.
1867 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1869 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1871 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1873 * config/nds32/nds32.md: Adjust indention.
1875 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1877 * config/nds32/nds32.md (feature): New attribute.
1879 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1881 * config/nds32/nds32.md (subtype): New attribute.
1883 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1885 PR target/85203
1886 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1887 expansion to perform a bitwise AND of the argument followed by a
1888 boolean negation of the result.
1890 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1892 PR rtl-optimization/84878
1893 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1894 the basic block. Assert the use reference is not artificial and that
1895 it has an associated insn.
1897 2018-04-04 Michael Matz <matz@suse.de>
1899 * builtins.c (compute_objsize): Pass correct operand
1900 to array_at_struct_end_p.
1902 2018-04-04 Richard Biener <rguenther@suse.de>
1904 PR lto/85176
1905 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1906 from contexts for DINFO_LEVEL_TERSE and below.
1908 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1910 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1911 resiter_operand condition.
1912 * config/nds32/nds32.md (*move<mode>): Ditto.
1914 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1915 Monk Chiang <sh.chiang04@gmail.com>
1917 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1919 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1921 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1923 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1924 Kito Cheng <kito.cheng@gmail.com>
1926 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1927 nds32_cond_code_str, output_cond_branch,
1928 output_cond_branch_compare_zero, nds32_expand_cbranch,
1929 nds32_expand_cstore, nds32_expand_movcc,
1930 nds32_output_cbranchsi4_equality_zero,
1931 nds32_output_cbranchsi4_equality_reg,
1932 nds32_output_cbranchsi4_equality_reg_or_const_int,
1933 nds32_output_cbranchsi4_greater_less_zero: New functions.
1934 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1935 nds32_expand_cstore, nds32_expand_movcc,
1936 nds32_output_cbranchsi4_equality_zero,
1937 nds32_output_cbranchsi4_equality_reg,
1938 nds32_output_cbranchsi4_equality_reg_or_const_int,
1939 nds32_output_cbranchsi4_greater_less_zero): Declare.
1940 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1941 nds32_rimm11s_operand): New predicates.
1942 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1943 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1944 patterns.
1946 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1948 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1949 * config/nds32/nds32.md: Ditto.
1950 * config/nds32/pipelines.md: Ditto.
1952 2018-04-04 Richard Biener <rguenther@suse.de>
1954 PR tree-optimization/85168
1955 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1956 propagating abnormals.
1958 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1960 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1962 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1963 Kito Cheng <kito.cheng@gmail.com>
1965 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1966 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1967 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1968 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1969 * config/nds32/nds32.md (sibcall_internal): New.
1970 (sibcall_register): Remove.
1971 (sibcall_immediate): Remove.
1972 (sibcall_value_internal): New.
1973 (sibcall_value_register): Remove.
1974 (sibcall_value_immediate): Remove.
1975 * config/nds32/predicates.md (nds32_general_register_operand): New.
1976 (nds32_call_address_operand): New.
1978 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1980 PR rtl-optimization/85167
1981 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1982 bb_defs if *split_p, instead preinitialize it to NULL.
1984 PR tree-optimization/85156
1985 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1986 evaluating the argument multiple times.
1988 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1990 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1991 than vector.
1992 (_mm_cvtpd_ps): Likewise.
1993 (_mm_cvttpd_epi32): Likewise.
1994 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1995 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1996 vector, pixel, and bool following altivec.h include.
1998 2018-04-03 Martin Sebor <msebor@redhat.com>
2000 * doc/extend.texi (Common Function Attributes): Clarify.
2001 (const attribute): Likewise.
2002 (pure attribute): Likewise.
2004 2018-04-03 Jakub Jelinek <jakub@redhat.com>
2006 PR target/85169
2007 * config/i386/i386.c (ix86_expand_vector_set): Use
2008 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
2010 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
2012 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
2013 instructions when changing rounding bits to preserve precision bits
2014 in the x87 control word.
2016 2018-04-03 Martin Liska <mliska@suse.cz>
2018 PR tree-optimization/82491
2019 * rtl.h (strip_offset_and_add): Replace += suboffset with
2020 poly_uint64 () + suboffset.
2022 2018-03-29 Martin Liska <mliska@suse.cz>
2023 Martin Jambor <mjambor@suse.cz>
2025 PR ipa/84947
2026 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
2027 param_type is not an integral or pointer type.
2029 2018-04-03 Richard Biener <rguenther@suse.de>
2031 * sese.h (recompute_all_dominators): Remove.
2033 2018-04-02 Martin Sebor <msebor@redhat.com>
2035 * doc/invoke.texi (-Wrestrict): Fix typos.
2037 2018-04-02 Jim Wilson <jimw@sifive.com>
2039 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
2040 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
2041 (<optab>di3, <optab>si3_extend): Likewise.
2042 (<optab>si3_mask, <optab>si3_mask_1): New.
2043 (<optab>di3_mask, <optab>di3_mask_1): New.
2044 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
2045 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
2046 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
2048 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
2050 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
2051 example.
2053 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
2055 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
2056 (nds32_canonicalize_comparison): New function.
2058 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2059 Kito Cheng <kito.cheng@gmail.com>
2060 Kuan-Lin Chen <kuanlinchentw@gmail.com>
2062 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
2063 * config/nds32/constants.md (unspec_volatile_element): Add
2064 UNSPEC_VOLATILE_RELAX_GROUP.
2065 * config/nds32/nds32-relax-opt.c: New file.
2066 * config/nds32/nds32-predicates.c
2067 (nds32_symbol_load_store_p): New function.
2068 * config/nds32/nds32-protos.h
2069 (nds32_symbol_load_store_p): Declare function.
2070 (make_pass_nds32_relax_opt): Declare new rtl pass function.
2071 * config/nds32/nds32.c
2072 (nds32_register_pass): New function to register pass.
2073 (nds32_register_passes): New function to register passes.
2074 * config/nds32/nds32.md (relax_group): New pattern.
2075 * config/nds32/nds32.opt (mrelax-hint): New option.
2076 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
2078 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
2080 * config/nds32/t-nds32: Modify files dependency.
2082 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2084 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
2085 (PROFILE_HOOK): Define its implementation.
2087 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
2089 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
2090 type and 32-bit size.
2092 2018-04-01 Jakub Jelinek <jakub@redhat.com>
2094 PR middle-end/85090
2095 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
2096 (V_128_256): New mode iterator.
2097 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
2098 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
2099 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
2100 of V.
2101 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
2102 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
2104 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
2106 PR target/83315
2107 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
2108 NaN inputs correctly.
2110 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
2112 PR target/80546
2113 * config/rs6000/vsx.md (??r): New mode attribute.
2114 (*vsx_mov<mode>_64bit): Use it.
2115 (*vsx_mov<mode>_32bit): Likewise.
2117 2018-03-30 Martin Sebor <msebor@redhat.com>
2119 PR tree-optimization/84818
2120 * builtins.c (check_access): Use warning_n.
2122 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2124 PR target/83822
2125 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
2126 condition.
2127 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
2128 condition.
2130 2018-03-30 Julia Koval <julia.koval@intel.com>
2132 PR target/84413
2133 * x86-tune.def (movx, partial_reg_dependency): Enable for
2134 m_SKYLAKE_AVX512.
2136 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
2138 PR inline-asm/84985
2139 * lra-constraints.c (process_alt_operands): Move setting
2140 this_alternative_matches below.
2142 2018-03-29 Martin Liska <mliska@suse.cz>
2144 PR lto/84995.
2145 * doc/invoke.texi: Document how LTO works with debug info.
2146 Describe auto-load support of binutils. Mention 'x86-64'
2147 as valid option value of -march option.
2149 2018-03-29 Jakub Jelinek <jakub@redhat.com>
2151 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
2153 PR c/85094
2154 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
2155 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
2156 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
2157 checking.
2159 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2161 PR target/84912
2162 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
2163 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
2164 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
2165 for RS6000_BTM_POWERPC64.
2166 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
2167 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
2168 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
2169 definition.
2170 (DIVDE): Use it.
2171 (DIVDEU): Likewise.
2173 2018-03-28 Carl Love <cel@us.ibm.com>
2175 Revert
2176 2017-09-27 Carl Love <cel@us.ibm.com>
2178 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
2179 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
2180 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
2181 fctiw instruction.
2183 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2185 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
2186 instead of __vector bool.
2187 (_mm_max_pu8): Likewise.
2188 (_mm_min_pi16): Likewise.
2190 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
2192 PR target/84912
2193 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
2194 (DIVWEUO): Likewise.
2195 (DIVDEO): Likewise.
2196 (DIVDEUO): Likewise.
2197 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
2198 DIVWEUO and DIVDEUO.
2199 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
2200 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
2201 (div_extend): Likewise.
2202 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
2203 builtin function.
2204 (__builtin_divweuo): Likewise.
2205 (__builtin_divdeo): Likewise.
2206 (__builtin_divdeuo): Likewise.
2208 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2210 PR target/85095
2211 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
2212 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
2214 PR tree-optimization/82004
2215 * gimple-match-head.c (optimize_pow_to_exp): New function.
2216 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
2217 Don't fold to exp if optimize_pow_to_exp is false.
2219 2018-03-28 Martin Liska <mliska@suse.cz>
2221 PR other/84819
2222 * calls.c (initialize_argument_information): Fix trailing space.
2223 * common.opt: Fix typo and provide better explanation for
2224 -fsanitize-coverage option.
2225 * config/i386/i386.opt: Fix typo.
2227 2018-03-28 Jakub Jelinek <jakub@redhat.com>
2228 Martin Liska <mliska@suse.cz>
2230 PR sanitizer/85081
2231 * gimplify.c (asan_poison_variable): Don't do the check for
2232 gimplify_omp_ctxp here.
2233 (gimplify_decl_expr): Do it here.
2234 (gimplify_target_expr): Likewise.
2236 2018-03-28 Martin Liska <mliska@suse.cz>
2238 PR target/84988
2239 * config/i386/i386.c (ix86_function_arg_advance): Do not call
2240 chkp_type_bounds_count if MPX is not enabled.
2242 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
2244 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
2246 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
2248 PR target/84914
2249 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
2250 function to create the function decl for complex long double
2251 multiply and divide for -mabi=ieeelongdouble.
2252 (init_float128_ieee): Call it.
2254 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
2256 PR target/85044
2257 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
2258 -fcf-protection=branch -mibt.
2259 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
2261 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2263 PR target/81863
2264 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
2266 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
2268 PR target/85056
2269 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
2270 extern array declarations.
2272 2018-03-27 Richard Biener <rguenther@suse.de>
2274 PR middle-end/84067
2275 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
2276 explicit single_use checks.
2278 2018-03-27 Richard Biener <rguenther@suse.de>
2280 PR tree-optimization/85082
2281 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
2282 Valueize the VUSE.
2284 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
2286 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
2287 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
2288 Turn on fasynchronous-unwind-tables and funwind-tables.
2290 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
2292 PR target/85073
2293 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
2294 (*bmi_blsr_<mode>_ccz): Ditto.
2296 2018-03-26 Tom de Vries <tom@codesourcery.com>
2298 PR tree-optimization/85063
2299 * omp-general.c (offloading_function_p): New function. Factor out
2300 of ...
2301 * omp-offload.c (pass_omp_target_link::gate): ... here.
2302 * omp-general.h (offloading_function_p): Declare.
2303 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
2304 with attribute omp declare target for offloading functions.
2306 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
2308 PR tree-optimization/84005
2309 * tree-data-ref.h (get_base_for_alignment): Declare.
2310 * tree-data-ref.c (get_base_for_alignment_1): New function.
2311 (get_base_for_alignment): Likewise.
2312 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
2313 get_base_for_alignment to find a suitable base object, instead
2314 of always using drb->base_address.
2316 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2318 PR inline-asm/85022
2319 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
2320 known size by default.
2322 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
2324 PR inline-asm/85030
2325 * lra-constraints.c (process_alt_operands): Don't match BLKmode
2326 and non BLKmode operands.
2328 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2330 PR target/85026
2331 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
2332 Clean up attributes.
2334 2018-03-23 Richard Biener <rguenther@suse.de>
2336 PR debug/85020
2337 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
2338 we are going to emit early debug for LTO.
2340 2018-03-23 Jakub Jelinek <jakub@redhat.com>
2342 PR inline-asm/85034
2343 * function.c (match_asm_constraints_1): Don't optimize if input
2344 doesn't satisfy general_operand predicate for output's mode.
2346 PR inline-asm/85022
2347 * alias.c (write_dependence_p): Don't require for x_canonicalized
2348 non-VOIDmode if x has VOIDmode.
2350 PR sanitizer/85029
2351 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
2352 just don't try to optimize it rather than assert it never happens.
2354 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2356 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
2357 macro expansions for definition of ST_INTERNAL_<mode> and
2358 LD_INTERNAL_<mode> builtins.
2359 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
2360 Remove prototype.
2361 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
2362 function.
2363 (altivec_expand_st_builtin): Likewise.
2364 (altivec_expand_builtin): Remove calls to deleted functions.
2365 (rs6000_address_for_altivec): Delete this function.
2366 * config/rs6000/vector.md: Remove expands for
2367 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
2369 2018-03-22 Sudakshina Das <sudi.das@arm.com>
2371 PR target/84826
2372 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
2373 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
2374 re-computing once computed.
2375 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
2376 (arm_init_machine_status): Initialize
2377 machine->static_chain_stack_bytes.
2379 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2381 PR target/84760
2382 * doc/extend.texi: Add four new prototypes for vec_ld.
2383 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
2384 definitions for more logical presentation.
2385 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
2386 entries for V1TI variants of __builtin_altivec_ld builtin.
2387 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
2388 handling of V1TI variant of LVX icode pattern.
2389 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
2390 (rs6000_gimple_fold_builtin): Likewise.
2391 (altivec_init_builtins): Add code to define
2392 __builtin_altivec_lvx_v1ti function.
2394 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2396 PR inline-asm/84941
2397 * function.c (match_asm_constraints_1): Don't do the optimization
2398 if input isn't a REG, SUBREG, MEM or constant.
2400 2018-03-22 Tom de Vries <tom@codesourcery.com>
2402 PR tree-optimization/84956
2403 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
2404 bb_has_abnormal_pred.
2406 2018-03-22 Jakub Jelinek <jakub@redhat.com>
2408 PR sanitizer/85018
2409 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
2410 DECL_INITIAL (decl) to decl at the end.
2411 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
2412 adjust the comment.
2414 2018-03-21 Joseph Myers <joseph@codesourcery.com>
2416 * doc/extend.texi (__builtin_tgmath): Document when complex
2417 integer types are treated as _Complex _Float64.
2419 2018-03-21 Tom de Vries <tom@codesourcery.com>
2421 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
2423 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2425 PR tree-optimization/84960
2426 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
2427 if it is ENTRY block, move them into single succ of ENTRY in that case.
2429 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
2431 PR tree-optimization/84811
2432 * poly-int.h (poly_span_traits): Remove the T3 parameter and
2433 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
2434 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
2435 (known_subrange_p): Update accordingly. Cast each value involved
2436 in the size comparison, rather than casting the result of the
2437 subtraction.
2439 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2441 PR tree-optimization/84982
2442 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
2443 by flipping the least significant bit rather than all bits from
2444 bitpos to bitpos + bitsize - 1.
2446 2018-03-21 Nathan Sidwell <nathan@acm.org>
2448 * doc/extend.texi (Deprecated Features): Remove mention of
2449 long-deleted deprecations.
2451 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2453 PR jit/84288
2454 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
2455 * configure: Regenerate.
2457 2018-03-21 Tom de Vries <tom@codesourcery.com>
2459 PR tree-optimization/83126
2460 * tree-parloops.c (num_phis): New function.
2461 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
2463 2018-03-21 Nathan Sidwell <nathan@acm.org>
2465 * doc/extend.texi (Deprecated Features): Update deprecated flags,
2466 mention anon-struct/union members and trailing attributes.
2468 2018-03-21 Bin Cheng <bin.cheng@arm.com>
2470 PR tree-optimization/84969
2471 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
2472 builtin memset partitions if they set different rhs values.
2474 2018-03-21 Jakub Jelinek <jakub@redhat.com>
2476 PR rtl-optimization/84989
2477 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
2478 VEC_DUPLICATE with scalar result mode.
2480 2018-03-21 Martin Liska <mliska@suse.cz>
2482 PR ipa/84963
2483 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
2484 not intended return statement.
2486 2018-03-21 Martin Liska <mliska@suse.cz>
2488 PR target/84988
2489 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
2490 (chkp_find_bound_slots_1): Limit number of iterations.
2492 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
2494 PR target/84838
2495 * Minor grammar fixes for x86 options.
2497 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2499 PR debug/84875
2500 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
2501 holding REG_CFA_RESTORE notes, instead turn them into a USE.
2503 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
2505 PR target/83789
2506 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
2507 (altivec_lvx_<mode>_1op): Likewise.
2508 (altivec_stvx_<mode>_2op): Likewise.
2509 (altivec_stvx_<mode>_1op): Likewise.
2510 (altivec_lvx_<VM2:mode>): New define_expand.
2511 (altivec_stvx_<VM2:mode>): Likewise.
2512 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
2513 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2514 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
2515 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
2516 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
2517 (rs6000_gen_lvx): Likewise.
2518 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
2519 (altivec_expand_stv_builtin): Likewise.
2520 (altivec_expand_builtin): Likewise.
2521 * config/rs6000/vector.md: Likewise.
2523 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2525 PR target/82518
2526 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
2527 BYTES_BIG_ENDIAN.
2529 2018-03-20 Richard Biener <rguenther@suse.de>
2531 PR target/84986
2532 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
2533 sign-conversions as zero, fall back to standard scalar_stmt
2534 cost for the rest.
2536 2018-03-20 Martin Liska <mliska@suse.cz>
2538 PR ipa/84825
2539 * predict.c (rebuild_frequencies): Handle case when we have
2540 PROFILE_ABSENT, but flag_guess_branch_prob is false.
2542 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2544 PR target/84990
2545 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
2546 flag_section_anchors.
2547 * varasm.c (use_blocks_for_decl_p): Remove hack for
2548 dw2_force_const_mem.
2550 PR target/84845
2551 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
2552 to ...
2553 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
2554 be created, use lowpart_subreg of operands[0] rather than operands[0]
2555 itself.
2556 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
2557 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
2558 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
2559 and n constraint instead of aarch64_shift_imm_di and Usd.
2560 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
2561 (*aarch64_<optab>_reg_minus<mode>3): ... this.
2563 2018-03-20 Sudakshina Das <sudi.das@arm.com>
2565 PR target/82989
2566 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
2567 to favor GPR over NEON registers.
2568 (<shift>di3_neon): Likewise.
2570 2018-03-20 Tom de Vries <tom@codesourcery.com>
2572 PR target/84952
2573 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
2574 (nvptx_process_pars): Emit bar.sync asap and alap.
2576 2018-03-20 Tom de Vries <tom@codesourcery.com>
2578 PR target/84954
2579 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
2580 seen_label if seen_label is already set.
2582 2018-03-20 Jakub Jelinek <jakub@redhat.com>
2584 PR target/84945
2585 * config/i386/i386.c (fold_builtin_cpu): For features above 31
2586 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
2587 Use 1U instead of 1. Formatting fixes.
2589 PR c/84953
2590 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
2591 instead of TREE_TYPE (s1) for the return value.
2593 2018-03-19 Jakub Jelinek <jakub@redhat.com>
2595 PR tree-optimization/84946
2596 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
2597 bitsize + bitsize in poly_uint64 rather than poly_int64.
2599 PR sanitizer/78651
2600 * dwarf2asm.c: Include fold-const.c.
2601 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
2602 of decl rather than decl itself.
2604 PR rtl-optimization/84643
2605 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
2607 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
2609 PR sanitizer/78651
2610 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
2611 calling assemble_variable.
2613 2018-03-19 Sudakshina Das <sudi.das@arm.com>
2615 PR target/81647
2616 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
2617 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
2619 2018-03-19 Jim Wilson <jimw@sifive.com>
2621 PR bootstrap/84856
2622 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
2623 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
2624 (riscv_first_stack_step): Likewise.
2625 (riscv_option_override): Use STACK_BOUNDARY instead of
2626 MIN_STACK_BOUNDARY.
2627 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
2628 MIN_STACK_BOUNDARY.
2629 (BIGGEST_ALIGNMENT): Set to 128.
2630 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
2631 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
2632 STACK_BOUNDARY.
2634 2018-03-19 Richard Biener <rguenther@suse.de>
2636 PR tree-optimization/84933
2637 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
2638 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
2640 2018-03-19 Richard Biener <rguenther@suse.de>
2642 PR tree-optimization/84859
2643 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
2644 (cond_if_else_store_replacement): Perform sinking operation on
2645 single-store BBs regardless of MAX_STORES_TO_SINK setting.
2646 Generalize what a BB with a single eligible store is.
2648 2018-03-19 Richard Biener <rguenther@suse.de>
2650 PR tree-optimization/84929
2651 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
2652 chrec_is_positive against non-chrec arg.
2654 2018-03-19 Tamar Christina <tamar.christina@arm.com>
2656 PR target/84711
2657 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
2659 2018-03-18 Martin Liska <mliska@suse.cz>
2661 PR rtl-optimization/84635
2662 * regrename.c (build_def_use): Use matches_mode only when
2663 matches >= 0.
2665 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
2667 PR tree-optimization/84913
2668 * tree-vect-loop.c (vectorizable_reduction): Don't try to
2669 vectorize chains of COND_EXPRs.
2671 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2673 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
2675 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2677 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
2679 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
2681 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
2683 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2684 Kito Cheng <kito.cheng@gmail.com>
2686 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
2687 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
2688 (nds32_adjust_reg_alloc_order): New function.
2689 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
2691 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
2693 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
2694 nds32_print_operand, nds32_print_operand_address): Use
2695 HOST_WIDE_INT_PRINT_DEC instead.
2697 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
2699 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
2701 2018-03-17 Jakub Jelinek <jakub@redhat.com>
2703 PR target/84902
2704 * config/i386/i386.c (initial_ix86_tune_features,
2705 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
2706 unsigned long long.
2707 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
2708 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
2709 rather than 1u << ix86_tune. Formatting fix.
2710 (ix86_option_override_internal): Change ix86_arch_mask from
2711 unsigned int to unsigned HOST_WIDE_INT, initialize to
2712 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
2713 (ix86_function_specific_restore): Likewise.
2715 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2717 PR target/84899
2718 * postreload.c (reload_combine_recognize_pattern): Perform
2719 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
2720 truncate_int_for_mode the result for the destination's mode.
2722 PR c/84909
2723 * hsa-gen.c (mem_type_for_type): Fix comment typo.
2724 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
2725 Likewise.
2726 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2727 Likewise.
2729 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
2731 PR target/84876
2732 * lra-assigns.c (lra_split_hard_reg_for): Don't use
2733 regno_allocno_class_array and sorted_pseudos.
2734 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
2735 insns where regno is used.
2737 2018-03-16 Martin Liska <mliska@suse.cz>
2739 PR ipa/84833
2740 * multiple_target.c (create_dispatcher_calls): Redirect
2741 reference in the symbol table.
2743 2018-03-16 Martin Liska <mliska@suse.cz>
2745 PR ipa/84722
2746 * multiple_target.c (create_dispatcher_calls): Redirect also
2747 an alias.
2749 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2751 PR c++/79937
2752 PR c++/82410
2753 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
2754 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
2755 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
2757 2018-03-16 Julia Koval <julia.koval@intel.com>
2759 * doc/invoke.texi (Skylake Server): Add CLWB.
2760 Cannonlake): Remove CLWB.
2762 2018-03-16 Jakub Jelinek <jakub@redhat.com>
2764 PR tree-optimization/84841
2765 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
2766 1 << 3.
2767 (FLOAT_ONE_CONST_TYPE): Define.
2768 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
2769 (sort_by_operand_rank): Put entries with higher constant_type last
2770 rather than first to match comments.
2772 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
2774 * config/nios2/nios2.md (movsi_internal): Fix thinko in
2775 split predicate.
2777 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2779 PR c++/79085
2780 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
2781 check and use address of target always.
2783 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
2785 PR target/84574
2786 * config/i386/i386.c (indirect_thunk_needed): Update comments.
2787 (indirect_thunk_bnd_needed): Likewise.
2788 (indirect_thunks_used): Likewise.
2789 (indirect_thunks_bnd_used): Likewise.
2790 (indirect_return_needed): New.
2791 (indirect_return_bnd_needed): Likewise.
2792 (output_indirect_thunk_function): Add a bool argument for
2793 function return.
2794 (output_indirect_thunk_function): Don't generate alias for
2795 function return thunk.
2796 (ix86_code_end): Call output_indirect_thunk_function to generate
2797 function return thunks.
2798 (ix86_output_function_return): Set indirect_return_bnd_needed
2799 and indirect_return_needed instead of indirect_thunk_bnd_needed
2800 and indirect_thunk_needed.
2802 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
2804 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
2805 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
2806 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
2808 2018-03-15 David Malcolm <dmalcolm@redhat.com>
2809 Paul Hua <paul.hua.gm@gmail.com>
2811 PR c/84852
2812 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
2814 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
2816 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
2817 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
2818 resp. SFmode cases.
2820 2018-03-15 Tamar Christina <tamar.christina@arm.com>
2822 PR target/84711
2823 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
2824 instead of GET_MODE_SIZE when comparing Units.
2826 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
2828 PR target/68256
2829 * varasm.c (hash_section): Return an unchangeble hash value
2830 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
2831 Return !aarch64_can_use_per_function_literal_pools_p ().
2833 2018-03-15 Jakub Jelinek <jakub@redhat.com>
2835 PR target/84860
2836 * optabs.c (emit_conditional_move): Pass address of cmode's copy
2837 rather than address of cmode as last argument to prepare_cmp_insn.
2839 2018-03-15 Julia Koval <julia.koval@intel.com>
2841 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
2842 F_AVX512VNNI, F_AVX512BITALG): New.
2844 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
2846 PR target/83451
2847 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
2848 insn for floating-point loads and stores.
2850 2018-03-14 Carl Love <cel@us.ibm.com>
2852 * config/rs6000/rs6000-c.c: Add macro definitions for
2853 ALTIVEC_BUILTIN_VEC_PERMXOR.
2854 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2855 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2856 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2857 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2858 UNSPEC_VPERMXOR.
2859 * config/doc/extend.texi: Add prototypes for vec_permxor.
2861 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2863 PR c/84852
2864 * diagnostic-show-locus.c (class layout_point): Convert m_line
2865 from int to linenum_type.
2866 (line_span::comparator): Use linenum "compare" function when
2867 comparing line numbers.
2868 (test_line_span): New function.
2869 (layout_range::contains_point): Convert param "row" from int to
2870 linenum_type.
2871 (layout_range::intersects_line_p): Likewise.
2872 (layout::will_show_line_p): Likewise.
2873 (layout::print_source_line): Likewise.
2874 (layout::should_print_annotation_line_p): Likewise.
2875 (layout::print_annotation_line): Likewise.
2876 (layout::print_leading_fixits): Likewise.
2877 (layout::annotation_line_showed_range_p): Likewise.
2878 (struct line_corrections): Likewise for field m_row.
2879 (line_corrections::line_corrections): Likewise for param "row".
2880 (layout::print_trailing_fixits): Likewise.
2881 (layout::get_state_at_point): Likewise.
2882 (layout::get_x_bound_for_row): Likewise.
2883 (layout::print_line): Likewise.
2884 (diagnostic_show_locus): Likewise for locals "last_line" and
2885 "row".
2886 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2887 * input.c (selftest::test_linenum_comparisons): New function.
2888 (selftest::input_c_tests): Call it.
2889 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2890 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2891 * selftest.h (ASSERT_GT): New macro.
2892 (ASSERT_GT_AT): New macro.
2893 (ASSERT_LT): New macro.
2894 (ASSERT_LT_AT): New macro.
2896 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2898 PR rtl-optimization/84780
2899 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2901 2018-03-14 Martin Liska <mliska@suse.cz>
2903 * tree.c (record_node_allocation_statistics): Use
2904 get_stats_node_kind.
2905 (get_stats_node_kind): New function extracted from
2906 record_node_allocation_statistics.
2907 (free_node): Use get_stats_node_kind.
2909 2018-03-14 Richard Biener <rguenther@suse.de>
2911 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2912 that the value-set of ANTIC_IN doesn't grow.
2914 Revert
2915 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2916 member.
2917 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2918 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2920 2018-03-14 Julia Koval <julia.koval@intel.com>
2922 * config.gcc (icelake-client, icelake-server): New.
2923 (icelake): Remove.
2924 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2925 (initial_ix86_arch_features): Ditto.
2926 (PTA_SKYLAKE): Add SGX.
2927 (PTA_ICELAKE): Remove.
2928 (PTA_ICELAKE_CLIENT): New.
2929 (PTA_ICELAKE_SERVER): New.
2930 (ix86_option_override_internal): Split up icelake on icelake client and
2931 icelake server.
2932 (get_builtin_code_for_version): Ditto.
2933 (fold_builtin_cpu): Ditto.
2934 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2935 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2936 * config/i386/i386.h (processor_type): Ditto.
2937 * doc/invoke.texi: Ditto.
2939 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2941 PR sanitizer/83392
2942 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2943 INTEGER_CST offset, add it together with bitpos / 8 and
2944 sign extend based on POINTER_SIZE.
2946 PR target/84844
2947 Revert
2948 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2950 PR target/78090
2951 * config/i386/constraints.md (Yc): New register constraint.
2952 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2953 Use Yc constraint for alternative 2 of operand 0. Remove
2954 preferred_for_speed attribute.
2956 2018-03-14 Richard Biener <rguenther@suse.de>
2958 PR tree-optimization/84830
2959 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2960 with the old one to avoid oscillations.
2962 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2964 PR target/83712
2965 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2966 pseudos.
2967 (assign_by_spills): Return a flag of reload assignment failure.
2968 Do not process the reload assignment failures. Do not spill other
2969 reload pseudos if they has the same reg class. Update n if
2970 necessary.
2971 (lra_assign): Add a return arg. Set up from the result of
2972 assign_by_spills call.
2973 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2974 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2975 usage_insns if it is not NULL.
2976 (spill_hard_reg_in_range): New function.
2977 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2978 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2979 function prototypes.
2980 (lra_assign): Change prototype.
2981 * lra.c (lra): Add code to deal with fails by splitting hard reg
2982 live ranges.
2984 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2986 * config/riscv/riscv.opt (mrelax): New option.
2987 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2988 "norelax" when riscv_mrelax is disabled.
2989 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2991 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2993 PR target/84743
2994 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2995 reassociation for int modes.
2997 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2999 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
3000 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
3001 for big-endian.
3002 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
3003 * config/aarch64/aarch64-sve.md
3004 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
3005 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
3006 (*extend<mode><Vwide>2): Rename to...
3007 (aarch64_sve_extend<mode><Vwide>2): ...this.
3008 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
3009 renaming the old pattern to...
3010 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
3011 unsigned packs.
3012 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
3013 define_expand, renaming the old pattern to...
3014 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
3015 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
3016 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
3017 account when deciding which SVE instruction the optab should use.
3018 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
3020 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3022 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
3023 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
3024 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
3025 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
3026 (tlsdesc_small_<mode>): Turn a define_expand and use
3027 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
3028 (tlsdesc_small_advsimd_<mode>): ...this.
3029 (tlsdesc_small_sve_<mode>): New pattern.
3031 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
3033 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
3034 (UNSPEC_UMUL_HIGHPART): New constants.
3035 (MUL_HIGHPART): New int iteraor.
3036 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
3037 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
3038 define_expand.
3039 (*<su>mul<mode>3_highpart): New define_insn.
3041 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
3043 PR lto/84805
3044 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
3045 incomplete types.
3047 2018-03-13 Martin Liska <mliska@suse.cz>
3049 PR ipa/84658.
3050 * (sem_item_optimizer::sem_item_optimizer): Initialize new
3051 vector.
3052 (sem_item_optimizer::~sem_item_optimizer): Release it.
3053 (sem_item_optimizer::merge_classes): Register variable aliases.
3054 (sem_item_optimizer::fixup_pt_set): New function.
3055 (sem_item_optimizer::fixup_points_to_sets): Likewise.
3056 * ipa-icf.h: Declare new variables and functions.
3058 2018-03-13 Jakub Jelinek <jakub@redhat.com>
3060 PR middle-end/84834
3061 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
3062 integer_pow2p@2 and test integer_pow2p in condition.
3063 (A < 0 ? C : 0): Similarly for @1.
3065 PR middle-end/84831
3066 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
3067 characters starting at p contain '\0' character, don't look beyond
3068 that.
3070 PR target/84827
3071 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
3072 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
3074 PR target/84828
3075 * reg-stack.c (change_stack): Change update_end var from int to
3076 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
3077 also call set_block_for_insn on the newly added insns and rescan.
3079 PR target/84786
3080 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
3081 on the last operand.
3083 PR c++/84704
3084 * tree.c (stabilize_reference_1): Return save_expr (e) for
3085 STATEMENT_LIST even if it doesn't have side-effects.
3087 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
3089 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
3091 2018-03-12 Renlin Li <renlin.li@arm.com>
3093 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
3094 aarch64_output_scalar_simd_mov_immediate.
3096 2018-03-12 Martin Sebor <msebor@redhat.com>
3098 PR tree-optimization/83456
3099 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
3100 for perfectly overlapping calls to memcpy.
3101 (gimple_fold_builtin_memory_chk): Same.
3102 (gimple_fold_builtin_strcpy): Handle no-warning.
3103 (gimple_fold_builtin_stxcpy_chk): Same.
3104 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
3106 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3108 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
3109 parameter. Use it for SFmode.
3110 (rs6000_function_arg_advance_1): Adjust.
3111 (rs6000_function_arg): Adjust.
3112 (rs6000_gimplify_va_arg): Pass false for that new parameter.
3114 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
3116 PR rtl-optimization/84169
3117 PR rtl-optimization/84780
3118 * combine.c (can_combine_p): Check for a 2-insn combination whether
3119 the destination register is used between the two insns, too.
3121 2018-03-12 Richard Biener <rguenther@suse.de>
3123 PR tree-optimization/84803
3124 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
3125 for refs DR analysis didn't process.
3127 2018-03-12 Richard Biener <rguenther@suse.de>
3129 PR tree-optimization/84777
3130 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
3131 force-vectorize loops ignore whether we are optimizing for size.
3133 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
3135 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
3136 (TARGET_MD_ASM_ADJUST): Define.
3138 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
3139 Kito Cheng <kito.cheng@gmail.com>
3140 Chung-Ju Wu <jasonwucj@gmail.com>
3142 * config/nds32/nds32.c (nds32_compute_stack_frame,
3143 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
3144 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
3145 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
3146 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
3147 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
3148 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
3149 * config/nds32/nds32.md (prologue, epilogue): Use macro
3150 NDS32_V3PUSH_AVAILABLE_P to do checking.
3152 2018-03-11 Jakub Jelinek <jakub@redhat.com>
3154 PR debug/58150
3155 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
3156 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
3157 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
3158 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
3159 addition of most attributes on !orig_type_die or the attribute not
3160 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
3162 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3163 Chung-Ju Wu <jasonwucj@gmail.com>
3165 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
3166 __NDS32_VH__ macro.
3167 * config/nds32/nds32.opt (mvh): New option.
3169 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3170 Chung-Ju Wu <jasonwucj@gmail.com>
3172 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
3173 function.
3174 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
3175 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
3176 definition.
3178 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
3179 Chung-Ju Wu <jasonwucj@gmail.com>
3181 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
3182 function.
3183 * config/nds32/nds32-multiple.md (strlensi): New pattern.
3184 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
3186 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
3187 Kito Cheng <kito.cheng@gmail.com>
3188 Chung-Ju Wu <jasonwucj@gmail.com>
3190 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
3191 UNSPEC_FFMISM and UNSPEC_FLMISM.
3192 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
3193 for ffb, ffmism and flmism.
3194 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
3195 (unspec_ffmism): Ditto.
3196 (unspec_flmism): Ditto.
3197 (nds32_expand_builtin_impl): Check if string extension is available.
3198 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
3199 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
3201 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
3203 Reverting patch:
3204 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3206 PR target/83712
3207 * lra-assigns.c (assign_by_spills): Return a flag of reload
3208 assignment failure. Do not process the reload assignment
3209 failures. Do not spill other reload pseudos if they has the same
3210 reg class.
3211 (lra_assign): Add a return arg. Set up from the result of
3212 assign_by_spills call.
3213 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3214 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3215 usage_insns if it is not NULL.
3216 (spill_hard_reg_in_range): New function.
3217 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3218 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3219 function prototypes.
3220 (lra_assign): Change prototype.
3221 * lra.c (lra): Add code to deal with fails by splitting hard reg
3222 live ranges.
3224 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
3226 PR target/84807
3227 * config/i386/i386.opt: Replace Enforcment with Enforcement.
3229 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
3231 PR debug/84620
3232 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
3233 (dw_val_node): Add val_symbolic_view.
3234 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
3235 (symview_upper_bound): New.
3236 (new_line_info_table): Initialize symviews_since_reset.
3237 (dwarf2out_source_line): Count symviews_since_reset and set
3238 symview_upper_bound.
3239 (dw_val_equal_p): Handle symview.
3240 (add_AT_symview): New.
3241 (print_dw_val): Handle symview.
3242 (attr_checksum, attr_checksum_ordered): Likewise.
3243 (same_dw_val_p, size_of_die): Likewise.
3244 (value_format, output_die): Likewise.
3245 (add_high_low_attributes): Use add_AT_symview for entry_view.
3246 (dwarf2out_finish): Reset symview_upper_bound, clear
3247 zero_view_p.
3249 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
3251 PR target/83969
3252 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
3253 Add strict argument and use it.
3254 (rs6000_split_multireg_move): Update for new strict argument.
3255 (mem_operand_gpr): Disallow all non-offsettable addresses.
3256 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
3258 2018-03-09 Jakub Jelinek <jakub@redhat.com>
3260 PR target/84772
3261 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
3262 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
3263 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
3265 PR c++/84767
3266 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
3267 decl, use remap_type if we want to use the type.
3269 2018-03-09 Martin Sebor <msebor@redhat.com>
3271 PR tree-optimization/84526
3272 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
3273 Remove dead code.
3274 (builtin_access::generic_overlap): Be prepared to handle non-array
3275 base objects.
3277 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
3279 PR rtl-optimization/84682
3280 * lra-constraints.c (process_address_1): Check is_address flag
3281 for address constraints.
3282 (process_alt_operands): Likewise.
3283 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
3284 preprocess_constraints.
3285 * recog.h (preprocess_constraints): Add oploc parameter.
3286 Adjust callers.
3287 * recog.c (preprocess_constraints): Test address_operand for
3288 CT_ADDRESS constraints.
3290 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
3292 PR target/83712
3293 * lra-assigns.c (assign_by_spills): Return a flag of reload
3294 assignment failure. Do not process the reload assignment
3295 failures. Do not spill other reload pseudos if they has the same
3296 reg class.
3297 (lra_assign): Add a return arg. Set up from the result of
3298 assign_by_spills call.
3299 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
3300 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
3301 usage_insns if it is not NULL.
3302 (spill_hard_reg_in_range): New function.
3303 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
3304 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
3305 function prototypes.
3306 (lra_assign): Change prototype.
3307 * lra.c (lra): Add code to deal with fails by splitting hard reg
3308 live ranges.
3310 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3312 PR target/83193
3313 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
3314 Accept complain bool parameter. Only emit errors if it is true.
3315 (arm_parse_cpu_option_name): Likewise.
3316 (arm_target_thumb_only): Adjust callers of the above.
3317 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
3318 prototype to take a default true bool parameter.
3319 (arm_parse_arch_option_name): Likewise.
3321 2018-03-09 David Malcolm <dmalcolm@redhat.com>
3322 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
3324 PR jit/64089
3325 PR jit/84288
3326 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
3327 * configure: Regenerate.
3328 * configure.ac ("linker --version-script option"): New.
3329 ("linker soname option"): New.
3331 2018-03-09 Richard Biener <rguenther@suse.de>
3333 PR tree-optimization/84775
3334 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
3335 immediate uses of predicate stmts and mark them modified.
3337 Revert
3338 PR tree-optimization/84178
3339 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3340 to caller.
3341 (version_loop_for_if_conversion): Delay update_ssa call.
3342 (tree_if_conversion): Delay update_ssa until after predicate
3343 insertion.
3345 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
3347 PR target/84763
3348 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
3349 when the function accesses prior frames.
3351 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3353 PR debug/84456
3354 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
3355 gen_llsym, otherwise call maybe_gen_llsym.
3357 PR inline-asm/84742
3358 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
3359 has ',' character inside of it.
3361 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3363 PR target/84748
3364 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
3365 as clobbering CC_REGNUM.
3367 2018-03-08 Richard Biener <rguenther@suse.de>
3369 PR middle-end/84552
3370 * tree-scalar-evolution.c: Include tree-into-ssa.h.
3371 (follow_copies_to_constant): Do not follow SSA names registered
3372 for update.
3374 2018-03-08 Richard Biener <rguenther@suse.de>
3376 PR tree-optimization/84178
3377 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
3378 to caller.
3379 (version_loop_for_if_conversion): Delay update_ssa call.
3380 (tree_if_conversion): Delay update_ssa until after predicate
3381 insertion.
3383 2018-03-08 David Malcolm <dmalcolm@redhat.com>
3385 PR tree-optimization/84178
3386 * tree-if-conv.c (release_bb_predicate): Remove the
3387 the assertion that the stmts have NULL use_ops.
3388 Discard the statements, asserting that they haven't
3389 yet been added to a BB.
3391 2018-03-08 Richard Biener <rguenther@suse.de>
3393 PR tree-optimization/84746
3394 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
3395 (phi_translate): Pass in destination ANTIC_OUT set.
3396 (phi_translate_1): Likewise. For a simplified result lookup
3397 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
3398 (phi_translate_set): Adjust.
3399 (do_pre_regular_insertion): Likewise.
3400 (do_pre_partial_partial_insertion): Likewise.
3402 2018-03-08 Martin Liska <mliska@suse.cz>
3404 PR gcov-profile/84735
3405 * doc/gcov.texi: Document usage of profile files.
3406 * gcov-io.h: Document changes in the format.
3408 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
3410 PR debug/84404
3411 PR debug/84408
3412 * dwarf2out.c (struct dw_line_info_table): Update comments for
3413 view == -1.
3414 (FORCE_RESET_NEXT_VIEW): New.
3415 (FORCE_RESETTING_VIEW_P): New.
3416 (RESETTING_VIEW_P): Check for -1 too.
3417 (ZERO_VIEW_P): Likewise.
3418 (new_line_info_table): Force-reset next view.
3419 (dwarf2out_begin_function): Likewise.
3420 (dwarf2out_source_line): Simplify zero_view_p initialization.
3421 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
3422 view directly. Omit view when omitting .loc at line 0.
3424 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3426 PR tree-optimization/84740
3427 * tree-switch-conversion.c (process_switch): Call build_constructors
3428 only if info.phi_count is non-zero.
3430 PR tree-optimization/84739
3431 * tree-tailcall.c (find_tail_calls): Check call arguments against
3432 DECL_ARGUMENTS (current_function_decl) rather than
3433 DECL_ARGUMENTS (func) when checking for tail recursion.
3435 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3437 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
3438 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
3439 Volker Reichelt's entry and add entries for people that perform
3440 GCC fuzzy testing and report numerous bugs.
3442 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
3444 PR target/82411
3445 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
3446 readonly data in sdata, if that is disabled.
3447 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
3448 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
3449 -mreadonly-in-sdata option.
3451 2018-03-07 Martin Sebor <msebor@redhat.com>
3453 PR tree-optimization/84468
3454 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
3455 basic block when looking for nul assignment.
3457 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
3459 PR target/84277
3460 * except.h (output_function_exception_table): Adjust prototype.
3461 * except.c (output_function_exception_table): Remove FNNAME parameter
3462 and add SECTION parameter. Ouput one part of the table at a time.
3463 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
3464 the first part of the exception table and emit unwind directives.
3465 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
3466 (i386_pe_seh_cold_init): Likewise.
3467 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
3468 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
3469 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
3470 (ix86_output_call_insn): Emit a nop in one more case for SEH.
3471 * config/i386/winnt.c: Include except.h.
3472 (struct seh_frame_state): Add reg_offset, after_prologue and
3473 in_cold_section fields.
3474 (i386_pe_seh_end_prologue): Set seh->after_prologue.
3475 (i386_pe_seh_cold_init): New function.
3476 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
3477 to seh->in_cold_section.
3478 (seh_emit_push): Record the offset of the push.
3479 (seh_emit_save): Record the offet of the save.
3480 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
3481 Test seh->after_prologue to disregard the epilogue.
3482 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
3483 (i386_pe_end_cold_function): New function.
3485 2018-03-07 Jakub Jelinek <jakub@redhat.com>
3487 PR fortran/84565
3488 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
3489 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
3491 PR c++/84704
3492 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
3493 on tmp_var.
3494 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
3495 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
3497 PR middle-end/84723
3498 * multiple_target.c: Include tree-inline.h and intl.h.
3499 (expand_target_clones): Diagnose and fail if node->definition and
3500 !tree_versionable_function_p (node->decl).
3502 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
3504 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
3505 sprint_ul.
3506 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
3507 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
3508 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
3510 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3512 PR target/84710
3513 * combine.c (try_combine): Use reg_or_subregno instead of handling
3514 just paradoxical SUBREGs and REGs.
3516 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
3518 * config/arc/arc.c (arc_finalize_pic): Remove function.
3519 (arc_must_save_register): We use single base PIC register, remove
3520 checks to save/restore the PIC register.
3521 (arc_expand_prologue): Likewise.
3522 * config/arc/arc-protos.h (arc_set_default_type_attributes):
3523 Remove.
3524 (arc_verify_short): Likewise.
3525 (arc_attr_type): Likewise.
3526 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
3527 (walk_stores): Likewise.
3528 (arc_address_cost): Make it static.
3529 (arc_verify_short): Likewise.
3530 (branch_dest): Likewise.
3531 (arc_attr_type): Likewise.
3532 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
3533 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
3534 (arc_final_prescan_insn): Remove inserting the nops due to
3535 hardware hazards. It is done in reorg step.
3536 (insn_length_variant_t): Remove.
3537 (insn_length_parameters_t): Likewise.
3538 (arc_insn_length_parameters): Likewise.
3539 (arc_get_insn_variants): Likewise.
3540 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
3542 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3544 PR inline-asm/84683
3545 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
3546 assertion failure.
3548 PR tree-optimization/84687
3549 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
3550 on new_node->decl.
3551 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
3553 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3555 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
3556 Rename to ppc_speculation_barrier.
3557 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
3558 __builtin_ppc_speculation_barrier.
3560 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3562 PR target/84700
3563 * combine.c (combine_simplify_rtx): Don't try to simplify if
3564 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
3565 are equal to x.
3567 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
3569 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
3570 to 32 bytes when compiling for POWER9.
3572 2018-03-05 Jakub Jelinek <jakub@redhat.com>
3574 PR target/84564
3575 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
3576 regparm >= 3 with no arg reg available also for calls with
3577 flag_force_indirect_call. Pass decl to ix86_function_regparm.
3579 PR target/84524
3580 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
3581 orig,vex.
3582 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
3584 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
3586 PR target/84264
3587 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
3589 2018-03-05 Richard Biener <rguenther@suse.de>
3591 PR tree-optimization/84486
3592 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
3593 When inserting a __builtin_assume_aligned call set the LHS
3594 SSA name alignment info accordingly.
3596 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
3598 PR tree-optimization/84114
3599 * config/aarch64/aarch64.c (aarch64_reassociation_width)
3600 Avoid reassociation of FLOAT_MODE addition.
3602 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
3604 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
3605 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
3606 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
3607 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
3608 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
3609 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
3610 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
3611 and -mwbnoinvd.
3612 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
3613 __builtin_ia32_wbinvd): New builtins.
3614 (SPECIAL_ARGS2): New.
3615 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
3616 (SPECIAL_ARGS2): New.
3617 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
3618 (ix86_valid_target_attribute_inner_p): Ditto.
3619 (ix86_init_mmx_sse_builtins): Add special_args2.
3620 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
3621 TARGET_WBNOINVD_P): New.
3622 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
3623 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
3624 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
3625 * config/i386/immintrin.h (_wbinvd): New intrinsic.
3626 * config/i386/pconfigintrin.h: New file.
3627 * config/i386/wbnoinvdintrin.h: Ditto.
3628 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
3629 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
3631 2018-03-05 Richard Biener <rguenther@suse.de>
3633 PR tree-optimization/84670
3634 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
3635 member.
3636 (BB_VISITED_WITH_VISITED_SUCCS): New define.
3637 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
3638 (compute_antic_aux): Only assert the number of values in ANTIC_IN
3639 doesn't grow if all successors (recursively) were visited at least
3640 once.
3642 2018-03-05 Richard Biener <rguenther@suse.de>
3644 PR tree-optimization/84650
3645 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
3646 if executed in the loop pipeline.
3648 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
3650 * doc/configfiles.texi (Configuration Files): Move info about
3651 conditionalizing $target-protos.h to...
3652 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
3653 differs from $target-protos.h.
3655 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
3656 Chung-Ju Wu <jasonwucj@gmail.com>
3658 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
3659 * config/nds32/nds32-multiple.md (setmemsi): Define.
3660 * config/nds32/nds32-memory-manipulation.c
3661 (nds32_gen_dup_4_byte_to_word_value): New.
3662 (emit_setmem_word_loop): New.
3663 (emit_setmem_byte_loop): New.
3664 (nds32_expand_setmem_loop): New.
3665 (nds32_expand_setmem_loop_v3m): New.
3666 (nds32_expand_setmem_unroll): New.
3667 (nds32_expand_setmem): New.
3669 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3670 Chung-Ju Wu <jasonwucj@gmail.com>
3672 * config/nds32/nds32-memory-manipulation.c
3673 (nds32_emit_load_store): New.
3674 (nds32_emit_post_inc_load_store): New.
3675 (nds32_emit_mem_move): New.
3676 (nds32_emit_mem_move_block): New.
3677 (nds32_expand_movmemsi_loop_unknown_size): New.
3678 (nds32_expand_movmemsi_loop_known_size): New.
3679 (nds32_expand_movmemsi_loop): New.
3680 (nds32_expand_movmemsi_unroll): New.
3681 (nds32_expand_movmemqi): Rename ...
3682 (nds32_expand_movmemsi): ... to this.
3683 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
3684 (movmemsi): ... to this.
3685 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
3686 (nds32_expand_movmemsi): ... to this.
3688 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3689 Monk Chiang <sh.chiang04@gmail.com>
3690 Chung-Ju Wu <jasonwucj@gmail.com>
3692 * config/nds32/nds32-protos.h
3693 (nds32_expand_load_multiple): New arguments.
3694 (nds32_expand_store_multiple): Ditto.
3695 (nds32_valid_multiple_load_store): Rename ...
3696 (nds32_valid_multiple_load_store_p): ... to this.
3697 * config/nds32/nds32-memory-manipulation.c
3698 (nds32_expand_load_multiple): Refine implementation.
3699 (nds32_expand_store_multiple): Ditto.
3700 * config/nds32/nds32-multiple.md
3701 (load_multiple): Update nds32_expand_load_multiple interface.
3702 (store_multiple): Update nds32_expand_store_multiple interface.
3703 * config/nds32/nds32-predicates.c
3704 (nds32_valid_multiple_load_store): Rename ...
3705 (nds32_valid_multiple_load_store_p): ... to this and refine
3706 implementation.
3707 * config/nds32/predicates.md
3708 (nds32_load_multiple_and_update_address_operation): New predicate.
3709 (nds32_store_multiple_and_update_address_operation): New predicate.
3711 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
3712 Chung-Ju Wu <jasonwucj@gmail.com>
3714 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
3715 (combo): New attribute.
3716 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
3718 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
3720 * config/nds32/nds32.opt: Change -mcmodel= default value.
3722 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
3723 Monk Chiang <sh.chiang04@gmail.com>
3724 Chung-Ju Wu <jasonwucj@gmail.com>
3726 * config/nds32/constants.md (unspec_element): New enum.
3727 * config/nds32/constraints.md (Umw): New constraint.
3728 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
3729 * config/nds32/nds32-intrinsic.md: Likewise.
3730 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
3731 (nds32_valid_smw_lwm_base_p): New.
3732 (nds32_output_smw_single_word): New.
3733 (nds32_output_lmw_single_word): New.
3734 (nds32_expand_unaligned_load): New.
3735 (nds32_expand_unaligned_store): New.
3736 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
3737 (nds32_output_smw_single_word): Declare.
3738 (nds32_output_lmw_single_word): Declare.
3739 (nds32_expand_unaligned_load): Declare.
3740 (nds32_expand_unaligned_store): Declare.
3741 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
3742 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
3743 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
3744 NDS32_BUILTIN_UASTORE_DW.
3745 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
3746 predicate.
3748 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
3749 Kito Cheng <kito.cheng@gmail.com>
3750 Chung-Ju Wu <jasonwucj@gmail.com>
3752 * config/nds32/nds32-intrinsic.c
3753 (nds32_expand_builtin_null_ftype_reg): Delete.
3754 (nds32_expand_builtin_reg_ftype_imm): Ditto.
3755 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
3756 (nds32_read_argument): New.
3757 (nds32_legitimize_target): Ditto.
3758 (nds32_legitimize_argument): Ditto.
3759 (nds32_check_constant_argument): Ditto.
3760 (nds32_expand_unop_builtin): Ditto.
3761 (nds32_expand_unopimm_builtin): Ditto.
3762 (nds32_expand_binop_builtin): Ditto.
3763 (nds32_builtin_decl_impl): Ditto.
3764 (builtin_description): Ditto.
3765 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
3766 (nds32_init_builtins_impl): Ditto.
3767 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
3768 (nds32_builtin_decl): New.
3769 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
3770 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
3772 2018-03-02 Jeff Law <law@redhat.com>
3774 * reorg.c (stop_search_p): Handle DEBUG_INSN.
3775 (redundant_insn, fill_simple_delay_slots): Likewise.
3776 (fill_slots_from_thread): Likewise.
3777 * resource.c (mark_referenced_resources): Likewise.
3778 (mark_set_resources, find_dead_or_set_registers): Likewise.
3780 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3782 * substring-locations.h (format_warning_va): Formatting fix for
3783 ATTRIBUTE_GCC_DIAG.
3784 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
3785 argument.
3786 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
3787 * substring-locations.c: Include intl.h.
3788 (format_warning_va): Turned into small wrapper around
3789 format_warning_n_va, renamed to ...
3790 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
3791 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
3792 use ngettext.
3793 (format_warning_at_substring_n): New function.
3794 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
3795 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
3796 format_warning_at_substring with just a shorter name instead of
3797 const function pointer.
3798 (fmtwarn_n): New function.
3799 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
3800 appropriate, get rid of all the fmtstr temporaries, move conditionals
3801 with G_() wrapped string literals directly into fmtwarn arguments,
3802 cast dir.len to (int), formatting fixes.
3804 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
3806 * doc/invoke.texi: Remove "Cilk Plus" references.
3808 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3809 Richard Biener <rguenther@suse.de>
3811 PR ipa/84628
3812 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
3813 for error or warning attributes if CALL_FROM_THUNK_P is set.
3814 Formatting fixes.
3816 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3818 PR target/56540
3819 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
3820 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
3822 PR target/56540
3823 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
3824 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
3826 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
3827 instead of -1U in last predictors element's probability member.
3829 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
3831 PR ipa/83983
3832 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
3833 arguments if they are comparable.
3835 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3837 PR tree-optimization/84634
3838 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
3839 masks and masked_loop_p with a single loop_masks, making sure it's
3840 null for bb vectorization.
3842 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
3844 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
3845 (vect_analyze_data_ref_access): Use loop->safe_len rather than
3846 loop->force_vectorize to check whether there is no alias.
3848 2018-03-02 Jakub Jelinek <jakub@redhat.com>
3850 PR target/84614
3851 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
3852 prototypes.
3853 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3854 comments.
3855 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3856 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3857 instead of a loop around prev_real_insn.
3858 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3859 prev_real_insn.
3861 PR inline-asm/84625
3862 * config/i386/i386.c (ix86_print_operand): Use conditional
3863 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3864 zero vector.
3866 2018-03-02 Richard Biener <rguenther@suse.de>
3868 PR tree-optimization/84427
3869 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3870 (bitmap_set_subtract_values): Rewrite to handle multiple
3871 exprs per value.
3872 (clean): Likewise.
3873 (prune_clobbered_mems): Likewise.
3874 (phi_translate): Take edge instead of pred/phiblock.
3875 (phi_translate_1): Likewise.
3876 (phi_translate_set): Likewise. Insert all translated
3877 exprs for a value into the set, keeping possibly multiple
3878 expressions per value.
3879 (compute_antic_aux): Adjust for phi_translate changes.
3880 When intersecting union the expressions and prune those
3881 not in the final value set, keeping possibly multiple
3882 expressions per value. Do not use value-insertion
3883 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3884 all expressions. Add verification that the value-sets
3885 only shrink during iteration.
3886 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3887 (do_pre_regular_insertion): Likewise.
3888 (do_pre_partial_partial_insertion): Likewise.
3890 2018-03-02 Richard Biener <rguenther@suse.de>
3892 PR target/82005
3893 * config/darwin.c (saved_debug_info_level): New static global.
3894 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3895 (darwin_asm_lto_end): Restore debug info generation settings.
3897 2018-03-01 Martin Liska <mliska@suse.cz>
3899 PR sanitizer/82484
3900 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3901 volatile arguments.
3903 2018-03-01 Richard Biener <rguenther@suse.de>
3905 PR debug/84645
3906 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3907 type annotation with LTO when debug was disabled at compile-time.
3909 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3911 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3912 XINT with INTVAL.
3913 (mips_final_postscan_insn): Likewise.
3915 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3917 PR rtl-optimization/84528
3918 * alias.c (init_alias_target): Add commentary.
3919 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3920 a unique base value if the frame pointer is not eliminated
3921 to the stack pointer.
3923 2018-03-01 Tom de Vries <tom@codesourcery.com>
3925 PR rtl-optimization/83327
3926 * lra-int.h (hard_regs_spilled_into): Declare.
3927 * lra.c (hard_regs_spilled_into): Define.
3928 (init_reg_info): Init hard_regs_spilled_into.
3929 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3930 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3931 (process_bb_lives): Handle hard_regs_spilled_into.
3932 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3933 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3935 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3937 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3938 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3939 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3940 * config/rs6000/aix72.h: New file.
3942 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3944 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3945 instead of warning_at with conditional singular and plural messages
3946 where possible.
3948 PR target/52991
3949 * stor-layout.c (update_alignment_for_field): For
3950 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3951 && !DECL_PACKED (field), do the alignment update, just use
3952 only desired_align instead of MAX (type_align, desired_align)
3953 as the alignment.
3954 (place_field): Don't do known_align < desired_align handling
3955 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3956 is non-NULL, instead do it after rli->prev_field handling and
3957 only if not within a bitfield word. For DECL_PACKED (field)
3958 use type_align of BITS_PER_UNIT.
3960 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3962 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3963 superfluous parentheses and trailing spaces.
3965 2018-02-28 Richard Biener <rguenther@suse.de>
3967 PR tree-optimization/84584
3968 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3969 SCoPs with fake exit edge.
3971 2018-02-28 Martin Liska <mliska@suse.cz>
3973 PR testsuite/84597
3974 * timevar.c (timer::print): Fix format to properly print 100%
3975 values.
3977 2018-02-28 Richard Biener <rguenther@suse.de>
3979 PR middle-end/84607
3980 * genmatch.c (capture_info::walk_match): Do not mark
3981 captured expressions without operands as expr_p given
3982 they act more like predicates and should be subject to
3983 "lost tail" side-effect preserving.
3985 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3987 PR rtl-optimization/81611
3988 * auto-inc-dec.c (attempt_change): Move dead note from
3989 mem_insn if it's the next use of regno
3990 (find_address): Take address use of reg holding
3991 non-incremented value. Add parm to limit search to the named
3992 reg only.
3993 (merge_in_block): Attempt to use a mem insn that is the next
3994 use of the original regno.
3996 2018-02-27 Martin Sebor <msebor@redhat.com>
3998 PR c++/83871
3999 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
4000 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
4002 2018-02-27 Martin Sebor <msebor@redhat.com>
4004 PR translation/84207
4005 * diagnostic-core.h (warning_n, error_n, inform_n): Change
4006 n argument to unsigned HOST_WIDE_INT.
4007 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
4008 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
4009 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
4010 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
4012 2018-02-27 Richard Biener <rguenther@suse.de>
4014 PR tree-optimization/84512
4015 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4016 Do not use the estimate returned from record_stmt_cost for
4017 the scalar iteration cost but sum properly using add_stmt_cost.
4019 2018-02-27 Richard Biener <rguenther@suse.de>
4021 PR tree-optimization/84466
4022 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4023 Adjust last change to less strictly validate use operands.
4025 2018-02-27 Martin Liska <mliska@suse.cz>
4027 PR gcov-profile/84548
4028 * gcov.c (process_file): Allow partial overlap and consider it
4029 also as group functions.
4030 (output_lines): Properly calculate range of lines for a group.
4032 2018-02-27 Martin Liska <mliska@suse.cz>
4034 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
4035 'ggc' suffixes. Change first column width.
4036 (timer::print): Fix formatting of the column.
4038 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
4040 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
4041 preserve inline entry blocks for the sake of debug inline
4042 entry point markers alone.
4043 (remove_unused_locals): Suggest in comments a better place to
4044 force the preservation of inline entry blocks that are
4045 otherwise unused, but do not preserve them.
4047 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4049 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
4051 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4053 PR target/84039
4054 * config/i386/constraints.md (Bs): Replace
4055 ix86_indirect_branch_register with
4056 TARGET_INDIRECT_BRANCH_REGISTER.
4057 (Bw): Likewise.
4058 * config/i386/i386.md (indirect_jump): Likewise.
4059 (tablejump): Likewise.
4060 (*sibcall_memory): Likewise.
4061 (*sibcall_value_memory): Likewise.
4062 Peepholes of indirect call and jump via memory: Likewise.
4063 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
4064 (*sibcall_value_GOT_32): Likewise.
4065 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4066 (GOT_memory_operand): Likewise.
4067 (call_insn_operand): Likewise.
4068 (sibcall_insn_operand): Likewise.
4069 (GOT32_symbol_operand): Likewise.
4070 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
4072 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4074 PR rtl-optimization/83496
4075 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
4076 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
4077 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
4078 redundant insn, if any.
4079 (relax_delay_slots): Likewise.
4080 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
4082 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
4084 PR tree-optimization/83965
4085 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
4086 that grouped statements are part of a reduction chain. Return
4087 true if the statement is not marked as a reduction itself but
4088 is part of a group.
4089 (vect_recog_dot_prod_pattern): Don't check whether the statement
4090 is part of a group here.
4091 (vect_recog_sad_pattern): Likewise.
4092 (vect_recog_widen_sum_pattern): Likewise.
4094 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
4096 PR debug/84545
4097 * final.c (rest_of_clean_state): Also look for calls inside sequences.
4099 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4101 PR target/84530
4102 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
4103 the bool argument.
4104 (ix86_output_indirect_function_return): New prototype.
4105 (ix86_split_simple_return_pop_internal): Likewise.
4106 * config/i386/i386.c (indirect_return_via_cx): New.
4107 (indirect_return_via_cx_bnd): Likewise.
4108 (indirect_thunk_name): Handle return va CX_REG.
4109 (output_indirect_thunk_function): Create alias for
4110 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
4111 (ix86_output_indirect_jmp): Remove the bool argument.
4112 (ix86_output_indirect_function_return): New function.
4113 (ix86_split_simple_return_pop_internal): Likewise.
4114 * config/i386/i386.md (*indirect_jump): Don't pass false
4115 to ix86_output_indirect_jmp.
4116 (*tablejump_1): Likewise.
4117 (simple_return_pop_internal): Change it to define_insn_and_split.
4118 Call ix86_split_simple_return_pop_internal to split it for
4119 -mfunction-return=.
4120 (simple_return_indirect_internal): Call
4121 ix86_output_indirect_function_return instead of
4122 ix86_output_indirect_jmp.
4124 2018-02-26 Jakub Jelinek <jakub@redhat.com>
4126 PR bootstrap/84405
4127 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
4128 memset and value initialization afterwards.
4130 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
4132 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
4134 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
4136 PR target/84521
4137 * common/config/aarch64/aarch64-common.c
4138 (aarch_option_optimization_table[]): Switch
4139 off fomit-frame-pointer
4141 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4142 Chung-Ju Wu <jasonwucj@gmail.com>
4144 * config/nds32/nds32-multiple.md (load_multiple): Disallow
4145 volatile memory.
4146 (store_multiple): Ditto.
4148 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
4150 * config.gcc: Add --with-cpu support for nds32 target.
4151 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
4152 * config/nds32/nds32.opt: Add -mcpu= option.
4154 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
4156 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
4157 isel=yes): Warn for these deprecated options.
4159 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
4161 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
4162 ISA_2_5_MASKS_EMBEDDED.
4164 2018-02-23 Jakub Jelinek <jakub@redhat.com>
4166 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4167 p->max as pointers rather than using iterative_hash_expr.
4169 2018-02-23 Carl Love <cel@us.ibm.com>
4171 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
4172 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
4173 BU_P8V_OVERLOAD_2.
4174 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
4175 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
4176 P8V_BUILTIN_VEC_VUNSIGNED2.
4178 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
4180 PR target/81572
4181 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
4182 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
4183 LRA_UNKNOWN_ALT.
4184 * lra-constraints.c (curr_insn_transform): Set up
4185 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
4186 LRA_UNKNOWN_ALT.
4187 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
4188 * lra-eliminations.c (spill_pseudos): Ditto.
4189 (process_insn_for_elimination): Ditto.
4190 * lra-lives.c (reg_early_clobber_p): Use the new macros.
4191 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
4192 LRA_NON_CLOBBERED_ALT.
4194 2018-02-22 Martin Sebor <msebor@redhat.com>
4196 PR tree-optimization/84480
4197 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
4198 to maybe_diag_stxncpy_trunc. Call it.
4199 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
4200 from gimple_fold_builtin_strcpy. Print inlining stack.
4201 (handle_builtin_stxncpy): Print inlining stack.
4202 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
4204 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
4206 PR target/84176
4207 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
4208 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
4209 and -fcheck-pointer-bounds are used together.
4210 (indirect_thunk_prefix): New enum.
4211 (indirect_thunk_need_prefix): New function.
4212 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
4213 "_nt" instead of "_bnd" for NOTRACK prefix.
4214 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
4215 (output_indirect_thunk_function): Likewise.
4216 (): Likewise.
4217 (ix86_code_end): Update output_indirect_thunk_function calls.
4218 (ix86_output_indirect_branch_via_reg): Replace
4219 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
4220 (ix86_output_indirect_branch_via_push): Likewise.
4221 (ix86_output_function_return): Likewise.
4222 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
4223 incompatible with -fcf-protection=branch and
4224 -fcheck-pointer-bounds.
4226 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4228 PR target/83335
4229 * config/aarch64/aarch64.c (aarch64_print_address_internal):
4230 Change gcc_assert call to output_operand_lossage.
4232 2018-02-22 Steve Ellcey <sellcey@cavium.com>
4234 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
4236 2018-02-22 DJ Delorie <dj@redhat.com>
4237 Sebastian Perta <sebastian.perta@renesas.com>
4238 Oleg Endo <olegendo@gcc.gnu.org>
4240 * config/rx/rx.c (rx_rtx_costs): New function.
4241 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
4243 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
4245 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
4247 2018-02-22 Martin Liska <mliska@suse.cz>
4249 PR driver/83193
4250 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
4251 Add "native" as a possible value.
4253 2018-02-22 Martin Liska <mliska@suse.cz>
4255 PR driver/83193
4256 * config/i386/i386.c (ix86_option_override_internal):
4257 Add "native" as a possible value for -march and -mtune.
4259 2018-02-22 Jakub Jelinek <jakub@redhat.com>
4261 PR target/84502
4262 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
4263 to all type variants.
4265 PR tree-optimization/84503
4266 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
4267 width as info->bitpos + info->bitsize - start.
4268 (merged_store_group::merge_overlapping): Simplify width computation.
4269 (check_no_overlap): New function.
4270 (imm_store_chain_info::try_coalesce_bswap): Compute expected
4271 start + width and last_order of the group, fail if check_no_overlap
4272 fails.
4273 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
4274 to group if check_no_overlap fails.
4276 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4278 * config/rs6000/altivec.md: Delete contraint arguments to
4279 define_expand, define_split, and define_peephole2, and in
4280 define_insn_and_split if always unused.
4281 * config/rs6000/darwin.md: Ditto.
4282 * config/rs6000/dfp.md: Ditto.
4283 * config/rs6000/rs6000.md: Ditto.
4284 * config/rs6000/sync.md: Ditto.
4285 * config/rs6000/vector.md: Ditto.
4286 * config/rs6000/vsx.md: Ditto.
4288 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
4290 * config/rs6000/altivec.md: Write output control strings as braced
4291 blocks instead of double-quoted strings.
4292 * config/rs6000/darwin.md: Ditto.
4293 * config/rs6000/rs6000.md: Ditto.
4294 * config/rs6000/vector.md: Ditto.
4295 * config/rs6000/vsx.md: Ditto.
4297 2018-02-21 Jason Merrill <jason@redhat.com>
4299 PR c++/84314 - ICE with templates and fastcall attribute.
4300 * attribs.c (build_type_attribute_qual_variant): Remove assert.
4302 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4304 * ipa-cp.c (determine_versionability): Fix comment typos.
4306 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
4308 PR c/84229
4309 * ipa-cp.c (determine_versionability): Do not version functions caling
4310 va_arg_pack.
4312 2018-02-21 Martin Liska <mliska@suse.cz>
4314 PR driver/83193
4315 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
4316 Add "native" as a possible value.
4317 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
4318 the macro when native cpu detection is available.
4320 2018-02-21 Martin Liska <mliska@suse.cz>
4322 PR driver/83193
4323 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
4324 Add "native" as a possible value.
4325 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
4326 when native cpu detection is available.
4328 2018-02-21 Jakub Jelinek <jakub@redhat.com>
4329 Martin Sebor <msebor@redhat.com>
4331 PR tree-optimization/84478
4332 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
4333 false.
4334 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
4335 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
4336 support which is conservatively correct, for 2 only stay conservative
4337 for maxlen. Formatting and comment capitalization fixes. Add STRICT
4338 argument to the 2 argument get_range_strlen, adjust 6 arg
4339 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
4340 false.
4341 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
4342 (gimple_fold_builtin_strlen): Pass true as last argument to
4343 get_range_strlen.
4345 2018-02-20 Martin Sebor <msebor@redhat.com>
4347 PR middle-end/84095
4348 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
4349 (builtin_memref::set_base_and_offset): Same. Handle inner references.
4350 (builtin_memref::builtin_memref): Factor out parts into
4351 set_base_and_offset and call it.
4353 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4355 PR middle-end/84406
4356 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
4357 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
4358 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
4359 search at the associated MODE_INT.
4361 2018-02-20 Jeff Law <law@redhat.com>
4363 PR middle-end/82123
4364 PR tree-optimization/81592
4365 PR middle-end/79257
4366 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
4367 for range data rather than using global data.
4368 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
4369 range data rather than using global data.
4370 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
4371 pass it to children as needed.
4372 (struct directive::fmtresult): Similarly.
4373 (struct directive::set_width): Similarly.
4374 (struct directive::set_precision): Similarly.
4375 (format_integer, format_directive, parse_directive): Similarly.
4376 (format_none): Accept unnamed vr_values parameter.
4377 (format_percent, format_floating, format_character): Similarly.
4378 (format_string, format_plain): Similarly.
4379 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
4380 the EVRP range analyzer for range data rather than using global data.
4381 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
4382 gimple-ssa-evrp-analyze.h
4383 (class sprintf_dom_walker): Add after_dom_children member function.
4384 Add evrp_range_analyzer member.
4385 (sprintf_dom_walker::before_dom_children): Call into the EVRP
4386 range analyzer as needed.
4387 (sprintf_dom_walker::after_dom_children): New member function.
4388 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
4389 if not optimizing.
4390 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
4391 (evrp_range_analyzer::pop_to_marker): Likewise.
4393 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
4395 PR tree-optimization/84419
4396 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
4397 with the required type if its current type is compatible but
4398 different.
4400 2018-02-20 Jakub Jelinek <jakub@redhat.com>
4402 PR middle-end/82004
4403 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
4404 after vectorization.
4406 2018-02-20 Martin Liska <mliska@suse.cz>
4408 PR driver/83193
4409 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
4410 possible values if we don't have a hint.
4412 2018-02-20 Martin Liska <mliska@suse.cz>
4414 PR c/84310
4415 PR target/79747
4416 * final.c (shorten_branches): Build align_tab array with one
4417 more element.
4418 * opts.c (finish_options): Add alignment option limit check.
4419 (MAX_CODE_ALIGN): Likewise.
4420 (MAX_CODE_ALIGN_VALUE): Likewise.
4421 * doc/invoke.texi: Document maximum allowed option value for
4422 all -falign-* options.
4424 2018-02-19 Jakub Jelinek <jakub@redhat.com>
4426 PR target/84146
4427 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
4428 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
4429 * var-tracking.c (emit_note_insn_var_location): Remove all references
4430 to NOTE_INSN_CALL_ARG_LOCATION.
4431 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
4432 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
4433 Use copy_rtx_if_shared.
4434 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
4435 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
4436 (dwarf2out_var_location): Remove handling of
4437 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
4438 on call_insn.
4439 * final.c (final_scan_insn): Remove all references to
4440 NOTE_INSN_CALL_ARG_LOCATION.
4441 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
4442 before dumping final insns.
4443 * except.c (emit_note_eh_region_end): Remove all references to
4444 NOTE_INSN_CALL_ARG_LOCATION.
4445 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
4446 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
4447 * config/arc/arc.c (hwloop_optimize): Likewise.
4448 * config/arm/arm.c (create_fix_barrier): Likewise.
4449 * config/s390/s390.c (s390_chunkify_start): Likewise.
4450 * config/sh/sh.c (find_barrier): Likewise.
4451 * config/i386/i386.c (rest_of_insert_endbranch,
4452 ix86_seh_fixup_eh_fallthru): Likewise.
4453 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
4454 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
4455 * config/frv/frv.c (frv_function_prologue): Likewise.
4456 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
4457 reg note.
4458 (note_outside_basic_block_p): Remove all references to
4459 NOTE_INSN_CALL_ARG_LOCATION.
4460 * gengtype.c (adjust_field_rtx_def): Likewise.
4461 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
4462 Likewise.
4463 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
4464 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
4466 PR c++/84444
4467 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
4468 is ADDR_EXPR.
4470 PR tree-optimization/84452
4471 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
4472 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
4473 is NULL.
4475 2018-02-19 Martin Liska <mliska@suse.cz>
4477 PR sanitizer/82183
4478 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
4480 2018-02-19 Martin Liska <mliska@suse.cz>
4481 Richard Sandiford <richard.sandiford@linaro.org>
4483 PR tree-optimization/82491
4484 * gimple-fold.c (get_base_constructor): Make earlier bail out
4485 to prevent ubsan.
4487 2018-02-19 Carl Love <cel@us.ibm.com>
4489 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
4490 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
4491 BU_P8V_OVERLOAD_1.
4492 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
4493 P8V_BUILTIN_VEC_NEG.
4495 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
4497 * config/rl78/rl78.md (movdf): New define expand.
4499 2018-02-19 Martin Liska <mliska@suse.cz>
4501 PR other/80589
4502 * doc/invoke.texi: Fix typo.
4503 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
4505 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
4507 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
4508 handle rs6000_single_float and rs6000_double_float specially for
4509 e500 family CPUs.
4511 2018-02-16 Jeff Law <law@redhat.com>
4513 * config/rx/rx.c (add_pop_cfi_notes): New function.;
4514 (pop_regs): Use it.
4516 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4518 PR ipa/84425
4519 * ipa-inline.c (inline_small_functions): Fix a typo.
4521 2018-02-16 Nathan Sidwell <nathan@acm.org>
4523 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
4525 2018-02-16 Carl Love <cel@us.ibm.com>
4527 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
4528 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
4529 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
4530 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
4531 expansion to P8V_BUILTIN_VEC_FLOAT2.
4533 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
4535 PR rtl-optimization/70023
4536 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
4537 src_regno into account.
4539 2018-02-16 Carl Love <cel@us.ibm.com>
4541 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
4542 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
4543 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
4544 * config/rs6000/rs6000.c: Remove case statements for
4545 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
4546 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
4547 and P9V_BUILTIN_VEC_VINSERT4B.
4548 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
4549 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
4550 * config/rs6000/vsx.md:
4551 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
4552 vec_insert4b.
4554 2018-02-16 Carl Love <cel@us.ibm.com>
4556 * config/rs6000/altivec.h: Add builtin names vec_extract4b
4557 vec_insert4b.
4558 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
4559 definitions.
4560 * config/rs6000/rs6000-c.c: Add the definitions for
4561 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
4562 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
4563 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
4564 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
4565 definition for insert4b and define insn *insert3b_internal.
4566 * doc/extend.texi: Add documentation for vec_extract4b.
4568 2018-02-16 Nathan Sidwell <nathan@acm.org>
4570 * doc/extend.texi (Backwards Compatibility): Mention friend
4571 injection. Note for-scope is deprecated.
4572 * doc/invoke.texi (-ffriend-injection): Deprecate.
4574 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
4576 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
4577 that moved to I2, also allow destinations that are a paradoxical
4578 subreg (instead of a normal reg).
4580 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
4582 PR target/83831
4583 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
4584 to QImode.
4586 2018-02-16 Richard Biener <rguenther@suse.de>
4588 PR tree-optimization/84037
4589 PR tree-optimization/84016
4590 PR target/82862
4591 * config/i386/i386.c (ix86_builtin_vectorization_cost):
4592 Adjust vec_construct for the fact we need additional higher latency
4593 128bit inserts for AVX256 and AVX512 vector builds.
4594 (ix86_add_stmt_cost): Scale vector construction cost for
4595 elementwise loads.
4597 2018-02-16 Richard Biener <rguenther@suse.de>
4599 PR tree-optimization/84417
4600 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
4601 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
4602 (non_rewritable_lvalue_p): Likewise, use poly-ints.
4604 2018-02-16 Martin Liska <mliska@suse.cz>
4606 PR sanitizer/84307
4607 * internal-fn.def (ASAN_CHECK): Set proper flags.
4608 (ASAN_MARK): Likewise.
4610 2018-02-16 Julia Koval <julia.koval@intel.com>
4612 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
4613 from PTA_CANNONLAKE.
4615 2018-02-16 Jakub Jelinek <jakub@redhat.com>
4617 PR target/84272
4618 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
4619 Use ++iter rather than iter++ for std::list iterators.
4620 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
4621 defer deleting them until all nodes in the forest are processed. Do
4622 free even leaf nodes. Change to_process into auto_vec.
4624 PR bootstrap/84405
4625 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
4626 * vec.h (vec_default_construct): Use memset instead of placement new
4627 if BROKEN_VALUE_INITIALIZATION is defined.
4628 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
4629 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
4630 is defined.
4632 PR rtl-optimization/83723
4633 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
4634 * lra.c (lra_substitute_pseudo): Likewise. If true, use
4635 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
4636 recursive calls.
4637 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
4638 callers.
4639 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
4641 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
4643 PR rtl-optimization/81443
4644 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
4645 from inner REGs to paradoxical SUBREGs.
4647 2018-02-16 Richard Biener <rguenther@suse.de>
4649 PR tree-optimization/84399
4650 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
4651 For operands we can analyze at their definition make sure we can
4652 analyze them at each use as well.
4654 2018-02-16 Richard Biener <rguenther@suse.de>
4656 PR tree-optimization/84190
4657 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
4658 volatile accesses if the decl isn't volatile.
4660 2018-02-15 Jason Merrill <jason@redhat.com>
4662 PR c++/84314 - ICE with templates and fastcall attribute.
4663 * attribs.c (build_type_attribute_qual_variant): Don't clobber
4664 TYPE_CANONICAL on an existing type.
4666 2018-02-15 Jakub Jelinek <jakub@redhat.com>
4668 PR tree-optimization/84383
4669 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
4670 dstoff nor call operand_equal_p if dstbase is NULL.
4672 PR tree-optimization/84334
4673 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
4674 also a CONSTANT_CLASS_P, punt.
4676 2018-02-14 Jim Wilson <jimw@sifive.com>
4678 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
4679 first SMALL_OPERAND check. New local min_second_step. Move assert
4680 to where locals are set. Add TARGET_RVC support.
4681 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
4683 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
4685 * doc/invoke.texi: Correct -Wformat-overflow code sample.
4687 2018-02-14 Martin Sebor <msebor@redhat.com>
4689 PR tree-optimization/83698
4690 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
4691 arrays constrain the offset range to their bounds.
4692 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
4693 (builtin_access::overlap): Avoid setting the size of overlap if it's
4694 already been set.
4695 (maybe_diag_overlap): Also consider arrays when deciding what values
4696 of offsets to include in diagnostics.
4698 2018-02-14 Martin Sebor <msebor@redhat.com>
4700 PR c/84108
4701 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
4702 that correspond to the kind of a declaration.
4704 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
4706 PR target/83984
4707 * config/pa/pa.md: Load address of PIC label using the linkage table
4708 if the label is nonlocal.
4710 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
4712 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
4713 warning message if user requests -maltivec=be.
4714 * doc/invoke.texi: Document deprecation of -maltivec=be.
4716 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
4718 PR target/84220
4719 * config/rs6000/rs6000-c.c: Update definitions for
4720 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
4721 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
4723 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4725 PR target/84239
4726 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
4727 add _get_ssp intrinsics. Remove argument from
4728 __builtin_ia32_rdssp[d|q].
4729 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
4730 * config/i386/i386-builtin.def: Remove argument from
4731 __builtin_ia32_rdssp[d|q].
4732 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
4733 ix86_expand_special_args_builtin for _rdssp[d|q].
4734 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
4735 Clear register before usage.
4736 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
4737 Add documentation for new _get_ssp and _inc_ssp intrinsics.
4739 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
4741 PR tree-optimization/84357
4742 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
4743 operand 1 of an ARRAY_REF too.
4745 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
4747 PR target/83831
4748 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
4749 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4750 declarations.
4751 (set_of_reg): New struct.
4752 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
4753 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
4754 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
4755 functions.
4756 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
4757 Split into bitclr, bitset, bitinvert patterns if appropriate.
4758 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
4759 use rx_fuse_in_memory_bitop.
4760 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
4761 to named insn, correct maximum insn length.
4763 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
4765 PR target/79242
4766 * machmode.def: Define a complex mode for PARTIAL_INT.
4767 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
4768 MODE_PARTIAL_INT.
4769 * doc/rtl.texi: Document CSPImode.
4770 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
4771 handling.
4772 (msp430_hard_regno_nregs_with_padding): Likewise.
4774 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
4776 PR target/84279
4777 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
4779 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
4781 PR rtl-optimization/84169
4782 * combine.c (try_combine): New variable split_i2i3. Set it to true if
4783 we generated a parallel as new i3 and we split that to new i2 and i3
4784 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
4785 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
4786 those to i2, not i1. Partially rewrite this scan code.
4788 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4790 PR c/82210
4791 * stor-layout.c (place_field): For variable length fields, adjust
4792 offset_align afterwards not just based on the field's alignment,
4793 but also on the size.
4795 PR middle-end/84309
4796 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
4797 of exps and logs in the use_exp2 case.
4799 2018-02-13 Jeff Law <law@redhat.com>
4801 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
4802 entry for "vector".
4804 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
4805 ARGS as unused.
4807 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
4809 PR debug/84342
4810 PR debug/84319
4811 * common.opt (gas-loc-support, gas-locview-support): New.
4812 (ginline-points, ginternal-reset-location-views): New.
4813 * doc/invoke.texi: Document them. Use @itemx where intended.
4814 (gvariable-location-views): Adjust.
4815 * target.def (reset_location_view): New.
4816 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
4817 (TARGET_RESET_LOCATION_VIEW): New.
4818 * doc/tm.texi: Rebuilt.
4819 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
4820 (dwarf2out_default_as_locview_support): New.
4821 (output_asm_line_debug_info): Use option variables.
4822 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
4823 (output_loc_list): Likewise.
4824 (add_high_low_attributes): Check option variables.
4825 Don't output entry view attribute in strict mode.
4826 (gen_inlined_subroutine_die): Check option variables.
4827 (dwarf2out_inline_entry): Likewise.
4828 (init_sections_and_labels): Likewise.
4829 (dwarf2out_early_finish): Likewise.
4830 (maybe_reset_location_view): New, from...
4831 (dwarf2out_var_location): ... here. Call it.
4832 * debug.h (dwarf2out_default_as_loc_support): Declare.
4833 (dwarf2out_default_as_locview_support): Declare.
4834 * hooks.c (hook_int_rtx_insn_0): New.
4835 * hooks.h (hook_int_rtx_insn_0): Declare.
4836 * toplev.c (process_options): Take -gas-loc-support and
4837 -gas-locview-support from dwarf2out. Enable
4838 -gvariable-location-views by default only with locview
4839 assembler support. Enable -ginternal-reset-location-views by
4840 default only if the target defines the corresponding hook.
4841 Enable -ginline-points by default if location views are
4842 enabled; force it disabled if statement frontiers are
4843 disabled.
4844 * tree-inline.c (expand_call_inline): Check option variables.
4845 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
4847 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
4849 PR tree-optimization/84321
4850 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
4851 handling. Also check whether the anti-range contains any values
4852 that satisfy the mask; switch to a VR_RANGE if not.
4854 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4856 PR sanitizer/84340
4857 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4859 2018-02-13 Martin Jambor <mjambor@suse.cz>
4861 PR c++/83990
4862 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4863 of call statements, also set location of a load to a temporary.
4865 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4867 * config/rl78/rl78.c (add_vector_labels): New function.
4868 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4869 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4870 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4871 which checks that no arguments are passed.
4872 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4873 * doc/extend.texi: Documentation for the new attribute.
4875 2018-02-13 Andreas Schwab <schwab@suse.de>
4877 * config/riscv/linux.h (CPP_SPEC): Define.
4879 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4881 PR target/84335
4882 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4883 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4884 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4885 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4886 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4887 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4888 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4889 temporarily for AES and PCLMUL builtins.
4891 PR tree-optimization/84339
4892 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4893 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4894 Formatting fixes.
4896 PR middle-end/84309
4897 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4898 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4899 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4900 inline function.
4901 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4902 inline function.
4903 * omp-simd-clone.h: New file.
4904 * omp-simd-clone.c: Include omp-simd-clone.h.
4905 (expand_simd_clones): No longer static.
4906 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4907 cgraph.h and omp-simd-clone.h.
4908 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4909 (vect_recog_widen_shift_pattern): Formatting fix.
4910 (vect_pattern_recog_1): Don't check optab for calls.
4912 PR target/84336
4913 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4914 operands[2] into a REG before using gen_lowpart on it.
4916 2018-02-12 Jeff Law <law@redhat.com>
4918 PR target/83760
4919 * config/sh/sh.c (find_barrier): Consider a sibling call
4920 a barrier as well.
4922 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4923 successfully back substituting a reg.
4925 2018-02-12 Richard Biener <rguenther@suse.de>
4927 PR tree-optimization/84037
4928 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4929 parameter, move visited init to caller.
4930 (vect_slp_analyze_operations): Separate cost from validity
4931 check, initialize visited once for all instances.
4932 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4933 for all instances.
4934 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4935 out an assert.
4936 (vect_model_promotion_demotion_cost): Likewise.
4937 (vectorizable_bswap): Guard cost modeling with !slp_node
4938 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4939 SLP stmts.
4940 (vectorizable_call): Likewise.
4941 (vectorizable_conversion): Likewise.
4942 (vectorizable_assignment): Likewise.
4943 (vectorizable_shift): Likewise.
4944 (vectorizable_operation): Likewise.
4945 (vectorizable_store): Likewise.
4946 (vectorizable_load): Likewise.
4947 (vectorizable_condition): Likewise.
4948 (vectorizable_comparison): Likewise.
4950 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4952 PR sanitizer/84307
4953 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4954 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4955 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4957 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4959 PR middle-end/83665
4960 * params.def (inline-min-speedup): Increase from 8 to 15.
4961 (max-inline-insns-auto): Decrease from 40 to 30.
4962 * ipa-split.c (consider_split): Add some buffer for function to
4963 be considered inlining candidate.
4964 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4965 default values.
4967 2018-02-12 Richard Biener <rguenther@suse.de>
4969 PR tree-optimization/84037
4970 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4971 matched stmts if we cannot swap the non-matched ones.
4973 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4975 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4976 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4977 _mm_maskz_scalef_round_ss): New intrinsics.
4978 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4979 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4980 __builtin_ia32_scalefss_round): Remove.
4981 (__builtin_ia32_scalefsd_mask_round,
4982 __builtin_ia32_scalefss_mask_round): New intrinsics.
4983 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4984 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4985 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4986 "<round_constraint>")): Changed to ...
4987 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4988 "<round_scalar_constraint>")): ... this.
4989 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4990 %0, %1, %2<round_op3>}"): Changed to ...
4991 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4992 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4993 %2<round_scalar_mask_op3>}"): ... this.
4994 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4996 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4998 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4999 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
5000 (_mm_maskz_sqrt_round_ss): New intrinsics.
5001 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
5002 (__builtin_ia32_sqrtsd_mask_round)
5003 (__builtin_ia32_sqrtss_mask_round): New builtins.
5004 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
5005 (__builtin_ia32_sqrtss_round): Remove.
5006 (__builtin_ia32_sqrtsd_mask_round)
5007 (__builtin_ia32_sqrtss_mask_round): New builtins.
5008 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
5009 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
5010 ((match_operand:VF_128 1 "vector_operand"
5011 "xBm,<round_constraint>")): Changed to ...
5012 ((match_operand:VF_128 1 "vector_operand"
5013 "xBm,<round_scalar_constraint>")): ... this.
5014 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
5015 %0, %2, %<iptr>1<round_op3>}): Changed to ...
5016 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
5017 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
5018 %<iptr>1<round_scalar_mask_op3>}): ... this.
5019 ((set_attr "prefix" "<round_prefix>")): Changed to ...
5020 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
5022 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
5024 PR target/84266
5025 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
5026 Cast vec_cmpeq result to correct type.
5027 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
5028 Cast vec_cmpgt result to correct type.
5030 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
5032 * final.c (final_scan_insn_1): Renamed from...
5033 (final_scan_insn): ... this. New wrapper, to recover
5034 seen from the outermost call in recursive ones.
5035 * config/sparc/sparc.c (output_return): Drop seen from call.
5036 (output_sibcall): Likewise.
5037 * config/visium/visium.c (output_branch): Likewise.
5039 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
5041 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
5042 function label.
5044 2018-02-10 Alan Modra <amodra@gmail.com>
5046 PR target/84300
5047 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
5048 Specify LR as an input.
5050 2018-02-10 Jakub Jelinek <jakub@redhat.com>
5052 PR sanitizer/83987
5053 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
5054 remove_member_access_dummy_vars): New functions.
5055 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
5056 lower_omp_1, execute_lower_omp): Use them.
5058 PR rtl-optimization/84308
5059 * shrink-wrap.c (spread_components): Release todo vector.
5061 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
5063 PR rtl-optimization/57193
5064 * ira-color.c (struct allocno_color_data): Add member
5065 conflict_allocno_hard_prefs.
5066 (update_conflict_allocno_hard_prefs): New.
5067 (bucket_allocno_compare_func): Add a preference based on
5068 conflict_allocno_hard_prefs.
5069 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
5070 (color_allocnos): Remove a dead code. Initiate
5071 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
5073 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5075 PR target/84226
5076 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
5077 constraint from =wa to wa. Avoid a subreg on the output operand,
5078 instead use a pseudo and subreg it in a move.
5079 (p9_xxbrd_<mode>): Changed to ...
5080 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
5081 (p9_xxbrd_v2df): New expander.
5082 (p9_xxbrw_<mode>): Changed to ...
5083 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
5084 (p9_xxbrw_v4sf): New expander.
5086 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5088 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
5090 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
5092 PR target/83926
5093 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
5094 multiply in 32-bit mode.
5095 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
5096 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
5097 mode.
5099 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
5101 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
5102 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
5103 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
5104 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
5106 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
5108 PR lto/84213
5109 * dwarf2out.c (is_trivial_indirect_ref): New function.
5110 (dwarf2out_late_global_decl): Do not generate a location
5111 attribute for variables that have a non-trivial DECL_VALUE_EXPR
5112 and that are not defined in the current unit.
5114 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5116 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
5117 instead of a libcall for UNORDERED.
5119 2018-02-09 Tamar Christina <tamar.christina@arm.com>
5121 PR target/82641
5122 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
5123 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
5125 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5127 PR target/PR84295
5128 * config/s390/s390.c (s390_set_current_function): Invoke
5129 s390_indirect_branch_settings also if fndecl didn't change.
5131 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5133 * config/rs6000/rs6000.md (blockage): Set length to zero.
5135 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
5137 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
5139 2018-02-09 Jakub Jelinek <jakub@redhat.com>
5141 PR sanitizer/84285
5142 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
5143 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
5144 -static-lib*san.
5146 PR debug/84252
5147 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
5148 PARALLEL incoming that failed vt_get_decl_and_offset check.
5150 PR middle-end/84237
5151 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
5152 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
5153 TREE_READONLY bit.
5154 (get_variable_section): For decls in named .bss* sections pass true as
5155 second argument to bss_initializer_p.
5157 2018-02-09 Marek Polacek <polacek@redhat.com>
5158 Jakub Jelinek <jakub@redhat.com>
5160 PR c++/83659
5161 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
5162 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
5163 Sync some changes from cxx_fold_indirect_ref.
5165 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
5167 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
5168 markers.
5169 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
5170 (BLOCK_INLINE_ENTRY_LABEL): New.
5171 (dwarf2out_var_location): Disregard inline entry markers.
5172 (inline_entry_data): New struct.
5173 (inline_entry_data_hasher): New hashtable type.
5174 (inline_entry_data_hasher::hash): New.
5175 (inline_entry_data_hasher::equal): New.
5176 (inline_entry_data_table): New variable.
5177 (add_high_low_attributes): Add DW_AT_entry_pc and
5178 DW_AT_GNU_entry_view attributes if a pending entry is found
5179 in inline_entry_data_table. Add old entry_pc attribute only
5180 if debug nonbinding markers are disabled.
5181 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
5182 markers are enabled.
5183 (block_within_block_p, dwarf2out_inline_entry): New.
5184 (dwarf2out_finish): Check that no entries remained in
5185 inline_entry_data_table.
5186 * final.c (reemit_insn_block_notes): Handle inline entry notes.
5187 (final_scan_insn, notice_source_line): Likewise.
5188 (rest_of_clean_state): Skip inline entry markers.
5189 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
5190 markers.
5191 * gimple.c (gimple_build_debug_inline_entry): New.
5192 * gimple.h (enum gimple_debug_subcode): Add
5193 GIMPLE_DEBUG_INLINE_ENTRY.
5194 (gimple_build_debug_inline_entry): Declare.
5195 (gimple_debug_inline_entry_p): New.
5196 (gimple_debug_nonbind_marker_p): Adjust.
5197 * insn-notes.def (INLINE_ENTRY): New.
5198 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
5199 inline entry marker notes.
5200 (print_insn): Likewise.
5201 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
5202 (INSN_DEBUG_MARKER_KIND): Likewise.
5203 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
5204 * tree-inline.c (expand_call_inline): Build and insert
5205 debug_inline_entry stmt.
5206 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
5207 inline entry blocks early, if nonbind markers are enabled.
5208 (dump_scope_block): Dump fragment info.
5209 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
5210 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
5211 (gimple_build_debug_inline_entry): New.
5212 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
5213 Enable/disable inline entry points too.
5214 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
5215 (DEBUG_INSN): Describe inline entry markers.
5217 * common.opt (gvariable-location-views): New.
5218 (gvariable-location-views=incompat5): New.
5219 * config.in: Rebuilt.
5220 * configure: Rebuilt.
5221 * configure.ac: Test assembler for view support.
5222 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
5223 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
5224 * dwarf2out.c (var_loc_view): New typedef.
5225 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
5226 (dwarf2out_locviews_in_attribute): New.
5227 (dwarf2out_locviews_in_loclist): New.
5228 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
5229 (enum dw_line_info_opcode): Add LI_adv_address.
5230 (struct dw_line_info_table): Add view.
5231 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
5232 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
5233 (zero_view_p): New variable.
5234 (ZERO_VIEW_P): New macro.
5235 (output_asm_line_debug_info): New.
5236 (struct var_loc_node): Add view.
5237 (add_AT_view_list, AT_loc_list): New.
5238 (add_var_loc_to_decl): Add view param. Test it against last.
5239 (new_loc_list): Add view params. Record them.
5240 (AT_loc_list_ptr): Handle loc and view lists.
5241 (view_list_to_loc_list_val_node): New.
5242 (print_dw_val): Handle dw_val_class_view_list.
5243 (size_of_die): Likewise.
5244 (value_format): Likewise.
5245 (loc_list_has_views): New.
5246 (gen_llsym): Set vl_symbol too.
5247 (maybe_gen_llsym, skip_loc_list_entry): New.
5248 (dwarf2out_maybe_output_loclist_view_pair): New.
5249 (output_loc_list): Output view list or entries too.
5250 (output_view_list_offset): New.
5251 (output_die): Handle dw_val_class_view_list.
5252 (output_dwarf_version): New.
5253 (output_compilation_unit_header): Use it.
5254 (output_skeleton_debug_sections): Likewise.
5255 (output_rnglists, output_line_info): Likewise.
5256 (output_pubnames, output_aranges): Update version comments.
5257 (output_one_line_info_table): Output view numbers in asm comments.
5258 (dw_loc_list): Determine current endview, pass it to new_loc_list.
5259 Call maybe_gen_llsym.
5260 (loc_list_from_tree_1): Adjust.
5261 (add_AT_location_description): Create view list attribute if
5262 needed, check it's absent otherwise.
5263 (convert_cfa_to_fb_loc_list): Adjust.
5264 (maybe_emit_file): Call output_asm_line_debug_info for test.
5265 (dwarf2out_var_location): Reset views as needed. Precompute
5266 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
5267 attribute. Set view.
5268 (new_line_info_table): Reset next view.
5269 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
5270 (dwarf2out_source_line): Likewise. Output view resets and labels to
5271 the assembler, or select appropriate line info opcodes.
5272 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
5273 (optimize_string_length): Catch it. Adjust.
5274 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
5275 dw_val_class_view_list, and remove it if no longer needed.
5276 (hash_loc_list): Hash view numbers.
5277 (loc_list_hasher::equal): Compare them.
5278 (optimize_location_lists): Check whether a view list symbol is
5279 needed, and whether the locview attribute is present, and
5280 whether they match. Remove the locview attribute if no longer
5281 needed.
5282 (index_location_lists): Call skip_loc_list_entry for test.
5283 (dwarf2out_finish): Call output_asm_line_debug_info for test.
5284 Use output_dwarf_version.
5285 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
5286 (struct dw_val_node): Add val_view_list.
5287 * final.c (SEEN_NEXT_VIEW): New.
5288 (set_next_view_needed): New.
5289 (clear_next_view_needed): New.
5290 (maybe_output_next_view): New.
5291 (final_start_function): Rename to...
5292 (final_start_function_1): ... this. Take pointer to FIRST,
5293 add SEEN parameter. Emit param bindings in the initial view.
5294 (final_start_function): Reintroduce SEEN-less interface.
5295 (final): Rename to...
5296 (final_1): ... this. Take SEEN parameter. Output final pending
5297 next view at the end.
5298 (final): Reintroduce seen-less interface.
5299 (final_scan_insn): Output pending next view before switching
5300 sections or ending a block. Mark the next view as needed when
5301 outputting variable locations. Notify debug backend of section
5302 changes, and of location view changes.
5303 (rest_of_handle_final): Adjust.
5304 * toplev.c (process_options): Autodetect value for debug variable
5305 location views option. Warn on incompat5 without -gdwarf-5.
5306 * doc/invoke.texi (gvariable-location-views): New.
5307 (gvariable-location-views=incompat5): New.
5308 (gno-variable-location-views): New.
5310 2018-02-08 David Malcolm <dmalcolm@redhat.com>
5312 PR tree-optimization/84136
5313 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
5314 that the result of find_edge is non-NULL.
5316 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5318 PR target/83008
5319 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
5320 storing integer register in SImode. Fix cost of 256 and 512
5321 byte aligned SSE register store.
5323 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
5325 * config/i386/i386.c (ix86_multiplication_cost): Fix
5326 multiplication cost for TARGET_AVX512DQ.
5328 2018-02-08 Marek Polacek <polacek@redhat.com>
5330 PR tree-optimization/84238
5331 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
5332 get_range_strlen.
5334 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5336 PR tree-optimization/84265
5337 * tree-vect-stmts.c (vectorizable_store): Don't treat
5338 VMAT_CONTIGUOUS accesses as grouped.
5339 (vectorizable_load): Likewise.
5341 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5343 PR tree-optimization/81635
5344 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
5345 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
5346 (test_round_for_mask): New functions.
5347 (wide_int_cc_tests): Call test_round_for_mask.
5348 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
5349 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
5350 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
5351 range returned by get_range_info.
5353 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
5355 PR ipa/81360
5356 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
5357 * symtab.c: Include builtins.h
5358 (symtab_node::output_to_lto_symbol_table_p): Move here
5359 from lto-streamer-out.c:output_symbol_p.
5360 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
5361 (output_symbol_p): Move all logic to symtab.c
5362 (produce_symtab): Update.
5364 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5366 * config/s390/s390-opts.h (enum indirect_branch): Define.
5367 * config/s390/s390-protos.h (s390_return_addr_from_memory)
5368 (s390_indirect_branch_via_thunk)
5369 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
5370 (enum s390_indirect_branch_type): Define.
5371 * config/s390/s390.c (struct s390_frame_layout, struct
5372 machine_function): Remove.
5373 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
5374 (indirect_branch_table_label_no, indirect_branch_table_name):
5375 Define variables.
5376 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
5377 (enum s390_indirect_branch_option): Define.
5378 (s390_return_addr_from_memory): New function.
5379 (s390_handle_string_attribute): New function.
5380 (s390_attribute_table): Add new attribute handler.
5381 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
5382 (s390_indirect_branch_via_thunk): New function.
5383 (s390_indirect_branch_via_inline_thunk): New function.
5384 (s390_function_ok_for_sibcall): When jumping via thunk disallow
5385 sibling call optimization for non z10 compiles.
5386 (s390_emit_call): Force indirect branch target to be a single
5387 register. Add r1 clobber for non-z10 compiles.
5388 (s390_emit_epilogue): Emit return jump via return_use expander.
5389 (s390_reorg): Handle JUMP_INSNs as execute targets.
5390 (s390_option_override_internal): Perform validity checks for the
5391 new command line options.
5392 (s390_indirect_branch_attrvalue): New function.
5393 (s390_indirect_branch_settings): New function.
5394 (s390_set_current_function): Invoke s390_indirect_branch_settings.
5395 (s390_output_indirect_thunk_function): New function.
5396 (s390_code_end): Implement target hook.
5397 (s390_case_values_threshold): Implement target hook.
5398 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
5399 macros.
5400 * config/s390/s390.h (struct s390_frame_layout)
5401 (struct machine_function): Move here from s390.c.
5402 (TARGET_INDIRECT_BRANCH_NOBP_RET)
5403 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
5404 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
5405 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
5406 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
5407 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
5408 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
5409 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
5410 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
5411 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
5412 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
5413 (mnemonic attribute): Add values which aren't recognized
5414 automatically.
5415 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
5416 pattern for branch conversion. Fix mnemonic attribute.
5417 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
5418 indirect branch via thunk if requested.
5419 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
5420 ("*indirect_jump"): Disable for branch conversion using out of
5421 line thunks.
5422 ("indirect_jump_via_thunk<mode>_z10")
5423 ("indirect_jump_via_thunk<mode>")
5424 ("indirect_jump_via_inlinethunk<mode>_z10")
5425 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
5426 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
5427 ("casesi_jump_via_inlinethunk<mode>_z10")
5428 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
5429 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
5430 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
5431 ("*indirect2_jump"): Disable for branch conversion.
5432 ("casesi_jump"): Turn into expander and expand patterns for branch
5433 conversion.
5434 ("return_use"): New expander.
5435 ("*return"): Emit return via thunk and rename it to ...
5436 ("*return<mode>"): ... this one.
5437 * config/s390/s390.opt: Add new options and and enum for the
5438 option values.
5440 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
5442 * lra-constraints.c (match_reload): Unconditionally use
5443 gen_lowpart_SUBREG, rather than selecting between that
5444 and equivalent gen_rtx_SUBREG code.
5446 2018-02-08 Richard Biener <rguenther@suse.de>
5448 PR tree-optimization/84233
5449 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
5450 changed flag instead of boguously re-using phi_inserted.
5452 2018-02-08 Martin Jambor <mjambor@suse.cz>
5454 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
5455 static local variables.
5457 2018-02-08 Richard Biener <rguenther@suse.de>
5459 PR tree-optimization/84278
5460 * tree-vect-stmts.c (vectorizable_store): When looking for
5461 smaller vector types to perform grouped strided loads/stores
5462 make sure the mode is supported by the target.
5463 (vectorizable_load): Likewise.
5465 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5467 * config/aarch64/aarch64.c (aarch64_components_for_bb):
5468 Increase LDP/STP opportunities by adding adjacent callee-saves.
5470 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
5472 PR rtl-optimization/84068
5473 PR rtl-optimization/83459
5474 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
5476 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
5478 PR tree-optimization/84224
5479 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
5480 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
5481 non-zero arguments.
5483 2018-02-07 Iain Sandoe <iain@codesourcery.com>
5485 PR target/84113
5486 * config/rs6000/altivec.md (*restore_world): Remove LR use.
5487 * config/rs6000/predicates.md (restore_world_operation): Adjust op
5488 count, remove one USE.
5490 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
5492 * doc/install.texi (Configuration): Document the
5493 --with-long-double-format={ibm,ieee} PowerPC configuration
5494 options.
5496 PR target/84154
5497 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
5498 Convert from define_expand to be define_insn_and_split. Rework
5499 float/double/_Float128 conversions to QI/HI/SImode to work with
5500 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
5501 conversions to QI/HImode types did a store and then a load to
5502 truncate the value. For conversions to VSX registers, don't split
5503 the insn, instead emit the code directly. Use the code iterator
5504 any_fix to combine signed and unsigned conversions.
5505 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
5506 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
5507 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5508 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
5509 (fix_<mode>di2_hw): Likewise.
5510 (fixuns_<mode>di2_hw): Likewise.
5511 (fix_<mode>si2_hw): Likewise.
5512 (fixuns_<mode>si2_hw): Likewise.
5513 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
5514 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
5515 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
5516 fix<uns>_trunc<SFDF:mode>si2_p8.
5517 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
5518 used.
5519 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
5520 (fix<uns>_<mode>_mem): Likewise.
5521 (fctiw<u>z_<mode>_mem): Likewise.
5522 (fix<uns>_<mode>_mem): Likewise.
5523 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
5524 the register allocator from doing a direct move to the GPRs to do
5525 a store, and instead use the ISA 3.0 store byte/half-word from
5526 vector register instruction. For IEEE 128-bit floating point,
5527 also optimize stores of 32-bit ints.
5528 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
5530 2018-02-07 Alan Hayward <alan.hayward@arm.com>
5532 * genextract.c (push_pathstr_operand): New function to support
5533 [a-zA-Z].
5534 (walk_rtx): Call push_pathstr_operand.
5535 (print_path): Support [a-zA-Z].
5537 2018-02-07 Richard Biener <rguenther@suse.de>
5539 PR tree-optimization/84037
5540 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
5541 (cse_and_gimplify_to_preheader): Declare.
5542 (vect_get_place_in_interleaving_chain): Likewise.
5543 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
5544 ivexpr_map.
5545 (_loop_vec_info::~_loop_vec_info): Delete it.
5546 (cse_and_gimplify_to_preheader): New function.
5547 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
5548 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
5549 (vectorizable_load): Likewise. For grouped stores always base
5550 the IV on the first element.
5551 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
5552 condition before gimplifying.
5554 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5556 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
5557 *DIV_EXPR and *MOD_EXPR.
5559 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
5561 PR target/84248
5562 * config/i386/i386.c (ix86_option_override_internal): Mask out
5563 the CF_SET bit when checking -fcf-protection.
5565 2018-02-07 Tom de Vries <tom@codesourcery.com>
5567 PR libgomp/84217
5568 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
5569 enough.
5571 2018-02-07 Richard Biener <rguenther@suse.de>
5573 PR tree-optimization/84204
5574 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
5575 this place.
5577 PR tree-optimization/84205
5578 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
5579 special-case isl_ast_op_zdiv_r.
5581 PR tree-optimization/84223
5582 * graphite-scop-detection.c (gather_bbs::before_dom_children):
5583 Only add conditions from within the region.
5584 (gather_bbs::after_dom_children): Adjust.
5586 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
5588 PR target/84209
5589 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
5590 * config/avr/avr.md: Only post-reload split REG-REG moves if
5591 either register is GENERAL_REG_P.
5593 2018-02-07 Jakub Jelinek <jakub@redhat.com>
5595 PR tree-optimization/84235
5596 * tree-ssa-scopedtables.c
5597 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
5598 if the subtraction is performed in floating point type where NaNs are
5599 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
5600 build 1. Formatting fix.
5602 2018-02-06 Jakub Jelinek <jakub@redhat.com>
5604 PR target/84146
5605 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
5606 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
5607 and skip it regardless of bb boundaries. Use CALL_P macro,
5608 don't test INSN_P (insn) together with CALL_P or JUMP_P check
5609 unnecessarily, formatting fix.
5611 2018-02-06 Michael Collison <michael.collison@arm.com>
5613 * config/arm/thumb2.md:
5614 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
5615 (*thumb_mov_notscc): Ditto.
5617 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
5619 PR target/84154
5620 * config/rs6000/rs6000.md (su code attribute): Use "u" for
5621 unsigned_fix, not "s".
5623 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5625 * configure.ac (gcc_fn_eh_frame_ro): New function.
5626 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
5627 correct .eh_frame permissions.
5628 * configure: Regenerate.
5630 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
5632 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
5633 irrelevant options.
5635 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5637 * config/rs6000/rs6000.c (rs6000_option_override_internal):
5638 Display warning message for -mno-speculate-indirect-jumps.
5640 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
5642 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
5643 Undocumented.
5644 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
5646 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
5648 PR tree-optimization/84225
5649 * tree-eh.c (find_trapping_overflow): Only call
5650 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
5652 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5654 PR target/84145
5655 * config/i386/i386.c: Reimplement the check of possible options
5656 -mibt/-mshstk conbination. Change error messages.
5657 * doc/invoke.texi: Fix a typo: remove extra '='.
5659 2018-02-06 Marek Polacek <polacek@redhat.com>
5661 PR tree-optimization/84228
5662 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
5664 2018-02-06 Tamar Christina <tamar.christina@arm.com>
5666 PR target/82641
5667 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
5668 emitted arch directives.
5669 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
5670 __ARM_FEATURE_COPROC before changing architectures.
5672 2018-02-06 Richard Biener <rguenther@suse.de>
5674 * config/i386/i386.c (print_reg): Fix typo.
5675 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
5677 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
5679 * configure: Regenerate.
5681 2018-02-05 Martin Sebor <msebor@redhat.com>
5683 PR tree-optimization/83369
5684 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
5685 inlining context.
5687 2018-02-05 Martin Liska <mliska@suse.cz>
5689 * doc/invoke.texi: Cherry-pick upstream r323995.
5691 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
5693 * ira.c (ira_init_register_move_cost): Adjust comment.
5695 2018-02-05 Martin Liska <mliska@suse.cz>
5697 PR gcov-profile/84137
5698 * doc/gcov.texi: Fix typo in documentation.
5700 2018-02-05 Martin Liska <mliska@suse.cz>
5702 PR gcov-profile/83879
5703 * doc/gcov.texi: Document necessity of --dynamic-list-data when
5704 using dlopen functionality.
5706 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
5708 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
5709 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
5710 _mm_maskz_range_ss, _mm_mask_range_round_ss,
5711 _mm_maskz_range_round_ss): New intrinsics.
5712 (__builtin_ia32_rangesd128_round)
5713 (__builtin_ia32_rangess128_round): Remove.
5714 (__builtin_ia32_rangesd128_mask_round,
5715 __builtin_ia32_rangess128_mask_round): New builtins.
5716 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
5717 __builtin_ia32_rangess128_round): Remove.
5718 (__builtin_ia32_rangesd128_mask_round,
5719 __builtin_ia32_rangess128_mask_round): New builtins.
5720 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
5721 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
5722 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
5723 "<round_saeonly_constraint>")): Changed to ...
5724 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
5725 "<round_saeonly_scalar_constraint>")): ... this.
5726 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
5727 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
5728 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
5729 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
5730 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
5732 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
5734 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
5735 options.
5736 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
5737 Remove all values except native, 8540 and 8548.
5739 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
5741 * config/i386/i386.c (ix86_output_function_return): Pass
5742 INVALID_REGNUM, instead of -1, as invalid register number to
5743 indirect_thunk_name and output_indirect_thunk.
5745 2018-02-02 Julia Koval <julia.koval@intel.com>
5747 * config.gcc: Add -march=icelake.
5748 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
5749 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
5750 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
5751 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
5752 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
5753 (processor_target_table): Add icelake.
5754 (ix86_option_override_internal): Handle new PTAs.
5755 (get_builtin_code_for_version): Handle icelake.
5756 (M_INTEL_COREI7_ICELAKE): New.
5757 (fold_builtin_cpu): Handle icelake.
5758 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
5759 * doc/invoke.texi: Add -march=icelake.
5761 2018-02-02 Julia Koval <julia.koval@intel.com>
5763 * config/i386/i386.c (ix86_option_override_internal): Change flags type
5764 to wide_int_bitmask.
5765 * wide-int-bitmask.h: New.
5767 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
5769 PR target/84066
5770 * config/i386/i386.md: Replace Pmode with word_mode in
5771 builtin_setjmp_setup and builtin_longjmp to support x32.
5773 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
5775 PR target/56010
5776 PR target/83743
5777 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
5778 #include "opts.h".
5779 (rs6000_supported_cpu_names): New static variable.
5780 (linux_cpu_translation_table): Likewise.
5781 (elf_platform) <cpu>: Define new static variable and use it.
5782 Translate kernel AT_PLATFORM name to canonical name if needed.
5783 Error if platform name is unknown.
5785 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
5787 PR target/84089
5788 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
5790 2018-02-01 Jeff Law <law@redhat.com>
5792 PR target/84128
5793 * config/i386/i386.c (release_scratch_register_on_entry): Add new
5794 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
5795 the scratch if RELEASE_VIA_POP is false.
5796 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
5797 If we have to save a temporary register, decrement SIZE appropriately.
5798 Pass new arguments to release_scratch_register_on_entry.
5799 (ix86_adjust_stack_and_probe): Likewise.
5800 (ix86_emit_probe_stack_range): Pass new arguments to
5801 release_scratch_register_on_entry.
5803 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
5805 PR rtl-optimization/84157
5806 * combine.c (change_zero_ext): Use REG_P predicate in
5807 front of HARD_REGISTER_P predicate.
5809 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
5811 * config/avr/avr.c (avr_option_override): Move disabling of
5812 -fdelete-null-pointer-checks to...
5813 * common/config/avr/avr-common.c (avr_option_optimization_table):
5814 ...here.
5816 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5818 PR tree-optimization/81635
5819 * tree-data-ref.c (split_constant_offset_1): For types that
5820 wrap on overflow, try to use range info to prove that wrapping
5821 cannot occur.
5823 2018-02-01 Renlin Li <renlin.li@arm.com>
5825 PR target/83370
5826 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
5827 TAILCALL_ADDR_REGS.
5828 (aarch64_register_move_cost): Likewise.
5829 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
5830 TAILCALL_ADDR_REGS.
5831 (REG_CLASS_NAMES): Likewise.
5832 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
5833 TAILCALL_ADDR_REGS. Remove IP registers.
5834 * config/aarch64/aarch64.md (Ucs): Update register constraint.
5836 2018-02-01 Richard Biener <rguenther@suse.de>
5838 * domwalk.h (dom_walker::dom_walker): Add additional constructor
5839 for specifying RPO order and allow NULL for that.
5840 * domwalk.c (dom_walker::dom_walker): Likewise.
5841 (dom_walker::walk): Handle NULL RPO order.
5842 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
5843 in RPO order.
5844 (rewrite_update_dom_walker): Likewise.
5845 (mark_def_dom_walker): Likewise.
5847 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5849 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
5850 (aarch64_maybe_expand_sve_subreg_move): Declare.
5851 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
5852 * config/aarch64/predicates.md (aarch64_any_register_operand): New
5853 predicate.
5854 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5855 that are semantically a reverse operation.
5856 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5857 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5858 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5859 functions.
5860 (aarch64_can_change_mode_class): For big-endian, forbid changes
5861 between two SVE modes if they have different element sizes.
5863 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5865 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5866 the TImode handling for big-endian targets.
5868 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5870 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5871 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5872 not just bytes.
5873 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5874 Remove BSWAP handing for big-endian targets and use the form of
5875 LD1RQ appropariate for the mode.
5877 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5879 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5880 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5881 duplicated element.
5883 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5885 PR tearget/83845
5886 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5887 check for operands that need to go through aarch64_sve_reload_be.
5889 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5891 PR tree-optimization/81661
5892 PR tree-optimization/84117
5893 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5894 * tree-eh.c: Include gimplify.h.
5895 (find_trapping_overflow, replace_trapping_overflow,
5896 rewrite_to_non_trapping_overflow): New functions.
5897 * tree-vect-loop.c: Include tree-eh.h.
5898 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5899 * tree-data-ref.c: Include tree-eh.h.
5900 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5902 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5904 PR rtl-optimization/84123
5905 * combine.c (change_zero_ext): Check if hard register satisfies
5906 can_change_dest_mode before calling gen_lowpart_SUBREG.
5908 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5910 PR target/82444
5911 * ira.c (ira_init_register_move_cost): Remove assert.
5913 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5915 PR rtl-optimization/84071
5916 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5917 * doc/tm.texi: Regenerate.
5919 2018-01-31 Richard Biener <rguenther@suse.de>
5921 PR tree-optimization/84132
5922 * tree-data-ref.c (analyze_miv_subscript): Properly
5923 check whether evolution_function_is_affine_multivariate_p
5924 before calling gcd_of_steps_may_divide_p.
5926 2018-01-31 Julia Koval <julia.koval@intel.com>
5928 PR target/83618
5929 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5930 * config/i386/i386.md (rdpid_rex64) New.
5931 (rdpid): Make 32bit only.
5933 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5935 PR lto/84105
5936 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5937 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5939 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5941 Revert
5942 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5944 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5946 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5948 PR rtl-optimization/84071
5949 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5950 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5952 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5954 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5955 (arc_attribute_table): Add 'aux' attribute.
5956 (arc_in_small_data_p): Consider aux like variables.
5957 (arc_is_aux_reg_p): New function.
5958 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5959 (arc_get_aux_arg): New function.
5960 (prepare_move_operands): Handle aux-register access.
5961 (arc_handle_aux_attribute): New function.
5962 * doc/extend.texi (ARC Variable attributes): Add subsection.
5964 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5966 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5967 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5968 (arc_attribute_table): Add 'uncached' attribute.
5969 (arc_print_operand): Print '.di' flag for uncached memory
5970 accesses.
5971 (arc_in_small_data_p): Do not consider for small data the uncached
5972 types.
5973 (arc_is_uncached_mem_p): New function.
5974 * config/arc/predicates.md (compact_store_memory_operand): Check
5975 for uncached memory accesses.
5976 (nonvol_nonimm_operand): Likewise.
5977 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5979 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5981 PR c/84100
5982 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5983 falign-loops=): Add Optimization flag.
5985 2018-01-30 Jeff Law <law@redhat.com>
5987 PR target/84064
5988 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5989 INT_REGISTERS_SAVED. Check it prior to calling
5990 get_scratch_register_on_entry.
5991 (ix86_adjust_stack_and_probe): Similarly.
5992 (ix86_emit_probe_stack_range): Similarly.
5993 (ix86_expand_prologue): Corresponding changes.
5995 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5997 PR target/40411
5998 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5999 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
6001 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
6003 PR target/84112
6004 * lra-constraints.c (curr_insn_transform): Process AND in the
6005 address.
6007 2018-01-30 Jakub Jelinek <jakub@redhat.com>
6009 PR rtl-optimization/83986
6010 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
6011 dependence against last_pending_memory_flush in addition to
6012 pending_jump_insns.
6014 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
6016 PR tree-optimization/81611
6017 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
6018 copies.
6020 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6022 PR target/83758
6023 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
6024 a reg rtx.
6026 2018-01-30 Richard Biener <rguenther@suse.de>
6027 Jakub Jelinek <jakub@redhat.com>
6029 PR tree-optimization/84111
6030 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
6031 inner loops added during recursion, as they don't have up-to-date
6032 SSA form.
6034 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6036 PR ipa/81360
6037 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
6038 (can_inline_edge_by_limits_p): ... here.
6039 (can_early_inline_edge_p, check_callers,
6040 update_caller_keys, update_callee_keys, recursive_inlining,
6041 add_new_edges_to_heap, speculation_useful_p,
6042 inline_small_functions,
6043 inline_small_functions, flatten_function,
6044 inline_to_all_callers_1): Update.
6046 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
6048 * profile-count.c (profile_count::combine_with_ipa_count): Handle
6049 zeros correctly.
6051 2018-01-30 Richard Biener <rguenther@suse.de>
6053 PR tree-optimization/83008
6054 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
6055 invariant and constant vector uses in stmts when they need
6056 more than one stmt.
6058 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6060 PR bootstrap/84017
6061 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
6062 * configure: Regenerate.
6064 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6066 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
6067 pattern.
6068 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
6069 Use gen_rtx_REG rather than gen_lowpart.
6071 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6073 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
6074 rather than 0 when creating partial subregs.
6076 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
6078 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
6079 of usage.
6081 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
6083 PR target/81550
6084 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
6085 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
6086 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
6087 flags. This restores the settings used before the 2017-07-24.
6088 Turning off pre increment/decrement/modify allows IVOPTS to
6089 optimize DF/SF loops where the index is an int.
6091 2018-01-29 Richard Biener <rguenther@suse.de>
6092 Kelvin Nilsen <kelvin@gcc.gnu.org>
6094 PR bootstrap/80867
6095 * tree-vect-stmts.c (vectorizable_call): Don't call
6096 targetm.vectorize_builtin_md_vectorized_function if callee is
6097 NULL.
6099 2018-01-22 Carl Love <cel@us.ibm.com>
6101 * doc/extend.tex: Fix typo in second arg in
6102 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
6104 2018-01-29 Richard Biener <rguenther@suse.de>
6106 PR tree-optimization/84086
6107 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
6108 (flush_ssaname_freelist): When SSA names were released reset
6109 the SCEV hash table.
6111 2018-01-29 Richard Biener <rguenther@suse.de>
6113 PR tree-optimization/84057
6114 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
6115 removed paths when removing edges.
6117 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
6119 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
6120 -mfunction-return=@var{choice}.
6122 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
6124 PR diagnostic/84034
6125 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
6126 Handle CR like TAB.
6127 (layout::print_source_line): Likewise.
6128 (test_get_line_width_without_trailing_whitespace): Add test cases.
6130 2018-01-27 Jakub Jelinek <jakub@redhat.com>
6132 PR middle-end/84040
6133 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
6134 debug insns.
6136 2018-01-26 Jim Wilson <jimw@sifive.com>
6138 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
6140 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
6141 specified.
6143 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6145 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
6146 and CMP + SUB-immediate -> SUBS.
6148 2018-01-26 Martin Sebor <msebor@redhat.com>
6150 PR tree-optimization/83896
6151 * tree-ssa-strlen.c (get_string_len): Rename...
6152 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
6153 Avoid assuming length is constant.
6154 (handle_char_store): Use HOST_WIDE_INT for string length.
6156 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
6158 PR target/81763
6159 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
6160 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
6162 2018-01-26 Richard Biener <rguenther@suse.de>
6164 PR rtl-optimization/84003
6165 * dse.c (record_store): Only record redundant stores when
6166 the earlier store aliases at least all accesses the later one does.
6168 2018-01-26 Jakub Jelinek <jakub@redhat.com>
6170 PR rtl-optimization/83985
6171 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
6172 REG_CFA_RESTORE insns.
6173 (delete_unmarked_insns): Don't ignore separate shrink wrapping
6174 REG_CFA_RESTORE insns here.
6176 PR c/83989
6177 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
6178 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
6180 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6182 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
6183 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
6184 (arc_init): Likewise.
6185 (arc_override_options): Likewise.
6186 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
6187 value.
6188 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
6189 support.
6190 * config/arc/arc.h (TARGET_DBNZ): Define.
6191 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
6192 properly set the tune attribute.
6193 (dbnz): Use TARGET_DBNZ guard.
6194 * config/arc/arc.opt (mtune): Add core3 option.
6196 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6198 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
6199 recognize new pic like addresses.
6200 (arc_delegitimize_address): Clean up.
6202 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6204 * config/arc/arc-arches.def: Option mrf16 valid for all
6205 architectures.
6206 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
6207 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
6208 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
6209 * config/arc/arc-tables.opt: Regenerate.
6210 * config/arc/arc.c (arc_conditional_register_usage): Handle
6211 reduced register file case.
6212 (arc_file_start): Set must have build attributes.
6213 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
6214 mrf16 option value.
6215 * config/arc/arc.opt (mrf16): Add new option.
6216 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
6217 * config/arc/genmultilib.awk: Handle new mrf16 option.
6218 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
6219 * config/arc/t-multilib: Regenerate.
6220 * doc/invoke.texi (ARC Options): Document mrf16 option.
6222 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6224 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
6225 * config/arc/arc.c (arc_handle_secure_attribute): New function.
6226 (arc_attribute_table): Add 'secure_call' attribute.
6227 (arc_print_operand): Print secure call operand.
6228 (arc_function_ok_for_sibcall): Don't optimize tail calls when
6229 secure.
6230 (arc_is_secure_call_p): New function. * config/arc/arc.md
6231 (call_i): Add support for sjli instruction.
6232 (call_value_i): Likewise.
6233 * config/arc/constraints.md (Csc): New constraint.
6235 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
6236 John Eric Martin <John.Martin@emmicro-us.com>
6238 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
6239 * config/arc/arc.c (_arc_jli_section): New struct.
6240 (arc_jli_section): New type.
6241 (rc_jli_sections): New static variable.
6242 (arc_handle_jli_attribute): New function.
6243 (arc_attribute_table): Add jli_always and jli_fixed attribute.
6244 (arc_file_end): New function.
6245 (TARGET_ASM_FILE_END): Define.
6246 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
6247 (arc_add_jli_section): New function.
6248 (jli_call_scan): Likewise.
6249 (arc_reorg): Call jli_call_scan.
6250 (arc_output_addsi): Remove 'S' from printing asm operand.
6251 (arc_is_jli_call_p): New function.
6252 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
6253 operand.
6254 (movhi_insn): Likewise.
6255 (movsi_insn): Likewise.
6256 (movsi_set_cc_insn): Likewise.
6257 (loadqi_update): Likewise.
6258 (load_zeroextendqisi_update): Likewise.
6259 (load_signextendqisi_update): Likewise.
6260 (loadhi_update): Likewise.
6261 (load_zeroextendhisi_update): Likewise.
6262 (load_signextendhisi_update): Likewise.
6263 (loadsi_update): Likewise.
6264 (loadsf_update): Likewise.
6265 (movsicc_insn): Likewise.
6266 (bset_insn): Likewise.
6267 (bxor_insn): Likewise.
6268 (bclr_insn): Likewise.
6269 (bmsk_insn): Likewise.
6270 (bicsi3_insn): Likewise.
6271 (cmpsi_cc_c_insn): Likewise.
6272 (movsi_ne): Likewise.
6273 (movsi_cond_exec): Likewise.
6274 (clrsbsi2): Likewise.
6275 (norm_f): Likewise.
6276 (normw): Likewise.
6277 (swap): Likewise.
6278 (divaw): Likewise.
6279 (flag): Likewise.
6280 (sr): Likewise.
6281 (kflag): Likewise.
6282 (ffs): Likewise.
6283 (ffs_f): Likewise.
6284 (fls): Likewise.
6285 (call_i): Remove 'S' asm letter, add jli instruction.
6286 (call_value_i): Likewise.
6287 * config/arc/arc.op (mjli-always): New option.
6288 * config/arc/constraints.md (Cji): New constraint.
6289 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
6290 operand.
6291 (subsf3_fpx): Likewise.
6292 (mulsf3_fpx): Likewise.
6293 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
6294 asm operand.
6295 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
6296 function attrbutes.
6297 * doc/invoke.texi (ARC): Document mjli-always option.
6299 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
6301 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
6302 avoid addition with 0 and use incw and decw where possible.
6304 2018-01-26 Richard Biener <rguenther@suse.de>
6306 PR tree-optimization/81082
6307 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
6308 association if it requires casting to unsigned.
6309 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
6310 from fold_plusminus_mult_expr to catch important cases late when
6311 range info is available.
6313 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6315 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
6316 * configure.ac (hidden_linkonce): New test.
6317 * configure: Regenerate.
6318 * config.in: Regenerate.
6320 2018-01-26 Julia Koval <julia.koval@intel.com>
6322 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
6323 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
6324 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
6325 _mm_mask_bitshuffle_epi64_mask): Fix type.
6326 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
6327 USI_FTYPE_V4DI_V4DI_USI): Remove.
6328 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
6329 __builtin_ia32_vpshufbitqmb256_mask,
6330 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
6331 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
6332 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
6334 2018-01-26 Alan Modra <amodra@gmail.com>
6336 PR target/84033
6337 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
6338 UNSPEC_VBPERMQ. Sort other unspecs.
6340 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
6342 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
6344 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
6346 PR middle-end/83055
6347 * predict.c (drop_profile): Do not push/pop cfun; update also
6348 node->count.
6349 (handle_missing_profiles): Fix logic looking for zero profiles.
6351 2018-01-25 Jakub Jelinek <jakub@redhat.com>
6353 PR middle-end/83977
6354 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
6355 on functions with #pragma omp declare simd or functions with simd
6356 attribute.
6357 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
6358 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
6359 Remove trailing \n from warning_at calls.
6361 2018-01-25 Tom de Vries <tom@codesourcery.com>
6363 PR target/84028
6364 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6365 for neutered workers.
6367 2018-01-24 Joseph Myers <joseph@codesourcery.com>
6369 PR target/68467
6370 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
6371 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
6373 2018-01-24 Jeff Law <law@redhat.com>
6375 PR target/83994
6376 * i386.c (get_probe_interval): Move to earlier point.
6377 (ix86_compute_frame_layout): If -fstack-clash-protection and
6378 the frame is larger than the probe interval, then use pushes
6379 to save registers rather than reg->mem moves.
6380 (ix86_expand_prologue): Remove conditional for int_registers_saved
6381 assertion.
6383 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
6385 PR target/84014
6386 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
6387 min/max for never referenced object.
6389 2018-01-24 Jakub Jelinek <jakub@redhat.com>
6391 PR middle-end/83977
6392 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
6393 here.
6394 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
6395 attributes from DECL_ATTRIBUTES (decl) without affecting
6396 DECL_ATTRIBUTES (current_function_decl).
6397 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
6398 functions with non-NULL DECL_ABSTRACT_ORIGIN.
6400 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
6402 PR tree-optimization/83979
6403 * fold-const.c (fold_comparison): Use constant_boolean_node
6404 instead of boolean_{true,false}_node.
6406 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
6408 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
6409 with zero counts.
6411 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6413 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6414 Simplify the clause that sets the length attribute.
6415 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6416 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
6417 clause that sets the length attribute.
6418 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6420 2018-01-24 Tom de Vries <tom@codesourcery.com>
6422 PR target/83589
6423 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
6424 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
6425 Add strict parameter.
6426 (prevent_branch_around_nothing): Insert dummy insn between branch to
6427 label and label with no ptx insn inbetween.
6428 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
6430 2018-01-24 Tom de Vries <tom@codesourcery.com>
6432 PR target/81352
6433 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
6434 for neutered threads in warp.
6435 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
6437 2018-01-24 Richard Biener <rguenther@suse.de>
6439 PR tree-optimization/83176
6440 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
6441 operands.
6443 2018-01-24 Richard Biener <rguenther@suse.de>
6445 PR tree-optimization/82819
6446 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
6447 code generating pluses that are no-ops in the target precision.
6449 2018-01-24 Richard Biener <rguenther@suse.de>
6451 PR middle-end/84000
6452 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
6454 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6456 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
6457 to merge probabilities.
6458 * predict.c (probably_never_executed): Also mark as cold functions
6459 with global 0 profile and guessed local profile.
6460 * profile-count.c (profile_probability::combine_with_count): New
6461 member function.
6462 * profile-count.h (profile_probability::operator*,
6463 profile_probability::operator*=, profile_probability::operator/,
6464 profile_probability::operator/=): Reduce precision to adjusted
6465 and set value to guessed on contradictory divisions.
6466 (profile_probability::combine_with_freq): Remove.
6467 (profile_probability::combine_wiht_count): Declare.
6468 (profile_count::force_nonzero):: Set to adjusted.
6469 (profile_count::probability_in):: Set quality to adjusted.
6470 * tree-ssa-tail-merge.c (replace_block_by): Use
6471 combine_with_count.
6473 2018-01-23 Andrew Waterman <andrew@sifive.com>
6474 Jim Wilson <jimw@sifive.com>
6476 * config/riscv/riscv.c (riscv_stack_boundary): New.
6477 (riscv_option_override): Set riscv_stack_boundary. Handle
6478 riscv_preferred_stack_boundary_arg.
6479 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
6480 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
6481 (STACK_BOUNDARY): Set to riscv_stack_boundary.
6482 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
6483 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
6484 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
6486 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
6488 PR target/83905
6489 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
6490 of struct ix86_frame.
6491 (ix86_expand_epilogue): Likewise. Add a local variable for
6492 the reg_save_offset field in struct ix86_frame.
6494 2018-01-23 Bin Cheng <bin.cheng@arm.com>
6496 PR tree-optimization/82604
6497 * tree-loop-distribution.c (enum partition_kind): New enum item
6498 PKIND_PARTIAL_MEMSET.
6499 (partition_builtin_p): Support above new enum item.
6500 (generate_code_for_partition): Ditto.
6501 (compute_access_range): Differentiate cases that equality can be
6502 proven at all loops, the innermost loops or no loops.
6503 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
6504 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
6505 (finalize_partitions, distribute_loop): Don't fuse partition of
6506 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
6507 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
6508 parloop is enabled.
6510 2018-01-23 Martin Liska <mliska@suse.cz>
6512 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
6513 order to ignore the predictor.
6514 (PRED_POLYMORPHIC_CALL): Likewise.
6515 (PRED_RECURSIVE_CALL): Likewise.
6517 2018-01-23 Martin Liska <mliska@suse.cz>
6519 * tree-profile.c (tree_profiling): Print function header to
6520 aware reader which function we are working on.
6521 * value-prof.c (gimple_find_values_to_profile): Do not print
6522 not interesting value histograms.
6524 2018-01-23 Martin Liska <mliska@suse.cz>
6526 * profile-count.h (enum profile_quality): Add
6527 profile_uninitialized as the first value. Do not number values
6528 as they are zero based.
6529 (profile_count::verify): Update sanity check.
6530 (profile_probability::verify): Likewise.
6532 2018-01-23 Nathan Sidwell <nathan@acm.org>
6534 * doc/invoke.texi (ffor-scope): Deprecate.
6536 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6538 PR tree-optimization/83510
6539 * domwalk.c (set_all_edges_as_executable): New function.
6540 (dom_walker::dom_walker): Convert bool param
6541 "skip_unreachable_blocks" to enum reachability. Move setup of
6542 edge flags to set_all_edges_as_executable and only do it when
6543 reachability is REACHABLE_BLOCKS.
6544 * domwalk.h (enum dom_walker::reachability): New enum.
6545 (dom_walker::dom_walker): Convert bool param
6546 "skip_unreachable_blocks" to enum reachability.
6547 (set_all_edges_as_executable): New decl.
6548 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
6549 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
6550 "reachability".
6551 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
6552 but converting true to REACHABLE_BLOCKS.
6553 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
6554 * tree-vrp.c
6555 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
6556 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
6557 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
6558 REACHABLE_BLOCKS.
6559 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
6560 if check_all_array_refs will be called.
6562 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6564 * tree.c (selftest::test_location_wrappers): Add more test
6565 coverage.
6567 2018-01-23 David Malcolm <dmalcolm@redhat.com>
6569 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
6570 (selftest::test_bit_in_range): Likewise.
6572 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6574 PR testsuite/83888
6575 * doc/sourcebuild.texi (vect_float): Say that the selector
6576 only describes the situation when -funsafe-math-optimizations is on.
6577 (vect_float_strict): Document.
6579 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
6581 PR tree-optimization/83965
6582 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
6583 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
6584 instead of checking only for a reduction.
6585 (vect_recog_widen_sum_pattern): Likewise.
6587 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6589 * predict.c (probably_never_executed): Only use precise profile info.
6590 (compute_function_frequency): Skip after inlining hack since we now
6591 have quality checking.
6593 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
6595 * profile-count.h (profile_probability::very_unlikely,
6596 profile_probability::unlikely, profile_probability::even): Set
6597 precision to guessed.
6599 2018-01-23 Richard Biener <rguenther@suse.de>
6601 PR tree-optimization/83963
6602 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
6603 Properly terminate dominator walk when crossing the exit edge not
6604 when visiting its source block.
6606 2018-01-23 Jakub Jelinek <jakub@redhat.com>
6608 PR c++/83918
6609 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
6610 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
6612 2018-01-22 Jakub Jelinek <jakub@redhat.com>
6614 PR tree-optimization/83957
6615 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
6616 semicolon after for body surrounded by braces.
6618 PR tree-optimization/83081
6619 * profile-count.h (profile_probability::split): New method.
6620 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
6621 Use profile_probability::split.
6622 (do_compare_rtx_and_jump): Fix adjustment of probabilities
6623 when splitting a single conditional jump into 2.
6625 2018-01-22 David Malcolm <dmalcolm@redhat.com>
6627 PR tree-optimization/69452
6628 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
6629 decl.
6631 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6633 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
6634 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
6635 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
6637 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6639 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
6640 * config/rl78/rl78.md (movdi): New define_expand.
6641 * config/rl78/rl78.c (rl78_split_movdi): New function.
6643 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
6645 PR target/83862
6646 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
6647 no longer used.
6648 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
6649 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
6650 128-bit to produce an UNSPEC move to get the double word with the
6651 signbit and then a shift directly to do signbit.
6652 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
6653 implementation with a new version that just does either a direct
6654 move or a regular move. Move memory interface to separate insns.
6655 Move insns so they are next to the expander.
6656 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
6657 with signbit move. Split big and little endian case.
6658 (signbit<mode>2_dm_mem_le): Likewise.
6659 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
6660 (signbit<mode>2_dm2): Likewise.
6662 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6664 * config/rl78/rl78.md (anddi3): New define_expand.
6666 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6668 * config/rl78/rl78.md (umindi3): New define_expand.
6670 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6672 * config/rl78/rl78.md (smindi3): New define_expand.
6674 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6676 * config/rl78/rl78.md (smaxdi3): New define_expand.
6678 2018-01-22 Carl Love <cel@us.ibm.com>
6680 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
6681 LVX_V1TI): Add macro expansion.
6682 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
6683 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
6684 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
6685 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
6686 Change check to determine if the instruction is a byte reversing
6687 entry. Fix typo in comment.
6688 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
6689 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
6690 Add def_builtin calls for new builtins.
6691 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
6692 Add define_insn expansion.
6694 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6696 * config/rl78/rl78.md (umaxdi3): New define_expand.
6698 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
6700 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
6701 for non-QImode registers.
6703 2018-01-22 Richard Biener <rguenther@suse.de>
6705 PR tree-optimization/83963
6706 * graphite-scop-detection.c (scop_detection::get_sese): Delay
6707 including the loop exit block.
6708 (scop_detection::merge_sese): Likewise.
6709 (scop_detection::add_scop): Do it here instead.
6711 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6713 * doc/sourcebuild.texi (arm_softfloat): Document.
6715 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
6717 PR gcc/77734
6718 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
6719 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
6720 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
6722 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6723 David Edelsohn <dje.gcc@gmail.com>
6725 PR target/83946
6726 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6727 Change "crset eq" to "crset 2".
6728 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6729 (*call_indirect_aix<mode>_nospec): Likewise.
6730 (*call_value_indirect_aix<mode>_nospec): Likewise.
6731 (*call_indirect_elfv2<mode>_nospec): Likewise.
6732 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
6733 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
6734 change assembly output from . to $.
6735 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6736 (indirect_jump<mode>_nospec): Change assembly output from . to $.
6737 (*tablejump<mode>_internal1_nospec): Likewise.
6739 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
6741 PR target/80870
6742 * config/sh/sh_optimize_sett_clrt.cc:
6743 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
6745 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
6747 PR tree-optimization/83940
6748 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
6749 offset_dt to vect_constant_def rather than vect_unknown_def_type.
6750 (vect_check_load_store_mask): Add a mask_dt_out parameter and
6751 use it to pass back the definition type.
6752 (vect_check_store_rhs): Likewise rhs_dt_out.
6753 (vect_build_gather_load_calls): Add a mask_dt argument and use
6754 it instead of a call to vect_is_simple_use.
6755 (vectorizable_store): Update calls to vect_check_load_store_mask
6756 and vect_check_store_rhs. Use the dt returned by the latter instead
6757 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
6758 instead of calls to vect_is_simple_use. Pass the scalar rather
6759 than the vector operand to vect_is_simple_use when handling
6760 second and subsequent copies of an rhs value.
6761 (vectorizable_load): Update calls to vect_check_load_store_mask
6762 and vect_build_gather_load_calls. Use the cached mask_dt and
6763 gs_info.offset_dt instead of calls to vect_is_simple_use.
6765 2018-01-20 Jakub Jelinek <jakub@redhat.com>
6767 PR middle-end/83945
6768 * tree-emutls.c: Include gimplify.h.
6769 (lower_emutls_2): New function.
6770 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
6771 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
6772 it before further processing.
6774 PR target/83930
6775 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
6776 UINTVAL (trueop1) instead of INTVAL (op1).
6778 2018-01-19 Jakub Jelinek <jakub@redhat.com>
6780 PR debug/81570
6781 PR debug/83728
6782 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
6783 INCOMING_FRAME_SP_OFFSET if not defined.
6784 (scan_trace): Add ENTRY argument. If true and
6785 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
6786 emit a note to adjust the CFA offset.
6787 (create_cfi_notes): Adjust scan_trace callers.
6788 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
6789 INCOMING_FRAME_SP_OFFSET in the CIE.
6790 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
6791 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
6792 Likewise.
6793 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
6794 * doc/tm.texi: Regenerated.
6796 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6798 PR rtl-optimization/83147
6799 * lra-constraints.c (remove_inheritance_pseudos): Use
6800 lra_substitute_pseudo_within_insn.
6802 2018-01-19 Tom de Vries <tom@codesourcery.com>
6803 Cesar Philippidis <cesar@codesourcery.com>
6805 PR target/83920
6806 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
6808 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
6810 PR target/83790
6811 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
6812 spaces for function labels.
6814 2018-01-19 Martin Liska <mliska@suse.cz>
6816 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
6817 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
6818 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
6819 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
6820 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
6821 (PRED_CONST_RETURN): Change from 69 to 65.
6822 (PRED_NULL_RETURN): Change from 91 to 71.
6823 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
6824 (PRED_LOOP_GUARD): Change from 66 to 73.
6826 2018-01-19 Martin Liska <mliska@suse.cz>
6828 * predict.c (predict_insn_def): Add new assert.
6829 (struct branch_predictor): Change type to signed integer.
6830 (test_prediction_value_range): Amend test to cover
6831 PROB_UNINITIALIZED.
6832 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
6833 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
6834 (PRED_LOOP_ITERATIONS_MAX): Likewise.
6835 (PRED_LOOP_IV_COMPARE): Likewise.
6836 * predict.h (PROB_UNINITIALIZED): Define new constant.
6838 2018-01-19 Martin Liska <mliska@suse.cz>
6840 * predict.c (dump_prediction): Add new format for
6841 analyze_brprob.py script which is enabled with -details
6842 suboption.
6843 * profile-count.h (precise_p): New function.
6845 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6847 PR tree-optimization/83922
6848 * tree-vect-loop.c (vect_verify_full_masking): Return false if
6849 there are no statements that need masking.
6850 (vect_active_double_reduction_p): New function.
6851 (vect_analyze_loop_operations): Use it when handling phis that
6852 are not in the loop header.
6854 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6856 PR tree-optimization/83914
6857 * tree-vect-loop.c (vectorizable_induction): Don't convert
6858 init_expr or apply the peeling adjustment for inductions
6859 that are nested within the vectorized loop.
6861 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6863 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6864 instead of NEG.
6866 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6868 PR sanitizer/81715
6869 PR testsuite/83882
6870 * function.h (gimplify_parameters): Add gimple_seq * argument.
6871 * function.c: Include gimple.h and options.h.
6872 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6873 for the added local temporaries if needed.
6874 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6875 if there are any parameter cleanups, wrap whole body into a
6876 try/finally with the cleanups.
6878 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6880 PR target/82964
6881 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6882 Use GET_MODE_CLASS for scalar floating point.
6884 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6886 PR ipa/82256
6887 patch by PaX Team
6888 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6889 Fix call of call_cgraph_insertion_hooks.
6891 2018-01-18 Martin Sebor <msebor@redhat.com>
6893 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6895 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6897 PR ipa/83619
6898 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6899 frequencies.
6901 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6903 PR other/70268
6904 * common.opt: (-ffile-prefix-map): New option.
6905 * opts.c (common_handle_option): Defer it.
6906 * opts-global.c (handle_common_deferred_options): Handle it.
6907 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6908 * file-prefix-map.h: New file.
6909 (remap_debug_filename, add_debug_prefix_map): ...here.
6910 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6911 * final.c (debug_prefix_map, add_debug_prefix_map
6912 remap_debug_filename): Move to...
6913 * file-prefix-map.c: New file.
6914 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6915 generalize, get rid of alloca(), use strrchr() instead of strchr().
6916 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6917 Implement in terms of add_prefix_map().
6918 (remap_macro_filename, remap_debug_filename): Implement in term of
6919 remap_filename().
6920 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6921 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6922 * dbxout.c: Include file-prefix-map.h.
6923 * varasm.c: Likewise.
6924 * vmsdbgout.c: Likewise.
6925 * xcoffout.c: Likewise.
6926 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6927 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6928 * doc/invoke.texi (-ffile-prefix-map): Document.
6929 (-fdebug-prefix-map): Update description.
6931 2018-01-18 Martin Liska <mliska@suse.cz>
6933 * config/i386/i386.c (indirect_thunk_name): Document that also
6934 lfence is emitted.
6935 (output_indirect_thunk): Document why both instructions
6936 (pause and lfence) are generated.
6938 2018-01-18 Richard Biener <rguenther@suse.de>
6940 PR tree-optimization/83887
6941 * graphite-scop-detection.c
6942 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6943 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6944 (scop_detection::merge_sese): Re-implement with a flood-fill
6945 algorithm that properly finds a SESE region if it exists.
6947 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6949 PR c/61240
6950 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6951 pointer_diff optimizations use view_convert instead of convert.
6953 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6955 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6956 Generate different code for -mno-speculate-indirect-jumps.
6957 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6958 (*call_indirect_aix<mode>): Disable for
6959 -mno-speculate-indirect-jumps.
6960 (*call_indirect_aix<mode>_nospec): New define_insn.
6961 (*call_value_indirect_aix<mode>): Disable for
6962 -mno-speculate-indirect-jumps.
6963 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6964 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6965 -mno-speculate-indirect-jumps.
6966 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6968 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6970 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6971 long double type, set the flags for noting the default long double
6972 type, even if we don't pass or return a long double type.
6974 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6976 PR ipa/83051
6977 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6978 failure.
6980 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6982 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6983 support for merge[hl].
6984 (fold_mergehl_helper): New helper function.
6985 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6986 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6987 (altivec_vmrglw_direct): Add xxmrglw insn.
6989 2018-01-17 Andrew Waterman <andrew@sifive.com>
6991 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6992 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6994 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6996 PR lto/83121
6997 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6998 call the lto_location_cache before reading the
6999 DECL_SOURCE_LOCATION of the types.
7001 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
7002 Richard Sandiford <richard.sandiford@linaro.org>
7004 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
7005 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
7006 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
7007 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
7008 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
7009 Add declaration.
7010 * config/aarch64/constraints.md (aarch64_movti_operand):
7011 Limit immediates.
7012 * config/aarch64/predicates.md (Uti): Add new constraint.
7014 2018-01-17 Carl Love <cel@us.ibm.com>
7016 * config/rs6000/vsx.md (define_expand xl_len_r,
7017 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
7018 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
7019 lxvll.
7020 (define_expand, define_insn): Move the shift left from the
7021 define_insn to the define_expand for lxvl and stxvl instructions.
7022 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
7023 and XL_LEN_R definitions to PURE.
7025 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
7027 * config/i386/i386.c (indirect_thunk_name): Declare regno
7028 as unsigned int. Compare regno with INVALID_REGNUM.
7029 (output_indirect_thunk): Ditto.
7030 (output_indirect_thunk_function): Ditto.
7031 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
7032 in the call to output_indirect_thunk_function.
7034 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
7036 PR middle-end/83884
7037 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
7038 rather than the size of inner_type to determine the stack slot size
7039 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
7041 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
7043 PR target/83546
7044 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
7045 to PTA_SILVERMONT.
7047 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
7049 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
7050 endian Linux systems to optionally enable multilibs for selecting
7051 the long double type if the user configured an explicit type.
7052 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
7053 have no long double multilibs if not defined.
7054 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
7055 warn if the user used -mabi={ieee,ibm}longdouble and we built
7056 multilibs for long double.
7057 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
7058 appropriate multilib option.
7059 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
7060 multilib options.
7061 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
7062 for building long double multilibs.
7063 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
7065 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
7067 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
7068 copies.
7070 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
7071 64 bits.
7072 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
7073 128 bits.
7075 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
7076 variables.
7078 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
7079 return value.
7081 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7083 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
7084 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
7086 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7088 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
7089 different rtl trees depending on TARGET_64BIT.
7090 (rs6000_gen_lvx): Likewise.
7092 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
7094 * config/visium/visium.md (nop): Tweak comment.
7095 (hazard_nop): Likewise.
7097 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7099 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
7100 -mspeculate-indirect-jumps.
7101 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
7102 for -mno-speculate-indirect-jumps.
7103 (*call_indirect_elfv2<mode>_nospec): New define_insn.
7104 (*call_value_indirect_elfv2<mode>): Disable for
7105 -mno-speculate-indirect-jumps.
7106 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
7107 (indirect_jump): Emit different RTL for
7108 -mno-speculate-indirect-jumps.
7109 (*indirect_jump<mode>): Disable for
7110 -mno-speculate-indirect-jumps.
7111 (*indirect_jump<mode>_nospec): New define_insn.
7112 (tablejump): Emit different RTL for
7113 -mno-speculate-indirect-jumps.
7114 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
7115 (tablejumpsi_nospec): New define_expand.
7116 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
7117 (tablejumpdi_nospec): New define_expand.
7118 (*tablejump<mode>_internal1): Disable for
7119 -mno-speculate-indirect-jumps.
7120 (*tablejump<mode>_internal1_nospec): New define_insn.
7121 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
7122 option.
7124 2018-01-16 Artyom Skrobov tyomitch@gmail.com
7126 * caller-save.c (insert_save): Drop unnecessary parameter. All
7127 callers updated.
7129 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7130 Richard Biener <rguenth@suse.de>
7132 PR libgomp/83590
7133 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
7134 return early, inline manually is_gimple_sizepos. Make sure if we
7135 call gimplify_expr we don't end up with a gimple constant.
7136 * tree.c (variably_modified_type_p): Don't return true for
7137 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
7138 * gimplify.h (is_gimple_sizepos): Remove.
7140 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7142 PR tree-optimization/83857
7143 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
7144 vectorizable_live_operation for pure SLP statements.
7145 (vectorizable_live_operation): Handle PHIs.
7147 2018-01-16 Richard Biener <rguenther@suse.de>
7149 PR tree-optimization/83867
7150 * tree-vect-stmts.c (vect_transform_stmt): Precompute
7151 nested_in_vect_loop_p since the scalar stmt may get invalidated.
7153 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7155 PR c/83844
7156 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
7157 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
7158 If off is not INTEGER_CST, issue a may not be aligned warning
7159 rather than isn't aligned. Use isn%'t rather than isn't.
7160 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
7161 into MULT_EXPR.
7162 <case MULT_EXPR>: Improve the case when bottom and one of the
7163 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
7164 operand, in that case check if the other operand is multiple of
7165 bottom divided by the INTEGER_CST operand.
7167 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7169 PR target/83858
7170 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
7171 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
7172 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
7173 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
7174 * config/pa/pa.c (pa_function_arg_advance): Likewise.
7175 (pa_function_arg, pa_arg_partial_bytes): Likewise.
7176 (pa_function_arg_size): New function.
7178 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7180 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
7181 in a separate statement.
7183 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
7185 PR tree-optimization/83847
7186 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
7187 group gathers and scatters.
7189 2018-01-16 Jakub Jelinek <jakub@redhat.com>
7191 PR rtl-optimization/86620
7192 * params.def (max-sched-ready-insns): Bump minimum value to 1.
7194 PR rtl-optimization/83213
7195 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
7196 to last if both are JUMP_INSNs.
7198 PR tree-optimization/83843
7199 * gimple-ssa-store-merging.c
7200 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
7201 store_immediate_info for bswap/nop orig_stores.
7203 2018-01-15 Andrew Waterman <andrew@sifive.com>
7205 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
7206 !TARGET_MUL.
7207 <UDIV>: Increase cost if !TARGET_DIV.
7209 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
7211 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
7212 (define_attr "cr_logical_3op"): New.
7213 (cceq_ior_compare): Adjust.
7214 (cceq_ior_compare_complement): Adjust.
7215 (*cceq_rev_compare): Adjust.
7216 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
7217 (is_cracked_insn): Adjust.
7218 (insn_must_be_first_in_group): Adjust.
7219 * config/rs6000/40x.md: Adjust.
7220 * config/rs6000/440.md: Adjust.
7221 * config/rs6000/476.md: Adjust.
7222 * config/rs6000/601.md: Adjust.
7223 * config/rs6000/603.md: Adjust.
7224 * config/rs6000/6xx.md: Adjust.
7225 * config/rs6000/7450.md: Adjust.
7226 * config/rs6000/7xx.md: Adjust.
7227 * config/rs6000/8540.md: Adjust.
7228 * config/rs6000/cell.md: Adjust.
7229 * config/rs6000/e300c2c3.md: Adjust.
7230 * config/rs6000/e500mc.md: Adjust.
7231 * config/rs6000/e500mc64.md: Adjust.
7232 * config/rs6000/e5500.md: Adjust.
7233 * config/rs6000/e6500.md: Adjust.
7234 * config/rs6000/mpc.md: Adjust.
7235 * config/rs6000/power4.md: Adjust.
7236 * config/rs6000/power5.md: Adjust.
7237 * config/rs6000/power6.md: Adjust.
7238 * config/rs6000/power7.md: Adjust.
7239 * config/rs6000/power8.md: Adjust.
7240 * config/rs6000/power9.md: Adjust.
7241 * config/rs6000/rs64.md: Adjust.
7242 * config/rs6000/titan.md: Adjust.
7244 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7246 * config/i386/predicates.md (indirect_branch_operand): Rewrite
7247 ix86_indirect_branch_register logic.
7249 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7251 * config/i386/constraints.md (Bs): Update
7252 ix86_indirect_branch_register check. Don't check
7253 ix86_indirect_branch_register with GOT_memory_operand.
7254 (Bw): Likewise.
7255 * config/i386/predicates.md (GOT_memory_operand): Don't check
7256 ix86_indirect_branch_register here.
7257 (GOT32_symbol_operand): Likewise.
7259 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7261 * config/i386/predicates.md (constant_call_address_operand):
7262 Rewrite ix86_indirect_branch_register logic.
7263 (sibcall_insn_operand): Likewise.
7265 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7267 * config/i386/constraints.md (Bs): Replace
7268 ix86_indirect_branch_thunk_register with
7269 ix86_indirect_branch_register.
7270 (Bw): Likewise.
7271 * config/i386/i386.md (indirect_jump): Likewise.
7272 (tablejump): Likewise.
7273 (*sibcall_memory): Likewise.
7274 (*sibcall_value_memory): Likewise.
7275 Peepholes of indirect call and jump via memory: Likewise.
7276 * config/i386/i386.opt: Likewise.
7277 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7278 (GOT_memory_operand): Likewise.
7279 (call_insn_operand): Likewise.
7280 (sibcall_insn_operand): Likewise.
7281 (GOT32_symbol_operand): Likewise.
7283 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7285 PR middle-end/83837
7286 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
7287 type rather than type addr's type points to.
7288 (expand_omp_atomic_mutex): Likewise.
7289 (expand_omp_atomic): Likewise.
7291 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
7293 PR target/83839
7294 * config/i386/i386.c (output_indirect_thunk_function): Use
7295 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
7296 for __x86_return_thunk.
7298 2018-01-15 Richard Biener <rguenther@suse.de>
7300 PR middle-end/83850
7301 * expmed.c (extract_bit_field_1): Fix typo.
7303 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7305 PR target/83687
7306 * config/arm/iterators.md (VF): New mode iterator.
7307 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
7308 Remove integer-related logic from pattern.
7309 (neon_vabd<mode>_3): Likewise.
7311 2018-01-15 Jakub Jelinek <jakub@redhat.com>
7313 PR middle-end/82694
7314 * common.opt (fstrict-overflow): No longer an alias.
7315 (fwrapv-pointer): New option.
7316 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
7317 also for pointer types based on flag_wrapv_pointer.
7318 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
7319 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
7320 opts->x_flag_wrapv got set.
7321 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
7322 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
7323 POINTER_TYPE_OVERFLOW_UNDEFINED.
7324 * match.pd: Likewise in address comparison pattern.
7325 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
7327 2018-01-15 Richard Biener <rguenther@suse.de>
7329 PR lto/83804
7330 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
7331 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
7332 Reset type names to their identifier if their TYPE_DECL doesn't
7333 have linkage (and thus is used for ODR and devirt).
7334 (save_debug_info_for_decl): Remove.
7335 (save_debug_info_for_type): Likewise.
7336 (add_tree_to_fld_list): Adjust.
7337 * tree-pretty-print.c (dump_generic_node): Make dumping of
7338 type names more robust.
7340 2018-01-15 Richard Biener <rguenther@suse.de>
7342 * BASE-VER: Bump to 8.0.1.
7344 2018-01-14 Martin Sebor <msebor@redhat.com>
7346 PR other/83508
7347 * builtins.c (check_access): Avoid warning when the no-warning bit
7348 is set.
7350 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
7352 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
7353 * ira-color (allocno_hard_regs_compare): Likewise.
7355 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
7357 PR target/83013
7358 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
7359 Use .pushsection/.popsection.
7361 2018-01-14 Martin Sebor <msebor@redhat.com>
7363 PR c++/81327
7364 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
7366 2018-01-14 Jakub Jelinek <jakub@redhat.com>
7368 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
7369 entry from extra_headers.
7370 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
7371 extra_headers, make the list bitwise identical to the i?86-*-* one.
7373 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7375 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
7376 -mcmodel=large with -mindirect-branch=thunk,
7377 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
7378 -mfunction-return=thunk-extern.
7379 * doc/invoke.texi: Document -mcmodel=large is incompatible with
7380 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
7381 -mfunction-return=thunk and -mfunction-return=thunk-extern.
7383 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7385 * config/i386/i386.c (print_reg): Print the name of the full
7386 integer register without '%'.
7387 (ix86_print_operand): Handle 'V'.
7388 * doc/extend.texi: Document 'V' modifier.
7390 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7392 * config/i386/constraints.md (Bs): Disallow memory operand for
7393 -mindirect-branch-register.
7394 (Bw): Likewise.
7395 * config/i386/predicates.md (indirect_branch_operand): Likewise.
7396 (GOT_memory_operand): Likewise.
7397 (call_insn_operand): Likewise.
7398 (sibcall_insn_operand): Likewise.
7399 (GOT32_symbol_operand): Likewise.
7400 * config/i386/i386.md (indirect_jump): Call convert_memory_address
7401 for -mindirect-branch-register.
7402 (tablejump): Likewise.
7403 (*sibcall_memory): Likewise.
7404 (*sibcall_value_memory): Likewise.
7405 Disallow peepholes of indirect call and jump via memory for
7406 -mindirect-branch-register.
7407 (*call_pop): Replace m with Bw.
7408 (*call_value_pop): Likewise.
7409 (*sibcall_pop_memory): Replace m with Bs.
7410 * config/i386/i386.opt (mindirect-branch-register): New option.
7411 * doc/invoke.texi: Document -mindirect-branch-register option.
7413 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7415 * config/i386/i386-protos.h (ix86_output_function_return): New.
7416 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
7417 set function_return_type.
7418 (indirect_thunk_name): Add ret_p to indicate thunk for function
7419 return.
7420 (output_indirect_thunk_function): Pass false to
7421 indirect_thunk_name.
7422 (ix86_output_indirect_branch_via_reg): Likewise.
7423 (ix86_output_indirect_branch_via_push): Likewise.
7424 (output_indirect_thunk_function): Create alias for function
7425 return thunk if regno < 0.
7426 (ix86_output_function_return): New function.
7427 (ix86_handle_fndecl_attribute): Handle function_return.
7428 (ix86_attribute_table): Add function_return.
7429 * config/i386/i386.h (machine_function): Add
7430 function_return_type.
7431 * config/i386/i386.md (simple_return_internal): Use
7432 ix86_output_function_return.
7433 (simple_return_internal_long): Likewise.
7434 * config/i386/i386.opt (mfunction-return=): New option.
7435 (indirect_branch): Mention -mfunction-return=.
7436 * doc/extend.texi: Document function_return function attribute.
7437 * doc/invoke.texi: Document -mfunction-return= option.
7439 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
7441 * config/i386/i386-opts.h (indirect_branch): New.
7442 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
7443 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
7444 with local indirect jump when converting indirect call and jump.
7445 (ix86_set_indirect_branch_type): New.
7446 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
7447 (indirectlabelno): New.
7448 (indirect_thunk_needed): Likewise.
7449 (indirect_thunk_bnd_needed): Likewise.
7450 (indirect_thunks_used): Likewise.
7451 (indirect_thunks_bnd_used): Likewise.
7452 (INDIRECT_LABEL): Likewise.
7453 (indirect_thunk_name): Likewise.
7454 (output_indirect_thunk): Likewise.
7455 (output_indirect_thunk_function): Likewise.
7456 (ix86_output_indirect_branch_via_reg): Likewise.
7457 (ix86_output_indirect_branch_via_push): Likewise.
7458 (ix86_output_indirect_branch): Likewise.
7459 (ix86_output_indirect_jmp): Likewise.
7460 (ix86_code_end): Call output_indirect_thunk_function if needed.
7461 (ix86_output_call_insn): Call ix86_output_indirect_branch if
7462 needed.
7463 (ix86_handle_fndecl_attribute): Handle indirect_branch.
7464 (ix86_attribute_table): Add indirect_branch.
7465 * config/i386/i386.h (machine_function): Add indirect_branch_type
7466 and has_local_indirect_jump.
7467 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
7468 to true.
7469 (tablejump): Likewise.
7470 (*indirect_jump): Use ix86_output_indirect_jmp.
7471 (*tablejump_1): Likewise.
7472 (simple_return_indirect_internal): Likewise.
7473 * config/i386/i386.opt (mindirect-branch=): New option.
7474 (indirect_branch): New.
7475 (keep): Likewise.
7476 (thunk): Likewise.
7477 (thunk-inline): Likewise.
7478 (thunk-extern): Likewise.
7479 * doc/extend.texi: Document indirect_branch function attribute.
7480 * doc/invoke.texi: Document -mindirect-branch= option.
7482 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
7484 PR ipa/83051
7485 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
7487 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
7489 * ipa-inline.c (want_inline_small_function_p): Return false if
7490 inlining has already failed with CIF_FINAL_ERROR.
7491 (update_caller_keys): Call want_inline_small_function_p before
7492 can_inline_edge_p.
7493 (update_callee_keys): Likewise.
7495 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
7497 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
7498 New function.
7499 (rs6000_quadword_masked_address_p): Likewise.
7500 (quad_aligned_load_p): Likewise.
7501 (quad_aligned_store_p): Likewise.
7502 (const_load_sequence_p): Add comment to describe the outer-most loop.
7503 (mimic_memory_attributes_and_flags): New function.
7504 (rs6000_gen_stvx): Likewise.
7505 (replace_swapped_aligned_store): Likewise.
7506 (rs6000_gen_lvx): Likewise.
7507 (replace_swapped_aligned_load): Likewise.
7508 (replace_swapped_load_constant): Capitalize argument name in
7509 comment describing this function.
7510 (rs6000_analyze_swaps): Add a third pass to search for vector loads
7511 and stores that access quad-word aligned addresses and replace
7512 with stvx or lvx instructions when appropriate.
7513 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
7514 New function prototype.
7515 (rs6000_quadword_masked_address_p): Likewise.
7516 (rs6000_gen_lvx): Likewise.
7517 (rs6000_gen_stvx): Likewise.
7518 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
7519 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
7520 when memory address is aligned.
7521 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
7522 this split to select lvx instruction when memory address is aligned.
7523 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
7524 instruction when memory address is aligned.
7525 (*vsx_le_perm_load_v16qi): Likewise.
7526 (four unnamed splitters): Modify to select the stvx instruction
7527 when memory is aligned.
7529 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
7531 * predict.c (determine_unlikely_bbs): Handle correctly BBs
7532 which appears in the queue multiple times.
7534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7535 Alan Hayward <alan.hayward@arm.com>
7536 David Sherwood <david.sherwood@arm.com>
7538 * tree-vectorizer.h (vec_lower_bound): New structure.
7539 (_loop_vec_info): Add check_nonzero and lower_bounds.
7540 (LOOP_VINFO_CHECK_NONZERO): New macro.
7541 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
7542 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
7543 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
7544 fields. Make seg_len the distance travelled, not including the
7545 access size.
7546 (dr_direction_indicator): Declare.
7547 (dr_zero_step_indicator): Likewise.
7548 (dr_known_forward_stride_p): Likewise.
7549 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
7550 tree-ssanames.h.
7551 (runtime_alias_check_p): Allow runtime alias checks with
7552 variable strides.
7553 (operator ==): Compare access_size and align.
7554 (prune_runtime_alias_test_list): Rework for new distinction between
7555 the access_size and seg_len.
7556 (create_intersect_range_checks_index): Likewise. Cope with polynomial
7557 segment lengths.
7558 (get_segment_min_max): New function.
7559 (create_intersect_range_checks): Use it.
7560 (dr_step_indicator): New function.
7561 (dr_direction_indicator): Likewise.
7562 (dr_zero_step_indicator): Likewise.
7563 (dr_known_forward_stride_p): Likewise.
7564 * tree-loop-distribution.c (data_ref_segment_size): Return
7565 DR_STEP * (niters - 1).
7566 (compute_alias_check_pairs): Update call to the dr_with_seg_len
7567 constructor.
7568 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
7569 (vect_preserves_scalar_order_p): New function, split out from...
7570 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
7571 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
7572 (vect_vfa_access_size): New function.
7573 (vect_vfa_align): Likewise.
7574 (vect_compile_time_alias): Take access_size_a and access_b arguments.
7575 (dump_lower_bound): New function.
7576 (vect_check_lower_bound): Likewise.
7577 (vect_small_gap_p): Likewise.
7578 (vectorizable_with_step_bound_p): Likewise.
7579 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
7580 depencies if the vectorization factor is 1. Convert the checks
7581 for nonzero steps into checks on the bounds of DR_STEP. Try using
7582 a bunds check for variable steps if the minimum required step is
7583 relatively small. Update calls to the dr_with_seg_len
7584 constructor and to vect_compile_time_alias.
7585 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
7586 function.
7587 (vect_loop_versioning): Call it.
7588 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
7589 when retrying.
7590 (vect_estimate_min_profitable_iters): Account for any bounds checks.
7592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * doc/sourcebuild.texi (vect_scatter_store): Document.
7597 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
7598 optabs.
7599 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
7600 Document.
7601 * genopinit.c (main): Add supports_vec_scatter_store and
7602 supports_vec_scatter_store_cached to target_optabs.
7603 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
7604 IFN_MASK_SCATTER_STORE.
7605 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
7606 functions.
7607 * internal-fn.h (internal_store_fn_p): Declare.
7608 (internal_fn_stored_value_index): Likewise.
7609 * internal-fn.c (scatter_store_direct): New macro.
7610 (expand_scatter_store_optab_fn): New function.
7611 (direct_scatter_store_optab_supported_p): New macro.
7612 (internal_store_fn_p): New function.
7613 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
7614 IFN_MASK_SCATTER_STORE.
7615 (internal_fn_mask_index): Likewise.
7616 (internal_fn_stored_value_index): New function.
7617 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
7618 for scatter stores.
7619 * optabs-query.h (supports_vec_scatter_store_p): Declare.
7620 * optabs-query.c (supports_vec_scatter_store_p): New function.
7621 * tree-vectorizer.h (vect_get_store_rhs): Declare.
7622 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
7623 true for scatter stores.
7624 (vect_gather_scatter_fn_p): Handle scatter stores too.
7625 (vect_check_gather_scatter): Consider using scatter stores if
7626 supports_vec_scatter_store_p.
7627 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
7628 scatter stores too.
7629 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7630 internal_fn_stored_value_index.
7631 (check_load_store_masking): Handle scatter stores too.
7632 (vect_get_store_rhs): Make public.
7633 (vectorizable_call): Use internal_store_fn_p.
7634 (vectorizable_store): Handle scatter store internal functions.
7635 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
7636 when deciding whether the end of the group has been reached.
7637 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
7638 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
7639 (mask_scatter_store<mode>): New insns.
7641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7642 Alan Hayward <alan.hayward@arm.com>
7643 David Sherwood <david.sherwood@arm.com>
7645 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
7646 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
7647 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
7648 function.
7649 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
7650 Use vect_truncate_gather_scatter_offset if we can't treat the
7651 operation as a normal gather load or scatter store.
7652 (get_group_load_store_type): Take the gather_scatter_info
7653 as argument. Try using a gather load or scatter store for
7654 single-element groups.
7655 (get_load_store_type): Update calls to get_group_load_store_type
7656 and vect_use_strided_gather_scatters_p.
7658 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
7663 optional tree argument.
7664 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
7665 null target hooks.
7666 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
7667 but continue to use the current value as a fallback.
7668 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
7669 to compare the updates.
7670 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
7671 (get_load_store_type): Use it when handling a strided access.
7672 (vect_get_strided_load_store_ops): New function.
7673 (vect_get_data_ptr_increment): Likewise.
7674 (vectorizable_load): Handle strided gather loads. Always pass
7675 a step to vect_create_data_ref_ptr and bump_vector_ptr.
7677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7678 Alan Hayward <alan.hayward@arm.com>
7679 David Sherwood <david.sherwood@arm.com>
7681 * doc/md.texi (gather_load@var{m}): Document.
7682 (mask_gather_load@var{m}): Likewise.
7683 * genopinit.c (main): Add supports_vec_gather_load and
7684 supports_vec_gather_load_cached to target_optabs.
7685 * optabs-tree.c (init_tree_optimization_optabs): Use
7686 ggc_cleared_alloc to allocate target_optabs.
7687 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
7688 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
7689 functions.
7690 * internal-fn.h (internal_load_fn_p): Declare.
7691 (internal_gather_scatter_fn_p): Likewise.
7692 (internal_fn_mask_index): Likewise.
7693 (internal_gather_scatter_fn_supported_p): Likewise.
7694 * internal-fn.c (gather_load_direct): New macro.
7695 (expand_gather_load_optab_fn): New function.
7696 (direct_gather_load_optab_supported_p): New macro.
7697 (direct_internal_fn_optab): New function.
7698 (internal_load_fn_p): Likewise.
7699 (internal_gather_scatter_fn_p): Likewise.
7700 (internal_fn_mask_index): Likewise.
7701 (internal_gather_scatter_fn_supported_p): Likewise.
7702 * optabs-query.c (supports_at_least_one_mode_p): New function.
7703 (supports_vec_gather_load_p): Likewise.
7704 * optabs-query.h (supports_vec_gather_load_p): Declare.
7705 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
7706 and memory_type field.
7707 (NUM_PATTERNS): Bump to 15.
7708 * tree-vect-data-refs.c: Include internal-fn.h.
7709 (vect_gather_scatter_fn_p): New function.
7710 (vect_describe_gather_scatter_call): Likewise.
7711 (vect_check_gather_scatter): Try using internal functions for
7712 gather loads. Recognize existing calls to a gather load function.
7713 (vect_analyze_data_refs): Consider using gather loads if
7714 supports_vec_gather_load_p.
7715 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
7716 (vect_get_gather_scatter_offset_type): Likewise.
7717 (vect_convert_mask_for_vectype): Likewise.
7718 (vect_add_conversion_to_patterm): Likewise.
7719 (vect_try_gather_scatter_pattern): Likewise.
7720 (vect_recog_gather_scatter_pattern): New pattern recognizer.
7721 (vect_vect_recog_func_ptrs): Add it.
7722 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
7723 internal_fn_mask_index and internal_gather_scatter_fn_p.
7724 (check_load_store_masking): Take the gather_scatter_info as an
7725 argument and handle gather loads.
7726 (vect_get_gather_scatter_ops): New function.
7727 (vectorizable_call): Check internal_load_fn_p.
7728 (vectorizable_load): Likewise. Handle gather load internal
7729 functions.
7730 (vectorizable_store): Update call to check_load_store_masking.
7731 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
7732 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
7733 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
7734 (aarch64_gather_scale_operand_d): New predicates.
7735 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
7736 (mask_gather_load<mode>): New insns.
7738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7739 Alan Hayward <alan.hayward@arm.com>
7740 David Sherwood <david.sherwood@arm.com>
7742 * optabs.def (fold_left_plus_optab): New optab.
7743 * doc/md.texi (fold_left_plus_@var{m}): Document.
7744 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
7745 * internal-fn.c (fold_left_direct): Define.
7746 (expand_fold_left_optab_fn): Likewise.
7747 (direct_fold_left_optab_supported_p): Likewise.
7748 * fold-const-call.c (fold_const_fold_left): New function.
7749 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
7750 * tree-parloops.c (valid_reduction_p): New function.
7751 (gather_scalar_reductions): Use it.
7752 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
7753 (vect_finish_replace_stmt): Declare.
7754 * tree-vect-loop.c (fold_left_reduction_fn): New function.
7755 (needs_fold_left_reduction_p): New function, split out from...
7756 (vect_is_simple_reduction): ...here. Accept reductions that
7757 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
7758 (vect_force_simple_reduction): Also store the reduction type in
7759 the assignment's STMT_VINFO_REDUC_TYPE.
7760 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
7761 (merge_with_identity): New function.
7762 (vect_expand_fold_left): Likewise.
7763 (vectorize_fold_left_reduction): Likewise.
7764 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
7765 scalar phi in place for it. Check for target support and reject
7766 cases that would reassociate the operation. Defer the transform
7767 phase to vectorize_fold_left_reduction.
7768 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
7769 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
7770 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
7772 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7774 * tree-if-conv.c (predicate_mem_writes): Remove redundant
7775 call to ifc_temp_var.
7777 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7778 Alan Hayward <alan.hayward@arm.com>
7779 David Sherwood <david.sherwood@arm.com>
7781 * target.def (legitimize_address_displacement): Take the original
7782 offset as a poly_int.
7783 * targhooks.h (default_legitimize_address_displacement): Update
7784 accordingly.
7785 * targhooks.c (default_legitimize_address_displacement): Likewise.
7786 * doc/tm.texi: Regenerate.
7787 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
7788 as an argument, moving assert of ad->disp == ad->disp_term to...
7789 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
7790 Try calling targetm.legitimize_address_displacement before expanding
7791 the address rather than afterwards, and adjust for the new interface.
7792 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
7793 Match the new hook interface. Handle SVE addresses.
7794 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
7795 new hook interface.
7797 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7799 * Makefile.in (OBJS): Add early-remat.o.
7800 * target.def (select_early_remat_modes): New hook.
7801 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
7802 * doc/tm.texi: Regenerate.
7803 * targhooks.h (default_select_early_remat_modes): Declare.
7804 * targhooks.c (default_select_early_remat_modes): New function.
7805 * timevar.def (TV_EARLY_REMAT): New timevar.
7806 * passes.def (pass_early_remat): New pass.
7807 * tree-pass.h (make_pass_early_remat): Declare.
7808 * early-remat.c: New file.
7809 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
7810 function.
7811 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
7813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7814 Alan Hayward <alan.hayward@arm.com>
7815 David Sherwood <david.sherwood@arm.com>
7817 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
7818 vfm1 with a bound_epilog parameter.
7819 (vect_do_peeling): Update calls accordingly, and move the prologue
7820 call earlier in the function. Treat the base bound_epilog as 0 for
7821 fully-masked loops and retain vf - 1 for other loops. Add 1 to
7822 this base when peeling for gaps.
7823 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
7824 with fully-masked loops.
7825 (vect_estimate_min_profitable_iters): Handle the single peeled
7826 iteration in that case.
7828 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7829 Alan Hayward <alan.hayward@arm.com>
7830 David Sherwood <david.sherwood@arm.com>
7832 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
7833 single-element interleaving even if the size is not a power of 2.
7834 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
7835 accesses for single-element interleaving if the group size is
7836 not a power of 2.
7838 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7839 Alan Hayward <alan.hayward@arm.com>
7840 David Sherwood <david.sherwood@arm.com>
7842 * doc/md.texi (fold_extract_last_@var{m}): Document.
7843 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
7844 * optabs.def (fold_extract_last_optab): New optab.
7845 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
7846 * internal-fn.c (fold_extract_direct): New macro.
7847 (expand_fold_extract_optab_fn): Likewise.
7848 (direct_fold_extract_optab_supported_p): Likewise.
7849 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
7850 * tree-vect-loop.c (vect_model_reduction_cost): Handle
7851 EXTRACT_LAST_REDUCTION.
7852 (get_initial_def_for_reduction): Do not create an initial vector
7853 for EXTRACT_LAST_REDUCTION reductions.
7854 (vectorizable_reduction): Leave the scalar phi in place for
7855 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7856 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7857 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7858 transform phase to vectorizable_condition.
7859 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7860 split out from...
7861 (vect_finish_stmt_generation): ...here.
7862 (vect_finish_replace_stmt): New function.
7863 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7864 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7865 pattern.
7866 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7868 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7869 Alan Hayward <alan.hayward@arm.com>
7870 David Sherwood <david.sherwood@arm.com>
7872 * doc/md.texi (extract_last_@var{m}): Document.
7873 * optabs.def (extract_last_optab): New optab.
7874 * internal-fn.def (EXTRACT_LAST): New internal function.
7875 * internal-fn.c (cond_unary_direct): New macro.
7876 (expand_cond_unary_optab_fn): Likewise.
7877 (direct_cond_unary_optab_supported_p): Likewise.
7878 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7879 loops using EXTRACT_LAST.
7880 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7881 (extract_last_<mode>): ...this optab.
7882 (vec_extract<mode><Vel>): Update accordingly.
7884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7885 Alan Hayward <alan.hayward@arm.com>
7886 David Sherwood <david.sherwood@arm.com>
7888 * target.def (empty_mask_is_expensive): New hook.
7889 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7890 * doc/tm.texi: Regenerate.
7891 * targhooks.h (default_empty_mask_is_expensive): Declare.
7892 * targhooks.c (default_empty_mask_is_expensive): New function.
7893 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7894 if the target says that empty masks are expensive.
7895 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7896 New function.
7897 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7900 Alan Hayward <alan.hayward@arm.com>
7901 David Sherwood <david.sherwood@arm.com>
7903 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7904 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7905 (vect_use_loop_mask_for_alignment_p): New function.
7906 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7907 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7908 niters_skip argument. Make sure that the first niters_skip elements
7909 of the first iteration are inactive.
7910 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7911 Update call to vect_set_loop_masks_directly.
7912 (get_misalign_in_elems): New function, split out from...
7913 (vect_gen_prolog_loop_niters): ...here.
7914 (vect_update_init_of_dr): Take a code argument that specifies whether
7915 the adjustment should be added or subtracted.
7916 (vect_update_init_of_drs): Likewise.
7917 (vect_prepare_for_masked_peels): New function.
7918 (vect_do_peeling): Skip prologue peeling if we're using a mask
7919 instead. Update call to vect_update_inits_of_drs.
7920 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7921 mask_skip_niters.
7922 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7923 alignment. Do not include the number of peeled iterations in
7924 the minimum threshold in that case.
7925 (vectorizable_induction): Adjust the start value down by
7926 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7927 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7928 Take the number of skipped iterations into account when calculating
7929 the loop bounds.
7930 * tree-vect-stmts.c (vect_gen_while_not): New function.
7932 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7933 Alan Hayward <alan.hayward@arm.com>
7934 David Sherwood <david.sherwood@arm.com>
7936 * doc/sourcebuild.texi (vect_fully_masked): Document.
7937 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7938 default value to 0.
7939 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7940 split out from...
7941 (vect_analyze_loop_2): ...here. Don't check the vectorization
7942 factor against the number of loop iterations if the loop is
7943 fully-masked.
7945 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7946 Alan Hayward <alan.hayward@arm.com>
7947 David Sherwood <david.sherwood@arm.com>
7949 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7950 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7951 (dump_groups): Update accordingly.
7952 (iv_use::mem_type): New member variable.
7953 (address_p): New function.
7954 (record_use): Add a mem_type argument and initialize the new
7955 mem_type field.
7956 (record_group_use): Add a mem_type argument. Use address_p.
7957 Remove obsolete null checks of base_object. Update call to record_use.
7958 (find_interesting_uses_op): Update call to record_group_use.
7959 (find_interesting_uses_cond): Likewise.
7960 (find_interesting_uses_address): Likewise.
7961 (get_mem_type_for_internal_fn): New function.
7962 (find_address_like_use): Likewise.
7963 (find_interesting_uses_stmt): Try find_address_like_use before
7964 calling find_interesting_uses_op.
7965 (addr_offset_valid_p): Use the iv mem_type field as the type
7966 of the addressed memory.
7967 (add_autoinc_candidates): Likewise.
7968 (get_address_cost): Likewise.
7969 (split_small_address_groups_p): Use address_p.
7970 (split_address_groups): Likewise.
7971 (add_iv_candidate_for_use): Likewise.
7972 (autoinc_possible_for_pair): Likewise.
7973 (rewrite_groups): Likewise.
7974 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7975 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7976 (get_alias_ptr_type_for_ptr_address): New function.
7977 (rewrite_use_address): Rewrite address uses in calls that were
7978 identified by find_address_like_use.
7980 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7981 Alan Hayward <alan.hayward@arm.com>
7982 David Sherwood <david.sherwood@arm.com>
7984 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7985 TARGET_MEM_REFs.
7986 * gimple-expr.h (is_gimple_addressable: Likewise.
7987 * gimple-expr.c (is_gimple_address): Likewise.
7988 * internal-fn.c (expand_call_mem_ref): New function.
7989 (expand_mask_load_optab_fn): Use it.
7990 (expand_mask_store_optab_fn): Likewise.
7992 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7993 Alan Hayward <alan.hayward@arm.com>
7994 David Sherwood <david.sherwood@arm.com>
7996 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7997 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7998 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7999 (cond_umax@var{mode}): Document.
8000 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
8001 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
8002 (cond_umin_optab, cond_umax_optab): New optabs.
8003 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
8004 (COND_IOR, COND_XOR): New internal functions.
8005 * internal-fn.h (get_conditional_internal_fn): Declare.
8006 * internal-fn.c (cond_binary_direct): New macro.
8007 (expand_cond_binary_optab_fn): Likewise.
8008 (direct_cond_binary_optab_supported_p): Likewise.
8009 (get_conditional_internal_fn): New function.
8010 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
8011 Cope with reduction statements that are vectorized as calls rather
8012 than assignments.
8013 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
8014 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
8015 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
8016 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
8017 (UNSPEC_COND_EOR): New unspecs.
8018 (optab): Add mappings for them.
8019 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
8020 (sve_int_op, sve_fp_op): New int attributes.
8022 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8023 Alan Hayward <alan.hayward@arm.com>
8024 David Sherwood <david.sherwood@arm.com>
8026 * optabs.def (while_ult_optab): New optab.
8027 * doc/md.texi (while_ult@var{m}@var{n}): Document.
8028 * internal-fn.def (WHILE_ULT): New internal function.
8029 * internal-fn.h (direct_internal_fn_supported_p): New override
8030 that takes two types as argument.
8031 * internal-fn.c (while_direct): New macro.
8032 (expand_while_optab_fn): New function.
8033 (convert_optab_supported_p): Likewise.
8034 (direct_while_optab_supported_p): New macro.
8035 * wide-int.h (wi::udiv_ceil): New function.
8036 * tree-vectorizer.h (rgroup_masks): New structure.
8037 (vec_loop_masks): New typedef.
8038 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
8039 and fully_masked_p.
8040 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
8041 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
8042 (vect_max_vf): New function.
8043 (slpeel_make_loop_iterate_ntimes): Delete.
8044 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
8045 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
8046 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
8047 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
8048 internal-fn.h, stor-layout.h and optabs-query.h.
8049 (vect_set_loop_mask): New function.
8050 (add_preheader_seq): Likewise.
8051 (add_header_seq): Likewise.
8052 (interleave_supported_p): Likewise.
8053 (vect_maybe_permute_loop_masks): Likewise.
8054 (vect_set_loop_masks_directly): Likewise.
8055 (vect_set_loop_condition_masked): Likewise.
8056 (vect_set_loop_condition_unmasked): New function, split out from
8057 slpeel_make_loop_iterate_ntimes.
8058 (slpeel_make_loop_iterate_ntimes): Rename to..
8059 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
8060 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
8061 (vect_do_peeling): Update call accordingly.
8062 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
8063 loops.
8064 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
8065 mask_compare_type, can_fully_mask_p and fully_masked_p.
8066 (release_vec_loop_masks): New function.
8067 (_loop_vec_info): Use it to free the loop masks.
8068 (can_produce_all_loop_masks_p): New function.
8069 (vect_get_max_nscalars_per_iter): Likewise.
8070 (vect_verify_full_masking): Likewise.
8071 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
8072 retries, and free the mask rgroups before retrying. Check loop-wide
8073 reasons for disallowing fully-masked loops. Make the final decision
8074 about whether use a fully-masked loop or not.
8075 (vect_estimate_min_profitable_iters): Do not assume that peeling
8076 for the number of iterations will be needed for fully-masked loops.
8077 (vectorizable_reduction): Disable fully-masked loops.
8078 (vectorizable_live_operation): Likewise.
8079 (vect_halve_mask_nunits): New function.
8080 (vect_double_mask_nunits): Likewise.
8081 (vect_record_loop_mask): Likewise.
8082 (vect_get_loop_mask): Likewise.
8083 (vect_transform_loop): Handle the case in which the final loop
8084 iteration might handle a partial vector. Call vect_set_loop_condition
8085 instead of slpeel_make_loop_iterate_ntimes.
8086 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
8087 (check_load_store_masking): New function.
8088 (prepare_load_store_mask): Likewise.
8089 (vectorizable_store): Handle fully-masked loops.
8090 (vectorizable_load): Likewise.
8091 (supportable_widening_operation): Use vect_halve_mask_nunits for
8092 booleans.
8093 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
8094 (vect_gen_while): New function.
8095 * config/aarch64/aarch64.md (umax<mode>3): New expander.
8096 (aarch64_uqdec<mode>): New insn.
8098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8099 Alan Hayward <alan.hayward@arm.com>
8100 David Sherwood <david.sherwood@arm.com>
8102 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
8103 (reduc_xor_scal_optab): New optabs.
8104 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
8105 (reduc_xor_scal_@var{m}): Document.
8106 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
8107 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
8108 internal functions.
8109 * fold-const-call.c (fold_const_call): Handle them.
8110 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
8111 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
8112 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
8113 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
8114 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
8115 (UNSPEC_XORV): New unspecs.
8116 (optab): Add entries for them.
8117 (BITWISEV): New int iterator.
8118 (bit_reduc_op): New int attributes.
8120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8121 Alan Hayward <alan.hayward@arm.com>
8122 David Sherwood <david.sherwood@arm.com>
8124 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
8125 * internal-fn.def (VEC_SHL_INSERT): New internal function.
8126 * optabs.def (vec_shl_insert_optab): New optab.
8127 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
8128 (duplicate_and_interleave): Likewise.
8129 * tree-vect-loop.c: Include internal-fn.h.
8130 (neutral_op_for_slp_reduction): New function, split out from
8131 get_initial_defs_for_reduction.
8132 (get_initial_def_for_reduction): Handle option 2 for variable-length
8133 vectors by loading the neutral value into a vector and then shifting
8134 the initial value into element 0.
8135 (get_initial_defs_for_reduction): Replace the code argument with
8136 the neutral value calculated by neutral_op_for_slp_reduction.
8137 Use gimple_build_vector for constant-length vectors.
8138 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
8139 but the first group_size elements have a neutral value.
8140 Use duplicate_and_interleave otherwise.
8141 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
8142 Update call to get_initial_defs_for_reduction. Handle SLP
8143 reductions for variable-length vectors by creating one vector
8144 result for each scalar result, with the elements associated
8145 with other scalar results stubbed out with the neutral value.
8146 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
8147 Require IFN_VEC_SHL_INSERT for double reductions on
8148 variable-length vectors, or SLP reductions that have
8149 a neutral value. Require can_duplicate_and_interleave_p
8150 support for variable-length unchained SLP reductions if there
8151 is no neutral value, such as for MIN/MAX reductions. Also require
8152 the number of vector elements to be a multiple of the number of
8153 SLP statements when doing variable-length unchained SLP reductions.
8154 Update call to vect_create_epilog_for_reduction.
8155 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
8156 and remove initial values.
8157 (duplicate_and_interleave): Make public.
8158 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
8159 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
8161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8162 Alan Hayward <alan.hayward@arm.com>
8163 David Sherwood <david.sherwood@arm.com>
8165 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
8166 (can_duplicate_and_interleave_p): New function.
8167 (vect_get_and_check_slp_defs): Take the vector of statements
8168 rather than just the current one. Remove excess parentheses.
8169 Restriction rejectinon of vect_constant_def and vect_external_def
8170 for variable-length vectors to boolean types, or types for which
8171 can_duplicate_and_interleave_p is false.
8172 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
8173 (duplicate_and_interleave): New function.
8174 (vect_get_constant_vectors): Use gimple_build_vector for
8175 constant-length vectors and suitable variable-length constant
8176 vectors. Use duplicate_and_interleave for other variable-length
8177 vectors. Don't defer the update when inserting new statements.
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 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
8184 min_profitable_iters doesn't go negative.
8186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8187 Alan Hayward <alan.hayward@arm.com>
8188 David Sherwood <david.sherwood@arm.com>
8190 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
8191 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
8192 * optabs.def (vec_mask_load_lanes_optab): New optab.
8193 (vec_mask_store_lanes_optab): Likewise.
8194 * internal-fn.def (MASK_LOAD_LANES): New internal function.
8195 (MASK_STORE_LANES): Likewise.
8196 * internal-fn.c (mask_load_lanes_direct): New macro.
8197 (mask_store_lanes_direct): Likewise.
8198 (expand_mask_load_optab_fn): Handle masked operations.
8199 (expand_mask_load_lanes_optab_fn): New macro.
8200 (expand_mask_store_optab_fn): Handle masked operations.
8201 (expand_mask_store_lanes_optab_fn): New macro.
8202 (direct_mask_load_lanes_optab_supported_p): Likewise.
8203 (direct_mask_store_lanes_optab_supported_p): Likewise.
8204 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
8205 parameter.
8206 (vect_load_lanes_supported): Likewise.
8207 * tree-vect-data-refs.c (strip_conversion): New function.
8208 (can_group_stmts_p): Likewise.
8209 (vect_analyze_data_ref_accesses): Use it instead of checking
8210 for a pair of assignments.
8211 (vect_store_lanes_supported): Take a masked_p parameter.
8212 (vect_load_lanes_supported): Likewise.
8213 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
8214 vect_store_lanes_supported and vect_load_lanes_supported.
8215 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
8216 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
8217 parameter. Don't allow gaps for masked accesses.
8218 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
8219 and vect_load_lanes_supported.
8220 (get_load_store_type): Take a masked_p parameter and update
8221 call to get_group_load_store_type.
8222 (vectorizable_store): Update call to get_load_store_type.
8223 Handle IFN_MASK_STORE_LANES.
8224 (vectorizable_load): Update call to get_load_store_type.
8225 Handle IFN_MASK_LOAD_LANES.
8227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8228 Alan Hayward <alan.hayward@arm.com>
8229 David Sherwood <david.sherwood@arm.com>
8231 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
8232 modes for SVE.
8233 * config/aarch64/aarch64-protos.h
8234 (aarch64_sve_struct_memory_operand_p): Declare.
8235 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
8236 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
8237 (VPRED, vpred): Handle SVE structure modes.
8238 * config/aarch64/constraints.md (Utx): New constraint.
8239 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
8240 (aarch64_sve_struct_nonimmediate_operand): New predicates.
8241 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
8242 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
8243 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
8244 structure modes. Split into pieces after RA.
8245 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
8246 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
8247 New patterns.
8248 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
8249 SVE structure modes.
8250 (aarch64_classify_address): Likewise.
8251 (sizetochar): Move earlier in file.
8252 (aarch64_print_operand): Handle SVE register lists.
8253 (aarch64_array_mode): New function.
8254 (aarch64_sve_struct_memory_operand_p): Likewise.
8255 (TARGET_ARRAY_MODE): Redefine.
8257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8258 Alan Hayward <alan.hayward@arm.com>
8259 David Sherwood <david.sherwood@arm.com>
8261 * target.def (array_mode): New target hook.
8262 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
8263 * doc/tm.texi: Regenerate.
8264 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
8265 * hooks.c (hook_optmode_mode_uhwi_none): New function.
8266 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
8267 targetm.array_mode.
8268 * stor-layout.c (mode_for_array): Likewise. Support polynomial
8269 type sizes.
8271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8272 Alan Hayward <alan.hayward@arm.com>
8273 David Sherwood <david.sherwood@arm.com>
8275 * fold-const.c (fold_binary_loc): Check the argument types
8276 rather than the result type when testing for a vector operation.
8278 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8280 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
8281 * doc/tm.texi: Regenerate.
8283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8284 Alan Hayward <alan.hayward@arm.com>
8285 David Sherwood <david.sherwood@arm.com>
8287 * doc/invoke.texi (-msve-vector-bits=): Document new option.
8288 (sve): Document new AArch64 extension.
8289 * doc/md.texi (w): Extend the description of the AArch64
8290 constraint to include SVE vectors.
8291 (Upl, Upa): Document new AArch64 predicate constraints.
8292 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
8293 enum.
8294 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
8295 (msve-vector-bits=): New option.
8296 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
8297 SVE when these are disabled.
8298 (sve): New extension.
8299 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
8300 modes. Adjust their number of units based on aarch64_sve_vg.
8301 (MAX_BITSIZE_MODE_ANY_MODE): Define.
8302 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
8303 aarch64_addr_query_type.
8304 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
8305 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
8306 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
8307 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
8308 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
8309 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
8310 (aarch64_simd_imm_zero_p): Delete.
8311 (aarch64_check_zero_based_sve_index_immediate): Declare.
8312 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8313 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8314 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8315 (aarch64_sve_float_mul_immediate_p): Likewise.
8316 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8317 rather than an rtx.
8318 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
8319 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
8320 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
8321 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
8322 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
8323 (aarch64_regmode_natural_size): Likewise.
8324 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
8325 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
8326 left one place.
8327 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
8328 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
8329 for VG and the SVE predicate registers.
8330 (V_ALIASES): Add a "z"-prefixed alias.
8331 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
8332 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
8333 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
8334 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
8335 (REG_CLASS_NAMES): Add entries for them.
8336 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
8337 and the predicate registers.
8338 (aarch64_sve_vg): Declare.
8339 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
8340 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
8341 (REGMODE_NATURAL_SIZE): Define.
8342 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
8343 SVE macros.
8344 * config/aarch64/aarch64.c: Include cfgrtl.h.
8345 (simd_immediate_info): Add a constructor for series vectors,
8346 and an associated step field.
8347 (aarch64_sve_vg): New variable.
8348 (aarch64_dbx_register_number): Handle VG and the predicate registers.
8349 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
8350 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
8351 (VEC_ANY_DATA, VEC_STRUCT): New constants.
8352 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
8353 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
8354 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
8355 (aarch64_get_mask_mode): New functions.
8356 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
8357 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
8358 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
8359 predicate modes and predicate registers. Explicitly restrict
8360 GPRs to modes of 16 bytes or smaller. Only allow FP registers
8361 to store a vector mode if it is recognized by
8362 aarch64_classify_vector_mode.
8363 (aarch64_regmode_natural_size): New function.
8364 (aarch64_hard_regno_caller_save_mode): Return the original mode
8365 for predicates.
8366 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
8367 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
8368 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
8369 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
8370 functions.
8371 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
8372 does not overlap dest if the function is frame-related. Handle
8373 SVE constants.
8374 (aarch64_split_add_offset): New function.
8375 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
8376 them aarch64_add_offset.
8377 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
8378 and update call to aarch64_sub_sp.
8379 (aarch64_add_cfa_expression): New function.
8380 (aarch64_expand_prologue): Pass extra temporary registers to the
8381 functions above. Handle the case in which we need to emit new
8382 DW_CFA_expressions for registers that were originally saved
8383 relative to the stack pointer, but now have to be expressed
8384 relative to the frame pointer.
8385 (aarch64_output_mi_thunk): Pass extra temporary registers to the
8386 functions above.
8387 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
8388 IP0 and IP1 values for SVE frames.
8389 (aarch64_expand_vec_series): New function.
8390 (aarch64_expand_sve_widened_duplicate): Likewise.
8391 (aarch64_expand_sve_const_vector): Likewise.
8392 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
8393 Handle SVE constants. Use emit_move_insn to move a force_const_mem
8394 into the register, rather than emitting a SET directly.
8395 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
8396 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
8397 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
8398 (offset_9bit_signed_scaled_p): New functions.
8399 (aarch64_replicate_bitmask_imm): New function.
8400 (aarch64_bitmask_imm): Use it.
8401 (aarch64_cannot_force_const_mem): Reject expressions involving
8402 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
8403 (aarch64_classify_index): Handle SVE indices, by requiring
8404 a plain register index with a scale that matches the element size.
8405 (aarch64_classify_address): Handle SVE addresses. Assert that
8406 the mode of the address is VOIDmode or an integer mode.
8407 Update call to aarch64_classify_symbol.
8408 (aarch64_classify_symbolic_expression): Update call to
8409 aarch64_classify_symbol.
8410 (aarch64_const_vec_all_in_range_p): New function.
8411 (aarch64_print_vector_float_operand): Likewise.
8412 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
8413 "vN" for FP registers with SVE modes. Handle (const ...) vectors
8414 and the FP immediates 1.0 and 0.5.
8415 (aarch64_print_address_internal): Handle SVE addresses.
8416 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
8417 (aarch64_regno_regclass): Handle predicate registers.
8418 (aarch64_secondary_reload): Handle big-endian reloads of SVE
8419 data modes.
8420 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
8421 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
8422 (aarch64_convert_sve_vector_bits): New function.
8423 (aarch64_override_options): Use it to handle -msve-vector-bits=.
8424 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
8425 rather than an rtx.
8426 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
8427 Handle SVE vector and predicate modes. Accept VL-based constants
8428 that need only one temporary register, and VL offsets that require
8429 no temporary registers.
8430 (aarch64_conditional_register_usage): Mark the predicate registers
8431 as fixed if SVE isn't available.
8432 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
8433 Return true for SVE vector and predicate modes.
8434 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
8435 rather than an unsigned int. Handle SVE modes.
8436 (aarch64_preferred_simd_mode): Update call accordingly. Handle
8437 SVE modes.
8438 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
8439 if SVE is enabled.
8440 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
8441 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
8442 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
8443 (aarch64_sve_float_mul_immediate_p): New functions.
8444 (aarch64_sve_valid_immediate): New function.
8445 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
8446 Explicitly reject structure modes. Check for INDEX constants.
8447 Handle PTRUE and PFALSE constants.
8448 (aarch64_check_zero_based_sve_index_immediate): New function.
8449 (aarch64_simd_imm_zero_p): Delete.
8450 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
8451 vector modes. Accept constants in the range of CNT[BHWD].
8452 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
8453 ask for an Advanced SIMD mode.
8454 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
8455 (aarch64_simd_vector_alignment): Handle SVE predicates.
8456 (aarch64_vectorize_preferred_vector_alignment): New function.
8457 (aarch64_simd_vector_alignment_reachable): Use it instead of
8458 the vector size.
8459 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
8460 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
8461 functions.
8462 (MAX_VECT_LEN): Delete.
8463 (expand_vec_perm_d): Add a vec_flags field.
8464 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
8465 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
8466 (aarch64_evpc_ext): Don't apply a big-endian lane correction
8467 for SVE modes.
8468 (aarch64_evpc_rev): Rename to...
8469 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
8470 (aarch64_evpc_rev_global): New function.
8471 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
8472 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
8473 MAX_VECT_LEN.
8474 (aarch64_evpc_sve_tbl): New function.
8475 (aarch64_expand_vec_perm_const_1): Update after rename of
8476 aarch64_evpc_rev. Handle SVE permutes too, trying
8477 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
8478 than aarch64_evpc_tbl.
8479 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
8480 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
8481 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
8482 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
8483 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
8484 (aarch64_expand_sve_vcond): New functions.
8485 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
8486 of aarch64_vector_mode_p.
8487 (aarch64_dwarf_poly_indeterminate_value): New function.
8488 (aarch64_compute_pressure_classes): Likewise.
8489 (aarch64_can_change_mode_class): Likewise.
8490 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
8491 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
8492 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
8493 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
8494 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
8495 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
8496 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
8497 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
8498 constraints.
8499 (Dn, Dl, Dr): Accept const as well as const_vector.
8500 (Dz): Likewise. Compare against CONST0_RTX.
8501 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
8502 of "vector" where appropriate.
8503 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
8504 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
8505 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
8506 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
8507 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
8508 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
8509 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
8510 (v_int_equiv): Extend to SVE modes.
8511 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
8512 mode attributes.
8513 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
8514 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
8515 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
8516 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
8517 (SVE_COND_FP_CMP): New int iterators.
8518 (perm_hilo): Handle the new unpack unspecs.
8519 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
8520 attributes.
8521 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
8522 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
8523 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
8524 (aarch64_equality_operator, aarch64_constant_vector_operand)
8525 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
8526 (aarch64_sve_nonimmediate_operand): Likewise.
8527 (aarch64_sve_general_operand): Likewise.
8528 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
8529 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
8530 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
8531 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
8532 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
8533 (aarch64_sve_float_arith_immediate): Likewise.
8534 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
8535 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
8536 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
8537 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
8538 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
8539 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
8540 (aarch64_sve_float_arith_operand): Likewise.
8541 (aarch64_sve_float_arith_with_sub_operand): Likewise.
8542 (aarch64_sve_float_mul_operand): Likewise.
8543 (aarch64_sve_vec_perm_operand): Likewise.
8544 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
8545 (aarch64_mov_operand): Accept const_poly_int and const_vector.
8546 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
8547 as well as const_vector.
8548 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
8549 in file. Use CONST0_RTX and CONSTM1_RTX.
8550 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
8551 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
8552 Use aarch64_simd_imm_zero.
8553 * config/aarch64/aarch64-sve.md: New file.
8554 * config/aarch64/aarch64.md: Include it.
8555 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
8556 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
8557 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
8558 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
8559 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
8560 (sve): New attribute.
8561 (enabled): Disable instructions with the sve attribute unless
8562 TARGET_SVE.
8563 (movqi, movhi): Pass CONST_POLY_INT operaneds through
8564 aarch64_expand_mov_immediate.
8565 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
8566 CNT[BHSD] immediates.
8567 (movti): Split CONST_POLY_INT moves into two halves.
8568 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
8569 Split additions that need a temporary here if the destination
8570 is the stack pointer.
8571 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
8572 (*add<mode>3_poly_1): New instruction.
8573 (set_clobber_cc): New expander.
8575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8577 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
8578 parameter and use it instead of GET_MODE_SIZE (innermode). Use
8579 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
8580 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
8581 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
8582 Change innermode from fixed_mode_size to machine_mode.
8583 (simplify_subreg): Update call accordingly. Handle a constant-sized
8584 subreg of a variable-length CONST_VECTOR.
8586 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
8587 Alan Hayward <alan.hayward@arm.com>
8588 David Sherwood <david.sherwood@arm.com>
8590 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
8591 (add_offset_to_base): New function, split out from...
8592 (create_mem_ref): ...here. When handling a scale other than 1,
8593 check first whether the address is valid without the offset.
8594 Add it into the base if so, leaving the index and scale as-is.
8596 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8598 PR c++/83778
8599 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
8600 fold_for_warn before checking if arg2 is INTEGER_CST.
8602 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
8604 * config/rs6000/predicates.md (load_multiple_operation): Delete.
8605 (store_multiple_operation): Delete.
8606 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
8607 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
8608 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
8609 guarded by TARGET_STRING.
8610 (rs6000_output_load_multiple): Delete.
8611 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
8612 OPTION_MASK_STRING / TARGET_STRING handling.
8613 (print_operand) <'N', 'O'>: Add comment that these are unused now.
8614 (const rs6000_opt_masks) <"string">: Change mask to 0.
8615 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
8616 (MASK_STRING): Delete.
8617 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
8618 parts. Simplify.
8619 (load_multiple): Delete.
8620 (*ldmsi8): Delete.
8621 (*ldmsi7): Delete.
8622 (*ldmsi6): Delete.
8623 (*ldmsi5): Delete.
8624 (*ldmsi4): Delete.
8625 (*ldmsi3): Delete.
8626 (store_multiple): Delete.
8627 (*stmsi8): Delete.
8628 (*stmsi7): Delete.
8629 (*stmsi6): Delete.
8630 (*stmsi5): Delete.
8631 (*stmsi4): Delete.
8632 (*stmsi3): Delete.
8633 (movmemsi_8reg): Delete.
8634 (corresponding unnamed define_insn): Delete.
8635 (movmemsi_6reg): Delete.
8636 (corresponding unnamed define_insn): Delete.
8637 (movmemsi_4reg): Delete.
8638 (corresponding unnamed define_insn): Delete.
8639 (movmemsi_2reg): Delete.
8640 (corresponding unnamed define_insn): Delete.
8641 (movmemsi_1reg): Delete.
8642 (corresponding unnamed define_insn): Delete.
8643 * config/rs6000/rs6000.opt (mno-string): New.
8644 (mstring): Replace by deprecation warning stub.
8645 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
8647 2018-01-12 Jakub Jelinek <jakub@redhat.com>
8649 * regrename.c (regrename_do_replace): If replacing the same
8650 reg multiple times, try to reuse last created gen_raw_REG.
8652 PR debug/81155
8653 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
8654 main to workaround a bug in GDB.
8656 2018-01-12 Tom de Vries <tom@codesourcery.com>
8658 PR target/83737
8659 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
8661 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
8663 PR rtl-optimization/80481
8664 * ira-color.c (get_cap_member): New function.
8665 (allocnos_conflict_by_live_ranges_p): Use it.
8666 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
8667 (setup_slot_coalesced_allocno_live_ranges): Ditto.
8669 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
8671 PR target/83628
8672 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
8673 (*saddl_se_1): Ditto.
8674 (*ssubsi_1): Ditto.
8675 (*ssubl_se_1): Ditto.
8677 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8679 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
8680 rather than wi::to_widest for DR_INITs.
8681 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
8682 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
8683 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
8684 INTEGER_CSTs.
8685 (vect_analyze_group_access_1): Note that here.
8687 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8689 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
8690 polynomial type sizes.
8692 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
8694 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
8695 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
8696 (gimple_add_tmp_var): Likewise.
8698 2018-01-12 Martin Liska <mliska@suse.cz>
8700 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
8701 (gimple_alloc_sizes): Likewise.
8702 (dump_gimple_statistics): Use PRIu64 in printf format.
8703 * gimple.h: Change uint64_t to int.
8705 2018-01-12 Martin Liska <mliska@suse.cz>
8707 * tree-core.h: Use uint64_t instead of int.
8708 * tree.c (tree_node_counts): Likewise.
8709 (tree_node_sizes): Likewise.
8710 (dump_tree_statistics): Use PRIu64 in printf format.
8712 2018-01-12 Martin Liska <mliska@suse.cz>
8714 * Makefile.in: As qsort_chk is implemented in vec.c, add
8715 vec.o to linkage of gencfn-macros.
8716 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
8717 passing the info to record_node_allocation_statistics.
8718 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
8719 and pass the info.
8720 * ggc-common.c (struct ggc_usage): Add operator== and use
8721 it in operator< and compare function.
8722 * mem-stats.h (struct mem_usage): Likewise.
8723 * vec.c (struct vec_usage): Remove operator< and compare
8724 function. Can be simply inherited.
8726 2018-01-12 Martin Jambor <mjambor@suse.cz>
8728 PR target/81616
8729 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
8730 * tree-ssa-math-opts.c: Include domwalk.h.
8731 (convert_mult_to_fma_1): New function.
8732 (fma_transformation_info): New type.
8733 (fma_deferring_state): Likewise.
8734 (cancel_fma_deferring): New function.
8735 (result_of_phi): Likewise.
8736 (last_fma_candidate_feeds_initial_phi): Likewise.
8737 (convert_mult_to_fma): Added deferring logic, split actual
8738 transformation to convert_mult_to_fma_1.
8739 (math_opts_dom_walker): New type.
8740 (math_opts_dom_walker::after_dom_children): New method, body moved
8741 here from pass_optimize_widening_mul::execute, added deferring logic
8742 bits.
8743 (pass_optimize_widening_mul::execute): Moved most of code to
8744 math_opts_dom_walker::after_dom_children.
8745 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
8746 * config/i386/i386.c (ix86_option_override_internal): Added
8747 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
8749 2018-01-12 Richard Biener <rguenther@suse.de>
8751 PR debug/83157
8752 * dwarf2out.c (gen_variable_die): Do not reset old_die for
8753 inline instance vars.
8755 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
8757 PR target/81819
8758 * config/rx/rx.c (rx_is_restricted_memory_address):
8759 Handle SUBREG case.
8761 2018-01-12 Richard Biener <rguenther@suse.de>
8763 PR tree-optimization/80846
8764 * target.def (split_reduction): New target hook.
8765 * targhooks.c (default_split_reduction): New function.
8766 * targhooks.h (default_split_reduction): Declare.
8767 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
8768 target requests first reduce vectors by combining low and high
8769 parts.
8770 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
8771 (get_vectype_for_scalar_type_and_size): Export.
8772 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
8773 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
8774 * doc/tm.texi: Regenerate.
8775 * config/i386/i386.c (ix86_split_reduction): Implement
8776 TARGET_VECTORIZE_SPLIT_REDUCTION.
8778 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8780 PR target/83368
8781 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
8782 in PIC mode except for TARGET_VXWORKS_RTP.
8783 * config/sparc/sparc.c: Include cfgrtl.h.
8784 (TARGET_INIT_PIC_REG): Define.
8785 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
8786 (sparc_pic_register_p): New predicate.
8787 (sparc_legitimate_address_p): Use it.
8788 (sparc_legitimize_pic_address): Likewise.
8789 (sparc_delegitimize_address): Likewise.
8790 (sparc_mode_dependent_address_p): Likewise.
8791 (gen_load_pcrel_sym): Remove 4th parameter.
8792 (load_got_register): Adjust call to above. Remove obsolete stuff.
8793 (sparc_expand_prologue): Do not call load_got_register here.
8794 (sparc_flat_expand_prologue): Likewise.
8795 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
8796 (sparc_use_pseudo_pic_reg): New function.
8797 (sparc_init_pic_reg): Likewise.
8798 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
8799 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
8801 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
8803 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
8804 Add item for branch_cost.
8806 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
8808 PR rtl-optimization/83565
8809 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
8810 not extend the result to a larger mode for rotate operations.
8811 (num_sign_bit_copies1): Likewise.
8813 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8815 PR target/40411
8816 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
8817 -symbolic.
8818 Use values-Xc.o for -pedantic.
8819 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
8821 2018-01-12 Martin Liska <mliska@suse.cz>
8823 PR ipa/83054
8824 * ipa-devirt.c (final_warning_record::grow_type_warnings):
8825 New function.
8826 (possible_polymorphic_call_targets): Use it.
8827 (ipa_devirt): Likewise.
8829 2018-01-12 Martin Liska <mliska@suse.cz>
8831 * profile-count.h (enum profile_quality): Use 0 as invalid
8832 enum value of profile_quality.
8834 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
8836 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
8837 -mext-string options.
8839 2018-01-12 Richard Biener <rguenther@suse.de>
8841 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
8842 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
8843 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
8844 Likewise.
8845 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
8847 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
8849 * configure.ac (--with-long-double-format): Add support for the
8850 configuration option to change the default long double format on
8851 PowerPC systems.
8852 * config.gcc (powerpc*-linux*-*): Likewise.
8853 * configure: Regenerate.
8854 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8855 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8856 used without modification.
8858 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8860 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8861 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8862 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8863 MISC_BUILTIN_SPEC_BARRIER.
8864 (rs6000_init_builtins): Likewise.
8865 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8866 enum value.
8867 (speculation_barrier): New define_insn.
8868 * doc/extend.texi: Document __builtin_speculation_barrier.
8870 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8872 PR target/83203
8873 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8874 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8875 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8876 iterators.
8877 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8878 integral modes instead of "ss" and "sd".
8879 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8880 vectors with 32-bit and 64-bit elements.
8881 (vecdupssescalarmodesuffix): New mode attribute.
8882 (vec_dup<mode>): Use it.
8884 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8886 PR target/83330
8887 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8888 frame if argument is passed on stack.
8890 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8892 PR target/82682
8893 * ree.c (combine_reaching_defs): Optimize also
8894 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8895 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8897 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8899 PR middle-end/83189
8900 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8902 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8904 PR middle-end/83718
8905 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8906 after they are computed.
8908 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8910 PR tree-optimization/83695
8911 * gimple-loop-linterchange.cc
8912 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8913 reset cached scev information after interchange.
8914 (pass_linterchange::execute): Remove call to scev_reset_htab.
8916 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8918 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8919 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8920 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8921 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8922 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8923 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8924 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8925 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8926 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8927 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8928 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8929 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8930 (V_lane_reg): Likewise.
8931 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8932 New define_expand.
8933 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8934 (vfmal_lane_low<mode>_intrinsic,
8935 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8936 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8937 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8938 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8939 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8940 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8942 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8944 * config/arm/arm-cpus.in (fp16fml): New feature.
8945 (ALL_SIMD): Add fp16fml.
8946 (armv8.2-a): Add fp16fml as an option.
8947 (armv8.3-a): Likewise.
8948 (armv8.4-a): Add fp16fml as part of fp16.
8949 * config/arm/arm.h (TARGET_FP16FML): Define.
8950 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8951 when appropriate.
8952 * config/arm/arm-modes.def (V2HF): Define.
8953 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8954 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8955 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8956 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8957 vfmsl_low, vfmsl_high): New set of builtins.
8958 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8959 (vfml_op): New code attribute.
8960 (VFMLHALVES): New int iterator.
8961 (VFML, VFMLSEL): New mode attributes.
8962 (V_reg): Define mapping for V2HF.
8963 (V_hi, V_lo): New mode attributes.
8964 (VF_constraint): Likewise.
8965 (vfml_half, vfml_half_selector): New int attributes.
8966 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8967 define_expand.
8968 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8969 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8970 New define_insn.
8971 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8972 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8973 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8974 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8975 documentation.
8976 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8977 Document new effective target and option set.
8979 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8981 * config/arm/arm-cpus.in (armv8_4): New feature.
8982 (ARMv8_4a): New fgroup.
8983 (armv8.4-a): New arch.
8984 * config/arm/arm-tables.opt: Regenerate.
8985 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8986 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8987 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8988 Add matching rules for -march=armv8.4-a and extensions.
8989 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8991 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8993 PR target/81821
8994 * config/rx/rx.md (BW): New mode attribute.
8995 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8997 2018-01-11 Richard Biener <rguenther@suse.de>
8999 PR tree-optimization/83435
9000 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
9001 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
9002 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
9004 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9005 Alan Hayward <alan.hayward@arm.com>
9006 David Sherwood <david.sherwood@arm.com>
9008 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
9009 field.
9010 (aarch64_classify_address): Initialize it. Track polynomial offsets.
9011 (aarch64_print_address_internal): Use it to check for a zero offset.
9013 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9014 Alan Hayward <alan.hayward@arm.com>
9015 David Sherwood <david.sherwood@arm.com>
9017 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
9018 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
9019 Return a poly_int64 rather than a HOST_WIDE_INT.
9020 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
9021 rather than a HOST_WIDE_INT.
9022 * config/aarch64/aarch64.h (aarch64_frame): Protect with
9023 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
9024 hard_fp_offset, frame_size, initial_adjust, callee_offset and
9025 final_offset from HOST_WIDE_INT to poly_int64.
9026 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
9027 to_constant when getting the number of units in an Advanced SIMD
9028 mode.
9029 (aarch64_builtin_vectorized_function): Check for a constant number
9030 of units.
9031 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
9032 GET_MODE_SIZE.
9033 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
9034 attribute instead of GET_MODE_NUNITS.
9035 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
9036 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
9037 GET_MODE_SIZE for fixed-size registers.
9038 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
9039 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
9040 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
9041 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
9042 (aarch64_print_operand, aarch64_print_address_internal)
9043 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
9044 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
9045 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
9046 Handle polynomial GET_MODE_SIZE.
9047 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
9048 wider than SImode without modification.
9049 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
9050 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
9051 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
9052 passing and returning SVE modes.
9053 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
9054 rather than GEN_INT.
9055 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
9056 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
9057 (aarch64_allocate_and_probe_stack_space): Likewise.
9058 (aarch64_layout_frame): Cope with polynomial offsets.
9059 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
9060 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
9061 polynomial offsets.
9062 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
9063 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
9064 poly_int64 rather than a HOST_WIDE_INT.
9065 (aarch64_get_separate_components, aarch64_process_components)
9066 (aarch64_expand_prologue, aarch64_expand_epilogue)
9067 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
9068 (aarch64_anchor_offset): New function, split out from...
9069 (aarch64_legitimize_address): ...here.
9070 (aarch64_builtin_vectorization_cost): Handle polynomial
9071 TYPE_VECTOR_SUBPARTS.
9072 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
9073 GET_MODE_NUNITS.
9074 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
9075 number of elements from the PARALLEL rather than the mode.
9076 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
9077 rather than GET_MODE_BITSIZE.
9078 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
9079 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
9080 (aarch64_expand_vec_perm_const_1): Handle polynomial
9081 d->perm.length () and d->perm elements.
9082 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
9083 Apply to_constant to d->perm elements.
9084 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
9085 polynomial CONST_VECTOR_NUNITS.
9086 (aarch64_move_pointer): Take amount as a poly_int64 rather
9087 than an int.
9088 (aarch64_progress_pointer): Avoid temporary variable.
9089 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
9090 the mode attribute instead of GET_MODE.
9092 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9093 Alan Hayward <alan.hayward@arm.com>
9094 David Sherwood <david.sherwood@arm.com>
9096 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
9097 x exists before using it.
9098 (aarch64_add_constant_internal): Rename to...
9099 (aarch64_add_offset_1): ...this. Replace regnum with separate
9100 src and dest rtxes. Handle the case in which they're different,
9101 including when the offset is zero. Replace scratchreg with an rtx.
9102 Use 2 additions if there is no spare register into which we can
9103 move a 16-bit constant.
9104 (aarch64_add_constant): Delete.
9105 (aarch64_add_offset): Replace reg with separate src and dest
9106 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
9107 Use aarch64_add_offset_1.
9108 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
9109 an rtx rather than an int. Take the delta as a poly_int64
9110 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
9111 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
9112 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
9113 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
9114 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
9115 and aarch64_add_sp.
9116 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
9117 aarch64_add_constant.
9119 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9121 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
9122 Use scalar_float_mode.
9124 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9126 * config/aarch64/aarch64-simd.md
9127 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
9128 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
9129 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
9130 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
9131 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
9132 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
9133 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
9134 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
9135 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
9136 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
9138 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9140 PR target/83514
9141 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
9142 targ_options->x_arm_arch_string is non NULL.
9144 2018-01-11 Tamar Christina <tamar.christina@arm.com>
9146 * config/aarch64/aarch64.h
9147 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
9149 2018-01-11 Sudakshina Das <sudi.das@arm.com>
9151 PR target/82096
9152 * expmed.c (emit_store_flag_force): Swap if const op0
9153 and change VOIDmode to mode of op0.
9155 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
9157 PR rtl-optimization/83761
9158 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
9159 than bytes to mode_for_size.
9161 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9163 PR middle-end/83189
9164 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
9165 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
9166 profile.
9168 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9170 PR middle-end/83575
9171 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
9172 when in layout mode.
9173 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
9174 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
9175 partition fixup.
9177 2018-01-10 Michael Collison <michael.collison@arm.com>
9179 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
9180 * config/aarch64/aarch64-option-extension.def: Add
9181 AARCH64_OPT_EXTENSION of 'fp16fml'.
9182 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9183 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
9184 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
9185 * config/aarch64/constraints.md (Ui7): New constraint.
9186 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
9187 (VFMLA_SEL_W): Ditto.
9188 (f16quad): Ditto.
9189 (f16mac1): Ditto.
9190 (VFMLA16_LOW): New int iterator.
9191 (VFMLA16_HIGH): Ditto.
9192 (UNSPEC_FMLAL): New unspec.
9193 (UNSPEC_FMLSL): Ditto.
9194 (UNSPEC_FMLAL2): Ditto.
9195 (UNSPEC_FMLSL2): Ditto.
9196 (f16mac): New code attribute.
9197 * config/aarch64/aarch64-simd-builtins.def
9198 (aarch64_fmlal_lowv2sf): Ditto.
9199 (aarch64_fmlsl_lowv2sf): Ditto.
9200 (aarch64_fmlalq_lowv4sf): Ditto.
9201 (aarch64_fmlslq_lowv4sf): Ditto.
9202 (aarch64_fmlal_highv2sf): Ditto.
9203 (aarch64_fmlsl_highv2sf): Ditto.
9204 (aarch64_fmlalq_highv4sf): Ditto.
9205 (aarch64_fmlslq_highv4sf): Ditto.
9206 (aarch64_fmlal_lane_lowv2sf): Ditto.
9207 (aarch64_fmlsl_lane_lowv2sf): Ditto.
9208 (aarch64_fmlal_laneq_lowv2sf): Ditto.
9209 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
9210 (aarch64_fmlalq_lane_lowv4sf): Ditto.
9211 (aarch64_fmlsl_lane_lowv4sf): Ditto.
9212 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
9213 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
9214 (aarch64_fmlal_lane_highv2sf): Ditto.
9215 (aarch64_fmlsl_lane_highv2sf): Ditto.
9216 (aarch64_fmlal_laneq_highv2sf): Ditto.
9217 (aarch64_fmlsl_laneq_highv2sf): Ditto.
9218 (aarch64_fmlalq_lane_highv4sf): Ditto.
9219 (aarch64_fmlsl_lane_highv4sf): Ditto.
9220 (aarch64_fmlalq_laneq_highv4sf): Ditto.
9221 (aarch64_fmlsl_laneq_highv4sf): Ditto.
9222 * config/aarch64/aarch64-simd.md:
9223 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
9224 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9225 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
9226 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
9227 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
9228 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
9229 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
9230 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
9231 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
9232 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
9233 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
9234 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
9235 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
9236 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
9237 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
9238 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
9239 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
9240 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
9241 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
9242 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
9243 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
9244 (vfmlsl_low_u32): Ditto.
9245 (vfmlalq_low_u32): Ditto.
9246 (vfmlslq_low_u32): Ditto.
9247 (vfmlal_high_u32): Ditto.
9248 (vfmlsl_high_u32): Ditto.
9249 (vfmlalq_high_u32): Ditto.
9250 (vfmlslq_high_u32): Ditto.
9251 (vfmlal_lane_low_u32): Ditto.
9252 (vfmlsl_lane_low_u32): Ditto.
9253 (vfmlal_laneq_low_u32): Ditto.
9254 (vfmlsl_laneq_low_u32): Ditto.
9255 (vfmlalq_lane_low_u32): Ditto.
9256 (vfmlslq_lane_low_u32): Ditto.
9257 (vfmlalq_laneq_low_u32): Ditto.
9258 (vfmlslq_laneq_low_u32): Ditto.
9259 (vfmlal_lane_high_u32): Ditto.
9260 (vfmlsl_lane_high_u32): Ditto.
9261 (vfmlal_laneq_high_u32): Ditto.
9262 (vfmlsl_laneq_high_u32): Ditto.
9263 (vfmlalq_lane_high_u32): Ditto.
9264 (vfmlslq_lane_high_u32): Ditto.
9265 (vfmlalq_laneq_high_u32): Ditto.
9266 (vfmlslq_laneq_high_u32): Ditto.
9267 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
9268 (AARCH64_FL_FOR_ARCH8_4): New.
9269 (AARCH64_ISA_F16FML): New ISA flag.
9270 (TARGET_F16FML): New feature flag for fp16fml.
9271 (doc/invoke.texi): Document new fp16fml option.
9273 2018-01-10 Michael Collison <michael.collison@arm.com>
9275 * config/aarch64/aarch64-builtins.c:
9276 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
9277 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9278 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
9279 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
9280 (AARCH64_ISA_SHA3): New ISA flag.
9281 (TARGET_SHA3): New feature flag for sha3.
9282 * config/aarch64/iterators.md (sha512_op): New int attribute.
9283 (CRYPTO_SHA512): New int iterator.
9284 (UNSPEC_SHA512H): New unspec.
9285 (UNSPEC_SHA512H2): Ditto.
9286 (UNSPEC_SHA512SU0): Ditto.
9287 (UNSPEC_SHA512SU1): Ditto.
9288 * config/aarch64/aarch64-simd-builtins.def
9289 (aarch64_crypto_sha512hqv2di): New builtin.
9290 (aarch64_crypto_sha512h2qv2di): Ditto.
9291 (aarch64_crypto_sha512su0qv2di): Ditto.
9292 (aarch64_crypto_sha512su1qv2di): Ditto.
9293 (aarch64_eor3qv8hi): Ditto.
9294 (aarch64_rax1qv2di): Ditto.
9295 (aarch64_xarqv2di): Ditto.
9296 (aarch64_bcaxqv8hi): Ditto.
9297 * config/aarch64/aarch64-simd.md:
9298 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
9299 (aarch64_crypto_sha512su0qv2di): Ditto.
9300 (aarch64_crypto_sha512su1qv2di): Ditto.
9301 (aarch64_eor3qv8hi): Ditto.
9302 (aarch64_rax1qv2di): Ditto.
9303 (aarch64_xarqv2di): Ditto.
9304 (aarch64_bcaxqv8hi): Ditto.
9305 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
9306 (vsha512h2q_u64): Ditto.
9307 (vsha512su0q_u64): Ditto.
9308 (vsha512su1q_u64): Ditto.
9309 (veor3q_u16): Ditto.
9310 (vrax1q_u64): Ditto.
9311 (vxarq_u64): Ditto.
9312 (vbcaxq_u16): Ditto.
9313 * config/arm/types.md (crypto_sha512): New type attribute.
9314 (crypto_sha3): Ditto.
9315 (doc/invoke.texi): Document new sha3 option.
9317 2018-01-10 Michael Collison <michael.collison@arm.com>
9319 * config/aarch64/aarch64-builtins.c:
9320 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
9321 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9322 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
9323 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
9324 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
9325 (AARCH64_ISA_SM4): New ISA flag.
9326 (TARGET_SM4): New feature flag for sm4.
9327 * config/aarch64/aarch64-simd-builtins.def
9328 (aarch64_sm3ss1qv4si): Ditto.
9329 (aarch64_sm3tt1aq4si): Ditto.
9330 (aarch64_sm3tt1bq4si): Ditto.
9331 (aarch64_sm3tt2aq4si): Ditto.
9332 (aarch64_sm3tt2bq4si): Ditto.
9333 (aarch64_sm3partw1qv4si): Ditto.
9334 (aarch64_sm3partw2qv4si): Ditto.
9335 (aarch64_sm4eqv4si): Ditto.
9336 (aarch64_sm4ekeyqv4si): Ditto.
9337 * config/aarch64/aarch64-simd.md:
9338 (aarch64_sm3ss1qv4si): Ditto.
9339 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
9340 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
9341 (aarch64_sm4eqv4si): Ditto.
9342 (aarch64_sm4ekeyqv4si): Ditto.
9343 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
9344 (sm3part_op): Ditto.
9345 (CRYPTO_SM3TT): Ditto.
9346 (CRYPTO_SM3PART): Ditto.
9347 (UNSPEC_SM3SS1): New unspec.
9348 (UNSPEC_SM3TT1A): Ditto.
9349 (UNSPEC_SM3TT1B): Ditto.
9350 (UNSPEC_SM3TT2A): Ditto.
9351 (UNSPEC_SM3TT2B): Ditto.
9352 (UNSPEC_SM3PARTW1): Ditto.
9353 (UNSPEC_SM3PARTW2): Ditto.
9354 (UNSPEC_SM4E): Ditto.
9355 (UNSPEC_SM4EKEY): Ditto.
9356 * config/aarch64/constraints.md (Ui2): New constraint.
9357 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
9358 * config/arm/types.md (crypto_sm3): New type attribute.
9359 (crypto_sm4): Ditto.
9360 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
9361 (vsm3tt1aq_u32): Ditto.
9362 (vsm3tt1bq_u32): Ditto.
9363 (vsm3tt2aq_u32): Ditto.
9364 (vsm3tt2bq_u32): Ditto.
9365 (vsm3partw1q_u32): Ditto.
9366 (vsm3partw2q_u32): Ditto.
9367 (vsm4eq_u32): Ditto.
9368 (vsm4ekeyq_u32): Ditto.
9369 (doc/invoke.texi): Document new sm4 option.
9371 2018-01-10 Michael Collison <michael.collison@arm.com>
9373 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
9374 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
9375 (AARCH64_FL_FOR_ARCH8_4): New.
9376 (AARCH64_FL_V8_4): New flag.
9377 (doc/invoke.texi): Document new armv8.4-a option.
9379 2018-01-10 Michael Collison <michael.collison@arm.com>
9381 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
9382 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
9383 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
9384 * config/aarch64/aarch64-option-extension.def: Add
9385 AARCH64_OPT_EXTENSION of 'sha2'.
9386 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
9387 (crypto): Disable sha2 and aes if crypto disabled.
9388 (crypto): Enable aes and sha2 if enabled.
9389 (simd): Disable sha2 and aes if simd disabled.
9390 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
9391 New flags.
9392 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
9393 (TARGET_SHA2): New feature flag for sha2.
9394 (TARGET_AES): New feature flag for aes.
9395 * config/aarch64/aarch64-simd.md:
9396 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
9397 conditional on TARGET_AES.
9398 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
9399 (aarch64_crypto_sha1hsi): Make pattern conditional
9400 on TARGET_SHA2.
9401 (aarch64_crypto_sha1hv4si): Ditto.
9402 (aarch64_be_crypto_sha1hv4si): Ditto.
9403 (aarch64_crypto_sha1su1v4si): Ditto.
9404 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
9405 (aarch64_crypto_sha1su0v4si): Ditto.
9406 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
9407 (aarch64_crypto_sha256su0v4si): Ditto.
9408 (aarch64_crypto_sha256su1v4si): Ditto.
9409 (doc/invoke.texi): Document new aes and sha2 options.
9411 2018-01-10 Martin Sebor <msebor@redhat.com>
9413 PR tree-optimization/83781
9414 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
9415 as string arrays.
9417 2018-01-11 Martin Sebor <msebor@gmail.com>
9418 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9420 PR tree-optimization/83501
9421 PR tree-optimization/81703
9423 * tree-ssa-strlen.c (get_string_cst): Rename...
9424 (get_string_len): ...to this. Handle global constants.
9425 (handle_char_store): Adjust.
9427 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
9428 Jim Wilson <jimw@sifive.com>
9430 * config/riscv/riscv-protos.h (riscv_output_return): New.
9431 * config/riscv/riscv.c (struct machine_function): New naked_p field.
9432 (riscv_attribute_table, riscv_output_return),
9433 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
9434 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
9435 (riscv_compute_frame_info): Only compute frame->mask if not a naked
9436 function.
9437 (riscv_expand_prologue): Add early return for naked function.
9438 (riscv_expand_epilogue): Likewise.
9439 (riscv_function_ok_for_sibcall): Return false for naked function.
9440 (riscv_set_current_function): New.
9441 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
9442 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
9443 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
9444 * doc/extend.texi (RISC-V Function Attributes): New.
9446 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
9448 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
9449 check for 128-bit long double before checking TCmode.
9450 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
9451 128-bit long doubles before checking TFmode or TCmode.
9452 (FLOAT128_IBM_P): Likewise.
9454 2018-01-10 Martin Sebor <msebor@redhat.com>
9456 PR tree-optimization/83671
9457 * builtins.c (c_strlen): Unconditionally return zero for the empty
9458 string.
9459 Use -Warray-bounds for warnings.
9460 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
9461 for non-constant array indices with COMPONENT_REF, arrays of
9462 arrays, and pointers to arrays.
9463 (gimple_fold_builtin_strlen): Determine and set length range for
9464 non-constant character arrays.
9466 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
9468 PR middle-end/81897
9469 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
9470 empty blocks.
9472 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
9474 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
9476 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9478 PR target/83399
9479 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
9480 VECTOR_MEM_ALTIVEC_OR_VSX_P.
9481 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
9482 indexed_or_indirect_operand predicate.
9483 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
9484 (*vsx_le_perm_load_v8hi): Likewise.
9485 (*vsx_le_perm_load_v16qi): Likewise.
9486 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
9487 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
9488 (*vsx_le_perm_store_v8hi): Likewise.
9489 (*vsx_le_perm_store_v16qi): Likewise.
9490 (eight unnamed splitters): Likewise.
9492 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
9494 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
9495 * config/rs6000/emmintrin.h: Likewise.
9496 * config/rs6000/mmintrin.h: Likewise.
9497 * config/rs6000/xmmintrin.h: Likewise.
9499 2018-01-10 David Malcolm <dmalcolm@redhat.com>
9501 PR c++/43486
9502 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
9503 "public_flag".
9504 * tree.c (tree_nop_conversion): Return true for location wrapper
9505 nodes.
9506 (maybe_wrap_with_location): New function.
9507 (selftest::check_strip_nops): New function.
9508 (selftest::test_location_wrappers): New function.
9509 (selftest::tree_c_tests): Call it.
9510 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
9511 (maybe_wrap_with_location): New decl.
9512 (EXPR_LOCATION_WRAPPER_P): New macro.
9513 (location_wrapper_p): New inline function.
9514 (tree_strip_any_location_wrapper): New inline function.
9516 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
9518 PR target/83735
9519 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
9520 stack_realign_offset for the largest alignment of stack slot
9521 actually used.
9522 (ix86_find_max_used_stack_alignment): New function.
9523 (ix86_finalize_stack_frame_flags): Use it. Set
9524 max_used_stack_alignment if we don't realign stack.
9525 * config/i386/i386.h (machine_function): Add
9526 max_used_stack_alignment.
9528 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
9530 * config/arm/arm.opt (-mbranch-cost): New option.
9531 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
9532 account.
9534 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
9536 PR target/83629
9537 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
9538 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
9540 2018-01-10 Richard Biener <rguenther@suse.de>
9542 PR debug/83765
9543 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
9544 early out so it also covers the case where we have a non-NULL
9545 origin.
9547 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9549 PR tree-optimization/83753
9550 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
9551 for non-strided grouped accesses if the number of elements is 1.
9553 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
9555 PR target/81616
9556 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
9557 * i386.h (TARGET_USE_GATHER): Define.
9558 * x86-tune.def (X86_TUNE_USE_GATHER): New.
9560 2018-01-10 Martin Liska <mliska@suse.cz>
9562 PR bootstrap/82831
9563 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
9564 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
9565 partitioning.
9566 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
9567 CLEANUP_NO_PARTITIONING is not set.
9569 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
9571 * doc/rtl.texi: Remove documentation of (const ...) wrappers
9572 for vectors, as a partial revert of r254296.
9573 * rtl.h (const_vec_p): Delete.
9574 (const_vec_duplicate_p): Don't test for vector CONSTs.
9575 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
9576 * expmed.c (make_tree): Likewise.
9578 Revert:
9579 * common.md (E, F): Use CONSTANT_P instead of checking for
9580 CONST_VECTOR.
9581 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9582 checking for CONST_VECTOR.
9584 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9586 PR middle-end/83575
9587 * predict.c (force_edge_cold): Handle in more sane way edges
9588 with no prediction.
9590 2018-01-09 Carl Love <cel@us.ibm.com>
9592 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
9593 V4SI, V4SF types.
9594 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
9595 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
9596 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
9597 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
9598 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
9599 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
9600 * config/rs6000/rs6000-protos.h: Add extern defition for
9601 rs6000_generate_float2_double_code.
9602 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
9603 function.
9604 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
9605 (float2_v2df): Add define_expand.
9607 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
9609 PR target/83628
9610 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
9611 op_mode in the force_to_mode call.
9613 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9615 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
9616 instead of checking each element individually.
9617 (aarch64_evpc_uzp): Likewise.
9618 (aarch64_evpc_zip): Likewise.
9619 (aarch64_evpc_ext): Likewise.
9620 (aarch64_evpc_rev): Likewise.
9621 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
9622 instead of checking each element individually. Return true without
9623 generating rtl if
9624 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
9625 whether all selected elements come from the same input, instead of
9626 checking each element individually. Remove calls to gen_rtx_REG,
9627 start_sequence and end_sequence and instead assert that no rtl is
9628 generated.
9630 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9632 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
9633 order of HIGH and CONST checks.
9635 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
9637 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
9638 if the destination isn't an SSA_NAME.
9640 2018-01-09 Richard Biener <rguenther@suse.de>
9642 PR tree-optimization/83668
9643 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
9644 move prologue...
9645 (canonicalize_loop_form): ... here, renamed from ...
9646 (canonicalize_loop_closed_ssa_form): ... this and amended to
9647 swap successor edges for loop exit blocks to make us use
9648 the RPO order we need for initial schedule generation.
9650 2018-01-09 Joseph Myers <joseph@codesourcery.com>
9652 PR tree-optimization/64811
9653 * match.pd: When optimizing comparisons with Inf, avoid
9654 introducing or losing exceptions from comparisons with NaN.
9656 2018-01-09 Martin Liska <mliska@suse.cz>
9658 PR sanitizer/82517
9659 * asan.c (shadow_mem_size): Add gcc_assert.
9661 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
9663 Don't save registers in main().
9665 PR target/83738
9666 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
9667 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
9668 * config/avr/avr.c (avr_set_current_function): Don't error if
9669 naked, OS_task or OS_main are specified at the same time.
9670 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
9671 OS_main.
9672 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
9673 attribute.
9674 * common/config/avr/avr-common.c (avr_option_optimization_table):
9675 Switch on -mmain-is-OS_task for optimizing compilations.
9677 2018-01-09 Richard Biener <rguenther@suse.de>
9679 PR tree-optimization/83572
9680 * graphite.c: Include cfganal.h.
9681 (graphite_transform_loops): Connect infinite loops to exit
9682 and remove fake edges at the end.
9684 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9686 * ipa-inline.c (edge_badness): Revert accidental checkin.
9688 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
9690 PR ipa/80763
9691 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
9692 symbols; not inline clones.
9694 2018-01-09 Jakub Jelinek <jakub@redhat.com>
9696 PR target/83507
9697 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
9698 hard registers. Formatting fixes.
9700 PR preprocessor/83722
9701 * gcc.c (try_generate_repro): Pass
9702 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
9703 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
9704 do_report_bug.
9706 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
9707 Kito Cheng <kito.cheng@gmail.com>
9709 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
9710 (riscv_leaf_function_p): Delete.
9711 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
9713 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9715 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
9716 function.
9717 (do_ifelse): New function.
9718 (do_isel): New function.
9719 (do_sub3): New function.
9720 (do_add3): New function.
9721 (do_load_mask_compare): New function.
9722 (do_overlap_load_compare): New function.
9723 (expand_compare_loop): New function.
9724 (expand_block_compare): Call expand_compare_loop() when appropriate.
9725 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
9726 option description.
9727 (-mblock-compare-inline-loop-limit): New option.
9729 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
9731 PR target/83677
9732 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
9733 Reverse order of second and third operands in first alternative.
9734 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
9735 of first and second elements in UNSPEC_VPERMR vector.
9736 (altivec_expand_vec_perm_le): Likewise.
9738 2018-01-08 Jeff Law <law@redhat.com>
9740 PR rtl-optimizatin/81308
9741 * tree-switch-conversion.c (cfg_altered): New file scoped static.
9742 (process_switch): If group_case_labels makes a change, then set
9743 cfg_altered.
9744 (pass_convert_switch::execute): If a switch is converted, then
9745 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
9747 PR rtl-optimization/81308
9748 * recog.c (split_all_insns): Conditionally cleanup the CFG after
9749 splitting insns.
9751 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
9753 PR target/83663 - Revert r255946
9754 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
9755 generation for cases where splatting a value is not useful.
9756 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
9757 across a vec_duplicate and a paradoxical subreg forming a vector
9758 mode to a vec_concat.
9760 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9762 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
9763 -march=armv8.3-a variants.
9764 * config/arm/t-multilib: Likewise.
9765 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
9767 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
9769 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
9770 to generate rtl.
9771 (cceq_ior_compare_complement): Give it a name so I can use it, and
9772 change boolean_or_operator predicate to boolean_operator so it can
9773 be used to generate a crand.
9774 (eqne): New code iterator.
9775 (bd/bd_neg): New code_attrs.
9776 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
9777 a single define_insn.
9778 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
9779 decrement (bdnzt/bdnzf/bdzt/bdzf).
9780 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
9781 with the new names of the branch decrement patterns, and added the
9782 names of the branch decrement conditional patterns.
9784 2018-01-08 Richard Biener <rguenther@suse.de>
9786 PR tree-optimization/83563
9787 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
9788 cache.
9790 2018-01-08 Richard Biener <rguenther@suse.de>
9792 PR middle-end/83713
9793 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
9795 2018-01-08 Richard Biener <rguenther@suse.de>
9797 PR tree-optimization/83685
9798 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
9799 references to abnormals.
9801 2018-01-08 Richard Biener <rguenther@suse.de>
9803 PR lto/83719
9804 * dwarf2out.c (output_indirect_strings): Handle empty
9805 skeleton_debug_str_hash.
9806 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
9808 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9810 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
9811 (emit_store_direct): Likewise.
9812 (arc_trampoline_adjust_address): Likewise.
9813 (arc_asm_trampoline_template): New function.
9814 (arc_initialize_trampoline): Use asm_trampoline_template.
9815 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
9816 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
9817 * config/arc/arc.md (flush_icache): Delete pattern.
9819 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
9821 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
9822 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
9823 munaligned-access.
9825 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9827 PR target/83681
9828 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
9829 by not USED_FOR_TARGET.
9830 (make_pass_resolve_sw_modes): Likewise.
9832 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
9834 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
9835 USED_FOR_TARGET.
9837 2018-01-08 Richard Biener <rguenther@suse.de>
9839 PR middle-end/83580
9840 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
9842 2018-01-08 Richard Biener <rguenther@suse.de>
9844 PR middle-end/83517
9845 * match.pd ((t * 2) / 2) -> t): Add missing :c.
9847 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
9849 PR middle-end/81897
9850 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
9851 basic blocks with a small number of successors.
9852 (convert_control_dep_chain_into_preds): Improve handling of
9853 forwarder blocks.
9854 (dump_predicates): Split apart into...
9855 (dump_pred_chain): ...here...
9856 (dump_pred_info): ...and here.
9857 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9858 (can_chain_union_be_invalidated_p): Improve check for invalidation
9859 of paths.
9860 (uninit_uses_cannot_happen): Avoid unnecessary if
9861 convert_control_dep_chain_into_preds yielded nothing.
9863 2018-01-06 Martin Sebor <msebor@redhat.com>
9865 PR tree-optimization/83640
9866 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9867 subtracting negative offset from size.
9868 (builtin_access::overlap): Adjust offset bounds of the access to fall
9869 within the size of the object if possible.
9871 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9873 PR rtl-optimization/83699
9874 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9875 extract_bit_field_as_subreg to cases in which the extracted
9876 value is also a vector.
9878 * lra-constraints.c (process_alt_operands): Test for the equivalence
9879 substitutions when detecting a possible reload cycle.
9881 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9883 PR debug/83480
9884 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9885 by default if flag_selective_schedling{,2}. Formatting fixes.
9887 PR rtl-optimization/83682
9888 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9889 if it has non-VECTOR_MODE element mode.
9890 (vec_duplicate_p): Likewise.
9892 PR middle-end/83694
9893 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9894 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9896 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9898 PR target/83604
9899 * config/i386/i386-builtin.def
9900 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9901 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9902 Require also OPTION_MASK_ISA_AVX512F in addition to
9903 OPTION_MASK_ISA_GFNI.
9904 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9905 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9906 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9907 to OPTION_MASK_ISA_GFNI.
9908 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9909 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9910 OPTION_MASK_ISA_AVX512BW.
9911 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9912 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9913 addition to OPTION_MASK_ISA_GFNI.
9914 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9915 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9916 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9917 to OPTION_MASK_ISA_GFNI.
9918 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9919 a requirement for all ISAs rather than any of them with a few
9920 exceptions.
9921 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9922 processing.
9923 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9924 bitmasks to be enabled with 3 exceptions, instead of requiring any
9925 enabled ISA with lots of exceptions.
9926 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9927 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9928 Change avx512bw in isa attribute to avx512f.
9929 * config/i386/sgxintrin.h: Add license boilerplate.
9930 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9931 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9932 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9933 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9934 defined.
9935 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9936 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9937 temporarily sse2 rather than sse if not enabled already.
9939 PR target/83604
9940 * config/i386/sse.md (VI248_VLBW): Rename to ...
9941 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9942 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9943 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9944 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9945 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9946 mode iterator instead of VI248_VLBW.
9948 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9950 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9951 (record_modified): Skip clobbers; add debug output.
9952 (param_change_prob): Use sreal frequencies.
9954 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9956 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9957 punt for user-aligned variables.
9959 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9961 * tree-chrec.c (chrec_contains_symbols): Return true for
9962 POLY_INT_CST.
9964 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9966 PR target/82439
9967 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9968 of (x|y) == x for BICS pattern.
9970 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9972 PR tree-optimization/83605
9973 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9974 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9975 can throw.
9977 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9979 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9980 * config/epiphany/rtems.h: New file.
9982 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9983 Uros Bizjak <ubizjak@gmail.com>
9985 PR target/83554
9986 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9987 QIreg_operand instead of register_operand predicate.
9988 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9989 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9990 comments instead of -fmitigate[-_]rop.
9992 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9994 PR bootstrap/81926
9995 * cgraphunit.c (symbol_table::compile): Switch to text_section
9996 before calling assembly_start debug hook.
9997 * run-rtl-passes.c (run_rtl_passes): Likewise.
9998 Include output.h.
10000 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10002 * tree-vrp.c (extract_range_from_binary_expr_1): Check
10003 range_int_cst_p rather than !symbolic_range_p before calling
10004 extract_range_from_multiplicative_op_1.
10006 2018-01-04 Jeff Law <law@redhat.com>
10008 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
10009 redundant test in assertion.
10011 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10013 * doc/rtl.texi: Document machine_mode wrapper classes.
10015 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10017 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
10018 using tree_to_uhwi.
10020 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10022 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
10023 the VEC_PERM_EXPR fold to fail.
10025 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10027 PR debug/83585
10028 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
10029 to switched_sections.
10031 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10033 PR target/83680
10034 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
10035 test for d.testing.
10037 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
10039 PR target/83387
10040 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
10041 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
10043 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10045 PR debug/83666
10046 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
10047 is BLKmode and bitpos not zero or mode change is needed.
10049 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
10051 PR target/83675
10052 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
10053 TARGET_VIS2.
10055 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
10057 PR target/83628
10058 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
10059 instead of MULT rtx. Update all corresponding splitters.
10060 (*saddl_se): Ditto.
10061 (*ssub<modesuffix>): Ditto.
10062 (*ssubl_se): Ditto.
10063 (*cmp_sadd_di): Update split patterns.
10064 (*cmp_sadd_si): Ditto.
10065 (*cmp_sadd_sidi): Ditto.
10066 (*cmp_ssub_di): Ditto.
10067 (*cmp_ssub_si): Ditto.
10068 (*cmp_ssub_sidi): Ditto.
10069 * config/alpha/predicates.md (const23_operand): New predicate.
10070 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
10071 Look for ASHIFT, not MULT inner operand.
10072 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
10074 2018-01-04 Martin Liska <mliska@suse.cz>
10076 PR gcov-profile/83669
10077 * gcov.c (output_intermediate_file): Add version to intermediate
10078 gcov file.
10079 * doc/gcov.texi: Document new field 'version' in intermediate
10080 file format. Fix location of '-k' option of gcov command.
10082 2018-01-04 Martin Liska <mliska@suse.cz>
10084 PR ipa/82352
10085 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
10087 2018-01-04 Jakub Jelinek <jakub@redhat.com>
10089 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
10091 2018-01-03 Martin Sebor <msebor@redhat.com>
10093 PR tree-optimization/83655
10094 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
10095 checking calls with invalid arguments.
10097 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 * tree-vect-stmts.c (vect_get_store_rhs): New function.
10100 (vectorizable_mask_load_store): Delete.
10101 (vectorizable_call): Return false for masked loads and stores.
10102 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
10103 instead of gimple_assign_rhs1.
10104 (vectorizable_load): Handle IFN_MASK_LOAD.
10105 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
10107 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10109 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
10110 split out from..,
10111 (vectorizable_mask_load_store): ...here.
10112 (vectorizable_load): ...and here.
10114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10116 * tree-vect-stmts.c (vect_build_all_ones_mask)
10117 (vect_build_zero_merge_argument): New functions, split out from...
10118 (vectorizable_load): ...here.
10120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10122 * tree-vect-stmts.c (vect_check_store_rhs): New function,
10123 split out from...
10124 (vectorizable_mask_load_store): ...here.
10125 (vectorizable_store): ...and here.
10127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10129 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
10130 split out from...
10131 (vectorizable_mask_load_store): ...here.
10133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10135 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
10136 (vect_model_store_cost): Take a vec_load_store_type instead of a
10137 vect_def_type.
10138 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
10139 (vect_model_store_cost): Take a vec_load_store_type instead of a
10140 vect_def_type.
10141 (vectorizable_mask_load_store): Update accordingly.
10142 (vectorizable_store): Likewise.
10143 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
10145 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10147 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
10148 IFN_MASK_LOAD calls here rather than...
10149 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
10151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10152 Alan Hayward <alan.hayward@arm.com>
10153 David Sherwood <david.sherwood@arm.com>
10155 * expmed.c (extract_bit_field_1): For vector extracts,
10156 fall back to extract_bit_field_as_subreg if vec_extract
10157 isn't available.
10159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10160 Alan Hayward <alan.hayward@arm.com>
10161 David Sherwood <david.sherwood@arm.com>
10163 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
10164 they are variable or constant sized.
10165 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
10166 slots for constant-sized data.
10168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10169 Alan Hayward <alan.hayward@arm.com>
10170 David Sherwood <david.sherwood@arm.com>
10172 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
10173 handling COND_EXPRs with boolean comparisons, try to find a better
10174 basis for the mask type than the boolean itself.
10176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10178 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
10179 is calculated and how it can be overridden.
10180 * genmodes.c (max_bitsize_mode_any_mode): New variable.
10181 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
10182 if defined.
10183 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
10184 if nonzero.
10186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10187 Alan Hayward <alan.hayward@arm.com>
10188 David Sherwood <david.sherwood@arm.com>
10190 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
10191 Remove the mode argument.
10192 (aarch64_simd_valid_immediate): Remove the mode and inverse
10193 arguments.
10194 * config/aarch64/iterators.md (bitsize): New iterator.
10195 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
10196 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
10197 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
10198 aarch64_simd_valid_immediate.
10199 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
10200 (aarch64_reg_or_bic_imm): Likewise.
10201 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
10202 with an insn_type enum and msl with a modifier_type enum.
10203 Replace element_width with a scalar_mode. Change the shift
10204 to unsigned int. Add constructors for scalar_float_mode and
10205 scalar_int_mode elements.
10206 (aarch64_vect_float_const_representable_p): Delete.
10207 (aarch64_can_const_movi_rtx_p)
10208 (aarch64_simd_scalar_immediate_valid_for_move)
10209 (aarch64_simd_make_constant): Update call to
10210 aarch64_simd_valid_immediate.
10211 (aarch64_advsimd_valid_immediate_hs): New function.
10212 (aarch64_advsimd_valid_immediate): Likewise.
10213 (aarch64_simd_valid_immediate): Remove mode and inverse
10214 arguments. Rewrite to use the above. Use const_vec_duplicate_p
10215 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
10216 and aarch64_float_const_representable_p on the result.
10217 (aarch64_output_simd_mov_immediate): Remove mode argument.
10218 Update call to aarch64_simd_valid_immediate and use of
10219 simd_immediate_info.
10220 (aarch64_output_scalar_simd_mov_immediate): Update call
10221 accordingly.
10223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10224 Alan Hayward <alan.hayward@arm.com>
10225 David Sherwood <david.sherwood@arm.com>
10227 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
10228 (mode_nunits): Likewise CONST_MODE_NUNITS.
10229 * machmode.def (ADJUST_NUNITS): Document.
10230 * genmodes.c (mode_data::need_nunits_adj): New field.
10231 (blank_mode): Update accordingly.
10232 (adj_nunits): New variable.
10233 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
10234 parameter.
10235 (emit_mode_size_inline): Set need_bytesize_adj for all modes
10236 listed in adj_nunits.
10237 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
10238 listed in adj_nunits. Don't emit case statements for such modes.
10239 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
10240 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
10241 nothing if adj_nunits is nonnull.
10242 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
10243 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
10244 (emit_mode_fbit): Update use of print_maybe_const_decl.
10245 (emit_move_size): Likewise. Treat the array as non-const
10246 if adj_nunits.
10247 (emit_mode_adjustments): Handle adj_nunits.
10249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10251 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
10252 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
10253 (VECTOR_MODES): Use it.
10254 (make_vector_modes): Take the prefix as an argument.
10256 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10257 Alan Hayward <alan.hayward@arm.com>
10258 David Sherwood <david.sherwood@arm.com>
10260 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
10261 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
10262 for MODE_VECTOR_BOOL.
10263 * machmode.def (VECTOR_BOOL_MODE): Document.
10264 * genmodes.c (VECTOR_BOOL_MODE): New macro.
10265 (make_vector_bool_mode): New function.
10266 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
10267 MODE_VECTOR_BOOL.
10268 * lto-streamer-in.c (lto_input_mode_table): Likewise.
10269 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
10270 Likewise.
10271 * stor-layout.c (int_mode_for_mode): Likewise.
10272 * tree.c (build_vector_type_for_mode): Likewise.
10273 * varasm.c (output_constant_pool_2): Likewise.
10274 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
10275 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
10276 for MODE_VECTOR_BOOL.
10277 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
10278 of mode class checks.
10279 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
10280 instead of a list of mode class checks.
10281 (expand_vector_scalar_condition): Likewise.
10282 (type_for_widest_vector_mode): Handle BImode as an inner mode.
10284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10285 Alan Hayward <alan.hayward@arm.com>
10286 David Sherwood <david.sherwood@arm.com>
10288 * machmode.h (mode_size): Change from unsigned short to
10289 poly_uint16_pod.
10290 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
10291 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10292 or if measurement_type is not polynomial.
10293 (fixed_size_mode::includes_p): Check for constant-sized modes.
10294 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
10295 return a poly_uint16 rather than an unsigned short.
10296 (emit_mode_size): Change the type of mode_size from unsigned short
10297 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
10298 (emit_mode_adjustments): Cope with polynomial vector sizes.
10299 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10300 for GET_MODE_SIZE.
10301 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10302 for GET_MODE_SIZE.
10303 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
10304 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
10305 * caller-save.c (setup_save_areas): Likewise.
10306 (replace_reg_with_saved_mem): Likewise.
10307 * calls.c (emit_library_call_value_1): Likewise.
10308 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
10309 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
10310 (gen_lowpart_for_combine): Likewise.
10311 * convert.c (convert_to_integer_1): Likewise.
10312 * cse.c (equiv_constant, cse_insn): Likewise.
10313 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
10314 (cselib_subst_to_values): Likewise.
10315 * dce.c (word_dce_process_block): Likewise.
10316 * df-problems.c (df_word_lr_mark_ref): Likewise.
10317 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
10318 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
10319 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
10320 (rtl_for_decl_location): Likewise.
10321 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
10322 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
10323 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
10324 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
10325 (expand_expr_real_1): Likewise.
10326 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
10327 (pad_below): Likewise.
10328 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10329 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
10330 * ira.c (get_subreg_tracking_sizes): Likewise.
10331 * ira-build.c (ira_create_allocno_objects): Likewise.
10332 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
10333 (ira_sort_regnos_for_alter_reg): Likewise.
10334 * ira-costs.c (record_operand_costs): Likewise.
10335 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
10336 (resolve_simple_move): Likewise.
10337 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
10338 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
10339 (lra_constraints): Likewise.
10340 (CONST_POOL_OK_P): Reject variable-sized modes.
10341 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
10342 (add_pseudo_to_slot, lra_spill): Likewise.
10343 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10344 * optabs-query.c (get_best_extraction_insn): Likewise.
10345 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10346 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
10347 (expand_mult_highpart, valid_multiword_target_p): Likewise.
10348 * recog.c (offsettable_address_addr_space_p): Likewise.
10349 * regcprop.c (maybe_mode_change): Likewise.
10350 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
10351 * regrename.c (build_def_use): Likewise.
10352 * regstat.c (dump_reg_info): Likewise.
10353 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
10354 (find_reloads, find_reloads_subreg_address): Likewise.
10355 * reload1.c (eliminate_regs_1): Likewise.
10356 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
10357 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
10358 (simplify_binary_operation_1, simplify_subreg): Likewise.
10359 * targhooks.c (default_function_arg_padding): Likewise.
10360 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
10361 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
10362 (verify_gimple_assign_ternary): Likewise.
10363 * tree-inline.c (estimate_move_cost): Likewise.
10364 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10365 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
10366 (get_address_cost_ainc): Likewise.
10367 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
10368 (vect_supportable_dr_alignment): Likewise.
10369 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10370 (vectorizable_reduction): Likewise.
10371 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
10372 (vectorizable_operation, vectorizable_load): Likewise.
10373 * tree.c (build_same_sized_truth_vector_type): Likewise.
10374 * valtrack.c (cleanup_auto_inc_dec): Likewise.
10375 * var-tracking.c (emit_note_insn_var_location): Likewise.
10376 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
10377 (ADDR_VEC_ALIGN): Likewise.
10379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10380 Alan Hayward <alan.hayward@arm.com>
10381 David Sherwood <david.sherwood@arm.com>
10383 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
10384 unsigned short.
10385 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
10386 or if measurement_type is polynomial.
10387 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
10388 * combine.c (make_extraction): Likewise.
10389 * dse.c (find_shift_sequence): Likewise.
10390 * dwarf2out.c (mem_loc_descriptor): Likewise.
10391 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
10392 (extract_bit_field, extract_low_bits): Likewise.
10393 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
10394 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
10395 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
10396 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
10397 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
10398 * reload.c (find_reloads): Likewise.
10399 * reload1.c (alter_reg): Likewise.
10400 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
10401 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
10402 * tree-if-conv.c (predicate_mem_writes): Likewise.
10403 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
10404 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
10405 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
10406 * valtrack.c (dead_debug_insert_temp): Likewise.
10407 * varasm.c (mergeable_constant_section): Likewise.
10408 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
10410 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10411 Alan Hayward <alan.hayward@arm.com>
10412 David Sherwood <david.sherwood@arm.com>
10414 * expr.c (expand_assignment): Cope with polynomial mode sizes
10415 when assigning to a CONCAT.
10417 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10418 Alan Hayward <alan.hayward@arm.com>
10419 David Sherwood <david.sherwood@arm.com>
10421 * machmode.h (mode_precision): Change from unsigned short to
10422 poly_uint16_pod.
10423 (mode_to_precision): Return a poly_uint16 rather than an unsigned
10424 short.
10425 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
10426 or if measurement_type is not polynomial.
10427 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
10428 in which the mode is already known to be a scalar_int_mode.
10429 * genmodes.c (emit_mode_precision): Change the type of mode_precision
10430 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
10431 initializer.
10432 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10433 for GET_MODE_PRECISION.
10434 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10435 for GET_MODE_PRECISION.
10436 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
10437 as polynomial.
10438 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
10439 (expand_field_assignment, make_extraction): Likewise.
10440 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
10441 (get_last_value): Likewise.
10442 * convert.c (convert_to_integer_1): Likewise.
10443 * cse.c (cse_insn): Likewise.
10444 * expr.c (expand_expr_real_1): Likewise.
10445 * lra-constraints.c (simplify_operand_subreg): Likewise.
10446 * optabs-query.c (can_atomic_load_p): Likewise.
10447 * optabs.c (expand_atomic_load): Likewise.
10448 (expand_atomic_store): Likewise.
10449 * ree.c (combine_reaching_defs): Likewise.
10450 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
10451 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
10452 * tree.h (type_has_mode_precision_p): Likewise.
10453 * ubsan.c (instrument_si_overflow): Likewise.
10455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10456 Alan Hayward <alan.hayward@arm.com>
10457 David Sherwood <david.sherwood@arm.com>
10459 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
10460 polynomial numbers of units.
10461 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
10462 (valid_vector_subparts_p): New function.
10463 (build_vector_type): Remove temporary shim and take the number
10464 of units as a poly_uint64 rather than an int.
10465 (build_opaque_vector_type): Take the number of units as a
10466 poly_uint64 rather than an int.
10467 * tree.c (build_vector_from_ctor): Handle polynomial
10468 TYPE_VECTOR_SUBPARTS.
10469 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
10470 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
10471 (build_vector_from_val): If the number of units is variable,
10472 use build_vec_duplicate_cst for constant operands and
10473 VEC_DUPLICATE_EXPR otherwise.
10474 (make_vector_type): Remove temporary is_constant ().
10475 (build_vector_type, build_opaque_vector_type): Take the number of
10476 units as a poly_uint64 rather than an int.
10477 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
10478 VECTOR_CST_NELTS.
10479 * cfgexpand.c (expand_debug_expr): Likewise.
10480 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
10481 (store_constructor, expand_expr_real_1): Likewise.
10482 (const_scalar_mask_from_tree): Likewise.
10483 * fold-const-call.c (fold_const_reduction): Likewise.
10484 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
10485 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
10486 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
10487 (fold_relational_const): Likewise.
10488 (native_interpret_vector): Likewise. Change the size from an
10489 int to an unsigned int.
10490 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
10491 TYPE_VECTOR_SUBPARTS.
10492 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
10493 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
10494 duplicating a non-constant operand into a variable-length vector.
10495 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
10496 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
10497 * ipa-icf.c (sem_variable::equals): Likewise.
10498 * match.pd: Likewise.
10499 * omp-simd-clone.c (simd_clone_subparts): Likewise.
10500 * print-tree.c (print_node): Likewise.
10501 * stor-layout.c (layout_type): Likewise.
10502 * targhooks.c (default_builtin_vectorization_cost): Likewise.
10503 * tree-cfg.c (verify_gimple_comparison): Likewise.
10504 (verify_gimple_assign_binary): Likewise.
10505 (verify_gimple_assign_ternary): Likewise.
10506 (verify_gimple_assign_single): Likewise.
10507 * tree-pretty-print.c (dump_generic_node): Likewise.
10508 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10509 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
10510 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
10511 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
10512 (vect_shift_permute_load_chain): Likewise.
10513 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
10514 (expand_vector_condition, optimize_vector_constructor): Likewise.
10515 (lower_vec_perm, get_compute_type): Likewise.
10516 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10517 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
10518 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
10519 (vect_recog_mask_conversion_pattern): Likewise.
10520 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
10521 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
10522 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10523 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
10524 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
10525 (vectorizable_shift, vectorizable_operation, vectorizable_store)
10526 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
10527 (supportable_widening_operation): Likewise.
10528 (supportable_narrowing_operation): Likewise.
10529 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
10530 Likewise.
10531 * varasm.c (output_constant): Likewise.
10533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10534 Alan Hayward <alan.hayward@arm.com>
10535 David Sherwood <david.sherwood@arm.com>
10537 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
10538 so that both the length == 3 and length != 3 cases set up their
10539 own permute vectors. Add comments explaining why we know the
10540 number of elements is constant.
10541 (vect_permute_load_chain): Likewise.
10543 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10544 Alan Hayward <alan.hayward@arm.com>
10545 David Sherwood <david.sherwood@arm.com>
10547 * machmode.h (mode_nunits): Change from unsigned char to
10548 poly_uint16_pod.
10549 (ONLY_FIXED_SIZE_MODES): New macro.
10550 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
10551 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
10552 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
10553 New typedefs.
10554 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
10555 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
10556 or if measurement_type is not polynomial.
10557 * genmodes.c (ZERO_COEFFS): New macro.
10558 (emit_mode_nunits_inline): Make mode_nunits_inline return a
10559 poly_uint16.
10560 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
10561 Use ZERO_COEFFS when emitting initializers.
10562 * data-streamer.h (bp_pack_poly_value): New function.
10563 (bp_unpack_poly_value): Likewise.
10564 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
10565 for GET_MODE_NUNITS.
10566 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
10567 for GET_MODE_NUNITS.
10568 * tree.c (make_vector_type): Remove temporary shim and make
10569 the real function take the number of units as a poly_uint64
10570 rather than an int.
10571 (build_vector_type_for_mode): Handle polynomial nunits.
10572 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
10573 * emit-rtl.c (const_vec_series_p_1): Likewise.
10574 (gen_rtx_CONST_VECTOR): Likewise.
10575 * fold-const.c (test_vec_duplicate_folding): Likewise.
10576 * genrecog.c (validate_pattern): Likewise.
10577 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
10578 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
10579 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
10580 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
10581 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
10582 * rtlanal.c (subreg_get_info): Likewise.
10583 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10584 (vect_grouped_load_supported): Likewise.
10585 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
10586 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
10587 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
10588 (simplify_const_unary_operation, simplify_binary_operation_1)
10589 (simplify_const_binary_operation, simplify_ternary_operation)
10590 (test_vector_ops_duplicate, test_vector_ops): Likewise.
10591 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
10592 instead of CONST_VECTOR_NUNITS.
10593 * varasm.c (output_constant_pool_2): Likewise.
10594 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
10595 explicit-encoded elements in the XVEC for variable-length vectors.
10597 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10599 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
10601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10602 Alan Hayward <alan.hayward@arm.com>
10603 David Sherwood <david.sherwood@arm.com>
10605 * coretypes.h (fixed_size_mode): Declare.
10606 (fixed_size_mode_pod): New typedef.
10607 * builtins.h (target_builtins::x_apply_args_mode)
10608 (target_builtins::x_apply_result_mode): Change type to
10609 fixed_size_mode_pod.
10610 * builtins.c (apply_args_size, apply_result_size, result_vector)
10611 (expand_builtin_apply_args_1, expand_builtin_apply)
10612 (expand_builtin_return): Update accordingly.
10614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10616 * cse.c (hash_rtx_cb): Hash only the encoded elements.
10617 * cselib.c (cselib_hash_rtx): Likewise.
10618 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
10619 CONST_VECTOR encoding.
10621 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10622 Jeff Law <law@redhat.com>
10624 PR target/83641
10625 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
10626 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
10627 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
10628 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
10630 PR target/83641
10631 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
10632 explicitly probe *sp in a noreturn function if there were any callee
10633 register saves or frame pointer is needed.
10635 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10637 PR debug/83621
10638 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
10639 BLKmode for ternary, binary or unary expressions.
10641 PR debug/83645
10642 * var-tracking.c (delete_vta_debug_insn): New inline function.
10643 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
10644 insns from get_insns () to NULL instead of each bb separately.
10645 Use delete_vta_debug_insn. No longer static.
10646 (vt_debug_insns_local, variable_tracking_main_1): Adjust
10647 delete_vta_debug_insns callers.
10648 * rtl.h (delete_vta_debug_insns): Declare.
10649 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
10650 instead of variable_tracking_main.
10652 2018-01-03 Martin Sebor <msebor@redhat.com>
10654 PR tree-optimization/83603
10655 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
10656 arguments past the endof the argument list in functions declared
10657 without a prototype.
10658 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
10659 Avoid checking when arguments are null.
10661 2018-01-03 Martin Sebor <msebor@redhat.com>
10663 PR c/83559
10664 * doc/extend.texi (attribute const): Fix a typo.
10665 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
10666 issuing -Wsuggest-attribute for void functions.
10668 2018-01-03 Martin Sebor <msebor@redhat.com>
10670 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
10671 offset_int::from instead of wide_int::to_shwi.
10672 (maybe_diag_overlap): Remove assertion.
10673 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
10674 * gimple-ssa-sprintf.c (format_directive): Same.
10675 (parse_directive): Same.
10676 (sprintf_dom_walker::compute_format_length): Same.
10677 (try_substitute_return_value): Same.
10679 2018-01-03 Jeff Law <law@redhat.com>
10681 PR middle-end/83654
10682 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
10683 non-constant residual for zero at runtime and avoid probing in
10684 that case. Reorganize code for trailing problem to mirror handling
10685 of the residual.
10687 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10689 PR tree-optimization/83501
10690 * tree-ssa-strlen.c (get_string_cst): New.
10691 (handle_char_store): Call get_string_cst.
10693 2018-01-03 Martin Liska <mliska@suse.cz>
10695 PR tree-optimization/83593
10696 * tree-ssa-strlen.c: Include tree-cfg.h.
10697 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
10698 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
10699 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
10700 to false.
10701 (strlen_dom_walker::before_dom_children): Call
10702 gimple_purge_dead_eh_edges. Dump tranformation with details
10703 dump flags.
10704 (strlen_dom_walker::before_dom_children): Update call by adding
10705 new argument cleanup_eh.
10706 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
10708 2018-01-03 Martin Liska <mliska@suse.cz>
10710 PR ipa/83549
10711 * cif-code.def (VARIADIC_THUNK): New enum value.
10712 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
10713 thunks.
10715 2018-01-03 Jan Beulich <jbeulich@suse.com>
10717 * sse.md (mov<mode>_internal): Tighten condition for when to use
10718 vmovdqu<ssescalarsize> for TI and OI modes.
10720 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10722 Update copyright years.
10724 2018-01-03 Martin Liska <mliska@suse.cz>
10726 PR ipa/83594
10727 * ipa-visibility.c (function_and_variable_visibility): Skip
10728 functions with noipa attribure.
10730 2018-01-03 Jakub Jelinek <jakub@redhat.com>
10732 * gcc.c (process_command): Update copyright notice dates.
10733 * gcov-dump.c (print_version): Ditto.
10734 * gcov.c (print_version): Ditto.
10735 * gcov-tool.c (print_version): Ditto.
10736 * gengtype.c (create_file): Ditto.
10737 * doc/cpp.texi: Bump @copying's copyright year.
10738 * doc/cppinternals.texi: Ditto.
10739 * doc/gcc.texi: Ditto.
10740 * doc/gccint.texi: Ditto.
10741 * doc/gcov.texi: Ditto.
10742 * doc/install.texi: Ditto.
10743 * doc/invoke.texi: Ditto.
10745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10747 * vector-builder.h (vector_builder::m_full_nelts): Change from
10748 unsigned int to poly_uint64.
10749 (vector_builder::full_nelts): Update prototype accordingly.
10750 (vector_builder::new_vector): Likewise.
10751 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
10752 (vector_builder::operator ==): Likewise.
10753 (vector_builder::finalize): Likewise.
10754 * int-vector-builder.h (int_vector_builder::int_vector_builder):
10755 Take the number of elements as a poly_uint64 rather than an
10756 unsigned int.
10757 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
10758 from unsigned int to poly_uint64.
10759 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
10760 (vec_perm_indices::new_vector): Likewise.
10761 (vec_perm_indices::length): Likewise.
10762 (vec_perm_indices::nelts_per_input): Likewise.
10763 (vec_perm_indices::input_nelts): Likewise.
10764 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
10765 number of elements per input as a poly_uint64 rather than an
10766 unsigned int. Use the original encoding for variable-length
10767 vectors, rather than clamping each individual element.
10768 For the second and subsequent elements in each pattern,
10769 clamp the step and base before clamping their sum.
10770 (vec_perm_indices::series_p): Handle polynomial element counts.
10771 (vec_perm_indices::all_in_range_p): Likewise.
10772 (vec_perm_indices_to_tree): Likewise.
10773 (vec_perm_indices_to_rtx): Likewise.
10774 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
10775 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
10776 (tree_vector_builder::new_binary_operation): Handle polynomial
10777 element counts. Return false if we need to know the number
10778 of elements at compile time.
10779 * fold-const.c (fold_vec_perm): Punt if the number of elements
10780 isn't known at compile time.
10782 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10784 * vec-perm-indices.h (vec_perm_builder): Change element type
10785 from HOST_WIDE_INT to poly_int64.
10786 (vec_perm_indices::element_type): Update accordingly.
10787 (vec_perm_indices::clamp): Handle polynomial element_types.
10788 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10789 (vec_perm_indices::all_in_range_p): Likewise.
10790 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
10791 than shwi trees.
10792 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
10793 polynomial vec_perm_indices element types.
10794 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
10795 * fold-const.c (fold_vec_perm): Likewise.
10796 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
10797 * tree-vect-generic.c (lower_vec_perm): Likewise.
10798 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10799 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
10800 element type to HOST_WIDE_INT.
10802 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10803 Alan Hayward <alan.hayward@arm.com>
10804 David Sherwood <david.sherwood@arm.com>
10806 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
10807 rather than an int. Use plus_constant.
10808 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
10809 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
10811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10812 Alan Hayward <alan.hayward@arm.com>
10813 David Sherwood <david.sherwood@arm.com>
10815 * calls.c (emit_call_1, expand_call): Change struct_value_size from
10816 a HOST_WIDE_INT to a poly_int64.
10818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10819 Alan Hayward <alan.hayward@arm.com>
10820 David Sherwood <david.sherwood@arm.com>
10822 * calls.c (load_register_parameters): Cope with polynomial
10823 mode sizes. Require a constant size for BLKmode parameters
10824 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
10825 forces a parameter to be padded at the lsb end in order to
10826 fill a complete number of words, require the parameter size
10827 to be ordered wrt UNITS_PER_WORD.
10829 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10830 Alan Hayward <alan.hayward@arm.com>
10831 David Sherwood <david.sherwood@arm.com>
10833 * reload1.c (spill_stack_slot_width): Change element type
10834 from unsigned int to poly_uint64_pod.
10835 (alter_reg): Treat mode sizes as polynomial.
10837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10838 Alan Hayward <alan.hayward@arm.com>
10839 David Sherwood <david.sherwood@arm.com>
10841 * reload.c (complex_word_subreg_p): New function.
10842 (reload_inner_reg_of_subreg, push_reload): Use it.
10844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10845 Alan Hayward <alan.hayward@arm.com>
10846 David Sherwood <david.sherwood@arm.com>
10848 * lra-constraints.c (process_alt_operands): Reject matched
10849 operands whose sizes aren't ordered.
10850 (match_reload): Refer to this check here.
10852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10853 Alan Hayward <alan.hayward@arm.com>
10854 David Sherwood <david.sherwood@arm.com>
10856 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10857 that the mode size is in the set {1, 2, 4, 8, 16}.
10859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10860 Alan Hayward <alan.hayward@arm.com>
10861 David Sherwood <david.sherwood@arm.com>
10863 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10864 Use plus_constant instead of gen_rtx_PLUS.
10866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10867 Alan Hayward <alan.hayward@arm.com>
10868 David Sherwood <david.sherwood@arm.com>
10870 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10871 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10872 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10873 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10874 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10875 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10876 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10877 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10878 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10879 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10880 a poly_int64.
10881 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10882 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10883 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10884 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10885 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10886 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10887 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10888 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10889 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10890 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10891 function.
10892 * expr.c (emit_move_resolve_push): Treat the input and result
10893 of PUSH_ROUNDING as a poly_int64.
10894 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10895 (emit_push_insn): Likewise.
10896 * lra-eliminations.c (mark_not_eliminable): Likewise.
10897 * recog.c (push_operand): Likewise.
10898 * reload1.c (elimination_effects): Likewise.
10899 * rtlanal.c (nonzero_bits1): Likewise.
10900 * calls.c (store_one_arg): Likewise. Require the padding to be
10901 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10908 Use plus_constant instead of gen_rtx_PLUS.
10910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10911 Alan Hayward <alan.hayward@arm.com>
10912 David Sherwood <david.sherwood@arm.com>
10914 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10915 rather than an int.
10917 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10918 Alan Hayward <alan.hayward@arm.com>
10919 David Sherwood <david.sherwood@arm.com>
10921 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10922 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10923 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
10930 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10931 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10932 as a poly_uint64 too.
10934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10935 Alan Hayward <alan.hayward@arm.com>
10936 David Sherwood <david.sherwood@arm.com>
10938 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10941 Alan Hayward <alan.hayward@arm.com>
10942 David Sherwood <david.sherwood@arm.com>
10944 * combine.c (can_change_dest_mode): Handle polynomial
10945 REGMODE_NATURAL_SIZE.
10946 * expmed.c (store_bit_field_1): Likewise.
10947 * expr.c (store_constructor): Likewise.
10948 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10949 and polynomial REGMODE_NATURAL_SIZE.
10950 (gen_lowpart_common): Likewise.
10951 * reginfo.c (record_subregs_of_mode): Likewise.
10952 * rtlanal.c (read_modify_subreg_p): Likewise.
10954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10955 Alan Hayward <alan.hayward@arm.com>
10956 David Sherwood <david.sherwood@arm.com>
10958 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10959 numbers of elements.
10961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10962 Alan Hayward <alan.hayward@arm.com>
10963 David Sherwood <david.sherwood@arm.com>
10965 * match.pd: Cope with polynomial numbers of vector elements.
10967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10968 Alan Hayward <alan.hayward@arm.com>
10969 David Sherwood <david.sherwood@arm.com>
10971 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10972 in a POINTER_PLUS_EXPR.
10974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10975 Alan Hayward <alan.hayward@arm.com>
10976 David Sherwood <david.sherwood@arm.com>
10978 * omp-simd-clone.c (simd_clone_subparts): New function.
10979 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10980 (ipa_simd_modify_function_body): Likewise.
10982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10983 Alan Hayward <alan.hayward@arm.com>
10984 David Sherwood <david.sherwood@arm.com>
10986 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10987 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10988 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10989 (expand_vector_condition, vector_element): Likewise.
10990 (subparts_gt): New function.
10991 (get_compute_type): Use subparts_gt.
10992 (count_type_subparts): Delete.
10993 (expand_vector_operations_1): Use subparts_gt instead of
10994 count_type_subparts.
10996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10997 Alan Hayward <alan.hayward@arm.com>
10998 David Sherwood <david.sherwood@arm.com>
11000 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
11001 (vect_compile_time_alias): ...this new function. Do the calculation
11002 on poly_ints rather than trees.
11003 (vect_prune_runtime_alias_test_list): Update call accordingly.
11005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11006 Alan Hayward <alan.hayward@arm.com>
11007 David Sherwood <david.sherwood@arm.com>
11009 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
11010 numbers of units.
11011 (vect_schedule_slp_instance): 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 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
11018 constant and extern definitions for variable-length vectors.
11019 (vect_get_constant_vectors): Note that the number of units
11020 is known to be constant.
11022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11023 Alan Hayward <alan.hayward@arm.com>
11024 David Sherwood <david.sherwood@arm.com>
11026 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
11027 of units as polynomial. Choose between WIDE and NARROW based
11028 on multiple_p.
11030 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11031 Alan Hayward <alan.hayward@arm.com>
11032 David Sherwood <david.sherwood@arm.com>
11034 * tree-vect-stmts.c (simd_clone_subparts): New function.
11035 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
11037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11038 Alan Hayward <alan.hayward@arm.com>
11039 David Sherwood <david.sherwood@arm.com>
11041 * tree-vect-stmts.c (vectorizable_call): Treat the number of
11042 vectors as polynomial. Use build_index_vector for
11043 IFN_GOMP_SIMD_LANE.
11045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11046 Alan Hayward <alan.hayward@arm.com>
11047 David Sherwood <david.sherwood@arm.com>
11049 * tree-vect-stmts.c (get_load_store_type): Treat the number of
11050 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
11051 for variable-length vectors.
11052 (vectorizable_mask_load_store): Treat the number of units as
11053 polynomial, asserting that it is constant if the condition has
11054 already been enforced.
11055 (vectorizable_store, vectorizable_load): Likewise.
11057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11058 Alan Hayward <alan.hayward@arm.com>
11059 David Sherwood <david.sherwood@arm.com>
11061 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
11062 of units as polynomial. Punt if we can't tell at compile time
11063 which vector contains the final result.
11065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11066 Alan Hayward <alan.hayward@arm.com>
11067 David Sherwood <david.sherwood@arm.com>
11069 * tree-vect-loop.c (vectorizable_induction): Treat the number
11070 of units as polynomial. Punt on SLP inductions. Use an integer
11071 VEC_SERIES_EXPR for variable-length integer reductions. Use a
11072 cast of such a series for variable-length floating-point
11073 reductions.
11075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11076 Alan Hayward <alan.hayward@arm.com>
11077 David Sherwood <david.sherwood@arm.com>
11079 * tree.h (build_index_vector): Declare.
11080 * tree.c (build_index_vector): New function.
11081 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
11082 of units as polynomial, forcibly converting it to a constant if
11083 vectorizable_reduction has already enforced the condition.
11084 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
11085 to create a {1,2,3,...} vector.
11086 (vectorizable_reduction): Treat the number of units as polynomial.
11087 Choose vectype_in based on the largest scalar element size rather
11088 than the smallest number of units. Enforce the restrictions
11089 relied on above.
11091 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11092 Alan Hayward <alan.hayward@arm.com>
11093 David Sherwood <david.sherwood@arm.com>
11095 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
11096 number of units as polynomial.
11098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11099 Alan Hayward <alan.hayward@arm.com>
11100 David Sherwood <david.sherwood@arm.com>
11102 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
11103 * target.def (autovectorize_vector_sizes): Return the vector sizes
11104 by pointer, using vector_sizes rather than a bitmask.
11105 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
11106 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
11107 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
11108 Likewise.
11109 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
11110 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
11111 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
11112 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
11113 * omp-general.c (omp_max_vf): Likewise.
11114 * omp-low.c (omp_clause_aligned_alignment): Likewise.
11115 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
11116 * tree-vect-loop.c (vect_analyze_loop): Likewise.
11117 * tree-vect-slp.c (vect_slp_bb): Likewise.
11118 * doc/tm.texi: Regenerate.
11119 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
11120 to a poly_uint64.
11121 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
11122 the vector size as a poly_uint64 rather than an unsigned int.
11123 (current_vector_size): Change from an unsigned int to a poly_uint64.
11124 (get_vectype_for_scalar_type): Update accordingly.
11125 * tree.h (build_truth_vector_type): Take the size and number of
11126 units as a poly_uint64 rather than an unsigned int.
11127 (build_vector_type): Add a temporary overload that takes
11128 the number of units as a poly_uint64 rather than an unsigned int.
11129 * tree.c (make_vector_type): Likewise.
11130 (build_truth_vector_type): Take the number of units as a poly_uint64
11131 rather than an unsigned int.
11133 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11134 Alan Hayward <alan.hayward@arm.com>
11135 David Sherwood <david.sherwood@arm.com>
11137 * target.def (get_mask_mode): Take the number of units and length
11138 as poly_uint64s rather than unsigned ints.
11139 * targhooks.h (default_get_mask_mode): Update accordingly.
11140 * targhooks.c (default_get_mask_mode): Likewise.
11141 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
11142 * doc/tm.texi: Regenerate.
11144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11145 Alan Hayward <alan.hayward@arm.com>
11146 David Sherwood <david.sherwood@arm.com>
11148 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
11149 * omp-general.c (omp_max_vf): Likewise.
11150 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
11151 (expand_omp_simd): Handle polynomial safelen.
11152 * omp-low.c (omplow_simd_context): Add a default constructor.
11153 (omplow_simd_context::max_vf): Change from int to poly_uint64.
11154 (lower_rec_simd_input_clauses): Update accordingly.
11155 (lower_rec_input_clauses): Likewise.
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 * tree-vectorizer.h (vect_nunits_for_cost): New function.
11162 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
11163 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
11164 (vect_analyze_slp_cost): Likewise.
11165 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
11166 (vect_model_load_cost): Likewise.
11168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11169 Alan Hayward <alan.hayward@arm.com>
11170 David Sherwood <david.sherwood@arm.com>
11172 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
11173 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
11174 from an unsigned int * to a poly_uint64_pod *.
11175 (calculate_unrolling_factor): New function.
11176 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
11178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11179 Alan Hayward <alan.hayward@arm.com>
11180 David Sherwood <david.sherwood@arm.com>
11182 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
11183 from an unsigned int to a poly_uint64.
11184 (_loop_vec_info::slp_unrolling_factor): Likewise.
11185 (_loop_vec_info::vectorization_factor): Change from an int
11186 to a poly_uint64.
11187 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
11188 (vect_get_num_vectors): New function.
11189 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
11190 (vect_get_num_copies): Use vect_get_num_vectors.
11191 (vect_analyze_data_ref_dependences): Change max_vf from an int *
11192 to an unsigned int *.
11193 (vect_analyze_data_refs): Change min_vf from an int * to a
11194 poly_uint64 *.
11195 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11196 than an unsigned HOST_WIDE_INT.
11197 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11198 (vect_analyze_data_ref_dependence): Change max_vf from an int *
11199 to an unsigned int *.
11200 (vect_analyze_data_ref_dependences): Likewise.
11201 (vect_compute_data_ref_alignment): Handle polynomial vf.
11202 (vect_enhance_data_refs_alignment): Likewise.
11203 (vect_prune_runtime_alias_test_list): Likewise.
11204 (vect_shift_permute_load_chain): Likewise.
11205 (vect_supportable_dr_alignment): Likewise.
11206 (dependence_distance_ge_vf): Take the vectorization factor as a
11207 poly_uint64 rather than an unsigned HOST_WIDE_INT.
11208 (vect_analyze_data_refs): Change min_vf from an int * to a
11209 poly_uint64 *.
11210 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
11211 vfm1 as a poly_uint64 rather than an int. Make the same change
11212 for the returned bound_scalar.
11213 (vect_gen_vector_loop_niters): Handle polynomial vf.
11214 (vect_do_peeling): Likewise. Update call to
11215 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
11216 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
11217 be constant.
11218 * tree-vect-loop.c (vect_determine_vectorization_factor)
11219 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
11220 (vect_get_known_peeling_cost): Likewise.
11221 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
11222 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
11223 (vect_transform_loop): Likewise. Use the lowest possible VF when
11224 updating the upper bounds of the loop.
11225 (vect_min_worthwhile_factor): Make static. Return an unsigned int
11226 rather than an int.
11227 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
11228 polynomial unroll factors.
11229 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
11230 (vect_make_slp_decision): Likewise.
11231 (vect_supported_load_permutation_p): Likewise, and polynomial
11232 vf too.
11233 (vect_analyze_slp_cost): Handle polynomial vf.
11234 (vect_slp_analyze_node_operations): Likewise.
11235 (vect_slp_analyze_bb_1): Likewise.
11236 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
11237 than an unsigned HOST_WIDE_INT.
11238 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
11239 (vectorizable_load): Handle polynomial vf.
11240 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
11241 a poly_uint64.
11242 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
11244 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11245 Alan Hayward <alan.hayward@arm.com>
11246 David Sherwood <david.sherwood@arm.com>
11248 * match.pd: Handle bit operations involving three constants
11249 and try to fold one pair.
11251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
11253 * tree-vect-loop-manip.c: Include gimple-fold.h.
11254 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
11255 niters_maybe_zero parameters. Handle other cases besides a step of 1.
11256 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
11257 Add a path that uses a step of VF instead of 1, but disable it
11258 for now.
11259 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
11260 and niters_no_overflow parameters. Update calls to
11261 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
11262 Create a new SSA name if the latter choses to use a ste other
11263 than zero, and return it via niters_vector_mult_vf_var.
11264 * tree-vect-loop.c (vect_transform_loop): Update calls to
11265 vect_do_peeling, vect_gen_vector_loop_niters and
11266 slpeel_make_loop_iterate_ntimes.
11267 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
11268 (vect_gen_vector_loop_niters): Update declarations after above changes.
11270 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
11272 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
11273 128-bit round to integer instructions.
11274 (ceil<mode>2): Likewise.
11275 (btrunc<mode>2): Likewise.
11276 (round<mode>2): Likewise.
11278 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11280 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
11281 unaligned VSX load/store on P8/P9.
11282 (expand_block_clear): Allow the use of unaligned VSX
11283 load/store on P8/P9.
11285 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11287 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
11288 New function.
11289 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
11290 swap associated with both a load and a store.
11292 2018-01-02 Andrew Waterman <andrew@sifive.com>
11294 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
11295 * config/riscv/riscv.md (clear_cache): Use it.
11297 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
11299 * web.c: Remove out-of-date comment.
11301 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11303 * expr.c (fixup_args_size_notes): Check that any existing
11304 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
11305 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
11306 (emit_single_push_insn): ...here.
11308 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11310 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
11311 (const_vector_encoded_nelts): New function.
11312 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
11313 (const_vector_int_elt, const_vector_elt): Declare.
11314 * emit-rtl.c (const_vector_int_elt_1): New function.
11315 (const_vector_elt): Likewise.
11316 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
11317 of CONST_VECTOR_ELT.
11319 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11321 * expr.c: Include rtx-vector-builder.h.
11322 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
11323 directly on the tree encoding.
11324 (const_vector_from_tree): Likewise.
11325 * optabs.c: Include rtx-vector-builder.h.
11326 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
11327 sequence of "u" values.
11328 * vec-perm-indices.c: Include rtx-vector-builder.h.
11329 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
11330 directly on the vec_perm_indices encoding.
11332 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11334 * doc/rtl.texi (const_vector): Describe new encoding scheme.
11335 * Makefile.in (OBJS): Add rtx-vector-builder.o.
11336 * rtx-vector-builder.h: New file.
11337 * rtx-vector-builder.c: Likewise.
11338 * rtl.h (rtx_def::u2): Add a const_vector field.
11339 (CONST_VECTOR_NPATTERNS): New macro.
11340 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
11341 (CONST_VECTOR_DUPLICATE_P): Likewise.
11342 (CONST_VECTOR_STEPPED_P): Likewise.
11343 (CONST_VECTOR_ENCODED_ELT): Likewise.
11344 (const_vec_duplicate_p): Check for a duplicated vector encoding.
11345 (unwrap_const_vec_duplicate): Likewise.
11346 (const_vec_series_p): Check for a non-duplicated vector encoding.
11347 Say that the function only returns true for integer vectors.
11348 * emit-rtl.c: Include rtx-vector-builder.h.
11349 (gen_const_vec_duplicate_1): Delete.
11350 (gen_const_vector): Call gen_const_vec_duplicate instead of
11351 gen_const_vec_duplicate_1.
11352 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
11353 (gen_const_vec_duplicate): Use rtx_vector_builder.
11354 (gen_const_vec_series): Likewise.
11355 (gen_rtx_CONST_VECTOR): Likewise.
11356 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
11357 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11358 Build a new vector rather than modifying a CONST_VECTOR in-place.
11359 (handle_special_swappables): Update call accordingly.
11360 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
11361 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
11362 Build a new vector rather than modifying a CONST_VECTOR in-place.
11363 (handle_special_swappables): Update call accordingly.
11365 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11367 * simplify-rtx.c (simplify_const_binary_operation): Use
11368 CONST_VECTOR_ELT instead of XVECEXP.
11370 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11372 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
11373 the selector elements to be different from the data elements
11374 if the selector is a VECTOR_CST.
11375 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
11376 ssizetype for the selector.
11378 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11380 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
11381 before testing each element individually.
11382 * tree-vect-generic.c (lower_vec_perm): Likewise.
11384 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11386 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
11387 * selftest-run-tests.c (selftest::run_tests): Call it.
11388 * vector-builder.h (vector_builder::operator ==): New function.
11389 (vector_builder::operator !=): Likewise.
11390 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
11391 (vec_perm_indices::all_from_input_p): New function.
11392 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
11393 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
11394 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
11395 instead of reading the VECTOR_CST directly. Detect whether both
11396 vector inputs are the same before constructing the vec_perm_indices,
11397 and update the number of inputs argument accordingly. Use the
11398 utility functions added above. Only construct sel2 if we need to.
11400 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11402 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
11403 the broadcast of the low byte.
11404 (expand_mult_highpart): Use an explicit encoding for the permutes.
11405 * optabs-query.c (can_mult_highpart_p): Likewise.
11406 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
11407 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11408 (vectorizable_bswap): Likewise.
11409 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
11410 explicit encoding for the power-of-2 permutes.
11411 (vect_permute_store_chain): Likewise.
11412 (vect_grouped_load_supported): Likewise.
11413 (vect_permute_load_chain): Likewise.
11415 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11417 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
11418 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
11419 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
11420 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
11421 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
11422 (vect_gen_perm_mask_any): Likewise.
11424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11426 * int-vector-builder.h: New file.
11427 * vec-perm-indices.h: Include int-vector-builder.h.
11428 (vec_perm_indices): Redefine as an int_vector_builder.
11429 (auto_vec_perm_indices): Delete.
11430 (vec_perm_builder): Redefine as a stand-alone class.
11431 (vec_perm_indices::vec_perm_indices): New function.
11432 (vec_perm_indices::clamp): Likewise.
11433 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
11434 (vec_perm_indices::new_vector): New function.
11435 (vec_perm_indices::new_expanded_vector): Update for new
11436 vec_perm_indices class.
11437 (vec_perm_indices::rotate_inputs): New function.
11438 (vec_perm_indices::all_in_range_p): Operate directly on the
11439 encoded form, without computing elided elements.
11440 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
11441 encoding. Update for new vec_perm_indices class.
11442 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
11443 the given vec_perm_builder.
11444 (expand_vec_perm_var): Update vec_perm_builder constructor.
11445 (expand_mult_highpart): Use vec_perm_builder instead of
11446 auto_vec_perm_indices.
11447 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
11448 vec_perm_indices instead of auto_vec_perm_indices. Use a single
11449 or double series encoding as appropriate.
11450 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
11451 vec_perm_indices instead of auto_vec_perm_indices.
11452 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11453 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11454 (vect_permute_store_chain): Likewise.
11455 (vect_grouped_load_supported): Likewise.
11456 (vect_permute_load_chain): Likewise.
11457 (vect_shift_permute_load_chain): Likewise.
11458 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11459 (vect_transform_slp_perm_load): Likewise.
11460 (vect_schedule_slp_instance): Likewise.
11461 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11462 (vectorizable_mask_load_store): Likewise.
11463 (vectorizable_bswap): Likewise.
11464 (vectorizable_store): Likewise.
11465 (vectorizable_load): Likewise.
11466 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
11467 vec_perm_indices instead of auto_vec_perm_indices. Use
11468 tree_to_vec_perm_builder to read the vector from a tree.
11469 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
11470 vec_perm_builder instead of a vec_perm_indices.
11471 (have_whole_vector_shift): Use vec_perm_builder and
11472 vec_perm_indices instead of auto_vec_perm_indices. Leave the
11473 truncation to calc_vec_perm_mask_for_shift.
11474 (vect_create_epilog_for_reduction): Likewise.
11475 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
11476 from auto_vec_perm_indices to vec_perm_indices.
11477 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11478 instead of changing individual elements.
11479 (aarch64_vectorize_vec_perm_const): Use new_vector to install
11480 the vector in d.perm.
11481 * config/arm/arm.c (expand_vec_perm_d::perm): Change
11482 from auto_vec_perm_indices to vec_perm_indices.
11483 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
11484 instead of changing individual elements.
11485 (arm_vectorize_vec_perm_const): Use new_vector to install
11486 the vector in d.perm.
11487 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
11488 Update vec_perm_builder constructor.
11489 (rs6000_expand_interleave): Likewise.
11490 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
11491 (rs6000_expand_interleave): Likewise.
11493 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11495 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
11496 to qimode could truncate the indices.
11497 * optabs.c (expand_vec_perm_var): Likewise.
11499 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11501 * Makefile.in (OBJS): Add vec-perm-indices.o.
11502 * vec-perm-indices.h: New file.
11503 * vec-perm-indices.c: Likewise.
11504 * target.h (vec_perm_indices): Replace with a forward class
11505 declaration.
11506 (auto_vec_perm_indices): Move to vec-perm-indices.h.
11507 * optabs.h: Include vec-perm-indices.h.
11508 (expand_vec_perm): Delete.
11509 (selector_fits_mode_p, expand_vec_perm_var): Declare.
11510 (expand_vec_perm_const): Declare.
11511 * target.def (vec_perm_const_ok): Replace with...
11512 (vec_perm_const): ...this new hook.
11513 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
11514 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
11515 * doc/tm.texi: Regenerate.
11516 * optabs.def (vec_perm_const): Delete.
11517 * doc/md.texi (vec_perm_const): Likewise.
11518 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
11519 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
11520 expand_vec_perm for constant permutation vectors. Assert that
11521 the mode of variable permutation vectors is the integer equivalent
11522 of the mode that is being permuted.
11523 * optabs-query.h (selector_fits_mode_p): Declare.
11524 * optabs-query.c: Include vec-perm-indices.h.
11525 (selector_fits_mode_p): New function.
11526 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
11527 is defined, instead of checking whether the vec_perm_const_optab
11528 exists. Use targetm.vectorize.vec_perm_const instead of
11529 targetm.vectorize.vec_perm_const_ok. Check whether the indices
11530 fit in the vector mode before using a variable permute.
11531 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
11532 vec_perm_indices instead of an rtx.
11533 (expand_vec_perm): Replace with...
11534 (expand_vec_perm_const): ...this new function. Take the selector
11535 as a vec_perm_indices rather than an rtx. Also take the mode of
11536 the selector. Update call to shift_amt_for_vec_perm_mask.
11537 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
11538 Use vec_perm_indices::new_expanded_vector to expand the original
11539 selector into bytes. Check whether the indices fit in the vector
11540 mode before using a variable permute.
11541 (expand_vec_perm_var): Make global.
11542 (expand_mult_highpart): Use expand_vec_perm_const.
11543 * fold-const.c: Includes vec-perm-indices.h.
11544 * tree-ssa-forwprop.c: Likewise.
11545 * tree-vect-data-refs.c: Likewise.
11546 * tree-vect-generic.c: Likewise.
11547 * tree-vect-loop.c: Likewise.
11548 * tree-vect-slp.c: Likewise.
11549 * tree-vect-stmts.c: Likewise.
11550 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
11551 Delete.
11552 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
11553 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
11554 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
11555 (aarch64_vectorize_vec_perm_const): ...this new function.
11556 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11557 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11558 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
11559 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
11560 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11561 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11562 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
11563 into...
11564 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
11565 check for NEON modes.
11566 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
11567 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
11568 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
11569 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
11570 into...
11571 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
11572 the old VEC_PERM_CONST conditions.
11573 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
11574 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
11575 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
11576 (ia64_vectorize_vec_perm_const_ok): Merge into...
11577 (ia64_vectorize_vec_perm_const): ...this new function.
11578 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
11579 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
11580 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
11581 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
11582 * config/mips/mips.c (mips_expand_vec_perm_const)
11583 (mips_vectorize_vec_perm_const_ok): Merge into...
11584 (mips_vectorize_vec_perm_const): ...this new function.
11585 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
11586 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
11587 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
11588 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
11589 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
11590 (rs6000_expand_vec_perm_const): Delete.
11591 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
11592 Delete.
11593 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11594 (altivec_expand_vec_perm_const_le): Take each operand individually.
11595 Operate on constant selectors rather than rtxes.
11596 (altivec_expand_vec_perm_const): Likewise. Update call to
11597 altivec_expand_vec_perm_const_le.
11598 (rs6000_expand_vec_perm_const): Delete.
11599 (rs6000_vectorize_vec_perm_const_ok): Delete.
11600 (rs6000_vectorize_vec_perm_const): New function.
11601 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11602 an element count and rtx array.
11603 (rs6000_expand_extract_even): Update call accordingly.
11604 (rs6000_expand_interleave): Likewise.
11605 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
11606 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
11607 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
11608 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
11609 (rs6000_expand_vec_perm_const): Delete.
11610 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
11611 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11612 (altivec_expand_vec_perm_const_le): Take each operand individually.
11613 Operate on constant selectors rather than rtxes.
11614 (altivec_expand_vec_perm_const): Likewise. Update call to
11615 altivec_expand_vec_perm_const_le.
11616 (rs6000_expand_vec_perm_const): Delete.
11617 (rs6000_vectorize_vec_perm_const_ok): Delete.
11618 (rs6000_vectorize_vec_perm_const): New function. Remove stray
11619 reference to the SPE evmerge intructions.
11620 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
11621 an element count and rtx array.
11622 (rs6000_expand_extract_even): Update call accordingly.
11623 (rs6000_expand_interleave): Likewise.
11624 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
11625 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
11626 new function.
11627 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
11629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11631 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
11632 vector mode and that that mode matches the mode of the data
11633 being permuted.
11634 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
11635 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
11636 directly using expand_vec_perm_1 when forcing selectors into
11637 registers.
11638 (expand_vec_perm_var): New function, split out from expand_vec_perm.
11640 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11642 * optabs-query.h (can_vec_perm_p): Delete.
11643 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
11644 * optabs-query.c (can_vec_perm_p): Split into...
11645 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
11646 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
11647 particular selector is valid.
11648 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
11649 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
11650 (vect_grouped_load_supported): Likewise.
11651 (vect_shift_permute_load_chain): Likewise.
11652 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
11653 (vect_transform_slp_perm_load): Likewise.
11654 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
11655 (vectorizable_bswap): Likewise.
11656 (vect_gen_perm_mask_checked): Likewise.
11657 * fold-const.c (fold_ternary_loc): Likewise. Don't take
11658 implementations of variable permutation vectors into account
11659 when deciding which selector to use.
11660 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
11661 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
11662 with a false third argument.
11663 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
11664 to test whether the constant selector is valid and can_vec_perm_var_p
11665 to test whether a variable selector is valid.
11667 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11669 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
11670 * optabs-query.c (can_vec_perm_p): Likewise.
11671 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
11672 instead of vec_perm_indices.
11673 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
11674 (vect_gen_perm_mask_checked): Likewise,
11675 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
11676 (vect_gen_perm_mask_checked): Likewise,
11678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
11680 * optabs-query.h (qimode_for_vec_perm): Declare.
11681 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
11682 (qimode_for_vec_perm): ...this new function.
11683 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
11685 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11687 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
11688 does not have a conditional at the top.
11690 2018-01-02 Richard Biener <rguenther@suse.de>
11692 * ipa-inline.c (big_speedup_p): Fix expression.
11694 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11696 PR target/81616
11697 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
11698 for generic 4->6.
11700 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
11702 PR target/81616
11703 Generic tuning.
11704 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
11705 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
11706 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
11707 cond_taken_branch_cost 3->4.
11709 2018-01-01 Jakub Jelinek <jakub@redhat.com>
11711 PR tree-optimization/83581
11712 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
11713 TODO_cleanup_cfg if any changes have been made.
11715 PR middle-end/83608
11716 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
11717 convert_modes if target mode has the right side, but different mode
11718 class.
11720 PR middle-end/83609
11721 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
11722 last argument when extracting from CONCAT. If either from_real or
11723 from_imag is NULL, use expansion through memory. If result is not
11724 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
11725 the parts directly to inner mode, if even that fails, use expansion
11726 through memory.
11728 PR middle-end/83623
11729 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
11730 check for bswap in mode rather than HImode and use that in expand_unop
11731 too.
11732 \f
11733 Copyright (C) 2018 Free Software Foundation, Inc.
11735 Copying and distribution of this file, with or without modification,
11736 are permitted in any medium without royalty provided the copyright
11737 notice and this notice are preserved.