1 2018-07-17 Jeff Law <law@redhat.com>
3 PR tree-optimization/86010
4 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
6 * config/mips/mips.c (vr4130_align_insns): Update for recent
7 changes to label_to_alignment.
9 * config/frv/frv.c (frv_label_align): Update for recent changes
10 to label_to_alignment.
12 * config/nios2/nios2.c (nios2_label_align): Update for recent
13 changes which dropped ALIGN_LABELS_LOG.
15 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
17 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
18 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
20 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
22 * config/arc/arc.c (arc_label_align): Use align_labels instead of
23 deprecated align_labels_log.
25 2018-07-17 Richard Biener <rguenther@suse.de>
28 * dwarf2out.c (init_sections_and_labels): Always generate
29 a debug_line_str_section for early LTO debug.
30 (dwarf2out_finish): Reset debug_line_str_hash output early.
31 Bump counter for extra dwarf5 .debug_loc labels to not conflict
33 (dwarf2out_early_finish): Output debug_line_str.
35 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
37 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
38 index register on z196 or later.
40 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
42 * config/s390/s390.c (s390_default_align): Set default function
44 (s390_override_options_after_change): Call s390_default align.
45 (s390_option_override_internal): Call s390_default_align.
46 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
48 2018-07-17 Jakub Jelinek <jakub@redhat.com>
51 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
52 to _looptemp_ clauses, other than the first two.
54 2018-07-17 Martin Liska <mliska@suse.cz>
56 * opts.c: Do not enable OPT_falign_* for -Os.
58 2018-07-17 Martin Liska <mliska@suse.cz>
60 * align.h (MAX_CODE_ALIGN): New.
61 (MAX_CODE_ALIGN_VALUE): New.
62 * common/config/i386/i386-common.c (ix86_handle_option):
63 (MAX_CODE_ALIGN): Moved to align.h.
64 * final.c (MAX_CODE_ALIGN): Likewise.
65 * opts.c (parse_and_check_align_values):
66 (MAX_CODE_ALIGN): Likewise.
67 (MAX_CODE_ALIGN_VALUE): Likewise.
69 2018-07-17 Martin Liska <mliska@suse.cz>
71 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
72 in order to fulfil coding style.
73 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
74 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
75 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
76 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
77 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
78 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
79 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
80 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
81 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
83 2018-07-17 Martin Liska <mliska@suse.cz>
86 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
87 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
88 align_flags of label_to_alignment.
89 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
91 * config/m68k/m68k.c: Do not use removed align_labels_value and
93 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
94 (LOOP_ALIGN): Likewise.
95 (LABEL_ALIGN): Likewise.
96 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
97 Remove not used macro.
98 (rs6000_loop_align): Change return type to align_flags.
99 (rs6000_loop_align_max_skip): Remove.
100 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
101 Change return type to align_flags.
102 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
103 Remove not used macro.
104 (rs6000_loop_align): Change return type to align_flags.
105 (rs6000_loop_align_max_skip): Remove.
106 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
107 * config/rx/rx-protos.h (rx_align_for_label): Make it
109 * config/rx/rx.c (rx_align_for_label): Change return type
111 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
113 into align_flags class.
114 (LABEL_ALIGN): Likewise.
115 (LOOP_ALIGN): Likewise.
116 * config/s390/s390.c (s390_label_align): Use align_flags
118 (s390_asm_output_function_label): Likewise.
119 * config/sh/sh.c (sh_override_options_after_change):
120 Use align_flags class directly without macros.
121 (find_barrier): Likewise.
122 (barrier_align): Likewise.
123 (sh_loop_align): Likewise.
124 * config/spu/spu.c (spu_option_override):
125 Use align_flags_tuple::get_value instead of removed macros.
126 (spu_sched_init): Likewise.
127 * config/spu/spu.h (GTY): Likewise.
128 * config/visium/visium.c (visium_option_override):
129 Set "8" as default secondary alignment.
130 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
131 in order to guarantee secondary alignment of 8.
132 * coretypes.h: Include align.h header file.
133 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
134 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
135 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
136 * doc/tm.texi.in: Likewise.
137 * final.c (struct label_alignment): Remove not used structure.
138 (LABEL_ALIGN): Change type to align_flags.
139 (LOOP_ALIGN): Likewise.
140 (JUMP_ALIGN): Likewise.
141 (default_loop_align_max_skip): Remove.
142 (default_label_align_max_skip): Likewise.
143 (default_jump_align_max_skip): Likewise.
144 (default_label_align_after_barrier_max_skip):
145 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
146 (LABEL_TO_MAX_SKIP): Remove.
147 (label_to_alignment): Return align_flags type instead of integer.
148 (label_to_max_skip): Remove.
149 (align_fuzz): Use align_flags type.
150 (compute_alignments): Use align_flags type and use align_flags::max
151 to combine multiple alignments.
152 (grow_label_align): Grow vec instead of C array.
153 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
154 (shorten_branches): Use align_flags type and use align_flags::max
155 to combine multiple alignments.
156 (final_scan_insn_1): Remove usage of secondary alignment that comes
157 from label alignment, but instead use proper secondary alignment
158 which is computed in grow_label_align.
159 * flags.h (struct align_flags_tuple): Move to align.h.
160 (struct align_flags): Likewise.
161 (state_align_loops): Rename to align_loops.
162 (state_align_jumps): Rename to align_jumps.
163 (state_align_labels): Rename to align_labels.
164 (state_align_functions): Rename to align_functions.
165 (align_loops_log): Remove.
166 (align_jumps_log): Remove.
167 (align_labels_log): Remove.
168 (align_functions_log): Remove.
169 (align_loops_max_skip): Remove.
170 (align_jumps_max_skip): Remove.
171 (align_labels_max_skip): Remove.
172 (align_functions_max_skip): Remove.
173 (align_loops_value): Remove.
174 (align_jumps_value): Remove.
175 (align_labels_value): Remove.
176 (align_functions_value): Remove.
177 * output.h (label_to_alignment): Change return type to align_flags.
178 (label_to_max_skip): Remove.
179 * target.def: Remove loop_align_max_skip, label_align_max_skip,
180 jump_align_max_skip macros.
181 * targhooks.h (default_loop_align_max_skip): Remove.
182 (default_label_align_max_skip): Likewise.
183 (default_jump_align_max_skip): Likewise.
184 (default_label_align_after_barrier_max_skip): Remove.
185 * toplev.c (read_log_maxskip): Use ::normalize function.
186 (parse_N_M): Remove not used argument and also call ::normalize.
187 (parse_alignment_opts): Do not pass unused arguments.
188 * varasm.c (assemble_start_function): Use directly align_functions
189 instead of removed macros.
190 * system.h: Do not poison removed macros.
192 2018-07-17 Jakub Jelinek <jakub@redhat.com>
195 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
196 and cond temporaries don't have reference type if iterator has
197 pointer type. For init use &for_pre_body instead of pre_p if
198 for_pre_body is non-empty.
200 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
202 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
203 double-double modes to SFmode directly directly.
204 (trunc<mode>sf2_fprs): Delete.
206 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
208 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
209 for conversions between IFmode and the decimal floating point modes.
210 (init_float128_ieee): Use the correct names for conversions between
211 KFmode and the decimal floating point modes.
213 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
215 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
216 for the conversions between TDmode and IFmode.
217 (init_float128_ieee): Use more correct names for the conversions
218 between TDmode and KFmode.
220 2018-07-16 Jakub Jelinek <jakub@redhat.com>
222 PR tree-optimization/86526
223 * builtins.c (expand_builtin_memcmp): Formatting fixes.
224 (inline_expand_builtin_string_cmp): Likewise.
225 (inline_string_cmp): Likewise. Use c_readstr instead of
226 builtin_memcpy_read_str. Add unit_mode temporary.
228 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
231 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
232 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
234 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
236 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
237 Alphabetize prototypes of built-in functions, separating out
238 built-in functions that are listed in this section but should be
241 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
244 * expmed.c (emit_store_flag): Do not emit setcc followed by a
245 conditional move when trapping comparison was split to a
246 non-trapping one (and vice versa).
248 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
250 * config/s390/s390.c (s390_function_profiler): Generate nops
251 instead of profiler call sequences.
252 * config/s390/s390.opt: Add the new option.
254 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
256 * config/s390/s390.c (s390_function_profiler): Generate
257 __mcount_loc section.
258 * config/s390/s390.opt: Add the new option.
260 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
262 * common.opt: Add the new warning.
263 * config/s390/s390.c (s390_function_profiler): Emit "brasl
264 %r0,__fentry__" when -mfentry is specified.
265 (s390_option_override_internal): Disallow -mfentry for 31-bit
267 * config/s390/s390.opt: Add the new option.
269 2018-07-16 Richard Biener <rguenther@suse.de>
272 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
273 for function-local FUNCTION_DECL and RESULT_DECL immediately.
275 2018-07-16 Martin Liska <mliska@suse.cz>
278 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
281 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
283 * config/arc/arcHS.md: Update ARCHS scheduling rules.
285 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
287 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
289 * config/arc/arc-cpus.def (hs4x): New cpu.
291 * config/arc/arc-tables.opt: Regenerate.
292 * config/arc/arc.c (arc_sched_issue_rate): New function.
293 (TARGET_SCHED_ISSUE_RATE): Define.
294 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
295 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
297 (attr tune): Add ARCHS4x tune values.
298 (attr tune_dspmpy): Define.
299 (*tst): Correct instruction type.
300 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
301 * config/arc/arcHS4x.md: New file.
302 * config/arc/fpu.md: Update instruction type attributes.
303 * config/arc/t-multilib: Regenerate.
305 2018-07-16 Tom de Vries <tdevries@suse.de>
308 * var-tracking.c (vt_initialize): Fix pre_dec handling.
310 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
312 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
315 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
317 PR tree-optimization/86514
318 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
319 conversion to a boolean type from a type with greater precision.
321 2018-07-16 Tom de Vries <tdevries@suse.de>
323 * var-tracking.c (vt_initialize): Print adjusted insn slim if
324 dump_flags request TDF_SLIM.
326 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
328 * fold-const.c (int_const_binop_1): Abstract...
329 (wide_int_binop): ...wide int code here.
330 (poly_int_binop): ...poly int code here.
331 Abstract the rest of int_const_binop_1 into int_const_binop.
332 * fold-const.h (wide_int_binop): New.
333 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
334 Remove useless PLUS/MINUS_EXPR case.
335 (zero_nonzero_bits_from_vr): Move wide int code...
336 (zero_nonzero_bits_from_bounds): ...here.
337 (extract_range_from_binary_expr_1): Move mask optimization code...
338 (range_easy_mask_min_max): ...here.
339 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
340 (range_easy_mask_min_max): New.
342 2018-07-15 Jeff Law <law@redhat.com>
345 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
348 2018-07-14 Jim Wilson <jimw@sifive.com>
350 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
352 2018-07-14 Paul Koning <ni1d@arrl.net>
354 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
356 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
358 * lto-streamer-out.c (copy_function_or_variable): Dump info about
361 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
362 Steve Munroe <munroesj52@gmail.com>
364 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
365 (_mm_andnot_si128): Likewise.
366 (_mm_or_si128): Likewise.
367 (_mm_xor_si128): Likewise.
369 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
372 * builtins.c (expand_builtin_memcmp): Inline the calls first
373 when result_eq is false.
374 (expand_builtin_strcmp): Inline the calls first.
375 (expand_builtin_strncmp): Likewise.
376 (inline_string_cmp): New routine. Expand a string compare
377 call by using a sequence of char comparison.
378 (inline_expand_builtin_string_cmp): New routine. Inline expansion
379 a call to str(n)cmp/memcmp.
380 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
382 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
384 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
386 * config/arm/driver-arm.c: Include arm-native.h.
387 (host_detect_local_cpu): Use auto-generated data tables.
388 (vendors, arm_cpu_table): Delete. Move part information to ...
389 * config/arm/arm-cpus.in: ... here.
390 * config/arm/parsecpu.awk (gen_native): New function.
391 (vendor, part): New CPU fields.
392 (END): Add support for building the native CPU detection tables.
393 * config/arm/t-arm (arm-native.h): Add build rule.
394 (driver-arm.o): Add dependency on arm-native.h.
396 2018-07-13 Richard Biener <rguenther@suse.de>
399 * match.pd (addr1 - addr2): Allow either of the operand to
402 2018-07-13 Tom de Vries <tdevries@suse.de>
404 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
407 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
409 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
410 arrays instead of numbered variables.
412 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
414 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
415 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
417 2018-07-13 Richard Biener <rguenther@suse.de>
420 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
421 instead of get_context_die.
423 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
424 Richard Biener <rguenther@suse.de>
427 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
428 that the loop latch destination where phi is defined.
430 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
432 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
433 (riscv_expand_epilogue): Add assertion to check interrupt mode.
434 (riscv_set_current_function): Extract getting interrupt type to new
436 (riscv_get_interrupt_type): New function.
437 (riscv_merge_decl_attributes): New function, checking interrupt type is
439 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
441 2018-07-12 Paul Koning <ni1d@arrl.net>
443 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
446 2018-07-12 Paul Koning <ni1d@arrl.net>
448 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
449 zero reference, add doloop_end instead.
450 * doc/md.texi (decrement_and_branch_until_zero): Remove.
451 (Looping patterns): Remove decrement_and_branch_until_zero. Add
452 detail for doloop_end.
454 2018-07-12 Martin Sebor <msebor@redhat.com>
457 * attribs.c (decl_attributes): Reject conflicting attributes before
458 calling attribute handlers.
460 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
462 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
464 (gcc::dump_manager::get_dump_file_name): likewise.
465 (dump_begin): Likewise.
466 * dumpfile.h (dump_begin): Update prototype.
467 (gcc::dump_manager::get_dump_file_name,
468 gcc::dump_manager::get_dump_file_name): Update prototype.
470 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
472 * internal-fn.h (vectorizable_internal_fn_p): New function.
473 * tree-vect-slp.c (compatible_calls_p): Likewise.
474 (vect_build_slp_tree_1): Remove nops argument. Handle calls
475 to internal functions.
476 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
478 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
480 * fold-const.h (inverse_conditions_p): Declare.
481 * fold-const.c (inverse_conditions_p): New function.
482 * match.pd: Use inverse_conditions_p. Add folds of view_converts
483 that test the inverse condition of a conditional internal function.
484 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
485 * internal-fn.c (internal_fn_mask_index): Handle conditional
487 (vectorized_internal_fn_supported_p): New function.
488 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
489 (any_pred_load_store): Replace with...
490 (need_to_predicate): ...this new variable.
491 (redundant_ssa_names): New variable.
492 (ifcvt_can_use_mask_load_store): Move initial checks to...
493 (ifcvt_can_predicate): ...this new function. Handle tree codes
494 for which a conditional internal function exists.
495 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
496 instead of ifcvt_can_use_mask_load_store. Update after variable
498 (predicate_load_or_store): New function, split out from
499 predicate_mem_writes.
500 (check_redundant_cond_expr): New function.
501 (value_available_p): Likewise.
502 (predicate_rhs_code): Likewise.
503 (predicate_mem_writes): Rename to...
504 (predicate_statements): ...this. Use predicate_load_or_store
505 and predicate_rhs_code.
506 (combine_blocks, tree_if_conversion): Update after above name changes.
507 (ifcvt_local_dce): Handle redundant_ssa_names.
508 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
509 general conditional functions.
510 * tree-vect-stmts.c (vectorizable_call): Likewise.
512 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
513 Alan Hayward <alan.hayward@arm.com>
514 David Sherwood <david.sherwood@arm.com>
516 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
517 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
518 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
519 plus and minus and convert them into IFN_COND_FMA-based sequences.
520 (convert_mult_to_fma): Handle conditional plus and minus.
522 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
524 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
525 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
526 (cond_fnms_optab): New optabs.
527 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
529 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
530 * internal-fn.h (get_conditional_internal_fn): Declare.
531 (get_unconditional_internal_fn): Likewise.
532 * internal-fn.c (cond_ternary_direct): New macro.
533 (expand_cond_ternary_optab_fn): Likewise.
534 (direct_cond_ternary_optab_supported_p): Likewise.
535 (FOR_EACH_COND_FN_PAIR): Likewise.
536 (get_conditional_internal_fn): New function.
537 (get_unconditional_internal_fn): Likewise.
538 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
539 (gimple_match_op::gimple_match_op): Add a new overload for 5
541 (gimple_match_op::set_op): Likewise.
542 (gimple_resimplify5): Declare.
543 * genmatch.c (decision_tree::gen): Generate simplifications for
545 * gimple-match-head.c (gimple_simplify): Define an overload for
546 5 operands. Handle calls with 5 arguments in the top-level overload.
547 (convert_conditional_op): Handle conversions from unconditional
548 internal functions to conditional ones.
549 (gimple_resimplify5): New function.
550 (build_call_internal): Pass a fifth operand.
551 (maybe_push_res_to_seq): Likewise.
552 (try_conditional_simplification): Try converting conditional
553 internal functions to unconditional internal functions.
554 Handle 3-operand unconditional forms.
555 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
556 Define ternary equivalents of the current rules for binary conditional
558 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
560 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
561 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
562 (optab): Handle them.
563 (SVE_COND_FP_TERNARY): New int iterator.
564 (sve_fmla_op, sve_fmad_op): New int attributes.
565 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
566 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
567 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
569 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
571 * target.def (preferred_else_value): New target hook.
572 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
573 * doc/tm.texi: Regenerate.
574 * targhooks.h (default_preferred_else_value): Declare.
575 * targhooks.c (default_preferred_else_value): New function.
576 * internal-fn.h (conditional_internal_fn_code): Declare.
577 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
578 (get_conditional_internal_fn): Use it.
579 (conditional_internal_fn_code): New function.
580 * gimple-match.h (gimple_match_cond): New struct.
581 (gimple_match_op): Add a cond member function.
582 (gimple_match_op::gimple_match_op): Update all forms to take a
584 * genmatch.c (expr::gen_transform): Use the same condition as res_op
585 for the suboperation, but don't specify a particular else_value.
586 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
587 (visit_nary_op, visit_reference_op_load): Pass
588 gimple_match_cond::UNCOND to the gimple_match_op constructor.
589 * gimple-match-head.c: Include tree-eh.h
590 (convert_conditional_op): New function.
591 (maybe_resimplify_conditional_op): Likewise.
592 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
593 (gimple_resimplify2): Likewise.
594 (gimple_resimplify3): Likewise.
595 (gimple_resimplify4): Likewise.
596 (maybe_push_res_to_seq): Return null for conditional operations.
597 (try_conditional_simplification): New function.
598 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
600 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
602 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
604 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
606 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
608 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
610 (hash_tree): Do not hash DECL_FCONTEXT
611 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
612 Do not stream DECL_FCONTEXT.
613 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
614 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
616 2018-07-12 Richard Biener <rguenther@suse.de>
619 * dwarf2out.c (gen_block_die): Only output blocks when they have
620 at least one !DECL_IGNORED_P variable.
622 2018-07-12 Richard Biener <rguenther@suse.de>
625 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
626 Remove -mieee-fp handling.
628 2018-07-12 Richard Biener <rguenther@suse.de>
630 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
631 left-over from last patch.
633 2018-07-12 Jakub Jelinek <jakub@redhat.com>
635 PR tree-optimization/86492
636 * gimple-ssa-store-merging.c
637 (imm_store_chain_info::coalesce_immediate_stores): Call
638 check_no_overlap even for the merge_overlapping case. Formatting fix.
640 2018-07-12 Richard Biener <rguenther@suse.de>
643 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
644 move possibly trapping operations into the conditional.
646 2018-07-12 Richard Biener <rguenther@suse.de>
648 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
649 (vn_lookup_simplify_result): Remove recursion limit applied
651 (vn_nary_build_or_lookup_1): Adjust.
652 (try_to_simplify): Likewise.
653 * gimple-match-head.c (gimple_resimplify1): Instead apply one
655 (gimple_resimplify2): Likewise.
656 (gimple_resimplify3): Likewise.
657 (gimple_resimplify4): Likewise.
659 2018-07-11 Jakub Jelinek <jakub@redhat.com>
661 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
662 Use __mmask64 type instead of __mmask8 for __M argument.
663 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
664 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
666 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
667 __mmask16 for __M argument.
668 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
669 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
670 to __mmask16 instead of __mmask8.
671 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
672 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
673 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
674 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
675 instead of __mmask16 for __U argument.
676 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
677 __mmask16 instead of __mmask8 for __U argument.
678 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
680 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
682 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
684 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
686 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
688 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
690 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
692 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
694 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
696 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
698 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
699 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
700 return type as well as __M argument type and all casts from __mmask8
702 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
703 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
704 return type as well as __M argument type and all casts from __mmask8
706 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
707 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
708 return type as well as __M argument type and all casts from __mmask8
710 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
711 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
712 return type as well as __M argument type and all casts from __mmask8
714 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
715 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
718 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
720 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
721 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
724 2018-07-11 Paul Koning <ni1d@arrl.net>
726 * doc/md.texi (define_subst): Document how multiple occurrences of
727 the same argument in the replacement pattern are handled.
729 2018-07-11 Paul Koning <ni1d@arrl.net>
731 * doc/extend.texi (Common Variable Attributes): Move "mode" into
733 (Common Type Attributes): Add "mode" attribute.
735 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
737 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
738 stream DECL_ORIGINAL_TYPE.
739 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
740 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
741 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
742 Do not walk original type.
743 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
745 (write_ts_decl_non_common_tree_pointers): Do not stream
747 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
748 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
750 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
752 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
753 thread twice from the same starting edge.
755 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
757 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
758 * gimple.c (gimple_call_nonnull_result_p): ...here...
759 (gimple_call_nonnull_arg): ...and here.
760 * gimple.h (gimple_call_nonnull_result_p): New.
761 (gimple_call_nonnull_arg): New.
763 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
765 * config/arm/arm-cpus.in: Move information from fpu field of each
766 cpu definition to the isa field.
767 * config/arm/parsecpu.awk (fpu): Delete match rule.
768 (gen_comm_data): Don't add bits from the CPU's FPU entry.
770 2018-07-11 Richard Biener <rguenther@suse.de>
773 * dwarf2out.c (init_sections_and_labels): Use
774 output_asm_line_debug_info consistently.
775 (dwarf2out_early_finish): Likewise.
776 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
779 2018-07-11 Richard Biener <rguenther@suse.de>
781 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
782 Rework father_bb setting in a way to avoid propagating constants
783 multiple times on a loop body.
785 2018-07-10 Mark Wielaard <mark@klomp.org>
788 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
790 2018-07-10 Richard Biener <rguenther@suse.de>
792 * hash-map.h (hash_map::iterator::operator*): Return
793 references to key and value.
795 2018-07-10 Jakub Jelinek <jakub@redhat.com>
798 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
799 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
800 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
801 (gimplify_omp_for): For composite loops, move outer
802 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
803 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
804 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
805 TREE_LIST for both the original class iterator and the "last" helper
806 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
807 loop, remember has_decl_expr from outer composite loops for the
808 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
810 2018-07-09 Martin Sebor <msebor@redhat.com>
814 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
815 accessing implicitly initialized array elements.
816 * expr.c (string_constant): Handle string initializers of
817 character arrays within aggregates.
818 * gimple-fold.c (fold_array_ctor_reference): Add argument.
819 Store element offset. As a special case, handle zero size.
820 (fold_nonarray_ctor_reference): Same.
821 (fold_ctor_reference): Add argument. Store subobject offset.
822 * gimple-fold.h (fold_ctor_reference): Add argument.
824 2018-07-09 Paul Koning <ni1d@arrl.net>
826 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
827 (pdp11_insn_cost): New function.
828 (pdp11_md_asm_adjust): New function.
829 (TARGET_INVALID_WITHIN_DOLOOP): Define.
830 (pdp11_rtx_costs): Update to match machine better.
831 (output_addr_const_pdp11): Correct format mismatch warnings.
832 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
833 * config/pdp11/pdp11.md: General change to add base_cost and/or
834 length attributes for use by new pdp11_insn_cost function.
835 (MIN_BRANCH): Correct definition.
837 (doloop_end): Use standard pattern name for looping pattern.
838 (doloop_end_nocc): New.
839 (movsf): Add another constraint alternative.
840 (zero_extendqihi2): Add constraint alternatives for not in place
842 (zero_extendhisi2): Remove.
843 (shift patterns): Add CC handling variants.
847 (define_peephole2): New peephole to recognize mov that sets CC for
850 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
852 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
853 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
856 2018-07-09 Jakub Jelinek <jakub@redhat.com>
859 * real.c (real_nextafter): Return true if result is denormal.
861 2018-07-09 Martin Liska <mliska@suse.cz>
863 * common.opt: Add back wrongly removed attribute.
865 2018-07-09 Richard Biener <rguenther@suse.de>
868 * dwarf2out.c (gen_block_die): For an early generated DIE
869 always output high/low PC attributes.
871 2018-07-09 Tom de Vries <tdevries@suse.de>
873 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
874 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
875 onto VAR_DECL with abstract origin.
877 2018-07-07 Jim Wilson <jimw@sifive.com>
879 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
881 2018-07-07 Tom de Vries <tdevries@suse.de>
883 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
885 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
887 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
889 (combine_bound): Use wide-int overflow calculation instead of
891 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
893 * fold-const.c (int_const_binop_2): Same.
894 (extract_muldiv_1): Same.
895 (fold_div_compare): Same.
896 (fold_abs_const): Same.
898 * poly-int.h (add): Same.
902 * predict.c (predict_iv_comparison): Same.
903 * profile-count.c (slow_safe_scale_64bit): Same.
904 * simplify-rtx.c (simplify_const_binary_operation): Same.
905 * tree-chrec.c (tree_fold_binomial): Same.
906 * tree-data-ref.c (split_constant_offset_1): Same.
907 * tree-if-conv.c (idx_within_array_bound): Same.
908 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
909 * tree-ssa-phiopt.c (minmax_replacement): Same.
910 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
911 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
912 * vr-values.c (vr_values::adjust_range_with_scev): Same.
913 * wide-int.cc (wi::add_large): Same.
914 (wi::mul_internal): Same.
915 (wi::sub_large): Same.
916 (wi::divmod_internal): Same.
917 * wide-int.h: Change overflow type to overflow_type for neg, add,
918 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
919 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
920 mul_internal, divmod_internal.
921 (overflow_type): New enum.
922 (accumulate_overflow): New.
924 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
926 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
927 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
929 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
931 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
932 argument is checked for zero before entering loop, avoid checking again.
934 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
936 * gimplify.h (generic_expr_could_trap_p): Set as global function.
937 * gimplify.h (generic_expr_could_trap_p): Likwise.
938 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
940 2018-07-06 Jakub Jelinek <jakub@redhat.com>
942 PR tree-optimization/86401
943 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
944 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
945 (fold_bit_and_mask): ... here. New helper function for match.pd.
946 * fold-const.h (fold_bit_and_mask): Declare.
947 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
949 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
952 * target.def (translate_mode_attribute): New hook.
953 * targhooks.h (default_translate_mode_attribute): Declare.
954 * targhooks.c (default_translate_mode_attribute): New function.
955 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
956 * doc/tm.texi: Regenerate.
957 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
958 (rs6000_translate_mode_attribute): New function.
960 2018-07-06 Paul Koning <ni1d@arrl.net>
962 * doc/md.texi (define_split): Document DONE and FAIL.
963 (define_peephole2): Ditto.
965 2018-07-05 Jeff Law <law@redhat.com>
967 PR tree-optimization/86010
968 * tree-ssa-dse.c (compute_trims): More aggressively trim at
969 both the head and tail of mem* and str* calls.
971 2018-07-05 Jim Wilson <jimw@sifive.com>
973 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
975 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
977 * config/aarch64/aarch64-simd.md: correct flags text for
978 MIN_EXPR replacement.
980 2018-07-05 James Clarke <jrtc27@jrtc27.com>
982 * configure: Regenerated.
984 2018-07-05 Carl Love <cel@us.ibm.com>
986 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
987 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
988 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
989 VSX_BUILTIN_DOUBLEL_V4SF.
991 2018-07-05 Martin Sebor <msebor@redhat.com>
994 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
995 than its domain to compute its the upper bound of a char array.
997 2018-07-05 Nathan Sidwell <nathan@acm.org>
999 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
1000 * doc/cpp.texi: Update comment.
1001 * doc/tm.texi: Rebuilt.
1002 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
1003 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
1004 * doc/extend.texi (Backwards Compatibility): Clarify it is system
1005 headers affected by extern "C".
1006 * system.h: Poison NO_IMPLICIT_EXTERN_C.
1007 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
1008 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
1009 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
1010 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
1011 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
1012 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
1013 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
1014 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
1015 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
1016 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
1017 config/sparc/sp64-elf.h, config/spu/spu.h,
1018 config/stormy16/stormy16.h, config/v850/v850.h,
1019 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1020 define NO_IMPLICIT_EXTERN_C.
1021 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1023 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1026 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1027 instead of GET_MODE_SIZE when comparing Units.
1029 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1032 * rtlanal.c (set_noop_p): Constrain on mode change,
1033 include hard-reg-set.h
1035 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1037 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1039 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1042 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1045 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1047 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1049 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1052 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1054 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1056 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1059 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1061 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1063 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1064 neg_*_op* variables.
1066 2018-07-04 Martin Liska <mliska@suse.cz>
1068 * tree-switch-conversion.c: Define
1069 max_ratio_for_speed and max_ratio_for_size constants.
1071 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1072 Martin Liska <mliska@suse.cz>
1077 * common.opt: Rename align options with 'str_' prefix.
1078 * common/config/i386/i386-common.c (set_malign_value): New
1080 (ix86_handle_option): Use it to set -falign-* options/
1081 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1082 type from int to string.
1083 * config/aarch64/aarch64.c: Update default values from int
1085 * config/alpha/alpha.c (alpha_override_options_after_change):
1087 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1088 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1089 max skip conditionally.
1090 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1091 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1092 max skip conditionally.
1093 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1094 max skip conditionally.
1095 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1096 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1097 max skip conditionally.
1098 * config/i386/i386.c (struct ptt): Change type from int to
1100 (ix86_default_align): Set default values.
1101 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1102 max skip conditionally.
1103 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1104 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1105 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1106 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1107 max skip conditionally.
1108 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1109 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1110 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1111 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1112 max skip conditionally.
1113 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1114 * config/ia64/ia64.c (ia64_option_override): Set default values
1115 for alignment options.
1116 * config/m68k/m68k.c: Handle new str_align_* options.
1117 * config/mips/mips.c (mips_set_compression_mode): Change
1119 (mips_option_override): Set default values for options.
1120 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1122 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1124 * config/rx/rx.c (rx_option_override): Likewise.
1125 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1126 (LABEL_ALIGN): Use align_labels_log.
1127 (LOOP_ALIGN): Use align_loops_align.
1128 * config/s390/s390.c (s390_asm_output_function_label): Use new
1130 * config/sh/sh.c (sh_override_options_after_change):
1131 Change type of constants.
1132 * config/spu/spu.c (spu_sched_init): Likewise.
1133 * config/sparc/sparc.c (sparc_option_override): Set default
1135 * config/visium/visium.c (visium_option_override): Likewise.
1136 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1137 emit p2align format with last argument if it's not needed.
1138 * doc/invoke.texi: Document extended format of -falign-*.
1139 * final.c: Use align_labels alignment.
1140 * flags.h (struct target_flag_state): Change type to use
1142 (struct align_flags_tuple): New.
1143 (struct align_flags): Likewise.
1144 (align_loops_log): Redefine macro to use new types.
1145 (align_loops_max_skip): Redefine macro to use new types.
1146 (align_jumps_log): Redefine macro to use new types.
1147 (align_jumps_max_skip): Redefine macro to use new types.
1148 (align_labels_log): Redefine macro to use new types.
1149 (align_labels_max_skip): Redefine macro to use new types.
1150 (align_functions_log): Redefine macro to use new types.
1151 (align_loops): Redefine macro to use new types.
1152 (align_jumps): Redefine macro to use new types.
1153 (align_labels): Redefine macro to use new types.
1154 (align_functions): Redefine macro to use new types.
1155 (align_functions_max_skip): Redefine macro to use new types.
1156 (align_loops_value): New macro.
1157 (align_jumps_value): New macro.
1158 (align_labels_value): New macro.
1159 (align_functions_value): New macro.
1160 * function.c (invoke_set_current_function_hook): Propagate
1161 alignment values from flags to global variables default in
1163 * ipa-icf.c (sem_function::equals_wpa): Use
1164 cl_optimization_option_eq instead of memcmp.
1165 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1167 (cl_optimization_stream_in): Likewise.
1168 * optc-save-gen.awk: Support strings in cl_optimization.
1169 * opth-gen.awk: Likewise.
1170 * opts.c (finish_options): Remove error checking of invalid
1172 (MAX_CODE_ALIGN): Remove.
1173 (MAX_CODE_ALIGN_VALUE): Likewise.
1174 (parse_and_check_align_values): New function.
1175 (check_alignment_argument): Likewise.
1176 (common_handle_option): Use check_alignment_argument.
1177 * opts.h (parse_and_check_align_values): Declare.
1178 * toplev.c (init_alignments): Remove.
1179 (read_log_maxskip): New.
1180 (parse_N_M): Likewise.
1181 (parse_alignment_opts): Likewise.
1182 (backend_init_target): Remove usage of init_alignments.
1183 * toplev.h (parse_alignment_opts): Declare.
1184 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1186 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1187 * tree.c (cl_option_hasher::equal): New.
1188 * varasm.c: Use new global macros.
1190 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1192 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1193 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1194 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1195 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1196 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1197 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1198 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1199 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1200 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1201 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1202 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1203 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1204 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1206 2018-07-04 Martin Liska <mliska@suse.cz>
1207 Jonathan Wakely <jwakely@redhat.com>
1209 * coverage.c: Use correct type.
1210 * doc/invoke.texi: Language correction.
1212 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1215 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1216 ENDBRANCH for non-tail call which may return via indirect branch.
1217 * doc/extend.texi: Document indirect_return attribute.
1219 2018-07-03 Martin Sebor <msebor@redhat.com>
1221 PR tree-optimization/86274
1222 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1224 (format_floating): Correct handling of infinities and NaNs.
1226 2018-07-03 Martin Sebor <msebor@redhat.com>
1228 * print-tree.c (print_real_cst): New function.
1229 (print_node_brief): Call it.
1230 (print_node): Ditto.
1232 2018-07-03 Jeff Law <law@redhat.com>
1234 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1235 into a single pattern.
1237 * config/h8300/h8300.md (ors code_iterator): New.
1238 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1239 a single pattern and single splitter.
1240 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1241 (iorqi3_1, xorqi3_1): Likewise.
1242 (iorqi3, xorqi3 expanders): Similarly.
1244 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1245 (movmd_internal) into a single pattern using the P mode iterator.
1246 (movmd splitters): Similarly.
1247 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1248 (movsd splitters): Similarly.
1250 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1251 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1252 right table based on the size of the operand.
1253 * config/h8300/h8300.md (length_table): Corresponding changes. All
1254 references to "addb", "addw" and "addl" changed to "add".
1255 (btst patterns): Merge two variants into a single pattern.
1256 (tstqi, tsthi): Likewise.
1257 (addhi3_incdec, addsi3_incdec): Likewise.
1258 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1259 (mulhi3, mulsi3): Likewise.
1260 (udivhi3, udivsi3): Likewise.
1261 (divhi3, divsi3): Likewise.
1262 (andorqi3, andorhi3, andorsi3): Likewise.
1264 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1267 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1268 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1270 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1272 PR tree-optimization/85694
1273 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1274 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1276 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1277 (<u>avg<mode>3_ceil): New patterns.
1279 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1281 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1282 scan-tree-dump directive.
1283 * gcc.dg/vect/slp-perm-2.c: Likewise.
1284 * gcc.dg/vect/slp-perm-3.c: Likewise.
1285 * gcc.dg/vect/slp-perm-5.c: Likewise.
1286 * gcc.dg/vect/slp-perm-6.c: Likewise.
1287 * gcc.dg/vect/slp-perm-7.c: Likewise.
1288 * gcc.dg/vect/slp-perm-8.c: Likewise.
1290 2018-07-03 Marek Polacek <polacek@redhat.com>
1293 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1295 2018-07-03 Richard Biener <rguenther@suse.de>
1298 * tree-ssa-structalias.c (find_func_clobbers): Properly
1299 handle indirect calls.
1301 2018-07-03 Jeff Law <law@redhat.com>
1303 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1304 (shifts): New code iterator.
1305 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1306 expander. Fix HImode handling on H8/SX.
1307 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1308 (subqi3, subhi3, subsi3 expanders): Likewise.
1309 (andqi3, andhi3, andsi3 expanders): Likewise.
1310 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1311 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1312 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1313 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1314 (zero_extendqihi2, zero_extendqisi2): Likewise.
1315 (extendqihi2, extendqisi2): Likewise.
1316 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1317 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1318 (rotlqi3_1, rotlhi3_1): Likewise.
1319 (logicalhi3_sn, logicalsi3_sn): Likewise.
1320 (logicalhi3, logicalsi3): Likewise.
1322 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1324 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1325 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1326 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1327 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1328 type to append_pattern_def_seq instead of creating a stmt_vec_info
1330 (build_mask_conversion): Likewise. Remove vinfo argument.
1331 (vect_add_conversion_to_patterm): Likewise, renaming to...
1332 (vect_add_conversion_to_pattern): ...this.
1333 (vect_recog_mask_conversion_pattern): Update call to
1334 build_mask_conversion. Pass the vector type to
1335 append_pattern_def_seq here too.
1336 (vect_recog_gather_scatter_pattern): Update call to
1337 vect_add_conversion_to_pattern.
1339 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1341 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1342 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1343 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1344 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1345 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1346 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1347 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1348 append_pattern_def_seq instead of new_pattern_def_seq.
1349 (vect_recog_divmod_pattern): Do both of the above.
1350 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1353 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1355 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1356 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1357 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1358 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1359 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1360 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1361 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1362 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1363 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1364 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1365 parameter with a single stmt_vec_info.
1366 (vect_recog_func_ptr): Likewise.
1367 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1368 (vect_try_gather_scatter_pattern): ...this.
1369 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1370 the stmt_vec_info of the statement to be matched. Don't clear
1371 STMT_VINFO_RELATED_STMT.
1372 (vect_pattern_recog): Update call accordingly.
1374 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1376 PR tree-optimization/85694
1377 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1378 (uavgM3_ceil): Document new optabs.
1379 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1380 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1382 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1383 (savg_ceil_optab): New optabs.
1384 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1385 (vect_vect_recog_func_ptrs): Add it.
1386 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1387 constant directly from the associated lhs.
1389 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1391 * tree-vect-patterns.c (vect_split_statement): New function.
1392 (vect_convert_input): Use it to try to split an existing cast.
1394 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1396 * poly-int.h (print_hex): New function.
1397 * dumpfile.h (dump_dec, dump_hex): Declare.
1398 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1399 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1400 min_input_precision, operation_precision and operation_sign.
1401 * tree-vect-patterns.c (vect_get_range_info): New function.
1402 (vect_same_loop_or_bb_p, vect_single_imm_use)
1403 (vect_operation_fits_smaller_type): Delete.
1404 (vect_look_through_possible_promotion): Add an optional
1405 single_use_p parameter.
1406 (vect_recog_over_widening_pattern): Rewrite to use new
1407 stmt_vec_info infomration. Handle one operation at a time.
1408 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1409 (vect_truncatable_operation_p, vect_set_operation_type)
1410 (vect_set_min_input_precision): New functions.
1411 (vect_determine_min_output_precision_1): Likewise.
1412 (vect_determine_min_output_precision): Likewise.
1413 (vect_determine_precisions_from_range): Likewise.
1414 (vect_determine_precisions_from_users): Likewise.
1415 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1416 (vect_vect_recog_func_ptrs): Put over_widening first.
1418 (vect_pattern_recog): Call vect_determine_precisions.
1420 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1422 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1423 statements that have been replaced by further pattern statements.
1424 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1426 2018-07-03 Richard Biener <rguenther@suse.de>
1428 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1429 always set *dt. Dump vectype in vectype overload.
1430 * dumpfile.h (dump_gimple_expr): New function.
1431 (dump_gimple_expr_loc): Likewise.
1432 * dumpfile.c (dump_gimple_expr): New function.
1433 (dump_gimple_expr_loc): Likewise.
1435 2018-07-02 Jeff Law <law@redhat.com>
1437 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1438 the H8/300, H8/300H and H8/S variants into a single pattern.
1439 (movhi_h8300, movqi_h8300hs): Similarly.
1440 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1441 (QHI mode iterator): New.
1443 * config/h8300/h8300.md: Remove trailing whitespace.
1445 2018-07-02 Jim Wilson <jimw@sifive.com>
1447 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1448 instead of emit_insn for interrupt returns.
1449 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1450 (riscv_sret, riscv_uret): Likewise.
1452 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1454 * pretty-print.c (selftest::test_pp_format): Move save and restore
1455 of quotes to class auto_fix_quotes, and add an instance.
1456 * selftest.c: Include "intl.h".
1457 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1458 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1459 * selftest.h (selftest::auto_fix_quotes): New class.
1461 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1463 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1464 (aarch64_sve_prepare_conditional_op): Remove.
1465 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1466 Allow aarch64_simd_reg_or_zero as select operand; remove
1467 the aarch64_sve_prepare_conditional_op call.
1468 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1469 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1470 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1471 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1472 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1473 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1474 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1475 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1476 and a splitters to match all of the *_any patterns.
1477 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1479 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1480 (SVE_COND_FP_BINARY_REV): Remove.
1481 (sve_int_op_rev, sve_fp_op_rev): New.
1482 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1483 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1484 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1485 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1486 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1487 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1488 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1489 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1490 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1492 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1493 Remove match_dup 1 from the inner unspec.
1494 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1496 * config/aarch64/aarch64.md (movprfx): New attr.
1497 (length): Default movprfx to 8.
1498 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1499 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1500 (*<su>mul<SVE_I>3_highpart): Likewise.
1501 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1502 (*v<ASHIFT><SVE_I>3): Likewise.
1503 (*<su><MAXMIN><SVE_I>3): Likewise.
1504 (*<su><MAXMIN><SVE_F>3): Likewise.
1505 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1506 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1507 (*div<SVE_F>4): Likewise.
1509 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1511 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1514 2018-07-02 Richard Biener <rguenther@suse.de>
1516 PR tree-optimization/86363
1517 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1518 memset argument refers to a non-variable address.
1520 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1522 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1523 {PLUS,MINUS}_EXPR code to...
1524 (adjust_symbolic_bound): ...here,
1525 (combine_bound): ...here,
1526 (set_value_range_with_overflow): ...and here.
1528 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1530 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1532 (extract_range_from_abs_expr): ...here.
1534 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1536 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1537 -fno-omit-frame-pointer when not optimizing.
1539 2018-07-02 Martin Liska <mliska@suse.cz>
1542 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1543 (propagate_nothrow): Likewise.
1545 2018-07-02 Martin Liska <mliska@suse.cz>
1548 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1550 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1552 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1553 function in r262149, changing "loc" param from source_location to
1554 const dump_location_t &.
1555 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1556 declaration, as above.
1558 2018-07-01 Paul Koning <ni1d@arrl.net>
1560 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1561 -munit-asm, -mgnu-asm, -mdec-asm.
1562 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1563 (pdp11_output_labelref): New.
1564 (pdp11_output_def): New.
1565 (pdp11_output_addr_vec_elt): New.
1566 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1567 %# and %@ format codes.
1568 (pdp11_option_override): New.
1569 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1570 (pdp11_output_ident): New.
1571 (pdp11_asm_named_section): New.
1572 (pdp11_asm_init_sections): New.
1573 (pdp11_file_start): New.
1574 (pdp11_file_end): New.
1575 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1576 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1577 %o, like %c but octal.
1578 (pdp11_option_override): New.
1579 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1581 (DATA_SECTION_ASM_OP): Ditto.
1582 (READONLY_DATA_SECTION_ASM_OP): New.
1583 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1584 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1585 (ASM_OUTPUT_LABELREF): Ditto.
1586 (ASM_OUTPUT_DEF): Ditto.
1587 (ASM_OUTPUT_EXTERNAL): New.
1588 (ASM_OUTPUT_SOURCE_FILENAME): New.
1589 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1590 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1591 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1592 %# and %@ format codes.
1593 * config/pdp11/pdp11.opt (mgnu-asm): New.
1594 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1595 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1596 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1598 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1600 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1601 dereferencing path[] beyond its length.
1603 (debug_all_paths): New.
1604 (rewire_first_differing_edge): New.
1605 (adjust_paths_after_duplication): New.
1606 (duplicate_thread_path): Call adjust_paths_after_duplication.
1608 (thread_through_all_blocks): Add new argument to
1609 duplicate_thread_path.
1611 2018-06-30 Jim Wilson <jimw@sifive.com>
1613 * config/riscv/predicates.md (p2m1_shift_operand): New.
1614 (high_mask_shift_operand): New.
1615 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1616 pattern using p2m1_shift_operand.
1617 (lshsi3_zero_extend_3+2): New combiner pattern using
1618 high_mask_shift_operand.
1620 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1622 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1624 (vect_recog_rotate_pattern): ...here.
1625 (vect_convert_input): Try to insert casts of invariants in the
1627 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1628 preheader to be empty.
1630 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1632 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1633 vector type. If given, install it in the new statement's
1635 (vect_element_precision): New function.
1636 (vect_unpromoted_value): New struct.
1637 (vect_unpromoted_value::vect_unpromoted_value): New function.
1638 (vect_unpromoted_value::set_op): Likewise.
1639 (vect_look_through_possible_promotion): Likewise.
1640 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1641 (vect_widened_op_tree, vect_convert_input): Likewise.
1642 (vect_convert_inputs, vect_convert_output): Likewise.
1643 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1644 to handle the optional cast of the multiplication result and
1645 vect_widened_op_tree to detect the widened multiplication itself.
1646 Do not require the input and output of promotion casts to have
1647 the same sign, but base the signedness of the operation on the
1648 input rather than the result. If the pattern includes two
1649 promotions, check that those promotions have the same sign.
1650 Do not restrict the MULT_EXPR handling to a double-width result;
1651 handle quadruple-width results and wider. Use vect_convert_inputs
1652 to convert the inputs to the common type.
1653 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1654 to handle the optional cast of the ABS result. Also allow a sign
1655 change or a sign extension between the ABS and MINUS.
1656 Use vect_widened_op_tree to detect the widened subtraction and use
1657 vect_convert_inputs to convert the inputs to the common type.
1658 (vect_handle_widen_op_by_const): Delete.
1659 (vect_recog_widen_op_pattern): New function.
1660 (vect_recog_widen_mult_pattern): Use it.
1661 (vect_recog_widen_shift_pattern): Likewise.
1662 (vect_recog_widen_sum_pattern): Use
1663 vect_look_through_possible_promotion to handle the promoted
1666 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1668 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1669 the containing gimple_seq *.
1670 * gimple-iterator.h (gsi_for_stmt): Declare it.
1671 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1672 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1673 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1674 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1675 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1677 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1679 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1680 statement being replaced is part of an existing pattern
1681 definition sequence, inserting the new pattern statements before
1683 (vect_pattern_recog_1): Don't return a bool. If the statement
1684 is already part of a pattern, instead apply pattern matching
1685 to the pattern definition statements. Don't clear the
1686 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1687 (vect_pattern_recog): Don't break after the first match;
1688 continue processing the pattern definition statements instead.
1689 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1691 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1693 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1694 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1695 (vect_recog_widen_sum_pattern): Use it.
1697 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1699 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1700 phi is not a pattern statement and has not been replaced by
1701 a pattern statement.
1702 * tree-vect-patterns.c (type_conversion_p): Don't check
1703 STMT_VINFO_IN_PATTERN_P.
1704 (vect_recog_vector_vector_shift_pattern): Likewise.
1705 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1706 the pattern statement rather than the original statement; check
1707 directly for a WIDEN_MULT_EXPR here.
1708 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1709 vect_is_simple_use to return the pattern statement rather
1710 than the original statement; use is_pattern_stmt_p to check
1711 for such a pattern statement.
1712 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1713 to return the pattern statement rather than the original statement;
1714 don't do the same transformation here.
1715 (vect_is_simple_use): If the defining statement has been replaced
1716 by a pattern statement, return the pattern statement instead.
1717 Remove the corresponding (local) transformation from the vectype
1720 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1722 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1723 end and default to null.
1724 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1725 (vectorizable_reduction): Update calls accordingly, dropping the
1726 gimple ** argument if the passed-back statement isn't needed.
1727 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1728 (vect_recog_rotate_pattern): Likewise.
1729 (vect_recog_mask_conversion_pattern): Likewise.
1730 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1731 (vect_mask_constant_operand_p): Likewise.
1732 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1733 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1734 (get_group_load_store_type, get_load_store_type): Likewise.
1735 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1736 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1737 (vectorizable_conversion, vectorizable_assignment): Likewise.
1738 (vectorizable_shift, vectorizable_operation): Likewise.
1739 (vectorizable_store, vect_is_simple_cond): Likewise.
1740 (vectorizable_condition, vectorizable_comparison): Likewise.
1741 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1742 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1743 and move it to the end. Cope with null def_stmt_outs.
1745 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1747 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1749 2018-06-29 Jeff Law <law@redhat.com>
1751 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1752 displacements for TARGET_V850E2V3 and newer.
1753 (TARGET_LRA_P): Remove. Defaults to LRA now.
1754 * config/v850/v850.md (sign23byte_load): Remove.
1755 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1756 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1758 2018-06-29 Martin Liska <mliska@suse.cz>
1761 * coverage.c (coverage_init): Mangle full path name.
1762 * doc/invoke.texi: Document the change.
1763 * gcov-io.c (mangle_path): New.
1764 * gcov-io.h (mangle_path): Likewise.
1765 * gcov.c (mangle_name): Use mangle_path for path mangling.
1767 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1769 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1770 if starting source register is not even.
1772 2018-06-29 Martin Liska <mliska@suse.cz>
1774 PR tree-optimization/86263
1775 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1776 Make edge redirection.
1778 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1780 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1781 (dump_scope_depth): New variable.
1782 (get_dump_scope_depth): New function.
1783 (dump_begin_scope): New function.
1784 (dump_end_scope): New function.
1785 * dumpfile.h (get_dump_scope_depth): New declaration.
1786 (dump_begin_scope): New declaration.
1787 (dump_end_scope): New declaration.
1788 (class auto_dump_scope): New class.
1789 (AUTO_DUMP_SCOPE): New macro.
1790 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1793 2018-06-29 Richard Biener <rguenther@suse.de>
1795 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1796 compute_all_dependences succeeds.
1797 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1798 exceed --param loop-max-datarefs-for-datadeps.
1800 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1802 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1804 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1807 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1808 alternative 0 in preferred_for_speed attribute.
1810 2018-06-28 Paul Koning <ni1d@arrl.net>
1812 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1813 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1814 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1815 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1816 for shift insn patterns.
1818 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1820 * cgraph.c (cgraph_node::get_body): Replace assignments to
1821 "dump_file" with calls to set_dump_file.
1822 * dumpfile.c (alt_dump_file): Make static, and group with...
1823 (alt_flags): ...this definition.
1824 (dumps_are_enabled): New variable.
1825 (refresh_dumps_are_enabled): New function.
1826 (set_dump_file): New function.
1827 (set_alt_dump_file): New function.
1828 (gcc::dump_manager::dump_start): Replace assignments to
1829 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1831 (gcc::dump_manager::dump_finish): Likewise.
1832 * dumpfile.h (alt_dump_file): Delete decl.
1833 (dumps_are_enabled): New variable decl.
1834 (set_dump_file): New function decl.
1835 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1837 * tree-nested.c (lower_nested_functions): Replace assignments to
1838 "dump_file" with calls to set_dump_file.
1840 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1842 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1843 goto_locus of each outgoing edge of each basic block.
1845 2018-06-28 Richard Biener <rguenther@suse.de>
1847 * dwarf2out.c (decl_scope_table): Remove.
1848 (push_decl_scope): Likewise.
1849 (pop_decl_scope): Likewise.
1850 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1851 (gen_struct_or_union_type_die): Likewise.
1852 (gen_tagged_type_die): Likewise.
1853 (dwarf2out_init): Do not initialize decl_scope_table.
1854 (dwarf2out_c_finalize): Do not free it.
1856 2018-06-28 Richard Biener <rguenther@suse.de>
1858 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1859 deciding whether to not re-use a DIE.
1861 2018-06-28 Richard Biener <rguenther@suse.de>
1863 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1864 DW_AT_abstract_origin attribute.
1866 2018-06-28 Martin Liska <mliska@suse.cz>
1868 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1869 Use newly introduced constants.
1870 * tree-switch-conversion.h (struct jump_table_cluster):
1871 Define max_ratio_for_size and max_ratio_for_speed.
1873 2018-06-28 Martin Liska <mliska@suse.cz>
1875 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1876 Add new checking assert to catch invalid state.
1877 (jump_table_cluster::can_be_handled): Handle single case
1879 (jump_table_cluster::is_beneficial): Bail out for such case.
1880 (bit_test_cluster::find_bit_tests):
1881 Add new checking assert to catch invalid state.
1882 (bit_test_cluster::can_be_handled): Handle single case
1884 (bit_test_cluster::is_beneficial): Bail out for such case.
1885 (switch_decision_tree::analyze_switch_statement):
1888 2018-06-28 Martin Liska <mliska@suse.cz>
1890 * common.opt: Introduce -completion option.
1891 * gcc.c (driver_handle_option): Handle it.
1892 (driver::main): Print completions if completion
1894 * opt-suggestions.c (option_proposer::get_completions):
1896 (option_proposer::suggest_completion): Likewise.
1897 (option_proposer::find_param_completions): Likewise.
1898 (verify_autocompletions): Likewise.
1899 (test_completion_valid_options): Likewise.
1900 (test_completion_valid_params): Likewise.
1901 (in_completion_p): Likewise.
1902 (empty_completion_p): Likewise.
1903 (test_completion_partial_match): Likewise.
1904 (test_completion_garbage): Likewise.
1905 (opt_proposer_c_tests): Likewise.
1906 * opt-suggestions.h: Declare new functions.
1907 * opts.c (common_handle_option): Handle OPT__completion_.
1908 * selftest-run-tests.c (selftest::run_tests): Add
1909 opt_proposer_c_tests.
1910 * selftest.c (assert_str_startswith): New.
1911 * selftest.h (assert_str_startswith): Likewise.
1912 (opt_proposer_c_tests): New.
1913 (ASSERT_STR_STARTSWITH): Likewise.
1915 2018-06-28 Martin Liska <mliska@suse.cz>
1917 * Makefile.in: Add opt-suggestions.o.
1918 * gcc-main.c: Include opt-suggestions.h.
1919 * gcc.c (driver::driver): Likewise.
1920 (driver::~driver): Remove m_option_suggestions.
1921 (driver::build_option_suggestions): Moved to option_proposer.
1922 (driver::suggest_option): Likewise.
1923 (driver::handle_unrecognized_options): Use option_proposer.
1924 * gcc.h (class driver): Add new memver m_option_proposer.
1925 * opt-suggestions.c: New file.
1926 * opt-suggestions.h: New file.
1928 2018-06-28 Martin Liska <mliska@suse.cz>
1930 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1931 (auto_string_vec::~auto_string_vec): Likewise.
1933 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1935 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1936 prevent_decl_creation_for_types fields up and add reset_location field.
1937 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1938 statement if id->reset_location is true.
1939 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1940 id->reset_location is true.
1941 (copy_phis_for_bb): Force input_location on the arguments if
1942 id->reset_location is true.
1943 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1944 is set on the function to be inlined.
1946 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1948 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1950 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1952 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1953 registers for Pmode.
1954 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1955 hard registers for the clobbered pseudo.
1957 2018-06-27 Paul Koning <ni1d@arrl.net>
1959 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1960 mutually exclusive options.
1961 * config/pdp11/constraints.md (h): New constraint.
1962 (O): Update definition to match shift code generation.
1963 (D): New constraint.
1964 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1966 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1968 (output_jump): Change arguments.
1969 (pdp11_fixed_cc_regs): New function.
1970 (pdp11_cc_mode): Ditto.
1971 (pdp11_expand_shift): Ditto.
1972 (pdp11_assemble_shift): Ditto.
1973 (pdp11_small_shift): Ditto.
1974 (pdp11_branch_cost): Remove.
1975 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1977 (pdp11_register_move_cost): Update for CC registers.
1978 (pdp11_rtx_costs): Add case for LSHIFTRT.
1979 (pdp11_output_jump): Add CCNZ mode conditional branches.
1980 (notice_update_cc_on_set): Remove.
1981 (pdp11_cc_mode): New function.
1982 (simple_memory_operand): Correct pre/post decrement case.
1983 (no_side_effect_operand): New function.
1984 (pdp11_regno_reg_class): Add CC_REGS class.
1985 (pdp11_fixed_cc_regs): New function.
1986 (pdp11_small_shift): New function.
1987 (pdp11_expand_shift): New function to expand shift insns.
1988 (pdp11_assemble_shift): New function to output shifts.
1989 (pdp11_branch_cost): Remove.
1990 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1991 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1992 (WCHAR_TYPE): Ditto.
1993 (PTRDIFF_TYPE): Ditto.
1994 (ADJUST_INSN_LENGTH): New macro.
1995 (FIXED_REGISTERS): Add CC registers.
1996 (CALL_USED_REGISTERS): Ditto.
1998 (REG_CLASS_NAMES): Ditto.
1999 (REG_CLASS_CONTENTS): Ditto.
2000 (SELECT_CC_MODE): Use new function.
2001 (TARGET_FLAGS_REGNUM): New macro.
2002 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
2003 (cc0_reg_rtx): Remove.
2004 (CC_STATUS_MDEP): Remove.
2005 (CC_STATUS_MDEFP_INIT): Remove.
2006 (CC_IN_FPU): Remove.
2007 (NOTICE_UPDATE_CC): Remove.
2008 (REGISTER_NAMES): Add CC registers.
2009 (BRANCH_COST): Change to constant 1.
2010 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
2012 * config/pdp11/pdp11.opt (mbcopy): Remove.
2013 (mbcopy-builtin): Remove.
2014 (mbranch-cheap): Remove.
2015 (mbranch-expensive): Remove.
2016 * config/pdp11/predicates.md (expand_shift_operand): Update to
2017 match shift code generation.
2018 (ccnz_operator): New predicate.
2019 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2020 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2021 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2023 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2024 description of O constraint.
2026 2018-06-27 Jeff Law <law@redhat.com>
2027 Austin Law <austinklaw@gmail.com>
2029 * config/v850/v850.md (addsi3_set_flags): New pattern.
2030 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2031 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2032 (zero_extendhisi2_v850_set_flags): Likewise.
2033 (zero_extendqisi2_v850_set_flags): Likewise.
2034 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2035 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2036 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2038 * config/v850/v850-protos.h (notice_update_cc): Remove.
2039 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2040 (v850_print_operand): Handle 'D' and "d".
2041 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2042 Add handling of arithmetic/logical operations compared against zero.
2043 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2044 Do not look at v850_compare_op, instead get mode from last argument.
2045 (v850_gen_compare): Remove
2046 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2047 after reload for prologue insns.
2048 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2050 (construct_save_jarl): Likewise.
2051 (TARGET_FLAGS_REGNUM): Define.
2052 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2053 (NOTICE_UPDATE_CC): Remove.
2054 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2055 than cc0. Conditionalize on reload_completed.
2056 (cmpsi_insn, setfcc_insn): Likewise.
2057 (tst1 splitter): Turn into define_and_split which sets the flags
2059 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2060 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2061 (cstoresf4, cstoredf4): Clobber the flags.
2062 (cmpsi, cmpsf, cmpdf): Remove expanders.
2063 (setf_insn): Remove pattern.
2064 (addsi3): Turn into define_and_split which clobbers the flags after
2065 reload and a suitable pattern (addsi3_clobber_flags) for use after
2067 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2068 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2069 (ashrsi3, ashrsi3_v850e2): Likewise.
2070 (bins): Clobber the flags.
2071 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2072 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2073 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2074 (call_value_internal_short, call_value_internal_long): Likewise.
2075 (callt_save_interrupt, callt_return_interrupt): Likewise.
2076 (save_interrupt, return_interrupt): Likewise.
2077 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2078 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2079 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2080 (All FP comparisons): Only allow after reload has completed.
2082 (divh, divhu): Tweak output template.
2083 (branch_z_normal, branch_z_invert): Remove
2084 (branch_nz_normal, branch_nz_invert): Likewise.
2085 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2087 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2088 * config/v850/v850.c (notice_update_cc): Remove.
2089 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2090 (CC_NO_CARRY): Likewise.
2091 (NOTICE_UPDATE_CC): Define to nothing.
2092 * config/v850/v850.md: Remove block comment on cc0 handling
2093 Remove "cc" attribute from all patterns. Remove cc_status handling
2094 from all patterns. Minor formatting fixes.
2096 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2098 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2099 (cortex-a76.cortex-a55): Likewise.
2100 * config/aarch64/aarch64-tune.md: Regenerate.
2101 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2102 cortex-a76.cortex-a55.
2104 2018-06-27 Jeff Law <law@redhat.com>
2106 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2107 (MULTILIB_DIRNAMES): Similarly.
2109 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2111 * gimple.h (gimple_return_retbnd): Delete.
2112 (gimple_return_set_retbnd): Likewise.
2113 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2114 gimple_return_set_retbnd.
2115 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2116 gimple_return_retbnd and adjust.
2117 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2118 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2119 Explicitly return NULL in a couple more cases. Move assertion
2120 on debug statements and remove unreachable code.
2121 (reset_debug_binding): Do not test id->retbnd.
2122 (expand_call_inline): Do not set it.
2124 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2126 * configure.ac: Add --disable-gcov option.
2127 * configure: Regenerate.
2128 * Makefile.in: Honour @enable_gcov@.
2129 * doc/install.texi: Document --disable-gcov.
2131 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2133 * config/arm/arm-cpus.in (cortex-a76): New entry.
2134 (cortex-a76.cortex-a55): Likewise.
2135 * config/arm/arm-tables.opt: Regenerate.
2136 * config/arm/arm-tune.md: Likewise.
2137 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2138 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2139 cortex-a76.cortex-a55.
2141 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2144 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2146 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2148 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2150 (EPILOGUE_USES): Likewise.
2152 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2154 * tree-inline.c (remap_location): New function extracted from...
2155 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2156 (copy_phis_for_bb): ...here. Call remap_location.
2157 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2159 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2161 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2162 unaligned vsx for 16B memset.
2164 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2167 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2168 ieee128_float_type_node to long_double_type_node unless
2169 TARGET_LONG_DOUBLE_128 is set.
2171 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2173 * cfgloop.c (get_loop_location): Convert return type from
2174 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2175 by implicit construction from rtx_insn *, and using
2176 dump_user_location_t::from_function_decl for the fallback case.
2177 * cfgloop.h (get_loop_location): Convert return type from
2178 location_t to dump_user_location_t.
2179 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2180 dump_printf_loc to pass in a dump_location_t rather than a
2181 location_t, via the gimple stmt.
2182 * coverage.c (get_coverage_counts): Update calls to
2183 dump_printf_loc to pass in dump_location_t rather than a
2185 * doc/optinfo.texi (Dump types): Convert example of
2186 dump_printf_loc from taking "locus" to taking "insn". Update
2187 description of the "_loc" calls to cover dump_location_t.
2188 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
2190 (dump_user_location_t::dump_user_location_t): New constructors,
2191 from gimple *stmt and rtx_insn *.
2192 (dump_user_location_t::from_function_decl): New function.
2193 (dump_loc): Make static.
2194 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2195 const dump_location_t &.
2196 (dump_generic_expr_loc): Delete.
2197 (dump_printf_loc): Convert param "loc" from location_t to
2198 const dump_location_t &.
2199 (selftest::test_impl_location): New function.
2200 (selftest::dumpfile_c_tests): New function.
2201 * dumpfile.h: Include "profile-count.h".
2202 (class dump_user_location_t): New class.
2203 (struct dump_impl_location_t): New struct.
2204 (class dump_location_t): New class.
2205 (dump_printf_loc): Convert 2nd param from source_location to
2206 const dump_location_t &.
2207 (dump_generic_expr_loc): Delete.
2208 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2209 const dump_location_t &.
2210 * gimple-fold.c (fold_gimple_assign): Update call to
2211 dump_printf_loc to pass in a dump_location_t rather than a
2212 location_t, via the gimple stmt.
2213 (gimple_fold_call): Likewise.
2214 * gimple-loop-interchange.cc
2215 (loop_cand::analyze_iloop_reduction_var): Update for change to
2216 check_reduction_path.
2217 (tree_loop_interchange::interchange): Update for change to
2219 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2220 change in return-type of find_loop_location.
2221 (graphite_regenerate_ast_isl): Likewise.
2222 * graphite-optimize-isl.c (optimize_isl): Likewise.
2223 * graphite.c (graphite_transform_loops): Likewise.
2224 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2225 pass in a dump_location_t rather than a location_t, via the
2227 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2228 * ipa.c (walk_polymorphic_call_targets): Likewise.
2229 * loop-unroll.c (report_unroll): Convert "locus" param from
2230 location_t to dump_location_t.
2231 (decide_unrolling): Update for change to get_loop_location's
2233 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2234 location_t to dump_user_location_t.
2235 (grid_find_single_omp_among_assignments_1): Updates calls to
2236 dump_printf_loc to pass in a dump_location_t rather than a
2237 location_t, via the gimple stmt.
2238 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2239 location_t to dump_location_t. Updates calls to dump_printf_loc
2240 to pass in a dump_location_t rather than a location_t, via the
2242 (grid_inner_loop_gridifiable_p): Likewise.
2243 (grid_dist_follows_simple_pattern): Likewise.
2244 (grid_gfor_follows_tiling_pattern): Likewise.
2245 (grid_target_follows_gridifiable_pattern): Likewise.
2246 (grid_attempt_target_gridification): Convert initialization
2247 of local "grid" from memset to zero-initialization; FIXME: does
2248 this require C++11? Update call to dump_printf_loc to pass in a
2249 optinfo_location rather than a location_t, via the gimple stmt.
2250 * profile.c (read_profile_edge_counts): Updates call to
2251 dump_printf_loc to pass in a dump_location_t rather than a
2253 (compute_branch_probabilities): Likewise.
2254 * selftest-run-tests.c (selftest::run_tests): Call
2256 * selftest.h (dumpfile_c_tests): New decl.
2257 * tree-loop-distribution.c (pass_loop_distribution::execute):
2258 Update for change in return type of find_loop_location.
2259 * tree-parloops.c (parallelize_loops): Likewise.
2260 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2261 "locus" from location_t to dump_user_location_t.
2262 (canonicalize_loop_induction_variables): Likewise.
2263 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2264 for change in return type of find_loop_location.
2265 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2266 to dump_printf_loc to pass in a dump_location_t rather than a
2267 location_t, via the stmt.
2268 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2270 * tree-vect-loop-manip.c (find_loop_location): Convert return
2271 type from source_location to dump_user_location_t.
2272 (vect_do_peeling): Update for above change.
2273 (vect_loop_versioning): Update for change in type of
2275 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2276 from location_t to dump_user_location_t.
2277 (vect_estimate_min_profitable_iters): Update for change in type
2279 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2280 location_t to dump_location_t.
2281 (vect_slp_bb): Update for change in type of vect_location.
2282 * tree-vectorizer.c (vect_location): Convert from source_location
2283 to dump_user_location_t.
2284 (try_vectorize_loop_1): Update for change in vect_location's type.
2285 (vectorize_loops): Likewise.
2286 (increase_alignment): Likewise.
2287 * tree-vectorizer.h (vect_location): Convert from source_location
2288 to dump_user_location_t.
2289 (find_loop_location): Convert return type from source_location to
2290 dump_user_location_t.
2291 (check_reduction_path): Convert 1st param from location_t to
2292 dump_user_location_t.
2293 * value-prof.c (check_counter): Update call to dump_printf_loc to
2294 pass in a dump_user_location_t rather than a location_t; update
2295 call to error_at for change in type of "locus".
2296 (check_ic_target): Update call to dump_printf_loc to
2297 pass in a dump_user_location_t rather than a location_t, via the
2300 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2302 * config/s390/s390.h (enum processor_flags): Do not use
2303 default tune parameter when -march was specified.
2305 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2308 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2309 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2312 2018-06-26 Richard Biener <rguenther@suse.de>
2314 PR tree-optimization/86287
2316 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2317 (vect_analyze_loop): Initialize n_stmts.
2319 2018-06-26 Richard Biener <rguenther@suse.de>
2322 * fold-const.c (fold_convertible_p): Pointer extension
2325 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2328 * dwarf2out.c (loc_list_has_views): Adjust comments.
2329 (dw_loc_list): Split single cross-partition range with
2332 2018-06-25 Jeff Law <law@redhat.com>
2334 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2335 on -mbig-switch by default.
2337 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2339 (const_float_0_operand): Remove unused predicate.
2340 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2341 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2342 (recipsf2): New expander. Original pattern now called
2344 (recipdf2, recipdf2_insn): Similarly.
2345 (rsqrtsf2, rsqrtsf2_insn): Similarly
2346 (rsqrtdf2, rsqrtdf2_insn): Similarly
2348 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2350 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2351 Simplify logic for FreeBSD (twice).
2353 2018-06-25 Martin Sebor <msebor@redhat.com>
2355 PR tree-optimization/86204
2356 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2357 a strnlen result if it's less than the length of the string.
2359 2018-06-25 Martin Sebor <msebor@redhat.com>
2361 PR tree-optimization/85700
2362 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2363 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2364 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2366 2018-06-25 Martin Sebor <msebor@redhat.com>
2368 * doc/extend.texi (Zero-length arrays): Update and clarify.
2370 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2372 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2373 added IEEE/IBM long double multilib support on PowerPC little
2374 endian Linux systems.
2375 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2376 (MULTILIB_DEFAULTS): Likewise.
2377 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2379 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2380 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2381 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2383 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2386 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2387 (REORDER_45): Likewise.
2389 2018-06-25 Jeff Law <law@redhat.com>
2391 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2392 dividend to 32 bits. Adjust length.
2393 (udivmodhi4): Cleanup output template. Fix length.
2395 2018-06-25 Carl Love <cel@us.ibm.com>
2397 * config/rs6000/vsx.md: Change word selector to prefered location.
2399 2018-06-25 Richard Biener <rguenther@suse.de>
2401 PR tree-optimization/86304
2402 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2403 epilogue-if-converted loops as well.
2405 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2407 * lto-section-out.c (lto_begin_section): Do not print section
2408 name for noaddr and unnumbered dumps.
2410 2018-06-25 Richard Biener <rguenther@suse.de>
2412 * tree-vectorizer.h (struct vec_info_shared): New structure
2413 with parts split out from struct vec_info and loop_nest from
2414 struct _loop_vec_info.
2415 (struct vec_info): Adjust accordingly.
2416 (struct _loop_vec_info): Likewise.
2417 (LOOP_VINFO_LOOP_NEST): Adjust.
2418 (LOOP_VINFO_DATAREFS): Likewise.
2419 (LOOP_VINFO_DDRS): Likewise.
2420 (struct _bb_vec_info): Likewise.
2421 (BB_VINFO_DATAREFS): Likewise.
2422 (BB_VINFO_DDRS): Likewise.
2423 (struct _stmt_vec_info): Add dr_aux member.
2424 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2425 (DR_MISALIGNMENT_UNINITIALIZED): New.
2426 (set_dr_misalignment): Adjust.
2427 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2428 (vect_analyze_loop): Adjust prototype.
2429 (vect_analyze_loop_form): Likewise.
2430 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2431 Compute dependences lazily.
2432 (vect_record_base_alignments): Use shared datarefs/ddrs.
2433 (vect_verify_datarefs_alignment): Likewise.
2434 (vect_analyze_data_refs_alignment): Likewise.
2435 (vect_analyze_data_ref_accesses): Likewise.
2436 (vect_analyze_data_refs): Likewise.
2437 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2438 constructor parameter for shared part.
2439 (vect_analyze_loop_form): Pass in shared part and adjust.
2440 (vect_analyze_loop_2): Pass in storage for the number of
2441 stmts. Move loop nest finding to the caller. Compute
2443 (vect_analyze_loop): Pass in shared part.
2444 (vect_transform_loop): Verify shared datarefs are unchanged.
2445 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2446 constructor parameter for shared part.
2447 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2448 (vect_slp_bb): Verify shared datarefs are unchanged before
2450 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2452 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2453 DR_MISALIGNMENT_UNINITIALIZED.
2454 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2455 parameter for shared part.
2456 (vec_info::~vec_info): Adjust.
2457 (vec_info_shared::vec_info_shared): New.
2458 (vec_info_shared::~vec_info_shared): Likewise.
2459 (vec_info_shared::save_datarefs): Likewise.
2460 (vec_info_shared::check_datarefs): Likewise.
2461 (try_vectorize_loop_1): Construct shared part live for analyses
2462 of a single loop for multiple vector sizes.
2463 * tree-parloops.c (gather_scalar_reductions): Adjust.
2465 2018-06-25 Richard Biener <rguenther@suse.de>
2467 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2468 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2469 (vect_analyze_data_refs): Remove similar code from here and
2470 simplify accordingly.
2472 2018-06-25 Richard Biener <rguenther@suse.de>
2474 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2475 for reverse storage order accesses rather than asserting
2476 they cannot happen here.
2478 2018-06-25 Tom de Vries <tdevries@suse.de>
2481 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2482 Use data16 instead of .byte for insn prefix.
2484 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2487 * parser.c (make_char_string_pack): Pass this literal chars
2488 through cpp_interpret_string.
2489 (cp_parser_userdef_numeric_literal): Check the result of
2490 make_char_string_pack.
2492 2018-06-24 Maya Rashish <coypu@sdf.org>
2494 * ginclude/stddef.h: Simplify conditions around avoiding
2495 re-definition of __size_t.
2497 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2499 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2500 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2502 2018-06-22 Maya Rashish <coypu@sdf.org>
2504 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2505 made to OpenBSD and NetBSD through the years.
2507 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2509 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2510 behavior of vec_pack (vector double, vector double) to match
2511 behavior of vec_float2 (vector double, vector double).
2513 2018-06-22 Olivier Hainque <hainque@adacore.com>
2515 * gimplify.c (gimplify_function_tree): Prevent creation
2516 of a trampoline for the address of the current function
2517 passed to entry/exit instrumentation hooks.
2519 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2522 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2525 2018-06-22 Martin Liska <mliska@suse.cz>
2527 PR tree-optimization/86263
2528 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2529 Bail out if is_enabled is false.
2530 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2532 (jump_table_cluster::is_enabled): New function.
2534 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2536 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2537 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2538 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2539 (lto_input_ts_binfo_tree_pointers): Likewise.
2540 * tree-streamer-out.c (streamer_write_tree_bitfields,
2541 write_ts_binfo_tree_pointers): Likewise.
2542 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2544 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2546 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2548 2018-06-22 Martin Liska <mliska@suse.cz>
2550 * symbol-summary.h (get): Make it pure and inline move
2551 functionality from ::get function.
2552 (get): Remove and inline into ::get and ::get_create.
2553 (get_create): Move code from ::get function.
2555 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2558 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2559 -x assembler-with-cpp.
2561 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2563 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2564 _FILE_OFFSET_BITS=64 for C++.
2566 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2568 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2569 conversion insn that shows up when pr85657-3.c is compiled using
2570 IEEE 128-bit long double.
2571 (neg<mode>2_internal): Use the correct mode to check whether the
2572 mode is IBM extended.
2573 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2574 multiply and divide external functions from being created more
2577 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2579 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2581 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2582 the edge can be forwarded.
2583 (cfg_layout_merge_blocks): Likewise.
2585 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2587 * except.c (finish_eh_generation): Commit edge insertions only after
2588 the EH edges have been redirected from post-landing to landing pads.
2590 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2592 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2593 create_tmp_var_for to create the FRAME decl.
2594 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2596 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2598 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2599 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2600 debug statement when resetting its value.
2601 (expand_call_inline): Copy the locus of the call onto the assignment
2602 of the return value, if any. Use local variable in more cases.
2604 2018-06-21 Martin Liska <mliska@suse.cz>
2606 * ipa-pure-const.c (propagate_nothrow): Use
2607 funct_state_summaries->get.
2608 (dump_malloc_lattice): Likewise.
2609 (propagate_malloc): Likewise.
2611 2018-06-21 Richard Biener <rguenther@suse.de>
2613 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2614 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2615 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2617 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2618 BLOCK_ABSTRACT_ORIGIN unconditionally.
2620 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2622 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2624 * ipa-reference.c (ipa_reference_c_finalize): Delete
2625 ipa_ref_opt_sum_summaries and set it to NULL.
2627 2018-06-21 Tom de Vries <tdevries@suse.de>
2629 PR tree-optimization/85859
2630 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2631 test with comment from bb_no_side_effects_p.
2633 2018-06-21 Richard Biener <rguenther@suse.de>
2635 PR tree-optimization/86232
2636 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2637 max for constant niter.
2639 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2641 * config/aarch64/aarch64-simd.md
2642 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2644 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2646 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2647 Make opernads of the unspec commutative.
2649 2018-06-21 Richard Biener <rguenther@suse.de>
2651 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2652 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2653 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2654 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2656 (vect_record_base_alignments): Do not record base alignment
2657 for gathers or scatters.
2658 (vect_compute_data_ref_alignment): Drop return value that is always
2659 true. Bail out early for gathers or scatters.
2660 (vect_enhance_data_refs_alignment): Bail out early for gathers
2662 (vect_find_same_alignment_drs): Likewise.
2663 (vect_analyze_data_refs_alignment): Remove dead code.
2664 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2665 (vect_analyze_data_refs): For possible gathers or scatters do
2666 not create an alternate DR, just check their possible validity
2667 and mark them. Adjust DECL_NONALIASED handling to not rely
2669 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2670 update inits of gathers or scatters.
2671 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2672 Also copy gather/scatter flag to pattern vinfo.
2674 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2676 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2677 behavior of vec_packsu (vector unsigned long long, vector unsigned
2678 long long) to match behavior of vec_packs with same signature.
2680 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2681 Thomas Schwinge <thomas@codesourcery.com>
2682 Cesar Philippidis <cesar@codesourcery.com>
2684 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2685 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2686 (gimplify_adjust_omp_clauses): Likewise.
2687 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2688 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2689 (gimplify_omp_target_update): Update handling of acc update and
2691 * omp-low.c (install_var_field): Remove unused parameter
2692 base_pointers_restrict.
2693 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2694 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2696 (omp_target_base_pointers_restrict_p): Delete.
2697 (scan_omp_target): Update call to scan_sharing_clauses.
2698 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2700 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2701 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2702 (convert_local_omp_clauses): Likewise.
2703 * tree-pretty-print.c (dump_omp_clause): Likewise.
2704 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2706 (omp_clause_code_name): Likewise.
2708 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2711 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2714 PR tree-optimization/86231
2715 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2716 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2718 2018-06-20 Tom de Vries <tdevries@suse.de>
2720 PR tree-optimization/86097
2721 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2722 iv type if signedness of iv type is not the same as that of *nit.
2724 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2726 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2727 EDGE_EH edges, verify they are all EDGE_EH.
2729 2018-06-20 Maya Rashish <coypu@sdf.org>
2731 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2733 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2735 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2736 * config/aarch64/aarch64.c (xgene1_tunings): Add
2737 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2738 (aarch64_mode_valid_for_sched_fusion_p):
2739 Allow 16-byte modes.
2740 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2741 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2743 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2745 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2746 * config/aarch64/iterators.md (VQ2): New mode iterator.
2748 2018-06-20 Martin Liska <mliska@suse.cz>
2750 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2751 Change default ratio from 10 to 8.
2753 2018-06-20 Martin Liska <mliska@suse.cz>
2755 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2757 (bit_test_cluster::find_bit_tests): Likewise.
2758 (switch_decision_tree::analyze_switch_statement): Find clusters.
2759 * tree-switch-conversion.h (struct jump_table_cluster): Document
2762 2018-06-20 Martin Liska <mliska@suse.cz>
2764 * tree-switch-conversion.c (switch_conversion::collect):
2765 Record m_uniq property.
2766 (switch_conversion::expand): Bail out for special conditions.
2767 (group_cluster::~group_cluster): New.
2768 (group_cluster::group_cluster): Likewise.
2769 (group_cluster::dump): Likewise.
2770 (jump_table_cluster::emit): New.
2771 (switch_decision_tree::fix_phi_operands_for_edges): New.
2772 (struct case_node): Remove struct.
2773 (jump_table_cluster::can_be_handled): New.
2774 (case_values_threshold): Moved to header.
2775 (reset_out_edges_aux): Likewise.
2776 (jump_table_cluster::is_beneficial): New.
2777 (bit_test_cluster::can_be_handled): Likewise.
2778 (add_case_node): Remove.
2779 (bit_test_cluster::is_beneficial): New.
2780 (case_bit_test::cmp): New.
2781 (bit_test_cluster::emit): New.
2782 (expand_switch_as_decision_tree_p): Remove.
2783 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2784 (fix_phi_operands_for_edge): Likewise.
2785 (switch_decision_tree::analyze_switch_statement): New.
2786 (compute_cases_per_edge): Move ...
2787 (switch_decision_tree::compute_cases_per_edge): ... here.
2788 (try_switch_expansion): Likewise.
2789 (switch_decision_tree::try_switch_expansion): Likewise.
2790 (record_phi_operand_mapping): Likewise.
2791 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2792 (emit_case_decision_tree): Likewise.
2793 (switch_decision_tree::emit): Likewise.
2794 (balance_case_nodes): Likewise.
2795 (switch_decision_tree::balance_case_nodes): Likewise.
2796 (dump_case_nodes): Likewise.
2797 (switch_decision_tree::dump_case_nodes): Likewise.
2798 (emit_jump): Likewise.
2799 (switch_decision_tree::emit_jump): Likewise.
2800 (emit_cmp_and_jump_insns): Likewise.
2801 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2802 (emit_case_nodes): Likewise.
2803 (switch_decision_tree::emit_case_nodes): Likewise.
2804 (conditional_probability): Remove.
2805 * tree-switch-conversion.h (enum cluster_type): New.
2807 (struct cluster): Likewise.
2808 (cluster::cluster): Likewise.
2809 (struct simple_cluster): Likewise.
2810 (simple_cluster::simple_cluster): Likewise.
2811 (struct group_cluster): Likewise.
2812 (struct jump_table_cluster): Likewise.
2813 (struct bit_test_cluster): Likewise.
2814 (struct min_cluster_item): Likewise.
2815 (struct case_tree_node): Likewise.
2816 (case_tree_node::case_tree_node): Likewise.
2817 (jump_table_cluster::case_values_threshold): Likewise.
2818 (struct case_bit_test): Likewise.
2819 (struct switch_decision_tree): Likewise.
2820 (struct switch_conversion): Likewise.
2821 (switch_decision_tree::reset_out_edges_aux): Likewise.
2823 2018-06-20 Martin Liska <mliska@suse.cz>
2825 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2826 (hoist_edge_and_branch_if_true): Likewise.
2827 (expand_switch_using_bit_tests_p): Likewise.
2828 (struct case_bit_test): Likewise.
2829 (case_bit_test_cmp): Likewise.
2830 (emit_case_bit_tests): Likewise.
2831 (switch_conversion::switch_conversion): New class.
2832 (struct switch_conv_info): Remove old struct.
2833 (collect_switch_conv_info): More to ...
2834 (switch_conversion::collect): ... this.
2835 (check_range): Likewise.
2836 (switch_conversion::check_range): Likewise.
2837 (check_all_empty_except_final): Likewise.
2838 (switch_conversion::check_all_empty_except_final): Likewise.
2839 (check_final_bb): Likewise.
2840 (switch_conversion::check_final_bb): Likewise.
2841 (create_temp_arrays): Likewise.
2842 (switch_conversion::create_temp_arrays): Likewise.
2843 (free_temp_arrays): Likewise.
2844 (gather_default_values): Likewise.
2845 (switch_conversion::gather_default_values): Likewise.
2846 (build_constructors): Likewise.
2847 (switch_conversion::build_constructors): Likewise.
2848 (constructor_contains_same_values_p): Likewise.
2849 (switch_conversion::contains_same_values_p): Likewise.
2850 (array_value_type): Likewise.
2851 (switch_conversion::array_value_type): Likewise.
2852 (build_one_array): Likewise.
2853 (switch_conversion::build_one_array): Likewise.
2854 (build_arrays): Likewise.
2855 (switch_conversion::build_arrays): Likewise.
2856 (gen_def_assigns): Likewise.
2857 (switch_conversion::gen_def_assigns): Likewise.
2858 (prune_bbs): Likewise.
2859 (switch_conversion::prune_bbs): Likewise.
2860 (fix_phi_nodes): Likewise.
2861 (switch_conversion::fix_phi_nodes): Likewise.
2862 (gen_inbound_check): Likewise.
2863 (switch_conversion::gen_inbound_check): Likewise.
2864 (process_switch): Use the newly created class.
2865 (switch_conversion::expand): New.
2866 (switch_conversion::~switch_conversion): New.
2867 * tree-switch-conversion.h: New file.
2869 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2871 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2872 tree-vect-patterns.c.
2873 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2874 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2875 (vect_recog_sad_pattern): Likewise.
2876 (vect_recog_widen_sum_pattern): Likewise.
2877 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2878 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2879 (vect_recog_rotate_pattern): Likewise.
2880 (vect_recog_mult_pattern): Likewise.
2881 (vect_recog_vector_vector_shift_pattern): Likewise.
2882 (vect_recog_divmod_pattern): Likewise.
2883 (vect_recog_mixed_size_cond_pattern): Likewise.
2884 (vect_recog_bool_pattern): Likewise.
2885 (vect_recog_mask_conversion_pattern): Likewise.
2886 (vect_try_gather_scatter_pattern): Likewise.
2887 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2888 (vect_recog_over_widening_pattern): Likewise.
2889 (vect_recog_gather_scatter_pattern): Likewise.
2890 (vect_recog_func_ptr): Move from tree-vectorizer.h
2891 (vect_vect_recog_func_ptrs): Move further down the file.
2892 (vect_recog_func): Likewise. Remove the third argument.
2893 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2894 (vect_pattern_recog_1): Expect the pattern function to do any
2895 necessary target tests. Also expect it to provide a vector type.
2896 Remove the type_in handling.
2898 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2900 * tree-vect-patterns.c (vect_pattern_detected): New function.
2901 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2902 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2903 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2904 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2905 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2906 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2907 (vect_recog_mask_conversion_pattern)
2908 (vect_try_gather_scatter_pattern): Likewise.
2910 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2912 * tree-vect-patterns.c (vect_get_internal_def): New function.
2913 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2914 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2915 (search_type_for_mask_1): Use it.
2917 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2919 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2920 redundant WIDEN_SUM_EXPR handling.
2921 (vect_recog_sad_pattern): Likewise.
2923 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2925 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2926 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2927 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2928 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2929 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2931 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2933 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2934 use the stmt_vec_info of the original bb statement for the
2935 new zero assignment, even if the call is part of a pattern.
2937 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2939 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2940 that the sequence is attached to the original statement rather
2941 than the pattern statement.
2942 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2943 PATTERN_DEF_SEQ from the original statement rather than
2944 the main pattern statement.
2945 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2946 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2947 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2949 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2951 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2952 definition statements before the early exit for statements that aren't
2954 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2956 (vect_transform_loop): ...here. Process pattern definition
2957 statements without first checking whether the main pattern
2958 statement is live or relevant.
2960 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2962 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2963 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2965 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2967 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2968 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2969 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2970 (expand_block_compare): Change select_block_compare_mode call.
2971 (expand_strncmp_align_check): Use new functions, fix comment.
2972 (emit_final_str_compare_gpr): New function.
2973 (expand_strn_compare): Refactor and clean up code.
2974 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2976 2018-06-19 Tony Reix <tony.reix@atos.com>
2977 Damien Bergamini <damien.bergamini@atos.com>
2978 David Edelsohn <dje.gcc@gmail.com>
2980 * collect2.c (static_obj): New variable.
2981 (static_libs): New variable.
2982 (is_in_list): Uncomment declaration.
2983 (main): Track AIX libraries linked statically.
2984 (is_in_list): Uncomment definition.
2985 (scan_prog_file): Don't add AIX shared libraries initializer
2986 to constructor list if linking statically.
2988 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2990 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2992 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2994 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2996 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2999 2018-06-19 Martin Liska <mliska@suse.cz>
3001 * config/i386/i386.c (ix86_can_inline_p): Do not use
3002 ipa_fn_summaries::get_create.
3003 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
3005 (devirtualization_time_bonus): Likewise.
3006 (ipcp_propagate_stage): Likewise.
3007 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
3008 (edge_set_predicate): Likewise.
3009 (evaluate_conditions_for_known_args): Likewise.
3010 (evaluate_properties_for_edge): Likewise.
3011 (ipa_call_summary::reset): Tranform to ...
3012 (ipa_call_summary::~ipa_call_summary): ... this.
3013 (ipa_fn_summary::reset): Transform to ...
3014 (ipa_fn_summary::~ipa_fn_summary): ... this.
3015 (ipa_fn_summary_t::remove): Rename to ...
3016 (ipa_fn_summary_t::remove_callees): ... this.
3017 (ipa_fn_summary_t::duplicate): Use placement new
3018 instead of memory copy.
3019 (ipa_call_summary_t::duplicate): Likewise.
3020 (ipa_call_summary_t::remove): Remove.
3021 (dump_ipa_call_summary): Change get_create to get.
3022 (ipa_dump_fn_summary): Dump only when summary exists.
3023 (analyze_function_body): Use symbol_summary::get instead
3025 (compute_fn_summary): Likewise.
3026 (estimate_edge_devirt_benefit): Likewise.
3027 (estimate_edge_size_and_time): Likewise.
3028 (inline_update_callee_summaries): Likewise.
3029 (remap_edge_change_prob): Likewise.
3030 (remap_edge_summaries): Likewise.
3031 (ipa_merge_fn_summary_after_inlining): Likewise.
3032 (write_ipa_call_summary): Likewise.
3033 (ipa_fn_summary_write): Likewise.
3034 (ipa_free_fn_summary): Likewise.
3035 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3036 (struct ipa_call_summary): Likewise.
3037 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3039 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3040 (estimate_size_after_inlining): Likewise.
3041 (estimate_growth): Likewise.
3042 (growth_likely_positive): Likewise.
3043 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3044 (inline_call): Likewise.
3045 * ipa-inline.c (caller_growth_limits): Likewise.
3046 (can_inline_edge_p): Likewise.
3047 (can_inline_edge_by_limits_p): Likewise.
3048 (compute_uninlined_call_time): Likewise.
3049 (compute_inlined_call_time): Likewise.
3050 (want_inline_small_function_p): Likewise.
3051 (edge_badness): Likewise.
3052 (update_caller_keys): Likewise.
3053 (update_callee_keys): Likewise.
3054 (inline_small_functions): Likewise.
3055 (inline_to_all_callers_1): Likewise.
3056 (dump_overall_stats): Likewise.
3057 (early_inline_small_functions): Likewise.
3058 (early_inliner): Likewise.
3059 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3060 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3061 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3062 * ipa-split.c (execute_split_functions): Likewise.
3063 * symbol-summary.h: Likewise.
3064 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3066 2018-06-19 Richard Biener <rguenther@suse.de>
3068 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3069 (vectorize_loops): ... here. Fix dbgcnt handling.
3070 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3072 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3075 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3076 ieee128 argument takes up only one (vector) register, not two (floating
3079 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3081 * gimplify.c (gimplify_init_constructor): Really never clear for an
3082 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3084 2018-06-19 Richard Biener <rguenther@suse.de>
3086 PR tree-optimization/86179
3087 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3088 after failed recognition.
3090 2018-06-18 Martin Sebor <msebor@redhat.com>
3093 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3094 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3095 Handle integer subtraction.
3096 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3097 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3099 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3101 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3102 param from rtx to rtx_insn *.
3103 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3105 (frv_ifcvt_modify_insn): Likwise.
3106 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3107 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3108 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3109 as_a <rtx_insn *> cast to local "unprotected_region" once
3110 it's been established that it's not NULL or pc_rtx.
3111 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3112 param "sethi" from rtx to rtx_insn *.
3113 (nds32_group_float_insns): Likewise for param "insn".
3114 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3116 (vax_output_int_subtract): Likewise.
3117 * config/vax/vax.c (vax_output_int_add): Likewise for param
3119 (vax_output_int_subtract): Likewise.
3120 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3121 (emit_pattern_after): Likewise for param "after".
3122 (emit_insn_after): Likewise.
3123 (emit_jump_insn_after): Likewise.
3124 (emit_call_insn_after): Likewise.
3125 (emit_debug_insn_after): Likewise.
3126 (emit_pattern_before): Likewise for param "before".
3127 (emit_insn_before): Likewise.
3128 (emit_jump_insn_before): Likewise.
3129 * final.c (get_insn_template): Likewise for param "insn", removing
3131 * output.h (get_insn_template): Likewise for 2nd param.
3132 * rtl.h (emit_insn_before): Likewise.
3133 (emit_jump_insn_before): Likewise.
3134 (emit_debug_insn_before_noloc): Likewise.
3135 (emit_insn_after): Likewise.
3136 (emit_jump_insn_after): Likewise.
3137 (emit_call_insn_after): Likewise.
3138 (emit_debug_insn_after): Likewise.
3139 (set_insn_deleted): Likewise for param.
3141 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3144 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3145 floating point modes, so that IFmode is numerically greater than
3146 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3147 to declare the ordering. This prevents IFmode from being
3148 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3149 machine. Include rs6000-modes.h to share the fractional values
3150 between genmodes* and the rest of the compiler.
3154 * config/rs6000/rs6000-modes.h: New file.
3155 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3156 meaning of rs6000_long_double_size so that 126..128 selects an
3157 appropriate 128-bit floating point type.
3158 (rs6000_option_override_internal): Likewise.
3159 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3160 (TARGET_LONG_DOUBLE_128): Change the meaning of
3161 rs6000_long_double_size so that 126..128 selects an appropriate
3162 128-bit floating point type.
3163 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3164 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3165 source and destination to match the standard usage.
3166 (truncifkf2): Likewise.
3167 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3168 ISA 2.07 to use an explicit clobber, instead of passing in a
3170 (copysign<mode>3_soft): Likewise.
3172 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3174 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3175 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3176 (vect_slp_analyze_instance_dependence): Likewise.
3177 (vect_enhance_data_refs_alignment): Likewise.
3178 (vect_analyze_data_refs_alignment): Likewise.
3179 (vect_slp_analyze_and_verify_instance_alignment
3180 (vect_analyze_data_ref_accesses): Likewise.
3181 (vect_prune_runtime_alias_test_list): Likewise.
3182 (vect_analyze_data_refs): Likewise.
3183 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3184 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3185 (vect_analyze_scalar_cycles_1): Likewise.
3186 (vect_get_loop_niters): Likewise.
3187 (vect_analyze_loop_form_1): Likewise.
3188 (vect_update_vf_for_slp): Likewise.
3189 (vect_analyze_loop_operations): Likewise.
3190 (vect_analyze_loop): Likewise.
3191 (vectorizable_induction): Likewise.
3192 (vect_transform_loop): Likewise.
3193 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3194 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3195 (vect_make_slp_decision): Likewise.
3196 (vect_detect_hybrid_slp): Likewise.
3197 (vect_slp_analyze_operations): Likewise.
3198 (vect_slp_bb): Likewise.
3199 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3200 (vectorizable_bswap): Likewise.
3201 (vectorizable_call): Likewise.
3202 (vectorizable_simd_clone_call): Likewise.
3203 (vectorizable_conversion): Likewise.
3204 (vectorizable_assignment): Likewise.
3205 (vectorizable_shift): Likewise.
3206 (vectorizable_operation): Likewise.
3207 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3209 2018-06-18 Martin Sebor <msebor@redhat.com>
3211 PR tree-optimization/81384
3212 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3213 * builtins.c (expand_builtin_strnlen): New function.
3214 (expand_builtin): Call it.
3215 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3216 * builtins.def (BUILT_IN_STRNLEN): New.
3217 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3218 Warn for bounds in excess of maximum object size.
3219 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3220 single-value ranges. Handle strnlen.
3221 (handle_builtin_strlen): Handle strnlen.
3222 (strlen_check_and_optimize_stmt): Same.
3223 * doc/extend.texi (Other Builtins): Document strnlen.
3225 2018-06-18 Maya Rashish <coypu@sdf.org>
3227 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3228 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3229 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3231 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3233 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3235 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3237 * tree.c (escaped_string::escape): Replace cast to char * by
3238 const_cast<char *> (unescaped).
3240 2018-06-18 Nick Clifton <nickc@redhat.com>
3243 * tree.c (escaped_string): New class. Converts an unescaped
3244 string into its escaped equivalent.
3245 (warn_deprecated_use): Use the new class to convert the
3246 deprecation message, if present.
3247 (test_escaped_strings): New self test.
3248 (test_c_tests): Add test_escaped_strings.
3249 * doc/extend.texi (deprecated): Add a note that the
3250 deprecation message is affected by the -fmessage-length
3251 option, and that control characters will be escaped.
3252 (#pragma GCC error): Document this pragma.
3253 (#pragma GCC warning): Likewise.
3254 * doc/invoke.texi (-fmessage-length): Document this option's
3255 effect on the #warning and #error preprocessor directives and
3256 the deprecated attribute.
3258 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3260 * tree.c (decl_value_expr_lookup): Revert latest change.
3261 (decl_value_expr_insert): Likewise.
3263 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3265 * gimplify.c (nonlocal_vlas): Delete.
3266 (nonlocal_vla_vars): Likewise.
3267 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3269 (gimplify_body): Do not create and destroy nonlocal_vlas.
3270 * tree-nested.c: Include diagnostic.h.
3271 (use_pointer_in_frame): Tweak.
3272 (lookup_field_for_decl): Add assertion and declare the transformation.
3273 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3274 internal error when the reference is in a wrong context. Do not
3275 create a debug decl by default.
3276 (note_nonlocal_block_vlas): Delete.
3277 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3278 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3279 create a debug decl by default.
3280 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3281 call is in a wrong context.
3282 (fixup_vla_decls): New function.
3283 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3284 debug variables were created.
3285 * tree.c (decl_value_expr_lookup): Add checking assertion.
3286 (decl_value_expr_insert): Likewise.
3288 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3291 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3292 * tree-scalar-evolution.c (interpret_expr): Likewise.
3293 (expression_expensive_p): Likewise.
3294 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3295 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3296 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3297 (ssa_defined_by_minus_one_stmt_p): New.
3299 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3302 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3303 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3304 * dojump.c (do_jump): Likewise.
3305 * expr.c (expand_expr_real_2): Check operand type's sign.
3306 * fold-const.c (const_unop): Handle ABSU_EXPR.
3307 (fold_abs_const): Likewise.
3308 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3309 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3310 (strip_sign_op_1): Likesise.
3311 * match.pd: Add new pattern to generate ABSU_EXPR.
3312 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3313 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3314 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3315 * tree-inline.c (estimate_operator_cost): Likewise.
3316 * tree-pretty-print.c (dump_generic_node): Likewise.
3317 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3318 * tree.def (ABSU_EXPR): New.
3320 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3323 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3324 documented as preserved for backward compatibility only.
3325 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3327 PR rtl-optimization/86108
3328 * bb-reorder.c (create_forwarder_block): Renamed to ...
3329 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3330 jump from new landing pad to the second part.
3331 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3334 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3337 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3338 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3339 Only call store_expr for halves if the mode is the same.
3342 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3343 Fix up comment formatting.
3345 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3347 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3348 (typed_splay_tree::closure,
3349 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3350 (typed_splay_tree::typed_splay_tree,
3351 typed_splay_tree::operator =): Declared private.
3352 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3353 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3354 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3355 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3356 typed_splay_tree::splay_tree_splay,
3357 typed_splay_tree::splay_tree_foreach_helper,
3358 typed_splay_tree::splay_tree_insert,
3359 typed_splay_tree::splay_tree_remove,
3360 typed_splay_tree::splay_tree_lookup,
3361 typed_splay_tree::splay_tree_predecessor,
3362 typed_splay_tree::splay_tree_successor,
3363 typed_splay_tree::splay_tree_min,
3364 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3365 (typed_splay_tree::root, typed_splay_tree::comp,
3366 typed_splay_tree::delete_key,
3367 typed_splay_tree::delete_value): New data members.
3368 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3369 typed_splay_tree::remove.
3371 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3373 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3374 -mginv and -mno-ginv to the assembler.
3375 * config/mips/mips.opt (-mcrc): New option.
3377 * doc/invoke.text (-mcrc): Document.
3380 2018-06-15 Nick Clifton <nickc@redhat.com>
3383 * tree.c (escaped_string): New class. Converts an unescaped
3384 string into its escaped equivalent.
3385 (warn_deprecated_use): Use the new class to convert the
3386 deprecation message, if present.
3387 (test_escaped_strings): New self test.
3388 (test_c_tests): Add test_escaped_strings.
3389 * doc/extend.texi (deprecated): Add a note that the
3390 deprecation message is affected by the -fmessage-length
3391 option, and that control characters will be escaped.
3392 (#pragma GCC error): Document this pragma.
3393 (#pragma GCC warning): Likewise.
3394 * doc/invoke.texi (-fmessage-length): Document this option's
3395 effect on the #warning and #error preprocessor directives and
3396 the deprecated attribute.
3398 2018-06-15 Richard Biener <rguenther@suse.de>
3400 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3401 here, also noting vector size used.
3402 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3403 size used in MSG_OPTIMIZED_LOCATIONS dump.
3404 (pass_slp_vectorize::execute): Adjust.
3406 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3409 * config/arc/arc.c (arc_return_address_register): Fix
3412 2018-06-15 Richard Biener <rguenther@suse.de>
3415 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3416 leave useless conversion stripping to force_gimple_operand_gsi.
3417 (gimplify_build2): Likewise.
3418 (gimplify_build1): Likewise.
3420 2018-06-15 Richard Biener <rguenther@suse.de>
3423 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3424 before adjusting their block.
3426 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3428 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3429 multilibs for *-*-rtems*.
3430 * config/riscv/t-rtems: New file.
3432 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3435 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3436 unsigned_type_for returns NULL.
3439 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3440 subregs of multi-word pseudos unless the float mode has word size.
3442 2018-06-14 Richard Biener <rguenther@suse.de>
3445 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3446 duplicate and harmful type_hash_canon.
3447 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3449 2018-06-14 Richard Biener <rguenther@suse.de>
3452 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3455 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3457 * config/rtems.h (STDINT_LONG32): Define.
3459 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3460 Prachi Godbole <prachi.godbole@imgtec.com>
3462 * config/mips/mips-cpus.def: Define P6600.
3463 * config/mips/mips-tables.opt: Regenerate.
3464 * config/mips/mips.c (mips_ucbranch_type): New enum.
3465 (mips_rtx_cost_data): Add support for P6600.
3466 (mips_issue_rate): Likewise.
3467 (mips_multipass_dfa_lookahead): Likewise.
3468 (mips_avoid_hazard): Likewise.
3469 (mips_reorg_process_insns): Likewise.
3470 (mips_classify_branch_p6600): New function.
3471 * config/mips/mips.h (TUNE_P6600): New define.
3472 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3473 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3474 * config/mips/mips.md: Include p6600.md.
3475 (processor): Add p6600.
3476 * config/mips/p6600.md: New file.
3477 * doc/invoke.texi: Add p6600 to supported architectures.
3479 2018-06-13 Martin Sebor <msebor@redhat.com>
3481 PR tree-optimization/86114
3482 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3484 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3486 2018-06-13 Richard Biener <rguenther@suse.de>
3488 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3489 Properly set vector type of the intermediate stmt.
3490 * tree-vect-stmts.c (vectorizable_operation): The destination
3491 var always has vectype_out type.
3493 2018-06-13 Jeff Law <law@redhat.com>
3495 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3496 integer 0 for argument to print_rtl_with_bb.
3497 (rl78_reorg): Likewise.
3499 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3501 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3502 from rtx to rtx_insn *.
3503 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3505 (add_sched_insns_for_speculation): Likewise for local "target",
3506 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3507 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3508 from rtx_insn ** to rtx_code_label **.
3509 (reorg_emit_nops): Likewise.
3510 (c6x_reorg): Likewise for local "call_labels".
3511 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3513 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3514 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3515 the loops over LABEL_REFS.
3516 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3518 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3519 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3520 (split_branches): Strengthen local "olabel" from rtx to
3521 rtx_insn *, adding a safe_as_a cast.
3522 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3524 (add_insn_after): Likewise for first two params.
3525 (add_insn_before): Likewise.
3526 (remove_insn): Likewise for param.
3527 (emit_pattern_before_noloc): Likewise for second and third params.
3528 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3529 (emit_call_insn_before_noloc): Likewise.
3530 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3532 (emit_barrier_before): Likewise.
3533 (emit_label_before): Strengthen "label" param from "rtx" to
3534 "rtx_code_label *". Strengthen "before" param from "rtx" to
3536 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3538 (emit_pattern_after_noloc): Likewise.
3539 (emit_insn_after_noloc): Likewise.
3540 (emit_jump_insn_after_noloc): Likewise.
3541 (emit_call_insn_after_noloc): Likewise.
3542 (emit_debug_insn_after_noloc): Likewise.
3543 (emit_barrier_after): Likewise.
3544 (emit_label_after): Likewise for both params.
3545 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3546 "loc" param from "int" to "location_t".
3547 (emit_insn_after_setloc): Likewise.
3548 (emit_jump_insn_after_setloc): Likewise.
3549 (emit_call_insn_after_setloc): Likewise.
3550 (emit_debug_insn_after_setloc): Likewise.
3551 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3552 "loc" param from "int" to "location_t".
3553 (emit_pattern_before): Convert NULL_RTX to NULL.
3554 (emit_insn_before_setloc): Convert "loc" param from "int" to
3556 (emit_jump_insn_before_setloc): Likewise.
3557 (emit_call_insn_before_setloc): Likewise.
3558 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3559 rtx_insn *. Convert "loc" param from "int" to "location_t".
3560 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3561 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3562 Convert 3rd param from "int" to "location_t".
3563 (emit_barrier_before, emit_barrier_after, next_real_insn):
3564 Strengthen param from rtx to rtx_insn *.
3565 (emit_label_before): Strengthen 1st param from "rtx" to
3566 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3568 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3569 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3570 Strengthen 2nd param from "rtx" to "rtx_insn *".
3571 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3572 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3573 Likewise. Convert 3rd param from "int" to "location_t".
3574 (emit_label_after): Strengthen 1st param from "rtx" to
3576 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3578 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3579 from "rtx" to "rtx_insn *".
3581 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3583 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3584 bodies streamed in with -Q.
3585 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3586 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3587 * gimple-streamer-out.c: Include gimple-pretty-print.h
3588 (output_bb): Dump stmts streamed.
3589 * lto-section-out.c: Include print-tree.h
3590 (lto_begin_section): Dump sections created.
3591 (lto_output_decl_index): Dump decl encoded.
3592 * lto-streamer-out.c: Include print-tree.h
3593 (create_output_block): Dump output block created.
3594 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3595 (output_function): Dump function output.
3596 (output_constructor): Dump constructor streamed.
3597 (write_global_stream): Output indexes encoded.
3598 (produce_asm_for_decls): Dump streams encoded.
3599 * lto-streamer.c (streamer_dump_file): New global var.
3600 * lto-streamer.h (streamer_dump_file): Declare.
3601 * passes.c (ipa_write_summaries): Initialize streamer dump.
3602 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3605 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3608 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3609 offsets for register save directives. Emit a second batch of save
3610 directives, if need be, when the function accesses prior frames.
3612 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3614 * config/arc/fpu.md (fmasf4): Force operand to register.
3615 (fnmasf4): Likewise.
3617 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3619 * config/arc/arc-protos.h (arc_pad_return): Remove.
3620 * config/arc/arc.c (machine_function): Remove force_short_suffix
3622 (arc_print_operand): Adjust printing of '&'.
3623 (arc_verify_short): Remove conditional printing of short suffix.
3624 (arc_final_prescan_insn): Remove reference to size_reason.
3625 (pad_return): New function.
3626 (arc_reorg): Call pad_return.
3627 (arc_pad_return): Remove.
3628 (arc_init_machine_status): Remove reference to force_short_suffix.
3629 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3630 (attr length): When attribute iscompact is true force to 2
3631 regardless; in the case of maybe check if we want to force the
3632 instruction to have 4 bytes length.
3633 (nopv): Change it to generate 4 byte long nop as well.
3634 (blockage): New pattern.
3635 (simple_return): Remove call to arc_pad_return.
3636 (p_return_i): Likewise.
3638 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3640 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3642 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3644 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3647 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3649 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3650 for ARC700 and ARCv2.
3652 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3655 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3656 operands[2] instead of operands[1].
3659 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3661 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3662 case, check whether the outer register overlaps an unallocatable
3663 register, not just whether it fits the required class.
3665 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3667 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3669 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3670 (memrefs_conflict_p): Likewise.
3671 (init_alias_analysis): Likewise.
3672 * cfgexpand.c (expand_debug_expr): Likewise.
3673 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3674 * cse.c (fold_rtx): Likewise.
3675 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3676 * expr.c (emit_block_move_hints): Likewise.
3677 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3678 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3679 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3680 (emit_group_store): Likewise.
3681 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3682 to read the PRE/POST_MODIFY increment.
3683 * calls.c (store_one_arg): Use strip_offset.
3684 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3686 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3688 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3689 (simplify_binary_operation_1): Extend CONST_INT handling to
3691 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3692 than a HOST_WIDE_INT.
3693 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3695 (adjust_mems, add_stores): Update accodingly.
3696 (vt_canonicalize_addr): Track polynomial offsets.
3697 (emit_note_insn_var_location): Likewise.
3698 (vt_add_function_parameter): Likewise.
3699 (vt_initialize): Likewise.
3701 2018-06-12 Jeff Law <law@redhat.com>
3703 * config.gcc (alpha*-*-freebsd*): Remove.
3704 * config/alpha/freebsd.h: Remove.
3706 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3709 * spellcheck-tree.c (levenshtein_distance): Rename to...
3710 (get_edit_distance): ...this, and update for underlying renaming.
3711 * spellcheck-tree.h (levenshtein_distance): Rename to...
3712 (get_edit_distance): ...this.
3713 * spellcheck.c (levenshtein_distance): Rename to...
3714 (get_edit_distance): ...this. Convert from Levenshtein distance
3715 to Damerau-Levenshtein distance by supporting transpositions of
3716 adjacent characters. Rename "v1" to "v_next" and "v0" to
3718 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3719 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3720 update for underlying renaming.
3721 (selftest::levenshtein_distance_unit_test): Rename to...
3722 (selftest::test_get_edit_distance_unit): ...this, and update for
3723 underlying renaming.
3724 (selftest::test_find_closest_string): Add example from PR 69968
3725 where transposition helps
3726 (selftest::test_metric_conditions): Update for renaming.
3727 (selftest::test_metric_conditions): Likewise.
3728 (selftest::spellcheck_c_tests): Likewise.
3729 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3730 (get_edit_distance): ...this.
3731 (best_match::consider): Update for renaming.
3733 2018-06-12 Martin Sebor <msebor@redhat.com>
3735 PR tree-optimization/85259
3736 * builtins.c (compute_objsize): Handle constant offsets.
3737 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3738 true iff a warning has been issued.
3739 * gimple.h (gimple_nonartificial_location): New function.
3740 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3741 gimple_nonartificial_location and handle -Wno-system-headers.
3742 (handle_builtin_stxncpy): Same.
3744 2018-06-12 Martin Sebor <msebor@redhat.com>
3747 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3749 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3751 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3752 BUILTIN_VEC_XST entries for pointer to double and long long.
3754 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3757 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3759 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3762 2018-06-12 Martin Liska <mliska@suse.cz>
3764 * doc/options.texi: Document IntegerRange.
3766 2018-06-12 Martin Liska <mliska@suse.cz>
3768 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3769 * opt-functions.awk: Handle Deprecated flag.
3770 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3772 (read_cmdline_option): Report warning for a deprecated option.
3773 * opts.h (struct cl_option): Add new field cl_deprecated.
3774 (CL_ERR_DEPRECATED): New.
3776 2018-06-12 Martin Liska <mliska@suse.cz>
3778 * doc/options.texi: Document Deprecated option flag.
3780 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3782 * config/arc/arc-arch.h (arc_extras): New enum.
3783 (arc_cpu_t):Add field extra.
3784 (arc_cpu_types): Consider the extras.
3785 * config/arc/arc-cpus.def: Add extras info.
3786 * config/arc/arc-opts.h (processor_type): Consider extra field.
3787 * config/arc/arc.c (arc_override_options): Handle extra field.
3789 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3791 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3792 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3794 * config/arc/arc.c (arc_init): Update pic warning.
3795 (irq_range): Update irq range parsing warnings.
3796 (arc_override_options): Update various warning messages.
3797 (arc_handle_aux_attribute): Likewise.
3799 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3801 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3803 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3805 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3806 compiler messages directives.
3808 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3810 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3811 * config/mips/mips-tables.opt: Regenerate.
3812 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3814 * doc/invoke.texi: Document -march=i6500.
3816 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3818 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3819 (i6400_gpmul): Add cpu_unit.
3820 (i6400_gpdiv): Likewise.
3821 (i6400_msa_add_d): Update reservations.
3822 (i6400_msa_int_add) Likewise.
3823 (i6400_msa_short_logic3) Likewise.
3824 (i6400_msa_short_logic2) Likewise.
3825 (i6400_msa_short_logic) Likewise.
3826 (i6400_msa_move) Likewise.
3827 (i6400_msa_cmp) Likewise.
3828 (i6400_msa_short_float2) Likewise.
3829 (i6400_msa_div_d) Likewise.
3830 (i6400_msa_long_logic1) Likewise.
3831 (i6400_msa_long_logic2) Likewise.
3832 (i6400_msa_mult) Likewise.
3833 (i6400_msa_long_float2) Likewise.
3834 (i6400_msa_long_float4) Likewise.
3835 (i6400_msa_long_float5) Likewise.
3836 (i6400_msa_long_float8) Likewise.
3837 (i6400_fpu_fadd): Include frint type.
3838 (i6400_fpu_store): New define_insn_reservation.
3839 (i6400_fpu_load): Likewise.
3840 (i6400_fpu_move): Likewise.
3841 (i6400_fpu_fcmp): Likewise.
3842 (i6400_fpu_fmadd): Likewise.
3843 (i6400_int_mult): Include imul3nc type and update reservation.
3844 (i6400_int_div): Include idiv3 type and update reservation.
3845 (i6400_int_load): Update to check type not move_type.
3846 (i6400_int_store): Likewise.
3847 (i6400_int_prefetch): Set zero latency.
3849 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3851 * gcc.c: Document new %@{...} sequence.
3852 (LINK_COMMAND_SPEC): Use it for the -L switches.
3853 (cpp_unique_options): Use it for the -I switches.
3854 (at_file_argbuf): New global variable.
3855 (in_at_file): Likewise.
3856 (alloc_args): Create at_file_argbuf.
3857 (clear_args): Truncate at_file_argbuf.
3858 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3859 (open_at_file): New function.
3860 (close_at_file): Likewise.
3861 (create_at_file): Delete.
3862 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3865 (validate_switches_from_spec): Deal with %@{...} sequence.
3866 (validate_switches): Likewise.
3867 (driver::finalize): Call clear_args.
3869 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3871 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3873 2018-06-11 Martin Sebor <msebor@redhat.com>
3875 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3876 (Wno-class-memaccess): Add @opindex.
3877 (Wno-templates, Wno-multiple-inheritance): Same.
3878 (Wno-virtual-inheritance, Wno-namespaces): Same.
3879 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3880 (Wno-format-overflow, Wno-format-truncation): Same.
3881 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3882 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3883 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3884 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3885 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3886 (Wno-misspelled-isr): Same.
3888 2018-06-11 Martin Sebor <msebor@redhat.com>
3890 * PR tree-optimization/86083
3891 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3893 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3895 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3897 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3900 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3901 on the correct operand.
3902 (*movdi_internal64): Ditto.
3904 2018-06-11 Martin Liska <mliska@suse.cz>
3906 PR tree-optimization/86089
3907 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3909 2018-06-11 Julia Koval <julia.koval@intel.com>
3911 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3912 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3913 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3915 2018-06-11 Olivier Hainque <hainque@adacore.com>
3917 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3918 for Ada with strict dwarf2.
3920 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3923 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3926 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3928 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3930 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3932 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3935 2018-06-08 Martin Liska <mliska@suse.cz>
3937 * tree-cfg.h (debug_function): Fix argument type to match
3940 2018-06-08 Martin Liska <mliska@suse.cz>
3942 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3943 Remove usage of MPX-related (and removed) fields.
3944 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3946 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3948 * cfg.c (debug): Use TDF_NONE rather than 0.
3949 * cfghooks.c (debug): Likewise.
3950 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3951 (struct dump_option_value_info): Convert to...
3952 (struct kv_pair): ...this template type.
3953 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3955 (optinfo_verbosity_options): Likewise.
3956 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3958 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3959 than int for "optgroup_flags" param.
3960 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3962 (dump_dec): Likewise.
3963 (dump_finish): Use TDF_NONE rather than 0.
3964 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3965 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3966 than 0. Update for change to option_ptr.
3967 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3968 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3969 0. Update for changes to optinfo_verbosity_options and
3971 (opt_info_switch_p): Convert optgroup_flags from int to
3973 (dump_basic_block): Use dump_flags_t rather than int
3974 for "dump_kind" param.
3975 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3976 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3977 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3978 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3979 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3980 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3981 TDF_NONE): Convert from macros to...
3982 (enum dump_flag): ...this new enum.
3983 (dump_flags_t): Update to use enum.
3984 (operator|, operator&, operator~, operator|=, operator&=):
3985 Implement for dump_flags_t.
3986 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3987 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3988 Convert from macros to...
3989 (enum optgroup_flag): ...this new enum.
3990 (optgroup_flags_t): New typedef.
3991 (operator|, operator|=): Implement for optgroup_flags_t.
3992 (struct dump_file_info): Convert field "alt_flags" to
3993 dump_flags_t. Convert field "optgroup_flags" to
3995 (dump_basic_block): Use dump_flags_t rather than int for param.
3996 (dump_generic_expr_loc): Likewise.
3997 (dump_dec): Likewise.
3998 (dump_register): Convert param "optgroup_flags" to
4000 (opt_info_enable_passes): Likewise.
4001 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
4003 * gimple-pretty-print.c (debug): Likewise.
4004 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
4005 (merged_store_group::apply_stores): Likewise.
4006 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
4007 * gimple.c (verify_gimple_pp): Likewise.
4008 * graphite-poly.c (print_pbb_body): Likewise.
4009 * passes.c (pass_manager::register_one_dump_file): Convert
4010 local "optgroup_flags" to optgroup_flags_t.
4011 * print-tree.c (print_node): Use TDF_NONE rather than 0.
4013 (debug_body): Likewise.
4014 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
4015 to optgroup_flags_t.
4016 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
4018 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4019 (convert_mult_to_fma): Likewise.
4020 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4021 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4022 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4023 "dump_kind" to dump_flags_t.
4025 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4027 * config/rs6000/rs6000.c (min, max): Delete.
4029 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4031 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4032 -mabi=spe and -mabi=no-spe.
4034 2018-06-08 Martin Liska <mliska@suse.cz>
4036 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4037 where we expect an existing summary.
4039 2018-06-08 Martin Liska <mliska@suse.cz>
4041 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4042 * ipa-inline.h (estimate_edge_growth): Likewise.
4044 2018-06-08 Martin Liska <mliska@suse.cz>
4046 * cgraph.c (function_version_hasher::hash): Use
4047 cgraph_node::get_uid ().
4048 (function_version_hasher::equal):
4049 * cgraph.h (cgraph_node::get_uid): New method.
4050 * ipa-inline.c (update_caller_keys): Use
4051 cgraph_node::get_uid ().
4052 (update_callee_keys): Likewise.
4053 * ipa-utils.c (searchc): Likewise.
4054 (ipa_reduced_postorder): Likewise.
4055 * lto-cgraph.c (input_node): Likewise.
4056 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4057 * symbol-summary.h (symtab_insertion): Likewise.
4058 (symtab_removal): Likewise.
4059 (symtab_duplication): Likewise.
4060 * tree-pretty-print.c (dump_function_header): Likewise.
4061 * tree-sra.c (convert_callers_for_node): Likewise.
4063 2018-06-08 Martin Liska <mliska@suse.cz>
4065 * cgraph.c (symbol_table::create_edge): Always assign a new
4067 (symbol_table::free_edge): Do not recycle numbers.
4068 * cgraph.h (cgraph_edge::get): New method.
4069 * symbol-summary.h (symtab_removal): Use it.
4070 (symtab_duplication): Likewise.
4071 (call_summary::hashable_uid): Remove.
4073 2018-06-08 Martin Liska <mliska@suse.cz>
4075 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4076 (initialize_growth_caches): Remove.
4077 (free_growth_caches): Likewise.
4078 (do_estimate_edge_time): Use edge_growth_cache.
4079 (do_estimate_edge_size): Likewise.
4080 (do_estimate_edge_hints): Likewise.
4081 * ipa-inline.c (reset_edge_caches): Likewise.
4082 (recursive_inlining): Likewise.
4083 (inline_small_functions): Likewise.
4084 * ipa-inline.h (initialize_growth_caches): Remove.
4085 (estimate_edge_size): Likewise.
4086 (estimate_edge_time): Likewise.
4087 (estimate_edge_hints): Likewise.
4088 (reset_edge_growth_cache): Likewise.
4089 * symbol-summary.h (call_summary::remove): New method.
4091 2018-06-08 Martin Liska <mliska@suse.cz>
4093 * ipa-cp.c (class edge_clone_summary): New summary.
4094 (grow_edge_clone_vectors): Remove.
4095 (ipcp_edge_duplication_hook): Remove.
4096 (class edge_clone_summary_t): New call_summary class.
4097 (ipcp_edge_removal_hook): Remove.
4098 (edge_clone_summary_t::duplicate): New function.
4099 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4100 (create_specialized_node): Likewise.
4101 (ipcp_driver): Initialize edge_clone_summaries and do not
4104 2018-06-08 Martin Liska <mliska@suse.cz>
4106 * symbol-summary.h (get): New function.
4107 (call_summary::m_initialize_when_cloning): New class member.
4109 2018-06-08 Martin Liska <mliska@suse.cz>
4111 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4112 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4113 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4114 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4115 (remove_cgraph_node_from_order): Use the removed_nodes set.
4116 (do_per_function_toporder): Likwise.
4117 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4118 instead of summary_uid.
4119 (symtab_removal): Likewise.
4120 (symtab_duplication): Likewise.
4122 2018-06-08 Martin Liska <mliska@suse.cz>
4124 * ipa-cp.c (ipcp_store_bits_results): Use
4125 ipcp_transformation_sum.
4126 (ipcp_store_vr_results): Likewise.
4127 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4129 (ipcp_transformation_initialize): ... this.
4130 (ipa_set_node_agg_value_chain):
4131 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4132 (write_ipcp_transformation_info): Likewise.
4133 (read_ipcp_transformation_info): Likewise.
4134 (ipcp_update_bits): Likewise.
4135 (ipcp_update_vr): Likewise.
4136 (ipcp_transform_function): Likewise.
4137 * ipa-prop.h: Rename ipcp_transformation_summary to
4138 ipcp_transformation.
4139 (class ipcp_transformation_t): New function summary.
4140 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4141 (ipa_get_agg_replacements_for_node): Likewise.
4143 2018-06-08 Martin Liska <mliska@suse.cz>
4145 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4147 (class funct_state_summary_t): New function_summary class.
4148 (has_function_state): Remove.
4149 (get_function_state): Likewise.
4150 (set_function_state): Likewise.
4151 (add_new_function): Likewise.
4152 (funct_state_summary_t::insert): New function.
4153 (duplicate_node_data): Remove.
4154 (remove_node_data): Remove.
4155 (funct_state_summary_t::duplicate): New function.
4156 (register_hooks): Create new funct_state_summaries.
4157 (pure_const_generate_summary): Use it.
4158 (pure_const_write_summary): Likewise.
4159 (pure_const_read_summary): Likewise.
4160 (propagate_pure_const): Likewise.
4161 (propagate_nothrow): Likewise.
4162 (dump_malloc_lattice): Likewise.
4163 (propagate_malloc): Likewise.
4164 (execute): Do not register hooks, just remove summary
4166 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4169 2018-06-08 Martin Liska <mliska@suse.cz>
4171 * ipa-reference.c (remove_node_data): Remove.
4172 (duplicate_node_data): Likewise.
4173 (class ipa_ref_var_info_summary_t): New class.
4174 (class ipa_ref_opt_summary_t): Likewise.
4175 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4176 (get_reference_optimization_summary): Use
4177 ipa_ref_opt_sum_summaries.
4178 (set_reference_vars_info): Remove.
4179 (set_reference_optimization_summary): Likewise.
4180 (ipa_init): Create summaries.
4181 (init_function_info): Use function summary.
4182 (ipa_ref_opt_summary_t::duplicate): New function.
4183 (ipa_ref_opt_summary_t::remove): New function.
4184 (get_read_write_all_from_node): Fix GNU coding style.
4185 (propagate): Use function summary.
4186 (write_node_summary_p): Fix GNU coding style.
4187 (stream_out_bitmap): Likewise.
4188 (ipa_reference_read_optimization_summary): Use function summary.
4189 (ipa_reference_c_finalize): Do not release hooks.
4191 2018-06-08 Martin Liska <mliska@suse.cz>
4193 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4194 (analyze_function_body): Extract multiple calls of get_create.
4195 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4196 * ipa-inline.c (recursive_inlining): Use ::get method.
4197 * ipa-inline.h (estimate_edge_growth): Likewise.
4199 2018-06-08 Martin Liska <mliska@suse.cz>
4201 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4203 (hsa_function_summary::hsa_function_summary): Use the new enum
4205 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4206 * hsa-gen.c (hsa_get_host_function): Likewise.
4207 (get_brig_function_name): Likewise.
4208 * ipa-hsa.c (process_hsa_functions): Likewise.
4209 (ipa_hsa_write_summary): Likewise.
4210 * symbol-summary.h (symtab_duplication): Use ::get function/
4211 (get): New function.
4213 2018-06-08 Martin Liska <mliska@suse.cz>
4215 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4217 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4218 (hsa_register_kernel): Likewise.
4219 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4220 * hsa-gen.c (hsa_get_host_function): Likewise.
4221 (get_brig_function_name): Likewise.
4222 (generate_hsa): Likewise.
4223 (pass_gen_hsail::execute): Likewise.
4224 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4225 (devirtualization_time_bonus): Likewise.
4226 (ipcp_propagate_stage): Likewise.
4227 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4228 (edge_set_predicate): Likewise.
4229 (evaluate_conditions_for_known_args): Likewise.
4230 (evaluate_properties_for_edge): Likewise.
4231 (ipa_fn_summary::reset): Likewise.
4232 (ipa_fn_summary_t::duplicate): Likewise.
4233 (dump_ipa_call_summary): Likewise.
4234 (ipa_dump_fn_summary): Likewise.
4235 (analyze_function_body): Likewise.
4236 (compute_fn_summary): Likewise.
4237 (estimate_edge_devirt_benefit): Likewise.
4238 (estimate_edge_size_and_time): Likewise.
4239 (estimate_calls_size_and_time): Likewise.
4240 (estimate_node_size_and_time): Likewise.
4241 (inline_update_callee_summaries): Likewise.
4242 (remap_edge_change_prob): Likewise.
4243 (remap_edge_summaries): Likewise.
4244 (ipa_merge_fn_summary_after_inlining): Likewise.
4245 (ipa_update_overall_fn_summary): Likewise.
4246 (read_ipa_call_summary): Likewise.
4247 (inline_read_section): Likewise.
4248 (write_ipa_call_summary): Likewise.
4249 (ipa_fn_summary_write): Likewise.
4250 (ipa_free_fn_summary): Likewise.
4251 * ipa-hsa.c (process_hsa_functions): Likewise.
4252 (ipa_hsa_write_summary): Likewise.
4253 (ipa_hsa_read_section): Likewise.
4254 * ipa-icf.c (sem_function::merge): Likewise.
4255 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4256 (do_estimate_edge_time): Likewise.
4257 (estimate_size_after_inlining): Likewise.
4258 (estimate_growth): Likewise.
4259 (growth_likely_positive): Likewise.
4260 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4261 (inline_call): Likewise.
4262 * ipa-inline.c (caller_growth_limits): Likewise.
4263 (can_inline_edge_p): Likewise.
4264 (can_inline_edge_by_limits_p): Likewise.
4265 (compute_uninlined_call_time): Likewise.
4266 (compute_inlined_call_time): Likewise.
4267 (want_inline_small_function_p): Likewise.
4268 (edge_badness): Likewise.
4269 (update_caller_keys): Likewise.
4270 (update_callee_keys): Likewise.
4271 (recursive_inlining): Likewise.
4272 (inline_small_functions): Likewise.
4273 (inline_to_all_callers_1): Likewise.
4274 (dump_overall_stats): Likewise.
4275 (early_inline_small_functions): Likewise.
4276 (early_inliner): Likewise.
4277 * ipa-inline.h (estimate_edge_growth): Likewise.
4278 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4279 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4280 * ipa-prop.h (IPA_NODE_REF): Likewise.
4281 (IPA_EDGE_REF): Likewise.
4282 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4283 (propagate_malloc): Likewise.
4284 * ipa-split.c (execute_split_functions): Likewise.
4285 * symbol-summary.h: Rename get to get_create.
4287 (get_create): Likewise.
4288 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4290 2018-06-08 Martin Liska <mliska@suse.cz>
4292 * symbol-summary.h (release): Move definition out of class
4294 (symtab_removal): Likewise.
4295 (symtab_duplication): Likewise.
4297 2018-06-08 Martin Liska <mliska@suse.cz>
4299 * symbol-summary.h (function_summary): Move constructor
4300 implementation out of class declaration.
4301 (release): Likewise.
4302 (symtab_insertion): Likewise.
4303 (symtab_removal): Likewise.
4304 (symtab_duplication): Likewise.
4307 2018-06-08 Martin Liska <mliska@suse.cz>
4309 * Makefile.in: Remove support for MPX (macros, related functions,
4310 fields in cgraph_node, ...).
4311 * builtin-types.def (BT_BND): Likewise.
4312 (BT_FN_BND_CONST_PTR): Likewise.
4313 (BT_FN_CONST_PTR_BND): Likewise.
4314 (BT_FN_VOID_PTR_BND): Likewise.
4315 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4316 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4317 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4318 (expand_builtin_mempcpy_with_bounds): Likewise.
4319 (expand_builtin_memset_with_bounds): Likewise.
4320 (expand_builtin_memset_args): Likewise.
4321 (std_expand_builtin_va_start): Likewise.
4322 (expand_builtin): Likewise.
4323 (expand_builtin_with_bounds): Likewise.
4324 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4325 (DEF_LIB_BUILTIN_CHKP): Likewise.
4326 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4327 (DEF_CHKP_BUILTIN): Likewise.
4328 (BUILT_IN_MEMCPY): Likewise.
4329 (BUILT_IN_MEMMOVE): Likewise.
4330 (BUILT_IN_MEMPCPY): Likewise.
4331 (BUILT_IN_MEMSET): Likewise.
4332 (BUILT_IN_STPCPY): Likewise.
4333 (BUILT_IN_STRCAT): Likewise.
4334 (BUILT_IN_STRCHR): Likewise.
4335 (BUILT_IN_STRCPY): Likewise.
4336 (BUILT_IN_STRLEN): Likewise.
4337 (BUILT_IN_MEMCPY_CHK): Likewise.
4338 (BUILT_IN_MEMMOVE_CHK): Likewise.
4339 (BUILT_IN_MEMPCPY_CHK): Likewise.
4340 (BUILT_IN_MEMSET_CHK): Likewise.
4341 (BUILT_IN_STPCPY_CHK): Likewise.
4342 (BUILT_IN_STRCAT_CHK): Likewise.
4343 (BUILT_IN_STRCPY_CHK): Likewise.
4344 * calls.c (store_bounds): Likewise.
4345 (emit_call_1): Likewise.
4346 (special_function_p): Likewise.
4347 (maybe_warn_nonstring_arg): Likewise.
4348 (initialize_argument_information): Likewise.
4349 (finalize_must_preallocate): Likewise.
4350 (compute_argument_addresses): Likewise.
4351 (expand_call): Likewise.
4352 * cfgexpand.c (expand_call_stmt): Likewise.
4353 (expand_return): Likewise.
4354 (expand_gimple_stmt_1): Likewise.
4355 (pass_expand::execute): Likewise.
4356 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4357 (cgraph_node::remove): Likewise.
4358 (cgraph_node::dump): Likewise.
4359 (cgraph_node::verify_node): Likewise.
4360 * cgraph.h (chkp_function_instrumented_p): Likewise.
4361 (symtab_node::get_alias_target): Likewise.
4362 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4363 (cgraph_local_p): Likewise.
4364 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4365 (cgraph_edge::rebuild_references): Likewise.
4366 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4367 (walk_polymorphic_call_targets): Likewise.
4368 (cgraph_node::expand_thunk): Likewise.
4369 (symbol_table::output_weakrefs): Likewise.
4370 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4371 (ix86_handle_option): Likewise.
4372 * config/i386/constraints.md: Likewise.
4373 * config/i386/i386-builtin-types.def (BND): Likewise.
4377 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4378 (BDESC_FIRST): Likewise.
4380 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4381 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4382 * config/i386/i386.c (enum reg_class): Likewise.
4383 (ix86_target_string): Likewise.
4384 (ix86_option_override_internal): Likewise.
4385 (ix86_conditional_register_usage): Likewise.
4386 (ix86_valid_target_attribute_inner_p): Likewise.
4387 (ix86_set_indirect_branch_type): Likewise.
4388 (ix86_set_current_function): Likewise.
4389 (ix86_function_arg_regno_p): Likewise.
4390 (init_cumulative_args): Likewise.
4391 (ix86_function_arg_advance): Likewise.
4392 (ix86_function_arg): Likewise.
4393 (ix86_pass_by_reference): Likewise.
4394 (ix86_function_value_regno_p): Likewise.
4395 (ix86_function_value_1): Likewise.
4396 (ix86_function_value_bounds): Likewise.
4397 (ix86_return_in_memory): Likewise.
4398 (ix86_setup_incoming_vararg_bounds): Likewise.
4399 (ix86_va_start): Likewise.
4400 (indirect_thunk_need_prefix): Likewise.
4401 (print_reg): Likewise.
4402 (ix86_print_operand): Likewise.
4403 (ix86_expand_call): Likewise.
4404 (ix86_output_function_return): Likewise.
4405 (reg_encoded_number): Likewise.
4406 (BDESC_VERIFYS): Likewise.
4407 (ix86_init_mpx_builtins): Likewise.
4408 (ix86_init_builtins): Likewise.
4409 (ix86_emit_cmove): Likewise.
4410 (ix86_emit_move_max): Likewise.
4411 (ix86_expand_builtin): Likewise.
4412 (ix86_builtin_mpx_function): Likewise.
4413 (ix86_get_arg_address_for_bt): Likewise.
4414 (ix86_load_bounds): Likewise.
4415 (ix86_store_bounds): Likewise.
4416 (ix86_load_returned_bounds): Likewise.
4417 (ix86_store_returned_bounds): Likewise.
4418 (ix86_class_likely_spilled_p): Likewise.
4419 (ix86_hard_regno_mode_ok): Likewise.
4420 (x86_order_regs_for_local_alloc): Likewise.
4421 (ix86_mitigate_rop): Likewise.
4422 (ix86_bnd_prefixed_insn_p): Likewise.
4423 (ix86_mpx_bound_mode): Likewise.
4424 (ix86_make_bounds_constant): Likewise.
4425 (ix86_initialize_bounds): Likewise.
4426 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4427 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4428 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4429 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4430 (TARGET_CHKP_BOUND_MODE): Likewise.
4431 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4432 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4433 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4434 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4435 * config/i386/i386.h (TARGET_MPX): Likewise.
4436 (TARGET_MPX_P): Likewise.
4437 (VALID_BND_REG_MODE): Likewise.
4438 (FIRST_BND_REG): Likewise.
4439 (LAST_BND_REG): Likewise.
4440 (enum reg_class): Likewise.
4441 (BND_REG_P): Likewise.
4442 (BND_REGNO_P): Likewise.
4443 (BNDmode): Likewise.
4444 (ADJUST_INSN_LENGTH): Likewise.
4445 * config/i386/i386.md: Likewise.
4446 * config/i386/i386.opt: Likewise.
4447 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4448 (defined): Likewise.
4449 (LINK_MPX): Likewise.
4450 (MPX_SPEC): Likewise.
4451 (LIBMPX_SPEC): Likewise.
4452 (LIBMPXWRAPPERS_SPEC): Likewise.
4453 (CHKP_SPEC): Likewise.
4454 * config/i386/predicates.md: Likewise.
4455 * dbxout.c (dbxout_type): Likewise.
4456 * doc/extend.texi: Likewise.
4457 * doc/invoke.texi: Likewise.
4458 * doc/md.texi: Likewise.
4459 * doc/tm.texi: Likewise.
4460 * doc/tm.texi.in: Likewise.
4461 * dwarf2out.c (is_base_type): Likewise.
4462 (gen_formal_types_die): Likewise.
4463 (gen_subprogram_die): Likewise.
4464 (gen_type_die_with_usage): Likewise.
4465 (gen_decl_die): Likewise.
4466 (dwarf2out_late_global_decl): Likewise.
4467 * expr.c (expand_assignment): Likewise.
4468 (emit_storent_insn): Likewise.
4469 (store_expr_with_bounds): Likewise.
4470 (store_expr): Likewise.
4471 (expand_expr_real_1): Likewise.
4472 * expr.h (store_expr_with_bounds): Likewise.
4473 * function.c (use_register_for_decl): Likewise.
4474 (struct bounds_parm_data): Likewise.
4475 (assign_parms_augmented_arg_list): Likewise.
4476 (assign_parm_find_entry_rtl): Likewise.
4477 (assign_parm_is_stack_parm): Likewise.
4478 (assign_parm_load_bounds): Likewise.
4479 (assign_bounds): Likewise.
4480 (assign_parms): Likewise.
4481 (expand_function_start): Likewise.
4482 * gcc.c (CHKP_SPEC): Likewise.
4483 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4484 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4485 (wrestrict_dom_walker::check_call): Likewise.
4486 * gimple.c (gimple_build_call_from_tree): Likewise.
4487 * gimple.h (enum gf_mask): Likewise.
4488 (gimple_call_with_bounds_p): Likewise.
4489 (gimple_call_set_with_bounds): Likewise.
4490 * gimplify.c (gimplify_init_constructor): Likewise.
4491 * ipa-cp.c (initialize_node_lattices): Likewise.
4492 (propagate_constants_across_call): Likewise.
4493 (find_more_scalar_values_for_callers_subset): Likewise.
4494 * ipa-hsa.c (process_hsa_functions): Likewise.
4495 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4496 * ipa-icf.c (sem_function::merge): Likewise.
4497 * ipa-inline.c (early_inliner): Likewise.
4498 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4499 (warn_function_cold): Likewise.
4500 (propagate_pure_const): Likewise.
4501 * ipa-ref.h (enum GTY): Likewise.
4502 * ipa-split.c (find_retbnd): Likewise.
4503 (consider_split): Likewise.
4504 (split_function): Likewise.
4505 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4506 * ipa.c (walk_polymorphic_call_targets): Likewise.
4507 (symbol_table::remove_unreachable_nodes): Likewise.
4508 (process_references): Likewise.
4509 (cgraph_build_static_cdtor_1): Likewise.
4510 * lto-cgraph.c (lto_output_node): Likewise.
4511 (output_refs): Likewise.
4512 (compute_ltrans_boundary): Likewise.
4513 (input_overwrite_node): Likewise.
4514 (input_node): Likewise.
4515 (input_cgraph_1): Likewise.
4516 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4517 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4518 (class pass_chkp_instrumentation_passes): Likewise.
4519 (make_pass_chkp_instrumentation_passes): Likewise.
4520 * passes.def: Likewise.
4521 * rtl.h (struct GTY): Likewise.
4522 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4523 * stor-layout.c (layout_type): Likewise.
4524 * symtab.c: Likewise.
4525 * target.def: Likewise.
4526 * targhooks.c (default_chkp_bound_type): Likewise.
4527 (default_chkp_bound_mode): Likewise.
4528 (default_builtin_chkp_function): Likewise.
4529 (default_chkp_function_value_bounds): Likewise.
4530 (default_chkp_make_bounds_constant): Likewise.
4531 (default_chkp_initialize_bounds): Likewise.
4532 * targhooks.h (default_chkp_bound_type): Likewise.
4533 (default_chkp_bound_mode): Likewise.
4534 (default_builtin_chkp_function): Likewise.
4535 (default_chkp_function_value_bounds): Likewise.
4536 (default_chkp_make_bounds_constant): Likewise.
4537 (default_chkp_initialize_bounds): Likewise.
4538 * toplev.c (compile_file): Likewise.
4539 (process_options): Likewise.
4540 * tree-core.h (DEF_BUILTIN): Likewise.
4541 (DEF_BUILTIN_CHKP): Likewise.
4542 * tree-inline.c (declare_return_variable): Likewise.
4543 (remap_gimple_stmt): Likewise.
4544 (copy_bb): Likewise.
4545 (initialize_inlined_parameters): Likewise.
4546 (expand_call_inline): Likewise.
4547 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4548 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4549 (make_pass_ipa_chkp_produce_thunks): Likewise.
4550 (make_pass_chkp): Likewise.
4551 (make_pass_chkp_opt): Likewise.
4552 (make_pass_chkp_instrumentation_passes): Likewise.
4553 * tree-pretty-print.c (dump_generic_node): Likewise.
4554 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4555 * tree-ssa-dce.c (propagate_necessity): Likewise.
4556 (eliminate_unnecessary_stmts): Likewise.
4557 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4558 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4559 * tree-ssa-sccvn.h: Likewise.
4560 * tree-ssa-strlen.c (get_string_length): Likewise.
4561 (valid_builtin_call): Likewise.
4562 (adjust_last_stmt): Likewise.
4563 (handle_builtin_strchr): Likewise.
4564 (handle_builtin_strcpy): Likewise.
4565 (handle_builtin_stxncpy): Likewise.
4566 (handle_builtin_memcpy): Likewise.
4567 (handle_builtin_strcat): Likewise.
4568 (strlen_check_and_optimize_stmt): Likewise.
4569 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4570 * tree-streamer-in.c: Likewise.
4571 * tree-streamer.c (record_common_node): Likewise.
4572 * tree.c (tree_code_size): Likewise.
4573 (wide_int_to_tree_1): Likewise.
4574 (type_contains_placeholder_1): Likewise.
4575 (build_common_tree_nodes): Likewise.
4576 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4577 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4578 (POINTER_BOUNDS_P): Likewise.
4579 (BOUNDED_TYPE_P): Likewise.
4580 (BOUNDED_P): Likewise.
4581 (CALL_WITH_BOUNDS_P): Likewise.
4582 (pointer_bounds_type_node): Likewise.
4583 * value-prof.c (gimple_ic): Likewise.
4584 * var-tracking.c (vt_add_function_parameters): Likewise.
4585 * varasm.c (make_decl_rtl): Likewise.
4586 (assemble_start_function): Likewise.
4587 (output_constant): Likewise.
4588 (maybe_assemble_visibility): Likewise.
4589 * varpool.c (ctor_for_folding): Likewise.
4590 * chkp-builtins.def: Remove.
4591 * ipa-chkp.c: Remove.
4592 * ipa-chkp.h: Remove.
4593 * rtl-chkp.c: Remove.
4594 * rtl-chkp.h: Remove.
4595 * tree-chkp-opt.c: Remove.
4596 * tree-chkp.c: Remove.
4597 * tree-chkp.h: Remove.
4599 2018-06-07 Carl Love <cel@us.ibm.com>
4601 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4602 vextract_fp_from_shortl): Add BE support.
4604 2018-06-07 Paul Koning <ni1d@arrl.net>
4606 * compare-elim.c (try_merge_compare): Don't merge compare if
4607 address contains a side effect.
4608 (try_eliminate_compare): Likewise.
4610 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4612 * config.gcc: Support "tremont".
4613 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4614 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4616 * config/i386/i386.c (m_TREMONT): Define.
4617 (processor_target_table): Add "tremont".
4618 (PTA_TREMONT): Define.
4619 (ix86_lea_outperforms): Add TARGET_TREMONT.
4620 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4621 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4622 and M_INTEL_GOLDMONT_PLUS.
4623 (fold_builtin_cpu): Add "tremont".
4624 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4625 (ix86_option_override_internal): Add "tremont".
4626 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4627 (processor_type): Add PROCESSOR_TREMONT.
4628 * config/i386/x86-tune.def: Add m_TREMONT.
4629 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4631 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4633 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4634 symbol defined for msp430i* devices to be lower case.
4636 gcc/testsuite/gcc.target/msp430/
4637 * msp430i-device-symbol.c: New test.
4638 * msp430f-device-symbol.c: New test.
4639 * msp430.h: New test header file.
4641 2018-06-07 Richard Biener <rguenther@suse.de>
4643 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4644 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4645 Properly wrap signed arithmetic if overflow wraps.
4647 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4649 PR tree-optimization/69615
4650 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4651 of a cast from a same precision integral SSA_NAME in a bb dominated
4652 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4653 cast to utype if rhs2 has already a compatible type.
4655 2018-06-07 Richard Biener <rguenther@suse.de>
4657 PR tree-optimization/85935
4658 * graphite-scop-detection.c (find_params_in_bb): Analyze
4659 condition operands with respect to the correct loop. Assert
4660 the analysis doesn't fail.
4662 2018-06-04 Carl Love <cel@us.ibm.com>
4664 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4665 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4666 as it is slightly cheaper.
4667 (first_match_or_eos_index_<mode>):
4668 Calculate index using natural element order.
4669 (first_match_index_<mode>):
4670 Calculate index using natural element order.
4671 (first_match_or_eos_index_<mode>):
4672 Calculate index using natural order.
4673 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4674 for BE and LE modes.
4675 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4676 P9V_BUILTIN_VCLZLSBB_V16QI.
4677 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4680 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4682 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4683 indentation and line wrap for many prototypes. Add missing
4684 @smallexample directives around block of prototypes for vec_xl and
4687 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4689 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4690 track if we pass or return IEEE 128-bit floating point.
4691 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4692 C++ mangling that is compatible with GCC 8.1.
4693 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4694 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4695 floating point types.
4696 (rs6000_function_arg_advance_1): Likewise.
4697 (rs6000_mangle_type): Optionally generate mangled names that match
4698 what GCC 8.1 generated for IEEE 128-bit floating point types.
4699 (rs6000_globalize_decl_name): If we have an external function that
4700 passes or returns IEEE 128-bit types, generate a weak reference
4701 from the mangled name used in GCC 8.1 to the current mangled
4703 (rs6000_init_builtins): Make __ibm128 use the long double type if
4704 long double is IBM extended double. Make __float128 use the long
4705 double type if long double is IEEE 128-bit.
4708 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4709 macro for __ibm128 built-in functions.
4710 (PACK_IF): Add __ibm128 pack/unpack functions.
4711 (UNPACK_IF): Likewise.
4712 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4713 enable long double built-in functions if long double is IEEE
4714 128-bit floating point.
4715 (rs6000_invalid_builtin): Update long double built-in function
4717 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4718 functions, adjust the built-in function to use the long double
4719 built-in function if __ibm128 and long double are the same type.
4720 * doc/extend.texi (PowerPC builtins): Update documention for
4721 __builtin_{,un}pack_longdouble. Add documentation for
4722 __builtin_{,un}pack_ibm128.
4724 2018-06-06 Jim Wilson <jimw@sifive.com>
4726 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4727 (struct machine_function): New field interrupt_mode.
4728 (riscv_handle_type_attribute): New function. Add forward declaration.
4729 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4730 (riscv_expand_epilogue): Check interrupt_mode field.
4731 (riscv_set_current_function): Check interrupt attribute args and
4732 set interrupt_mode field.
4733 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4734 (riscv_sret, riscv_uret): New.
4735 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4736 new arguments to interrupt attribute.
4738 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4741 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4742 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4744 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4746 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4747 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4748 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4749 several redundant entries.
4751 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4753 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4754 type from "rtx" to "rtx_insn *".
4755 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4756 for local "call_insn", removing cast.
4757 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4759 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4761 PR tree-optimization/86066
4762 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4763 for BIT_INSERT_EXPR stores.
4765 2018-06-06 Richard Biener <rguenther@suse.de>
4767 PR tree-optimization/86062
4768 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4769 component refs ontop
4770 of to be offsetted base.
4772 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4774 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4775 to be static and remove check on interrupt attribute name.
4777 gcc/testsuite/gcc.target/msp430/
4778 * function-attributes-4.c: New test.
4779 * static-interrupts.c: New test.
4781 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4783 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4784 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4786 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4789 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4791 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4792 Remove second argument, change how error is called.
4793 (aarch64_layout_arg): Remove second argument from
4794 aarch64_err_no_fpadvsimd call.
4795 (aarch64_init_cumulative_args): Ditto.
4796 (aarch64_gimplify_va_arg_expr): Ditto.
4797 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4799 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4801 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4802 (*simple_return_indirect_internal<mode>): Rename from
4803 simple_return_indirect_internal. Use W mode iterator.
4804 (rstorssp): New expander.
4805 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4806 (clrssbsy): New expander.
4807 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4809 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4811 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4813 (cmse_check_pointed_object): Likewise.
4815 2018-06-05 Martin Liska <mliska@suse.cz>
4817 PR gcov-profile/47618
4818 * doc/invoke.texi: Document how -fprofile-dir format
4821 2018-06-05 Richard Biener <rguenther@suse.de>
4823 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4824 removal pretend DOM info isn't available so we do not update
4825 it and only remove edges, not dominated blocks. Actually free
4826 DOM info in case we removed something. Remove unreachable blocks.
4827 (mfb_keep_latches): Work with either DOM info or marked backedges.
4828 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4829 first. Mark backedges if DOM info isn't available.
4830 (Re-)compute DOM info after cleanup_control_flow_pre.
4832 2018-06-05 Richard Biener <rguenther@suse.de>
4834 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4835 (locus_discrim_hasher::hash): Adjust.
4836 (locus_discrim_hasher::equal): Likewise.
4837 (next_discriminator_for_locus): Work on line directly.
4838 (same_line_p): Pass in expanded locus1 as well.
4839 (assign_discriminators): Avoid redundant location expansions.
4841 2018-06-05 Richard Biener <rguenther@suse.de>
4843 PR tree-optimization/86046
4844 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4845 if required after clearing TREE_ADDRESSABLE.
4847 2018-06-05 Richard Biener <rguenther@suse.de>
4849 PR tree-optimization/86047
4850 * tree-ssa-loop.c (for_each_index): Glob handling of all
4851 decls and constants and really handle all of them.
4853 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4856 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4857 qualifier_void_pointer and qualifier_const_void_pointer.
4858 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4859 (arm_init_builtins): Handle the above.
4860 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4861 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4862 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4865 2018-06-05 Martin Liska <mliska@suse.cz>
4867 * auto-profile.c (read_autofdo_file): Do not use
4868 gcov_ctr_summary struct.
4869 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4870 * coverage.c (struct counts_entry): Likewise.
4871 (read_counts_file): Read just single summary entry.
4872 (get_coverage_counts): Use gcov_summary struct.
4873 * coverage.h (get_coverage_counts): Likewise.
4874 * gcov-dump.c (dump_working_sets): Likewise.
4875 (tag_summary): Dump just single summary.
4876 * gcov-io.c (gcov_write_summary): Write just histogram
4878 (gcov_read_summary): Read just single summary.
4879 (compute_working_sets): Use gcov_summary struct.
4880 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4881 of GCOV_COUNTERS_SUMMABLE.
4882 (GCOV_COUNTERS_SUMMABLE): Remove.
4883 (GCOV_FIRST_VALUE_COUNTER): Replace with
4884 GCOV_COUNTER_V_INTERVAL.
4885 (struct gcov_ctr_summary): Remove.
4886 (struct gcov_summary): Directly use fields of former
4888 (compute_working_sets): Use gcov_summary struct.
4889 * gcov.c (read_count_file): Do not use ctrs fields.
4890 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4892 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4894 * profile.c: Likewise.
4895 * profile.h: Likewise.
4897 2018-06-05 Martin Liska <mliska@suse.cz>
4899 PR gcov-profile/84846
4900 * gcov.c (output_lines): Print working directory only
4901 in intermediate format.
4903 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4905 * config/s390/s390-builtin-types.def: Add void function type.
4906 * config/s390/s390-builtins.def: Use the function type for the
4909 2018-06-04 Jim Wilson <jimw@sifive.com>
4911 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4913 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4914 and maybe_eh_return. Change regno to unsigned int. Use new args to
4915 handle EH_RETURN_DATA_REGNO registers properly.
4916 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4917 (riscv_expand_epilogue): Update comment. Change argument name and
4918 type. Update code to use new name and type. Pass new args to
4919 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4921 * config/riscv/riscv.md (NORMAL_RETURN): New.
4922 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4923 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4924 (eh_return): Call gen_eh_return_internal and emit barrier.
4925 (eh_return_internal): Call riscv_expand_epilogue.
4927 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4929 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4930 bit_insertion field and declare can_be_merged_into method.
4931 (merged_store_group::can_be_merged_into): New method.
4932 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4933 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4934 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4936 2018-06-04 Richard Biener <rguenther@suse.de>
4938 PR tree-optimization/85955
4939 * builtins.c (fold_builtin_sincos): Convert pointers to
4940 destination to appropriate type before dereferencing.
4942 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4944 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4946 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4948 * expr.c (expand_expr_real_1): Force the operand into memory if
4949 its TYPE_MODE is BLKmode and if there is no integer mode for
4950 the number of bits being extracted.
4952 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4956 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4957 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4959 2018-06-04 Richard Biener <rguenther@suse.de>
4961 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4962 (cleanup_tree_cfg_noloop): ... single caller. Do
4963 start_recording_case_labels later.
4965 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4967 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4968 to _IMMINTRIN_H_INCLUDED.
4969 * config/i386/pconfigintrin.h: Ditto.
4970 * config/i386/waitpkgintrin.h: Ditto.
4971 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4972 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4973 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4974 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4975 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4976 waitpkgintrin.h and cldemoteintrin.h.
4978 2018-06-04 Richard Biener <rguenther@suse.de>
4980 PR tree-optimization/86038
4981 * tracer.c (find_best_successor): Check probability for
4982 being initialized, bail out if not.
4984 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4987 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4988 of bits to ignore when comparing architectures.
4990 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4992 PR tree-optimization/69615
4993 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4994 maximum or minimum of the type, try to merge it also as if
4995 range1 is + [-, x - 1] or + [x + 1, -].
4998 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
5000 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
5002 PR tree-optimization/86034
5003 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
5004 the unsigned bitfield type in a bit insertion sequence if it does not
5005 have a larger precision than the bitfield size.
5006 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
5008 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
5010 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
5012 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5014 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
5015 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
5016 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
5017 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5019 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5021 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5022 Disable -fdelete-null-pointer-checks for ELF toolchain.
5024 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5025 Kito Cheng <kito.cheng@gmail.com>
5027 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5028 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5029 (nds32*-*-*): Add glibc and uclibc conditions.
5030 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5031 (TARGET_EXCEPT_UNWIND_INFO): Define.
5032 * config/nds32/elf.h: New file.
5033 * config/nds32/linux.h: New file.
5034 * config/nds32/nds32-elf.opt: New file.
5035 * config/nds32/nds32-linux.opt: New file.
5036 * config/nds32/nds32-fp-as-gp.c
5037 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5038 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5040 (nds32_asm_file_end): Ditto.
5041 (nds32_print_operand): Ditto.
5042 (nds32_insert_attributes): Ditto.
5043 (nds32_init_libfuncs): New function.
5044 (TARGET_HAVE_TLS): Define.
5045 (TARGET_INIT_LIBFUNCS): Define.
5046 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5048 (TARGET_ELF): Apply different mcmodel setting.
5049 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5050 been migrated into elf.h and linux.h files.
5051 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5052 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5053 (mcmodel): The content has been migrated into nds32-elf.opt and
5054 nds32-linux.opt files.
5055 * config/nds32/t-elf: New file.
5056 * config/nds32/t-linux: New file.
5058 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5059 Shiva Chen <shiva0217@gmail.com>
5061 * config/nds32/constants.md (unspec_volatile_element): Add
5062 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5063 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5065 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5066 (make_pass_nds32_fp_as_gp): Declare.
5067 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5069 (nds32_asm_function_end_prologue): Remove unused asm output.
5070 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5071 (nds32_asm_file_start): Output necessary fp_as_gp information.
5072 (nds32_option_override): Adjust register usage.
5073 (nds32_expand_prologue): Consider fp_as_gp situation.
5074 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5075 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5078 (simple_return): Ditto.
5079 (omit_fp_begin): Output special directive for fp_as_gp.
5080 (omit_fp_end): Output special directive for fp_as_gp.
5081 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5082 mforbid-fp-as-gp): New options.
5084 2018-06-01 Mark Wielaard <mark@klomp.org>
5086 * dwarf2out.c (dwarf2out_finish): Remove generation of
5087 DW_AT_loclists_base.
5089 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5091 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5092 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5093 (struct merged_store_group): Add bit_insertion field.
5094 (dump_char_array): Use standard hexadecimal format.
5095 (merged_store_group::merged_store_group): Set bit_insertion to false.
5096 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5097 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5098 also print the mask in the dump file.
5099 (pass_store_merging::gate): Minor tweak.
5100 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5101 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5102 stores with INTEGER_CST stores.
5103 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5104 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5105 and use it throughout. Generate bit insertion sequences if need be.
5106 (pass_store_merging::process_store): Remove redundant condition.
5107 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5109 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5111 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5112 the 128-bit floating point types. Fix function comment.
5114 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5116 * config/aarch64/aarch64-simd.md
5117 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5119 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5122 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5124 PR tree-optimization/85989
5125 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5127 (backprop::intersect_uses): Check it when deciding whether this
5128 is a backedge reference.
5129 (backprop::process_block): Add each phi to m_visited_phis
5130 after visiting it, then clear it at the end.
5132 2018-06-01 Richard Biener <rguenther@suse.de>
5134 * tree-vectorizer.h (vect_dr_stmt): New function.
5135 (vect_get_load_cost): Adjust.
5136 (vect_get_store_cost): Likewise.
5137 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5138 Use vect_dr_stmt instead of DR_SMTT.
5139 (vect_record_base_alignments): Likewise.
5140 (vect_calculate_target_alignment): Likewise.
5141 (vect_compute_data_ref_alignment): Likewise and make static.
5142 (vect_update_misalignment_for_peel): Likewise.
5143 (vect_verify_datarefs_alignment): Likewise.
5144 (vector_alignment_reachable_p): Likewise.
5145 (vect_get_data_access_cost): Likewise. Pass down
5146 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5147 (vect_get_peeling_costs_all_drs): Likewise.
5148 (vect_peeling_hash_get_lowest_cost): Likewise.
5149 (vect_enhance_data_refs_alignment): Likewise.
5150 (vect_find_same_alignment_drs): Likewise.
5151 (vect_analyze_data_refs_alignment): Likewise.
5152 (vect_analyze_group_access_1): Likewise.
5153 (vect_analyze_group_access): Likewise.
5154 (vect_analyze_data_ref_access): Likewise.
5155 (vect_analyze_data_ref_accesses): Likewise.
5156 (vect_vfa_segment_size): Likewise.
5157 (vect_small_gap_p): Likewise.
5158 (vectorizable_with_step_bound_p): Likewise.
5159 (vect_prune_runtime_alias_test_list): Likewise.
5160 (vect_analyze_data_refs): Likewise.
5161 (vect_supportable_dr_alignment): Likewise.
5162 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5163 (vect_gen_prolog_loop_niters): Likewise.
5164 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5165 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5167 (vect_recog_mask_conversion_pattern): Likewise.
5168 (vect_try_gather_scatter_pattern): Likewise.
5169 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5170 to vect_get_store_cost.
5171 (vect_get_store_cost): Get stmt_info instead of DR.
5172 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5173 (vect_get_load_cost): Get stmt_info instead of DR.
5175 2018-06-01 Richard Biener <rguenther@suse.de>
5178 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5179 vars wrapped in MEM_REFs.
5181 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5183 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5184 Fix subreg tests so that we only return a choice between
5185 GENERAL_REGS and FP_REGS if the original classes included both.
5187 2018-06-01 Richard Biener <rguenther@suse.de>
5190 * tree-ssa-structalias.c (get_function_part_constraint):
5191 Handle NULL fi->decl.
5192 (find_func_aliases_for_call): Properly handle indirect
5193 fi from direct call.
5194 (find_func_clobbers): Likewise.
5195 (ipa_pta_execute): Likewise.
5196 (create_variable_info_for): For functions that are ifunc_resolver
5197 resolve to a varinfo that contains the result of the resolver call.
5198 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5201 2018-05-31 Michael Collison <michael.collison@arm.com>
5203 * config/aarch64/aarch64.md:
5204 (*fix_to_zero_extenddfdi2): New pattern.
5205 * gcc.target/aarch64/fix_extend1.c: New testcase.
5207 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5211 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5212 and BUILT_IN_STRNCMP_EQ.
5213 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5214 BUILT_IN_STRNCMP_EQ.
5215 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5216 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5217 (gimple_fold_builtin): Likewise.
5218 * tree-ssa-strlen.c (compute_string_length): New function.
5219 (determine_min_obsize): New function.
5220 (handle_builtin_string_cmp): New function to handle calls to
5221 string compare functions.
5222 (strlen_optimize_stmt): Add handling to builtin string compare
5224 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5225 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5226 * tree.c (build_common_builtin_nodes): Add new defines of
5227 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5229 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5232 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5233 functions with naked attribute.
5235 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5237 * config/i386/sse.md (avx_vec_concat<mode>):
5238 Substitute concat_tg_mode mode attribute with xtg_mode.
5239 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5240 (concat_tg_mode): Remove mode attribute.
5242 2018-05-31 Martin Sebor <msebor@redhat.com>
5245 * calls.c (alloc_max_size): Correct a logic error/typo.
5246 Treat excessive arguments as infinite. Warn for invalid arguments.
5247 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5249 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5252 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5253 and movx for Haswell.
5255 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5256 Cesar Philippidis <cesar@codesourcery.com>
5259 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5260 when emitting error on private/firstprivate reductions.
5261 * omp-low.c (lower_omp_target): Avoid reference-type processing
5262 on pointers for firstprivate clause.
5264 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5266 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5269 * config/aarch64/aarch64-simd.md
5270 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5271 (aarch64_ld1_x3_<mode>): Likewise
5272 (aarch64_st1x2<VALLDIF:mode>): Likewise
5273 (aarch64_st1_x2_<mode>): Likewise
5274 (aarch64_st1x3<VALLDIF:mode>): Likewise
5275 (aarch64_st1_x3_<mode>): Likewise
5276 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5277 (vld1_s8_x3): Likewise.
5278 (vld1_u16_x3): Likewise.
5279 (vld1_s16_x3): Likewise.
5280 (vld1_u32_x3): Likewise.
5281 (vld1_s32_x3): Likewise.
5282 (vld1_u64_x3): Likewise.
5283 (vld1_s64_x3): Likewise.
5284 (vld1_f16_x3): Likewise.
5285 (vld1_f32_x3): Likewise.
5286 (vld1_f64_x3): Likewise.
5287 (vld1_p8_x3): Likewise.
5288 (vld1_p16_x3): Likewise.
5289 (vld1_p64_x3): Likewise.
5290 (vld1q_u8_x3): Likewise.
5291 (vld1q_s8_x3): Likewise.
5292 (vld1q_u16_x3): Likewise.
5293 (vld1q_s16_x3): Likewise.
5294 (vld1q_u32_x3): Likewise.
5295 (vld1q_s32_x3): Likewise.
5296 (vld1q_u64_x3): Likewise.
5297 (vld1q_s64_x3): Likewise.
5298 (vld1q_f16_x3): Likewise.
5299 (vld1q_f32_x3): Likewise.
5300 (vld1q_f64_x3): Likewise.
5301 (vld1q_p8_x3): Likewise.
5302 (vld1q_p16_x3): Likewise.
5303 (vld1q_p64_x3): Likewise.
5304 (vst1_s64_x2): Likewise.
5305 (vst1_u64_x2): Likewise.
5306 (vst1_f64_x2): Likewise.
5307 (vst1_s8_x2): Likewise.
5308 (vst1_p8_x2): Likewise.
5309 (vst1_s16_x2): Likewise.
5310 (vst1_p16_x2): Likewise.
5311 (vst1_s32_x2): Likewise.
5312 (vst1_u8_x2): Likewise.
5313 (vst1_u16_x2): Likewise.
5314 (vst1_u32_x2): Likewise.
5315 (vst1_f16_x2): Likewise.
5316 (vst1_f32_x2): Likewise.
5317 (vst1_p64_x2): Likewise.
5318 (vst1q_s8_x2): Likewise.
5319 (vst1q_p8_x2): Likewise.
5320 (vst1q_s16_x2): Likewise.
5321 (vst1q_p16_x2): Likewise.
5322 (vst1q_s32_x2): Likewise.
5323 (vst1q_s64_x2): Likewise.
5324 (vst1q_u8_x2): Likewise.
5325 (vst1q_u16_x2): Likewise.
5326 (vst1q_u32_x2): Likewise.
5327 (vst1q_u64_x2): Likewise.
5328 (vst1q_f16_x2): Likewise.
5329 (vst1q_f32_x2): Likewise.
5330 (vst1q_f64_x2): Likewise.
5331 (vst1q_p64_x2): Likewise.
5332 (vst1_s64_x3): Likewise.
5333 (vst1_u64_x3): Likewise.
5334 (vst1_f64_x3): Likewise.
5335 (vst1_s8_x3): Likewise.
5336 (vst1_p8_x3): Likewise.
5337 (vst1_s16_x3): Likewise.
5338 (vst1_p16_x3): Likewise.
5339 (vst1_s32_x3): Likewise.
5340 (vst1_u8_x3): Likewise.
5341 (vst1_u16_x3): Likewise.
5342 (vst1_u32_x3): Likewise.
5343 (vst1_f16_x3): Likewise.
5344 (vst1_f32_x3): Likewise.
5345 (vst1_p64_x3): Likewise.
5346 (vst1q_s8_x3): Likewise.
5347 (vst1q_p8_x3): Likewise.
5348 (vst1q_s16_x3): Likewise.
5349 (vst1q_p16_x3): Likewise.
5350 (vst1q_s32_x3): Likewise.
5351 (vst1q_s64_x3): Likewise.
5352 (vst1q_u8_x3): Likewise.
5353 (vst1q_u16_x3): Likewise.
5354 (vst1q_u32_x3): Likewise.
5355 (vst1q_u64_x3): Likewise.
5356 (vst1q_f16_x3): Likewise.
5357 (vst1q_f32_x3): Likewise.
5358 (vst1q_f64_x3): Likewise.
5359 (vst1q_p64_x3): Likewise.
5361 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5363 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5364 user_label_prefix to name.
5366 * tree-core.h: Update comment about the format of NAME string
5367 passed to handler in attribute_spec.
5369 * config/msp430/msp430.md: Remove erroneous subreg expression from
5370 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5371 zero_extend{q,h}isi2.
5373 2018-05-30 Borislav Petkov <bp@suse.de>
5375 * doc/extend.texi: Document some architecture specific
5376 constraints and sort entries.
5378 2018-05-30 Martin Sebor <msebor@redhat.com>
5381 * builtins.c (expand_builtin_stpcpy_1): New function.
5382 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5383 only if the former succeeds.
5385 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5387 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5390 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5392 * doc/invoke.texi (-flinker-output): Document
5394 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5396 * passes.c (ipa_write_summaries): Only modify statements if body
5398 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5399 incrementally linking.
5400 (ipa_passes): Likewise.
5401 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5402 pass down resolution info.
5403 * common.opt (flag_incremental_link): Update info.
5404 * gcc.c (plugin specs): Turn flinker-output=* to
5405 -plugin-opt=-linker-output-known
5406 * toplev.c (compile_file): Also cut compilation when doing incremental
5408 * flag-types. (enum lto_partition_model): Add
5409 LTO_LINKER_OUTPUT_NOLTOREL.
5410 (invoke.texi): Add -flinker-output docs.
5411 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5412 link same way as WPA; do not stream in dead initializers.
5414 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5415 darf2out_register_external_decl): Support incremental link.
5417 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5419 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5421 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5423 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5424 it down to simple_object_copy_lto_debug_sections.
5425 (run_gcc): Determine incremental LTO link time and configure
5426 lto1 into non-wpa mode, disable renaming of debug sections.
5428 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5430 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5431 descriptions of various incorrectly documented functions.
5433 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5436 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5437 address check not strict.
5439 2018-05-30 Richard Biener <rguenther@suse.de>
5441 PR tree-optimization/85964
5442 * tracer.c (better_p): Drop initialized count check, we only
5443 call the function with initialized counts now.
5444 (find_best_successor): Do find a best edge if one
5445 has uninitialized count.
5446 (find_best_predecessor): Likewise. Do BB frequency check only
5447 if count is initialized.
5449 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5451 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5452 (aarch64_ldrstr_offset_compare): New.
5453 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5454 load/store orderings.
5455 (aarch64_gen_adjusted_ldpstp): Likewise.
5457 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5459 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5460 Check for subset of GENERAL_REGS and FP_REGS.
5461 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5464 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5466 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5467 and wi::to_poly_offset. Add the current offset and then check
5468 whether the sum fits, rather than using an unchecked addition of
5469 a checked term. Check for a shwi rather than a uhwi.
5470 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5471 (store_constructor): Use poly_int_tree_p.
5472 (expand_expr_real_1): Likewise.
5473 * function.c (assign_temp): Likewise.
5474 * fold-const.c (const_binop): Use poly_int_tree_p and
5476 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5478 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5479 to_poly_offset for MEM offsets.
5480 * ipa-icf.c (sem_variable::equals): Likewise.
5481 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5482 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5483 wi::to_poly_offset for BIT_FIELD_REF offsets.
5484 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5486 * var-tracking.c (emit_note_insn_var_location): Use
5487 tree_to_poly_uint64.
5489 2018-05-29 Jim Wilson <jimw@sifive.com>
5491 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5493 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5496 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5497 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5499 (sse4_1_round<mode>2): Use nonimmediate_operand
5500 for operand 1 predicate.
5502 2018-05-29 Martin Sebor <msebor@redhat.com>
5503 Richard Biener <rguenther@suse.de>
5506 * calls.c (get_size_range): Call determine_value_range instead
5507 of get_value_range..
5508 * tree-vrp.h (determine_value_range): Declared new function.
5509 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5511 2018-05-29 Richard Biener <rguenther@suse.de>
5513 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5514 sure to use non-pattern stmts for get_earlier_stmt arguments.
5515 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5516 called on pattern stmts.
5517 (get_later_stmt): Likewise.
5519 2018-05-29 Martin Liska <mliska@suse.cz>
5521 PR gcov-profile/85759
5522 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5525 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5527 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5529 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5533 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5534 VEC_PACK_FLOAT_EXPR): New tree codes.
5535 * tree-pretty-print.c (op_code_prio): Handle
5536 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5537 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5538 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5539 * tree-inline.c (estimate_operator_cost): Likewise.
5540 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5541 * fold-const.c (const_binop): Likewise.
5542 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5543 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5544 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5545 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5546 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5547 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5548 * expr.c (expand_expr_real_2): Likewise.
5549 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5550 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5551 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5553 * optabs.c (expand_widen_pattern_expr): For
5554 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5555 sign from result type rather than operand's type.
5556 (expand_binop_directly): For vec_packu_float_optab and
5557 vec_packs_float_optab allow result type to be different from operand's
5559 * optabs-tree.c (optab_for_tree_code): Handle
5560 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5561 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5562 * tree-vect-generic.c (expand_vector_operations_1): Handle
5563 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5564 VEC_PACK_FLOAT_EXPR.
5565 * tree-vect-stmts.c (supportable_widening_operation): Handle
5567 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5568 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5569 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5570 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5571 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5573 (vec_pack<floatprefix>_float_<mode>): New expander.
5574 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5576 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5577 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5578 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5579 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5580 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5582 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5583 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5584 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5585 VEC_PACK_FLOAT_EXPR): Document.
5587 2018-05-29 Richard Biener <rguenther@suse.de>
5589 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5591 (stmt_vec_info_vec): Make pointer.
5592 (init_stmt_vec_info_vec): Remove.
5593 (free_stmt_vec_info_vec): Likewise.
5594 (set_stmt_vec_info_vec): New function.
5595 (free_stmt_vec_infos): Likewise.
5596 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5597 (set_vinfo_for_stmt): Likewise.
5598 (get_earlier_stmt): Likewise.
5599 (get_later_stmt): Likewise.
5600 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5601 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5602 (vec_info::~vec_info): Free stmt_vec_infos.
5603 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5604 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5605 (pass_slp_vectorize::execute): Likewise.
5606 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5607 (free_stmt_vec_info_vec): Likewise.
5608 (set_stmt_vec_info_vec): New function.
5609 (free_stmt_vec_infos): Likewise.
5610 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5611 the global stmt_vec_info_vec.
5612 * tree-parloops.c (gather_scalar_reductions): Use
5613 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5616 2018-05-29 Richard Biener <rguenther@suse.de>
5618 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5620 2018-05-29 Martin Liska <mliska@suse.cz>
5621 David Malcolm <dmalcolm@redhat.com>
5623 * vec.c (test_reverse): New.
5624 (vec_c_tests): Add new test.
5625 * vec.h (vl_ptr>::reverse): New function.
5627 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5629 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5631 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5634 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5636 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5638 2018-05-28 Richard Biener <rguenther@suse.de>
5640 PR tree-optimization/85933
5641 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5642 look at stmts marked as vectorizable.
5644 2018-05-28 Richard Biener <rguenther@suse.de>
5646 PR tree-optimization/85934
5647 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5648 vector boolean check before scalar optimization.
5650 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5652 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5655 2018-05-28 Mark Wielaard <mark@klomp.org>
5657 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5658 if it is an expression containing a minus sign.
5660 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5662 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5664 2018-05-27 Paul Koning <ni1d@arrl.net>
5666 * config/pdp11/pdp11.md (truncsihi2): Remove.
5668 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5669 Chung-Ju Wu <jasonwucj@gmail.com>
5671 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5673 (unaligned_store_dw): Ditto.
5674 * config/nds32/nds32-memory-manipulation.c
5675 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5676 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5677 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5678 (emit_setmem_word_loop): Rename to ...
5679 (emit_setmem_doubleword_loop): ... this.
5680 (nds32_gen_dup_4_byte_to_word_value): New function.
5681 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5682 (nds32_expand_setmem_loop): Refine implementation.
5683 (nds32_expand_setmem_loop_v3m): Ditto.
5684 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5687 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5689 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5691 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5693 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5694 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5695 machine->attr_no_prologue_p.
5696 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5697 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5698 (nds32_expand_epilogue): Consider attr_naked_p.
5699 (nds32_expand_epilogue_v3pop): Likewise.
5700 (nds32_can_use_return_insn): Likewise.
5701 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5702 attr_no_prologue_p fields.
5703 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5705 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5708 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5710 * config/i386/sse.md
5711 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5713 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5715 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5717 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5719 (*<floatsuffix>floatv2div2sf2): Rename to ...
5720 (*float<floatunssuffix>v2div2sf2): ... this.
5721 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5722 (float<floatunssuffix>v2div2sf2_mask): ... this.
5723 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5724 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5725 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5727 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5729 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5731 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5733 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5735 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5737 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5738 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5739 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5740 gen_ufix_truncv8dfv8si2.
5741 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5742 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5743 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5744 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5745 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5746 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5747 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5748 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5750 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5754 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5756 2018-05-25 Jim Wilson <jimw@sifive.com>
5758 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5759 * config/riscv/riscv.c (struct machine_function): Add
5760 interrupt_handler_p and attribute_checked_p fields.
5761 (riscv_attribute_table): Add interrupt.
5762 (riscv_interrupt_type_p): New.
5763 (riscv_save_reg_p): Save extra regs for interrupt handler.
5764 (riscv_use_save_libcall): Return false for interrupt handler.
5765 (riscv_first_stack_step): Add forward declaration.
5766 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5767 for interrupt handler with large frame. Use it for saved reg list.
5768 (riscv_expand_prologue): Move flag_stack_usage_info support to
5769 eliminate duplication.
5770 (riscv_expand_epilogue): Generate mret for interrupt handler.
5771 (riscv_epilogue_uses): New.
5772 (riscv_can_use_return_insn): Return false for interrupt handler.
5773 (riscv_function_ok_for_sibcall): Likewise.
5774 (riscv_set_current_function): Add interrupt handler support.
5775 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5776 * config/riscv/riscv.md (UNSPECV_MRET): New.
5778 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5780 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5782 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5784 PR tree-optimization/85712
5785 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5786 this candidate has already been replaced in-situ by a copy.
5788 2018-05-25 Jason Merrill <jason@redhat.com>
5790 PR c++/80485 - inline function non-zero address.
5791 * symtab.c (nonzero_address): Check DECL_COMDAT.
5793 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5796 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5797 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5798 extension of SImode operation. Use const123_operand predicate.
5799 (*saddsi_1): Remove.
5800 (*saddl_se_1): Ditto.
5802 (*ssubl_se_1): Ditto.
5803 * config/alpha/predicates.md (const123_operand): New predicate.
5804 * config/alpha/constraints.md (P): Use IN_RANGE.
5806 2018-05-25 Richard Biener <rguenther@suse.de>
5808 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5810 (ref_maybe_used_by_stmt_p): Likewise.
5811 (stmt_may_clobber_ref_p): Likewise.
5812 (stmt_may_clobber_ref_p_1): Likewise.
5813 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5815 (ref_maybe_used_by_stmt_p): Likewise.
5816 (stmt_may_clobber_ref_p): Likewise.
5817 (stmt_may_clobber_ref_p_1): Likewise.
5818 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5819 the alias oracle to disambiguate DRs with stmts DR analysis
5821 (vect_analyze_data_refs): Do not give up on not analyzable
5822 DRs for BB vectorization. Remove code truncating the dataref
5825 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5828 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5829 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5830 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5832 2018-05-25 Richard Biener <rguenther@suse.de>
5834 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5835 function, combining stmt data ref gathering and fatal analysis
5837 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5838 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5839 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5840 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5841 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5843 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5845 PR tree-optimization/85720
5846 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5847 SCC if all partitions are builtins.
5848 (version_loop_by_alias_check): New parameter. Generate cancelable
5849 runtime alias check if all partitions are builtins.
5850 (distribute_loop): Update call to above function.
5852 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5854 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5855 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5856 (parm_default_def_partition_arg): Ditto.
5857 (set_parm_default_def_partition): Ditto.
5858 (get_parm_default_def_partitions): Ditto and make it static.
5859 (get_undefined_value_partitions): Ditto and make it static.
5860 (remove_ssa_form): Refactor call to init_var_map here.
5861 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5862 computation for loop region.
5863 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5864 (register_default_def): Delete.
5865 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5866 (parm_default_def_partition_arg): Ditto.
5867 (set_parm_default_def_partition): Ditto.
5868 (get_parm_default_def_partitions): Ditto and make it static.
5869 (get_undefined_value_partitions): Ditto and make it static.
5870 (coalesce_with_default, coalesce_with_default): Update comment.
5871 (create_coalesce_list_for_region): New func factored out from
5872 create_outofssa_var_map.
5873 (populate_coalesce_list_for_outofssa): New func factored out from
5874 create_outofssa_var_map and coalesce_ssa_name.
5875 (create_outofssa_var_map): Delete.
5876 (coalesce_ssa_name): Refactor to support live range computation.
5877 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5878 (get_parm_default_def_partitions): Delete.
5879 (get_undefined_value_partitions): Ditto.
5880 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5881 computation for loop region.
5882 (new_tree_live_info, loe_visit_block): Ditto.
5883 (live_worklist, set_var_live_on_entry): Ditto.
5884 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5885 * tree-ssa-live.h (struct _var_map): New fields.
5886 (init_var_map): Change decl.
5887 (region_contains_p): New.
5889 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5891 * tree-ssa-live.h (live_merge_and_clear): Delete.
5893 2018-05-25 Richard Biener <rguenther@suse.de>
5896 * tree-dump.c (dequeue_and_dump): Remove access to removed
5897 operand 2 of a SWITCH_EXPR.
5899 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5901 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5902 multiplication and division.
5903 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5904 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5905 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5906 (cond_udiv_optab, cond_umod_optab): New optabs.
5907 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5908 (IFN_COND_RDIV): New internal functions.
5909 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5910 TRUNC_MOD_EXPR and RDIV_EXPR.
5911 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5912 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5914 (SVE_INT_BINARY): Include mult.
5915 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5916 (optab, sve_int_op): Handle mult.
5917 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5919 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5920 for SVE_INT_BINARY_SD.
5922 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5924 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5925 (optab, sve_int_op): Handle div and udiv.
5926 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5927 for SVE_INT_BINARY_SD.
5928 (*<optab><mode>3): New insn for the same.
5930 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5932 * tree-vect-patterns.c: Include predict.h.
5933 (vect_recog_divmod_pattern): Restrict check for division support
5934 to when optimizing for size.
5936 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5938 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5939 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5940 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5941 (gimple_match_op::set_op): Likewise.
5942 (gimple_resimplify4): Declare.
5943 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5944 (expr::gen_transform): Likewise.
5945 (decision_tree::gen): Generate a simplification routine for 4 operands.
5946 * gimple-match-head.c (gimple_simplify): Add an overload for
5947 4 operands. In the top-level function, handle up to 4 call
5948 arguments and call gimple_resimplify4.
5949 (gimple_resimplify4): New function.
5950 (build_call_internal): Pass a fourth operand.
5951 (maybe_push_to_seq): Likewise.
5952 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5953 Fold VEC_COND_EXPRs of an operation and a default value into
5954 an IFN_COND_* function if possible.
5955 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5957 (SVE_COND_FP_BINARY): Include them.
5958 (optab, sve_fp_op): Handle them.
5959 (SVE_INT_BINARY_REV): New code iterator.
5960 (SVE_COND_FP_BINARY_REV): New int iterator.
5961 (commutative): New int attribute.
5962 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5964 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5966 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5967 (*cond_<optab><mode>): New patterns for reversed operands.
5969 2018-05-25 Richard Biener <rguenther@suse.de>
5971 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5972 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5973 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5974 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5975 * tree-vect-data-refs.c (everywhere): Adjust users.
5976 * tree-vect-loop.c (everywhere): Likewise.
5977 * tree-vect-slp.c (everywhere): Likewise.
5978 * tree-vect-stmts.c (everywhere): Likewise.
5979 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5981 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5983 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5985 (gcc_cv_as_section_exclude): ... this.
5986 Try Solaris as #exclude syntax.
5987 * configure: Regenerate.
5988 * config.in: Regenerate.
5989 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5991 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5992 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5994 * varasm.c (default_elf_asm_named_section): Don't check if
5995 HAVE_GAS_SECTION_EXCLUDE is defined.
5997 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5999 * doc/md.texi: Update the documentation of the cond_* optabs
6000 to mention the new final operand. Fix GET_MODE_NUNITS call.
6001 Describe the scalar case too.
6002 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
6003 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
6005 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
6006 (get_conditional_internal_fn): Update comment.
6007 * tree-vect-loop.c (vectorizable_reduction): Pass the original
6008 accumulator value as a final argument to conditional functions.
6009 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
6010 a define_expand and add an "else" operand. Assert for now that
6011 the else operand is equal to operand 2. Use SVE_INT_BINARY and
6012 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
6013 (*cond_<optab><mode>): New patterns.
6014 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
6015 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6016 (UNSPEC_COND_EOR): Delete.
6017 (optab): Remove associated mappings.
6018 (SVE_INT_BINARY): New code iterator.
6019 (sve_int_op): Remove int attribute and add "minus" to the code
6021 (SVE_COND_INT_OP): Delete.
6022 (SVE_COND_FP_OP): Rename to...
6023 (SVE_COND_FP_BINARY): ...this.
6025 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6027 * optabs.c (can_reuse_operands_p): New function.
6028 (maybe_legitimize_operands): Try to reuse the results for
6031 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6033 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6034 Add {q} suffix to insn mnemonic.
6036 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6038 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6039 (msp430_warn_func_return): New.
6041 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6043 * fold-const.c (tree_nonzero_bits): New function.
6044 * fold-const.h (tree_nonzero_bits): Likewise.
6045 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6046 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6048 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6052 * varasm.c (assemble_alias): Check ifunc_resolver only on
6055 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6058 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6059 when memory input operand is handled.
6061 2018-05-24 Luis Machado <luis.machado@linaro.org>
6063 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6065 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6067 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6069 * match.pd: Delay FMA folds until after vectorization.
6071 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6074 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6075 address check not strict.
6077 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6079 * gimple-match.h (gimple_match_op): New class.
6080 (mprts_hook): Replace parameters with a gimple_match_op *.
6081 (maybe_build_generic_op): Likewise.
6082 (gimple_simplified_result_is_gimple_val): Replace parameters with
6083 a const gimple_match_op *.
6084 (gimple_simplify): Replace code_helper * and tree * parameters with
6085 a gimple_match_op * parameter.
6086 (gimple_resimplify1): Replace code_helper *, tree and tree *
6087 parameters with a gimple_match_op * parameter.
6088 (gimple_resimplify2): Likewise.
6089 (gimple_resimplify3): Likewise.
6090 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6091 parameters with a gimple_match_op * parameter.
6092 * gimple-match-head.c (gimple_simplify): Change prototypes of
6093 auto-generated functions to take a gimple_match_op * instead of
6094 separate code_helper * and tree * parameters. Make the same
6095 change in the top-level overload and update calls to the
6096 gimple_resimplify routines. Update calls to the auto-generated
6097 functions and to maybe_push_res_to_seq in the publicly-facing
6098 operation-specific gimple_simplify overloads.
6099 (gimple_match_op::MAX_NUM_OPS): Define.
6100 (gimple_resimplify1): Replace rcode and ops with a single res_op
6101 parameter. Update call to gimple_simplify.
6102 (gimple_resimplify2): Likewise.
6103 (gimple_resimplify3): Likewise.
6104 (mprts_hook): Replace parameters with a gimple_match_op *.
6105 (maybe_build_generic_op): Likewise.
6106 (build_call_internal): Replace type, nargs and ops with
6107 a gimple_match_op *.
6108 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6109 with a single gimple_match_op *. Update calls to mprts_hook,
6110 build_call_internal and gimple_simplified_result_is_gimple_val.
6111 Factor out code that is common to the tree_code and combined_fn cases.
6112 * genmatch.c (expr::gen_transform): Replace tem_code and
6113 tem_ops with a gimple_match_op called tem_op. Update calls
6114 to the gimple_resimplify functions and maybe_push_res_to_seq.
6115 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6116 res_ops. Update call to the gimple_resimplify functions.
6117 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6118 (decision_tree::gen): Make the functions take a gimple_match_op *
6119 called res_op instead of separate res_code and res_ops parameters.
6120 Update call accordingly.
6121 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6122 and ops with a single res_op parameter. Update calls to
6123 maybe_build_generic_op and maybe_push_res_to_seq.
6124 (fold_stmt_1): Update calls to gimple_simplify and
6125 replace_stmt_with_simplification.
6126 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6127 and gimple_simplified_result_is_gimple_val.
6128 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6130 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6131 with a gimple_match_op *.
6132 (vn_nary_build_or_lookup): Likewise. Update call to
6133 vn_nary_build_or_lookup_1.
6134 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6135 gimple_match_op *. Update calls to the gimple_resimplify routines
6136 and to gimple_simplified_result_is_gimple_val.
6137 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6138 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6139 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6140 (visit_nary_op): Likewise.
6141 (visit_reference_op_load): Likewise.
6143 2018-05-23 Luis Machado <luis.machado@linaro.org>
6145 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6146 modifier for printing the step amount.
6148 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6151 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6154 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6156 * doc/sourcebuild.texi (Endianness): New subsubsection.
6158 2018-05-23 Luis Machado <luis.machado@linaro.org>
6160 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6161 <prefetch_dynamic_strides>: New const bool field.
6162 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6163 prefetch_dynamic_strides.
6164 (exynosm1_prefetch_tune): Likewise.
6165 (thunderxt88_prefetch_tune): Likewise.
6166 (thunderx_prefetch_tune): Likewise.
6167 (thunderx2t99_prefetch_tune): Likewise.
6168 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6170 (aarch64_override_options_internal): Update to set
6171 PARAM_PREFETCH_DYNAMIC_STRIDES.
6172 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6173 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6174 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6175 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6176 prefetch-dynamic-strides setting.
6178 2018-05-23 Luis Machado <luis.machado@linaro.org>
6180 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6181 <minimum_stride>: New const int field.
6182 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6183 minimum_stride field defaulting to -1.
6184 (exynosm1_prefetch_tune): Likewise.
6185 (thunderxt88_prefetch_tune): Likewise.
6186 (thunderx_prefetch_tune): Likewise.
6187 (thunderx2t99_prefetch_tune): Likewise.
6188 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6189 <default_opt_level>: Set to 3.
6190 (aarch64_override_options_internal): Update to set
6191 PARAM_PREFETCH_MINIMUM_STRIDE.
6192 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6193 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6194 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6195 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6196 stride is constant and is below the minimum stride threshold.
6198 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6200 * config/arm/arm-cpus.in (mode26): Delete.
6201 (armv4): Delete mode26 reference.
6202 * config/arm/arm.c (arm_configure_build_target): Delete use of
6205 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6207 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6209 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6210 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6212 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6213 (floatunsdidf2): Ditto.
6215 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6217 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6218 (fixuns_trunc<mode>si2_avx512f): Ditto.
6219 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6220 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6221 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6223 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6225 PR rtl-optimization/79985
6226 * df-scan.c (df_insn_refs_collect): Remove special case for
6227 global registers and asm statements.
6229 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6231 * extend.texi (Global Register Variables): Rewrite the bullet list.
6232 Note that the register is available for allocation. Note that access
6233 via inline asm must use constraints. Add note about async-signal
6234 handlers. Remove paragraph about automagic register selection.
6236 2018-05-23 Richard Biener <rguenther@suse.de>
6238 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6239 of fixed offset from memset VN.
6241 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6243 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6245 (alloc_cand_and_find_basis): Initialize first_interp field.
6246 (slsr_process_mul): Modify first_interp field.
6247 (slsr_process_add): Likewise.
6248 (slsr_process_cast): Modify first_interp field for each new
6250 (slsr_process_copy): Likewise.
6251 (dump_candidate): Dump first_interp field.
6252 (replace_mult_candidate): Process all interpretations, not just
6254 (replace_rhs_if_not_dup): Likewise.
6255 (replace_one_candidate): Likewise.
6257 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6259 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6261 (aarch64_needs_frame_chain): New function.
6262 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6264 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6267 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6268 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6269 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6270 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6271 as true for strict-align.
6272 (aarch64_can_inline_p): Perform checks even when callee has no
6273 attributes to check for strict alignment.
6274 * doc/extend.texi (AArch64 Function Attributes): Document
6276 * doc/invoke.texi: (AArch64 Options): Likewise.
6278 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6280 PR tree-optimization/85853
6281 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6282 the handling of the root of the node to...
6283 (vect_slp_analyze_node_operations_1): ...this new function,
6284 and run the whole thing with the child nodes' def types
6285 set according to their SLP node's def type.
6287 2018-05-23 Richard Biener <rguenther@suse.de>
6290 * tree-data-ref.c (create_runtime_alias_checks): Defer
6291 and ignore overflow warnings.
6293 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6295 PR tree-optimization/85822
6296 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6299 2018-05-23 Richard Biener <rguenther@suse.de>
6301 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6302 memset constants via native_interpret_expr.
6304 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6307 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6309 (cgraph_node::create_alias): Likewise.
6310 (cgraph_node::get_availability): Check ifunc_resolver instead
6311 of looking up ifunc attribute.
6312 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6313 * varasm.c (do_assemble_alias): Likewise.
6314 (assemble_alias): Likewise.
6315 (default_binds_local_p_3): Likewise.
6316 * cgraph.h (cgraph_node): Add ifunc_resolver.
6317 (cgraph_node::only_called_directly_or_aliased_p): Return false
6319 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6321 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6322 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6323 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6324 instead of looking up ifunc attribute.
6326 2018-05-22 Luis Machado <luis.machado@linaro.org>
6328 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6330 2018-05-22 Martin Sebor <msebor@redhat.com>
6333 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6334 only when expasion succeeds.
6335 (expand_builtin_strcmp): Same.
6336 (expand_builtin_strncmp): Same.
6338 2018-05-22 Martin Sebor <msebor@redhat.com>
6340 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6342 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6343 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6345 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6346 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6348 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6349 Add check for legitimate_address.
6350 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6351 (aarch64_swap_ldrstr_operands): New.
6352 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6355 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6356 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6358 * config/aarch64/aarch64.md: New patterns to generate stp
6360 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6361 single words and double words.
6362 (load_pair_sw, load_pair_dw): Likewise.
6363 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6365 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6367 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6368 for different mode ldpstp and add peephole for merged zero stores.
6370 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6372 (aarch64_gen_store_pair): Rename calls to match new patterns.
6373 (aarch64_gen_load_pair): Rename calls to match new patterns.
6374 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6375 (load_pair<DREG:mode><DREG2:mode>): ... This.
6376 (store_pair<mode>): Rename to...
6377 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6378 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6380 (V_INT_EQUIV): Handle SImode.
6381 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6384 2018-05-22 Martin Sebor <msebor@redhat.com>
6387 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6388 or ajust the presumed bound on an operation to avoid unnecessary
6391 2018-05-22 Martin Sebor <msebor@redhat.com>
6393 PR tree-optimization/85826
6394 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6395 assuming that a DECL necesarily has a constant size.
6397 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6400 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6402 2018-05-22 Richard Biener <rguenther@suse.de>
6404 PR tree-optimization/85834
6405 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6406 non-constant and non-zero memset arguments.
6408 2018-05-22 Martin Liska <mliska@suse.cz>
6411 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6413 2018-05-22 Richard Biener <rguenther@suse.de>
6415 PR tree-optimization/85863
6416 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6417 comparisons when vectype is specified.
6418 (vectorizable_condition): Do not specify vectype for
6419 vect_is_simple_cond when SLP vectorizing.
6421 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6424 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6425 define __ibm128 as long double.
6426 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6427 as a distinct type when IEEE 128-bit support is enabled.
6428 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6429 128-bit types to use the correct functions.
6430 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6431 convert between 128-bit floating point types that have different
6432 modes but the same representation, instead of using gen_lowpart to
6434 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6436 (IFKF_reg): New attributes to give the register constraints for
6438 (extend<mode>tf2_internal): New insns to mark an explicit
6439 conversion between 128-bit floating point types that have a
6440 different mode but share the same representation.
6442 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6444 PR tree-optimization/85814
6445 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6446 a null return from get_strinfo when unsharing the next
6447 strinfo in the chain.
6449 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6452 * varasm.c (weak_finish): Clean up weak_decls.
6454 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6456 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6457 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6458 UNSPEC_UADALP values.
6459 * config/aarch64/iterators.md (ABAL): New int iterator.
6462 (sur): Add mappings for the above.
6463 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6465 (aarch64_<sur>abal<mode>_4): Likewise.
6466 (aarch64_<sur>adalp<mode>_3): Likewise.
6467 (<sur>sadv16qi): New define_expand.
6469 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6471 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6472 (*movdf_internal): Ditto.
6473 (*rcpsf2_sse): Ditto.
6474 (*rsqrtsf2_sse): Ditto.
6475 (*sqrt<mode>2_sse): Ditto.
6477 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6479 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6481 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6482 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6483 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6484 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6486 * config/aarch64/arm_neon.h: Likewise.
6487 * config/aarch64/iterators.md (VQ_I): New.
6489 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6491 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6493 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6496 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6498 * config/nds32/constraints.md (S): New constraint.
6499 * config/nds32/nds32.md (call_internal): Use constraint S.
6500 (call_value_internal): Likewise.
6501 (sibcall_internal): Likewise.
6502 (sibcall_value_internal): Likewise.
6504 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6505 Chung-Ju Wu <jasonwucj@gmail.com>
6507 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6510 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6511 Chung-Ju Wu <jasonwucj@gmail.com>
6513 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6514 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6515 (nds32_rtx_costs_impl): Simplify.
6516 (nds32_address_cost_impl): Simplify.
6517 (nds32_init_rtx_costs): New function.
6518 (nds32_rtx_costs_speed_prefer): Likewise.
6519 (nds32_rtx_costs_size_prefer): Likewise.
6520 (nds32_address_cost_speed_prefer): Likewise.
6521 (nds32_address_cost_speed_fwprop): Likewise.
6522 (nds32_address_cost_size_prefer): Likewise.
6523 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6524 * config/nds32/nds32.c (nds32_option_override): Use
6525 nds32_init_rtx_costs function.
6527 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6529 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6530 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6531 (TARGET_PIPELINE_N8): Likewise.
6532 (TARGET_PIPELINE_N10): Likewise.
6533 (TARGET_PIPELINE_N13): Likewise.
6534 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6536 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6538 * config/nds32/nds32-fpu.md: Update copyright year.
6540 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6542 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6544 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6546 * config/nds32/nds32.c
6547 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6548 * config/nds32/nds32.opt (minline-asm-r15): New option.
6550 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6552 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6554 * config/nds32/nds32.md (abssi2): New pattern.
6556 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6558 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6559 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6560 Merge insn pattern from sse_cvtsi2ss<round_name> and
6561 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6562 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6563 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6564 using SWI48 mode iterator.
6565 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6566 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6567 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6568 pattern from sse_cvttss2si<round_saeonly_name>
6569 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6570 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6571 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6572 using SWI48 mode iterator.
6573 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6574 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6575 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6576 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6577 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6578 using SWI48 mode iterator.
6579 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6580 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6581 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6582 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6583 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6584 SWI48 mode iterator.
6585 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6586 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6587 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6588 pattern from sse_cvttsd2si<round_saeonly_name>
6589 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6591 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6593 * config/nds32/nds32-md-auxiliary.c
6594 (nds32_valid_smw_lwm_base_p): Refine.
6595 (nds32_output_smw_single_word): Refine.
6596 (nds32_output_smw_double_word): New.
6597 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6599 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6601 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6602 (nds32_output_stack_pop): Refine.
6603 (nds32_expand_unaligned_load): Refine.
6604 (nds32_expand_unaligned_store): Refine.
6606 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6607 Chung-Ju Wu <jasonwucj@gmail.com>
6609 * config/nds32/constants.md: Add TP_REGNUM constant.
6610 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6611 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6613 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6614 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6615 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6616 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6617 and PIC code generation.
6618 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6620 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6622 * config/nds32/nds32.md: Support TLS and PIC.
6623 * config/nds32/nds32.c: Support TLS and PIC.
6624 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6625 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6628 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6630 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6631 mode with E_ prefix.
6633 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6634 Chung-Ju Wu <jasonwucj@gmail.com>
6636 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6637 * config/nds32/nds32-md-auxiliary.c
6638 (symbolic_reference_mentioned_p): New.
6639 (nds32_legitimize_ict_address): New.
6640 (nds32_expand_ict_move): New.
6641 (nds32_indirect_call_referenced_p): New.
6642 (nds32_symbol_binds_local_p): Delete.
6643 (nds32_long_call_p): Modify.
6644 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6645 * config/nds32/nds32-protos.h
6646 (symbolic_reference_mentioned_p): Declare.
6647 (nds32_legitimize_ict_address): Declare.
6648 (nds32_expand_ict_move): Declare.
6649 (nds32_indirect_call_referenced_p): Declare.
6650 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6651 (nds32_relax_group): Use nds32_ict_const_p as condition.
6652 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6653 (nds32_asm_file_start): Output ict_model directive in asm code.
6654 (nds32_legitimate_address_p): Consider indirect call.
6655 (nds32_print_operand): Consider indirect call.
6656 (nds32_print_operand_address): Consider indirect call.
6657 (nds32_insert_attributes): Handle "indirect_call" attribute.
6658 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6659 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6660 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6661 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6662 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6663 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6664 (TARGET_ICT_MODEL_SMALL): Define.
6665 (TARGET_ICT_MODEL_LARGE): Define.
6666 * config/nds32/nds32.md (movsi): Consider ict model.
6667 (call, call_value): Consider ict model.
6668 (sibcall, sibcall_value): Consider ict model.
6669 * config/nds32/nds32.opt (mict-model): New option.
6670 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6673 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6674 Monk Chiang <sh.chiang04@gmail.com>
6675 Jim Wilson <jimw@sifive.com>
6677 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6678 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6679 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6680 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6681 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6682 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6683 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6684 compute save_libcall_adjustment properly.
6685 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6686 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6687 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6688 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6689 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6690 (ABI_SPEC): Handle mabi=ilp32e.
6691 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6692 (RVE): Add RVE mask.
6693 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6694 <-march>: Add rv32e as an example.
6696 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6699 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6700 (intra_create_variable_infos): Handle C++ constructors.
6702 2018-05-18 Martin Liska <mliska@suse.cz>
6704 * passes.def: Remove a redundant pass.
6706 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6709 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6711 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6713 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6715 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6717 (armv2, armv2a, armv3, armv3m): Delete architectures.
6718 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6719 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6720 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6722 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6723 (*mulsidi3adddi): Likewise.
6724 (mulsidi3): Likewise.
6725 (*mulsidi3_nov6): Likewise.
6726 (umulsidi3): Likewise.
6727 (umulsidi3_nov6): Likewise.
6728 (umaddsidi4): Likewise.
6729 (*umulsidi3adddi): Likewise.
6730 (smulsi3_highpart): Likewise.
6731 (*smulsi3_highpart_nov6): Likewise.
6732 (umulsi3_highpart): Likewise.
6733 (*umulsi3_highpart_nov6): Likewise.
6734 * config/arm/arm.h (arm_arch3m): Delete.
6735 * config/arm/arm.c (arm_arch3m): Delete.
6736 (arm_option_override_internal): Update armv3-related comment.
6737 (arm_configure_build_target): Delete use of isa_bit_mode32.
6738 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6739 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6740 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6742 (mulusa3): Likewise.
6743 * config/arm/arm-protos.h (arm_arch3m): Delete.
6744 * config/arm/arm-tables.opt: Regenerate.
6745 * config/arm/arm-tune.md: Likewise.
6746 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6747 deleted architectures.
6749 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6751 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6752 (armv5t, armv5te): New features.
6753 (ARMv5, ARMv5e): Delete fgroups.
6754 (ARMv5t, ARMv5te): Adjust for above changes.
6756 (armv5, armv5e): Delete arches.
6757 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6759 (*call_reg_arm): Likewise.
6760 (*call_value_reg_armv5): Likewise.
6761 (*call_value_reg_arm): Likewise.
6762 (*call_symbol): Likewise.
6763 (*call_value_symbol): Likewise.
6764 (*sibcall_insn): Likewise.
6765 (*sibcall_value_insn): Likewise.
6767 (prefetch): Likewise.
6768 (define_split and define_peephole2 dependent on arm_arch5):
6770 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6772 (TARGET_ARM_QBIT): Likewise.
6773 (TARGET_DSP_MULTIPLY): Likewise.
6774 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6775 (arm_arch5, arm_arch5e): Delete.
6776 (arm_arch5t, arm_arch5te): Declare.
6777 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6778 (arm_arch5t): Declare.
6779 (arm_option_reconfigure_globals): Update for the above.
6780 (arm_options_perform_arch_sanity_checks): Update comment, replace
6781 use of arm_arch5 with arm_arch5t.
6782 (use_return_insn): Likewise.
6783 (arm_emit_call_insn): Likewise.
6784 (output_return_instruction): Likewise.
6785 (arm_final_prescan_insn): Likewise.
6786 (arm_coproc_builtin_available): Likewise.
6787 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6788 arm_arch5e with arm_arch5t and arm_arch5te.
6789 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6790 (arm_arch5t, arm_arch5te): Declare.
6791 * config/arm/arm-tables.opt: Regenerate.
6792 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6793 * config/arm/t-multilib: Likewise.
6794 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6795 instead of arm_arch5.
6796 (*call_reg_thumb1): Likewise.
6797 (*call_value_reg_thumb1_v5): Likewise.
6798 (*call_value_reg_thumb1): Likewise.
6799 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6801 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6803 2018-05-18 Martin Liska <mliska@suse.cz>
6805 PR gcov-profile/84846
6806 * doc/gcov.texi: Document -t option of gcov tool.
6808 2018-05-18 Martin Liska <mliska@suse.cz>
6810 PR gcov-profile/84846
6811 * gcov.c (print_usage): Add new -t option.
6812 (process_args): Handle the option.
6813 (generate_results): Use stdout as output when requested by
6816 2018-05-18 Martin Liska <mliska@suse.cz>
6818 PR gcov-profile/84846
6819 * coverage.c (coverage_init): Write PWD to .gcno file.
6820 * doc/gcov.texi: Document how working directory is printed.
6821 * gcov-dump.c (dump_gcov_file): Print PWD.
6822 * gcov.c (output_intermediate_file): Likewise.
6823 (read_graph_file): Read PWD string.
6824 (output_lines): Print PWD.
6826 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6829 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6830 for retval and return false if all args to phi are zero.
6832 2018-05-18 Richard Biener <rguenther@suse.de>
6834 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6836 (evrp_dom_walker::before_dom_children): Call it.
6838 2018-05-18 Richard Biener <rguenther@suse.de>
6840 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6841 results when processing array refs with variable index.
6843 2018-05-18 Toon Moene <toon@moene.org>
6845 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6846 directly after that of -floop-interchange. Indicate that both
6847 options are enabled by default when specifying -O3.
6849 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6851 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6852 iterator. Delete separate integer-mode vec_set<mode> expander.
6853 (aarch64_simd_vec_setv2di): Delete.
6854 (vec_setv2di): Delete.
6855 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6856 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6857 the "w, r" alternative.
6859 2018-05-18 Martin Liska <mliska@suse.cz>
6861 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6862 * tree-pass.h (make_pass_lower_switch_O0): New function.
6863 * tree-switch-conversion.c (node_has_low_bound): Remove.
6864 (node_has_high_bound): Likewise.
6865 (node_is_bounded): Likewise.
6866 (class pass_lower_switch): Make it a template type and create
6868 (pass_lower_switch::execute): Add template argument.
6869 (make_pass_lower_switch): New function.
6870 (make_pass_lower_switch_O0): New function.
6871 (do_jump_if_equal): Remove.
6872 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6873 all the hard work to tree optimization passes.
6875 2018-05-18 Martin Liska <mliska@suse.cz>
6877 * dbgcnt.c (limit_low): Renamed from limit.
6878 (limit_high): New variable.
6879 (dbg_cnt_is_enabled): Check for upper limit.
6880 (dbg_cnt): Adjust dumping.
6881 (dbg_cnt_set_limit_by_index): Add new argument for high
6883 (dbg_cnt_set_limit_by_name): Likewise.
6884 (dbg_cnt_process_single_pair): Parse new format.
6885 (dbg_cnt_process_opt): Use strtok.
6886 (dbg_cnt_list_all_counters): Remove 'value' and add
6888 * doc/invoke.texi: Document changes.
6890 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6892 * doc/sourcebuild.texi (scalar_all_fma): Document.
6893 * tree.def (FMA_EXPR): Delete.
6894 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6895 * internal-fn.c (ternary_direct): New macro.
6896 (expand_ternary_optab_fn): Likewise.
6897 (direct_ternary_optab_supported_p): Likewise.
6898 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6899 * builtins.c (fold_builtin_fma): Delete.
6900 (fold_builtin_3): Don't call it.
6901 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6902 * expr.c (expand_expr_real_2): Likewise.
6903 * fold-const.c (operand_equal_p): Likewise.
6904 (fold_ternary_loc): Likewise.
6905 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6906 * gimple.c (DEFTREECODE): Likewise.
6907 * gimplify.c (gimplify_expr): Likewise.
6908 * optabs-tree.c (optab_for_tree_code): Likewise.
6909 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6910 * tree-eh.c (operation_could_trap_p): Likewise.
6911 (stmt_could_throw_1_p): Likewise.
6912 * tree-inline.c (estimate_operator_cost): Likewise.
6913 * tree-pretty-print.c (dump_generic_node): Likewise.
6914 (op_code_prio): Likewise.
6915 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6916 * tree-ssa-operands.c (get_expr_operands): Likewise.
6917 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6918 * fold-const-call.h (fold_fma): Delete.
6919 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6920 CFN_FNMA and CFN_FNMS.
6922 * genmatch.c (combined_fn): New enum.
6923 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6924 (commutative_op): New function.
6925 (commutate): Use it. Handle more than 2 operands.
6926 (dt_operand::gen_gimple_expr): Use commutative_op.
6927 (parser::parse_expr): Allow :c to be used with non-binary
6928 operators if the commutative operand is known.
6929 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6930 CFN_FMS, CFN_FNMA and CFN_FNMS.
6931 (backprop::process_assign_use): Remove FMA_EXPR handling.
6932 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6933 (gen_hsa_fma): New function.
6934 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6935 IFN_FNMA and IFN_FNMS.
6936 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6937 * gimple-fold.h (follow_all_ssa_edges): Declare.
6938 * gimple-fold.c (follow_all_ssa_edges): New function.
6939 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6940 gimple_build interface and use follow_all_ssa_edges to fold the result.
6941 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6942 instead of checking for optabs directly.
6943 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6944 rather than FMA_EXPRs.
6945 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6946 call to IFN_FMA instead of an FMA_EXPR.
6948 2018-05-17 Jim Wilson <jimw@sifive.com>
6950 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6951 sign extended promoted subreg, and the range does not have the sign bit
6952 set, then do a sign extend.
6954 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6955 test, check for sign extended subreg and/or constant operands, and
6956 do a sign extend in that case.
6958 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6960 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6961 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6963 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6964 Change logics_shift_reg to logics_shift_imm.
6965 (thunderx2t99_fp_loadpair_basic): Delete.
6966 (thunderx2t99_fp_storepair_basic): Delete.
6967 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6968 (thunderx2t99_asimd_polynomial): Delete.
6969 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6970 and neon_fp_mul_d_scalar_q.
6971 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6972 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6973 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6974 (thunderx2t99_asimd_lut): Add missing tbl types.
6975 (thunderx2t99_asimd_ext): Delete.
6976 (thunderx2t99_asimd_load1_1_mult): Delete.
6977 (thunderx2t99_asimd_load1_2_mult): Delete.
6978 (thunderx2t99_asimd_load1_ldp): New.
6979 (thunderx2t99_asimd_load1): New.
6980 (thunderx2t99_asimd_load2): Add missing *load2* types.
6981 (thunderx2t99_asimd_load3): New.
6982 (thunderx2t99_asimd_load4): New.
6983 (thunderx2t99_asimd_store1_1_mult): Delete.
6984 (thunderx2t99_asimd_store1_2_mult): Delete.
6985 (thunderx2t99_asimd_store2_mult): Delete.
6986 (thunderx2t99_asimd_store2_onelane): Delete.
6987 (thunderx2t99_asimd_store_stp): New.
6988 (thunderx2t99_asimd_store1): New.
6989 (thunderx2t99_asimd_store2): New.
6990 (thunderx2t99_asimd_store3): New.
6991 (thunderx2t99_asimd_store4): New.
6993 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6995 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6996 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6998 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6999 Segher Boessenkool <segher@kernel.crashing.org>
7002 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
7005 2018-05-17 Richard Biener <rguenther@suse.de>
7007 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
7008 for pruning loop and prune defs feeding only already visited PHIs.
7010 2018-05-17 Richard Biener <rguenther@suse.de>
7012 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
7014 2018-05-17 Bin Cheng <bin.cheng@arm.com>
7015 Richard Biener <rguenther@suse.de>
7017 PR tree-optimization/85793
7018 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7019 for VMAT_ELEMENTWISE.
7021 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7023 * internal-fn.h (lookup_internal_fn): Declare
7024 * internal-fn.c (lookup_internal_fn): New function.
7025 * gimple.c (gimple_build_call_from_tree): Handle calls to
7027 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7028 internal function names.
7029 * tree-pretty-print.c (dump_generic_node): Likewise.
7030 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7032 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7034 * gimple-fold.h (gimple_build): Make the function forms take
7035 combined_fn rather than built_in_function.
7036 (gimple_simplify): Likewise.
7037 * gimple-match-head.c (gimple_simplify): Likewise.
7038 * gimple-fold.c (gimple_build): Likewise.
7039 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7040 rather than gimple_build_call_internal.
7041 (get_initial_defs_for_reduction): Likewise.
7042 (vect_create_epilog_for_reduction): Likewise.
7043 (vectorizable_live_operation): Likewise.
7045 2018-05-17 Martin Liska <mliska@suse.cz>
7047 * gimple-ssa-sprintf.c (format_directive): Do not use
7048 space in between 'G_' and '('.
7050 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7053 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7054 even if the mask is not all ones.
7057 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7059 (ix86_gimple_fold_builtin): Likewise.
7062 * config/i386/i386.c: Include tree-vector-builder.h.
7063 (ix86_vector_shift_count): New function.
7064 (ix86_fold_builtin): Fold shift builtins by scalar count.
7065 (ix86_gimple_fold_builtin): Likewise.
7067 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7068 _mm512_setzero): New intrinsics.
7070 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7071 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7073 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7074 code generation for cases where splatting a value is not useful.
7075 * simplify-rtx.c (simplify_ternary_operation): Simplify
7076 vec_merge across a vec_duplicate and a paradoxical subreg forming
7077 a vector mode to a vec_concat.
7079 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7081 * config.gcc: Support "goldmont-plus".
7082 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7084 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7085 PROCESSOR_GOLDMONT_PLUS.
7086 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7087 (processor_target_table): Add "goldmont-plus".
7088 (PTA_GOLDMONT_PLUS): Define.
7089 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7090 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7091 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7092 (fold_builtin_cpu): Add "goldmont-plus".
7093 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7094 (ix86_option_override_internal): Add "goldmont-plus".
7095 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7096 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7097 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7098 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7100 2018-05-17 Richard Biener <rguenther@suse.de>
7102 PR tree-optimization/85757
7103 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7104 remove defs that only feed that PHI from further processing.
7106 2018-05-16 Jim Wilson <jimw@sifive.com>
7108 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7110 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7112 2018-05-16 Mark Wielaard <mark@klomp.org>
7114 * dwarf2out.c (count_index_strings): New function.
7115 (output_indirect_strings): Call count_index_strings and generate
7116 header for dwarf_version >= 5.
7118 2018-05-16 Mark Wielaard <mark@klomp.org>
7120 * dwarf2out.c (dwarf_FORM): New function.
7121 (set_indirect_string): Use dwarf_FORM.
7122 (reset_indirect_string): Likewise.
7123 (size_of_die): Likewise.
7124 (value_format): Likewise.
7125 (output_die): Likewise.
7126 (add_skeleton_AT_string): Likewise.
7127 (output_macinfo_op): Likewise.
7128 (index_string): Likewise.
7129 (output_index_string_offset): Likewise.
7130 (output_index_string): Likewise.
7131 (count_index_strings): Likewise.
7133 2018-05-16 Carl Love <cel@us.ibm.com>
7135 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7136 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7138 2018-05-16 Martin Jambor <mjambor@suse.cz>
7140 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7141 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7143 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7145 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7146 (fnma<mode>4): Likewise.
7147 (fms<mode>4): Likewise.
7148 (fnms<mode>4): Likewise.
7149 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7150 (aarch64_fnma<mode>4): Likewise.
7151 (aarch64_fms<mode>4): Likewise.
7152 (aarch64_fnms<mode>4): Likewise.
7153 (aarch64_fnmadd<mode>4): Likewise.
7155 2018-05-16 Jason Merrill <jason@redhat.com>
7157 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7159 2018-05-16 Richard Biener <rguenther@suse.de>
7161 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7162 (dump_stmt_cost): Declare.
7163 (add_stmt_cost): Dump cost we add.
7164 (add_stmt_costs): New function.
7165 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7167 (vect_analyze_stmt): Adjust prototype.
7168 (vectorizable_condition): Likewise.
7169 (vectorizable_live_operation): Likewise.
7170 (vectorizable_reduction): Likewise.
7171 (vectorizable_induction): Likewise.
7172 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7173 cost vector to pass to vectorizable_ and record afterwards.
7174 (vect_model_reduction_cost): Take cost vector argument and adjust.
7175 (vect_model_induction_cost): Likewise.
7176 (vectorizable_reduction): Likewise.
7177 (vectorizable_induction): Likewise.
7178 (vectorizable_live_operation): Likewise.
7179 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7180 SLP_TREE_NUMBER_OF_VEC_STMTS.
7181 (vect_analyze_slp_cost_1): Remove.
7182 (vect_analyze_slp_cost): Likewise.
7183 (vect_slp_analyze_node_operations): Take visited args and
7184 a target cost vector. Avoid processing already visited stmt sets.
7185 (vect_slp_analyze_operations): Use a local cost vector to gather
7186 costs and register those of non-discarded instances.
7187 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7188 (vect_schedule_slp_instance): Remove copying of
7189 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7191 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7192 adding cost. Record cost entry location.
7193 (vect_prologue_cost_for_slp_op): Function to compute cost of
7194 a constant or invariant generated for SLP vect in the prologue,
7195 split out from vect_analyze_slp_cost_1.
7196 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7197 (vect_model_promotion_demotion_cost): Likewise.
7198 (vect_model_store_cost): Likewise, make static.
7199 (vect_model_load_cost): Likewise.
7200 (vectorizable_bswap): Add cost vector arg and adjust.
7201 (vectorizable_call): Likewise.
7202 (vectorizable_simd_clone_call): Likewise.
7203 (vectorizable_conversion): Likewise.
7204 (vectorizable_assignment): Likewise.
7205 (vectorizable_shift): Likewise.
7206 (vectorizable_operation): Likewise.
7207 (vectorizable_store): Likewise.
7208 (vectorizable_load): Likewise.
7209 (vectorizable_condition): Likewise.
7210 (vectorizable_comparison): Likewise.
7211 (can_vectorize_live_stmts): Likewise.
7212 (vect_analyze_stmt): Likewise.
7213 (vect_transform_stmt): Adjust calls to vectorizable_*.
7214 * tree-vectorizer.c: Include gimple-pretty-print.h.
7215 (dump_stmt_cost): New function.
7217 2018-05-16 Richard Biener <rguenther@suse.de>
7219 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7220 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7221 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7222 (check_name): New callback.
7223 (dse_classify_store): Track cycles via a visited bitmap of PHI
7224 defs and simplify handling of in-loop and across loop dead stores
7225 and properly fail for loop-variant refs. Handle byte-tracking with
7226 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7229 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7231 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7232 (vect_get_mask_type_for_stmt): Likewise.
7233 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7235 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7236 to determine the statement's vector type and the vector type that
7237 should be used for calculating nunits. Deal with cases in which
7238 the type has to be deferred.
7239 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7240 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7241 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7242 (vect_determine_vf_for_stmt): New functions, split out from...
7243 (vect_determine_vectorization_factor): ...here.
7244 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7245 (vect_get_mask_type_for_stmt): New functions, split out from
7246 vect_determine_vectorization_factor.
7248 2018-05-16 Richard Biener <rguenther@suse.de>
7250 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7251 verify the [VEC_]COND_EXPR embedded comparison.
7253 2018-05-15 Martin Sebor <msebor@redhat.com>
7255 PR tree-optimization/85753
7256 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7257 RECORD_TYPE in addition to ARRAY_TYPE.
7259 2018-05-15 Martin Sebor <msebor@redhat.com>
7262 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7264 2018-05-15 Richard Biener <rguenther@suse.de>
7266 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7267 add by_clobber_p one. Change algorithm to collect all defs
7268 representing uses we need to walk and try reducing them to
7269 a single one before failing.
7270 (dse_dom_walker::dse_optimize_stmt): Adjust.
7272 2018-05-13 Mark Wielaard <mark@klomp.org>
7274 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7275 (size_of_loc_descr): Likewise.
7276 (output_loc_operands): Likewise.
7277 (output_loc_operands_raw): Likewise.
7278 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7279 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7280 (hash_loc_operands): Likewise.
7281 (compare_loc_operands): Likewise.
7283 2018-05-14 Mark Wielaard <mark@klomp.org>
7285 * dwarf2out.c (count_index_addrs): New function.
7286 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7288 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7290 PR tree-optimization/83648
7291 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7292 return value as malloc candidate.
7294 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7297 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7298 param as true in call to suggest_attribute.
7300 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7302 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7303 -mreadonly-in-sdata.
7305 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7307 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7309 (aarch64_crypto_aesd_fused): Likewise.
7311 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7313 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7314 (movsi_aarch64): Likewise.
7315 (load_pairsi): Likewise.
7316 (load_pairdi): Likewise.
7317 (store_pairsi): Likewise.
7318 (store_pairdi): Likewise.
7319 (load_pairsf): Likewise.
7320 (load_pairdf): Likewise.
7321 (store_pairsf): Likewise.
7322 (store_pairdf): Likewise.
7323 (zero_extend): Likewise.
7324 (trunc): Swap alternatives.
7325 (fcvt_target): Add '?' to prefer w over r.
7327 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7330 * config/i386/i386.md: Disallow non-commutative arithmetics in
7331 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7332 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7333 in the peephole2 before it.
7335 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7337 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7338 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7339 (ix86_handle_option): Handle -mcldemote.
7340 * config.gcc: New header.
7341 * config/i386/cldemoteintrin.h: New file.
7342 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7343 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7345 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7346 OPTION_MASK_ISA_CLDEMOTE.
7347 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7348 (ix86_valid_target_attribute_inner_p): Ditto.
7349 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7350 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7351 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7352 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7353 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7355 * config/i386/i386.opt: Add -mcldemote.
7356 * config/i386/x86intrin.h: New header.
7357 * doc/invoke.texi: Add -mcldemote.
7359 2018-05-14 Richard Biener <rguenther@suse.de>
7361 * doc/match-and-simplify.texi: Adjust :s documentation.
7363 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7365 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7366 intended memcpy size.
7367 (REORDER_45): Likewise.
7369 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7371 * sort.cc: New file.
7372 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7373 * vec.c (qsort_chk): Use gcc_qsort.
7374 * Makefile.in (OBJS-libcommon): Add sort.o.
7375 (build/sort.o): New target. Use it...
7376 (BUILD_RTL): ... here, and...
7377 (build/gencfn-macros): ... here, and...
7378 (build/genmatch): ... here.
7380 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7381 Chung-Ju Wu <jasonwucj@gmail.com>
7383 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7384 * config/nds32/nds32-graywolf.md: New file.
7385 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7386 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7388 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7389 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7390 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7391 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7392 * config/nds32/pipelines.md: Include n15 settings.
7394 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7395 Chung-Ju Wu <jasonwucj@gmail.com>
7397 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7398 * config/nds32/nds32-n13.md: New file.
7399 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7400 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7402 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7403 * config/nds32/nds32.md (pipeline_model): Add n13.
7404 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7405 * config/nds32/pipelines.md: Include n13 settings.
7407 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7408 Chung-Ju Wu <jasonwucj@gmail.com>
7410 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7411 * config/nds32/nds32-n10.md: New file.
7412 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7413 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7415 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7416 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7417 * config/nds32/nds32.md (pipeline_model): Add n10.
7418 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7419 * config/nds32/pipelines.md: Include n10 settings.
7421 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7422 Kito Cheng <kito.cheng@gmail.com>
7423 Chung-Ju Wu <jasonwucj@gmail.com>
7425 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7426 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7427 Add enum values for DSP extension instructions.
7428 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7430 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7431 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7433 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7434 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7435 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7436 * config/nds32/nds32-intrinsic.md: Likewise.
7437 * config/nds32/nds32_intrinsic.h: Likewise.
7438 * config/nds32/nds32-md-auxiliary.c: Likewise.
7439 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7440 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7441 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7442 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7443 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7444 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7445 TYPE_DMAC in switch statement.
7446 * config/nds32/nds32.c: New checking and implementation for DSP
7447 extension instructions.
7448 * config/nds32/nds32.h: Likewise.
7449 * config/nds32/nds32.md: Likewise.
7450 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7451 * config/nds32/predicates.md: Implement new predicates for DSP
7454 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7456 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7457 Reformat alternatives and attributes so it is easier to identify
7458 which constraints/attributes go with which instruction.
7459 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7460 (mov<mode>_softfloat32, FMOVE64): Likewise.
7461 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7462 (mov<mode>_softfloat64, FMOVE64): Likewise.
7464 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7466 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7468 (Basic PowerPC Built-in Functions): The new name of the
7469 subsection previously known as "PowerPC Built-in Functions".
7470 (Basic PowerPC Built-in Functions Available on all Configurations):
7472 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7473 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7474 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7475 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7477 2018-05-11 Martin Jambor <mjambor@suse.cz>
7480 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7483 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7486 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7488 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7490 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7491 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7492 (ix86_handle_option): Handle -mwaitpkg.
7493 * config.gcc: New header.
7494 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7495 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7496 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7498 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7499 OPTION_MASK_ISA_WAITPKG.
7500 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7501 (ix86_option_override_internal): Add PTA_WAITPKG.
7502 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7503 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7504 IX86_BUILTIN_TPAUSE.
7505 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7506 __builtin_ia32_umwait and __builtin_ia32_tpause.
7507 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7508 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7509 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7510 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7511 UNSPECV_TPAUSE): New.
7512 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7513 * config/i386/i386.opt: Add -mwaitpkg.
7514 * config/i386/waitpkgintrin.h: New file.
7515 * config/i386/x86intrin.h: New header.
7516 * doc/invoke.texi: Add -mwaitpkg.
7518 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7521 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7523 (cortex-m0): Use armv6s-m isa.
7524 (cortex-m0plus): Likewise.
7525 (cortex-m1): Likewise.
7526 (cortex-m0.small-multiply): Likewise.
7527 (cortex-m0plus.small-multiply): Likewise.
7528 (cortex-m1.small-multiply): Likewise.
7530 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7531 Jakub Jelinek <jakub@redhat.com>
7533 PR tree-optimization/85692
7534 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7535 source permute as well.
7537 2018-05-11 Martin Liska <mliska@suse.cz>
7540 * doc/extend.texi: Document LLVM style format for no_sanitize
7543 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7545 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7546 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7547 (mode_supports_vsx_dform_quad): Likewise.
7548 (mode_supports_vmx_dform): Move these functions to be next to the
7549 other mode_supports functions.
7550 (mode_supports_dq_form): Likewise.
7551 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7552 mode_supports_dq_form.
7553 (reg_offset_addressing_ok_p): Likewise.
7554 (offsettable_ok_by_alignment): Likewise.
7555 (rs6000_legitimate_offset_address_p): Likewise.
7556 (legitimate_lo_sum_address_p): Likewise.
7557 (rs6000_legitimize_address): Likewise.
7558 (rs6000_legitimize_reload_address): Likewise.
7559 (rs6000_secondary_reload_inner): Likewise.
7560 (rs6000_preferred_reload_class): Likewise.
7561 (rs6000_output_move_128bit): Likewise.
7563 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7565 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7566 Generate SImode target register for null target.
7567 <case IX86_BUILTIN_XGETBV>: Ditto.
7568 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7569 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7571 2018-05-10 Carl Love <cel@us.ibm.com>
7573 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7574 dcbtt and dcbtstt if operands[2] is 0.
7576 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7579 * config/i386/sse.md (usadv64qi): New expander.
7581 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7583 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7584 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7585 -maltivec=be support.
7586 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7587 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7588 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7589 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7590 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7591 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7592 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7593 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7594 altivec_vsumsws): Adjust.
7595 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7596 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7597 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7599 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7600 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7601 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7602 (altivec_lve<VI_char>x): Delete expand.
7603 (*altivec_lve<VI_char>x_internal): Rename to...
7604 (altivec_lve<VI_char>x): ... this.
7605 (altivec_lvxl_<mode>): Delete expand.
7606 (*altivec_lvxl_<mode>_internal): Rename to ...
7607 (altivec_lvxl_<mode>): ... this.
7608 (altivec_stvxl_<mode>): Delete expand.
7609 (*altivec_stvxl_<mode>_internal): Rename to ...
7610 (altivec_stvxl_<mode>): ... this.
7611 (altivec_stve<VI_char>x): Delete expand.
7612 (*altivec_stve<VI_char>x_internal): Rename to ...
7613 (altivec_stve<VI_char>x): ... this.
7614 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7615 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7616 reduc_plus_scal_<mode>): Adjust.
7617 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7619 (rs6000_cpu_cpp_builtins): Adjust.
7620 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7621 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7622 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7623 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7624 -maltivec=be support.
7625 (rs6000_split_vec_extract_var): Adjust.
7626 (rs6000_split_v4si_init): Adjust.
7627 (swap_selector_for_mode): Delete.
7628 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7629 altivec_expand_stvex_be): Delete.
7630 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7631 -maltivec=be support.
7632 (rs6000_gimple_fold_builtin): Ditto.
7633 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7635 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7636 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7637 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7638 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7639 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7640 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7641 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7642 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7643 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7644 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7645 anonymous split): Adjust.
7646 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7647 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7649 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7651 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7652 when --with-gxx-include-dir is also specified.
7653 * configure: Regenerate.
7655 2018-05-09 Jim Wilson <jimw@sifive.com>
7658 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7659 * config/riscv/t-withmultilib: New.
7660 * config/riscv/withmultilib.h: New.
7661 * doc/install.texi: Document RISC-V --with-multilib-list support.
7663 2018-05-09 Richard Biener <rguenther@suse.de>
7665 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7667 (vect_bb_vectorization_profitable_p): Adjust. Compute
7668 actual scalar cost using the cost vector and the add_stmt_cost
7671 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7673 PR rtl-optimization/85645
7674 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7675 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7677 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7679 PR rtl-optimization/85645
7680 * shrink-wrap.c (spread_components): Return a boolean saying if
7681 anything was changed.
7682 (try_shrink_wrapping_separate): Iterate spread_components until
7683 nothing changes anymore.
7685 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7687 PR rtl-optimization/85645
7688 * regrename.c (build_def_use): Also kill the chains that include the
7689 destination of a REG_CFA_REGISTER note.
7691 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7693 PR rtl-optimization/85645
7694 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7695 insn that has a REG_CFA_REGISTER note.
7697 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7699 * cfgexpand.c (expand_clobber): New function.
7700 (expand_gimple_stmt_1): Use it.
7701 * tree-vect-stmts.c (vect_clobber_variable): New function,
7703 (vectorizable_simd_clone_call): ...here.
7704 (vectorizable_store): Emit a clobber either side of an
7705 IFN_STORE_LANES sequence.
7706 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7708 2018-05-09 Tom de Vries <tom@codesourcery.com>
7711 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7712 (define_insn "trap_if_false"): Add exit after trap.
7714 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7716 PR rtl-optimization/85638
7717 * bb-reorder.c: Include common/common-target.h.
7718 (create_forwarder_block): New function extracted from...
7719 (fix_up_crossing_landing_pad): ...here. Rename into...
7720 (dw2_fix_up_crossing_landing_pad): ...this.
7721 (sjlj_fix_up_crossing_landing_pad): New function.
7722 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7723 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7724 from both partitions and exit the loop after one iteration.
7726 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7729 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7731 (Basic PowerPC Built-in Functions): The new name of the
7732 subsection previously known as "PowerPC Built-in Functions".
7733 (Basic PowerPC Built-in Functions Available on all Configurations):
7735 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7737 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7739 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7741 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7744 2018-05-08 Jim Wilson <jimw@sifive.com>
7746 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7747 (LD_EMUL_SUFFIX): New.
7748 (LINK_SPEC): Use it.
7750 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7752 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7754 (Basic PowerPC Built-in Functions): The new name of the
7755 subsection previously known as "PowerPC Built-in Functions".
7756 (Basic PowerPC Built-in Functions Available on all Configurations):
7758 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7760 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7762 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7764 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7767 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7770 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7771 after cmpelim optimization.
7773 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7775 * config.gcc: Support "goldmont".
7776 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7777 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7779 * config/i386/i386.c (m_GOLDMONT): Define.
7780 (processor_target_table): Add "goldmont".
7781 (PTA_GOLDMONT): Define.
7782 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7783 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7784 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7785 (fold_builtin_cpu): Add "goldmont".
7786 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7787 (ix86_option_override_internal): Add "goldmont".
7788 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7789 (processor_type): Add PROCESSOR_GOLDMONT.
7790 * config/i386/i386.md: Add CPU "glm".
7791 * config/i386/glm.md: New file.
7792 * config/i386/x86-tune.def: Add m_GOLDMONT.
7793 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7795 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7798 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7800 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7801 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7802 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7805 * config/i386/i386.c (ix86_fold_builtin): Handle
7806 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7809 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7810 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7812 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7815 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7816 (check_arch): Likewise.
7817 (check_fpu): Return the result rather than printing it.
7818 (end arch): Fix operator precedence.
7819 (end cpu): Likewise.
7820 (END): Print the result from check_fpu.
7822 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7823 Alan Hayward <alan.hayward@arm.com>
7824 David Sherwood <david.sherwood@arm.com>
7826 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7827 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7828 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7829 (*fcmuo<mode>_and): New patterns.
7831 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7833 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7834 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7835 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7836 (cmp_op, sve_imm_con): New code attributes.
7837 (SVE_COND_INT_CMP, imm_con): Delete.
7838 (cmp_op): Remove above unspecs from int attribute.
7839 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7841 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7842 comparison-specific unspecs.
7843 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7844 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7845 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7846 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7847 (*vec_fcm<cmp_op><mode>): Rename to...
7848 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7849 (*vec_fcmuo<mode>): Rename to...
7850 (*fcmuo<mode>): ...this and adjust likewise.
7851 (*pred_fcm<cmp_op><mode>): New pattern.
7852 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7853 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7855 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7857 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7858 (aarch64_emit_sve_predicated_cond): New function.
7859 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7860 (aarch64_emit_unspec_cond_or): Replace with...
7861 (aarch64_emit_sve_or_conds): ...this new function. Use
7862 aarch64_emit_sve_ptrue_op for the individual comparisons and
7863 aarch64_emit_binop to OR them together.
7864 (aarch64_emit_inverted_unspec_cond): Replace with...
7865 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7866 aarch64_emit_sve_ptrue_op for the comparison and
7867 aarch64_emit_unop to invert the result.
7868 (aarch64_expand_sve_vec_cmp_float): Update after the above
7869 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7871 2018-05-07 Nathan Sidwell <nathan@acm.org>
7873 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7874 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7875 (Backwards Compatibility): Likewise.
7877 2018-05-07 Luis Machado <luis.machado@linaro.org>
7881 2018-05-07 Luis Machado <luis.machado@linaro.org>
7883 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7884 <prefetch_dynamic_strides>: New const bool field.
7885 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7886 prefetch_dynamic_strides.
7887 (exynosm1_prefetch_tune): Likewise.
7888 (thunderxt88_prefetch_tune): Likewise.
7889 (thunderx_prefetch_tune): Likewise.
7890 (thunderx2t99_prefetch_tune): Likewise.
7891 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7893 (aarch64_override_options_internal): Update to set
7894 PARAM_PREFETCH_DYNAMIC_STRIDES.
7895 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7896 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7897 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7898 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7899 prefetch-dynamic-strides setting.
7901 2018-05-07 Luis Machado <luis.machado@linaro.org>
7903 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7904 <minimum_stride>: New const int field.
7905 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7906 minimum_stride field.
7907 (exynosm1_prefetch_tune): Likewise.
7908 (thunderxt88_prefetch_tune): Likewise.
7909 (thunderx_prefetch_tune): Likewise.
7910 (thunderx2t99_prefetch_tune): Likewise.
7911 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7912 (aarch64_override_options_internal): Update to set
7913 PARAM_PREFETCH_MINIMUM_STRIDE.
7914 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7915 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7916 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7917 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7918 stride is constant and is below the minimum stride threshold.
7920 2018-05-07 Luis Machado <luis.machado@linaro.org>
7922 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7925 2018-05-07 Luis Machado <luis.machado@linaro.org>
7927 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7928 <prefetch_dynamic_strides>: New const bool field.
7929 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7930 prefetch_dynamic_strides.
7931 (exynosm1_prefetch_tune): Likewise.
7932 (thunderxt88_prefetch_tune): Likewise.
7933 (thunderx_prefetch_tune): Likewise.
7934 (thunderx2t99_prefetch_tune): Likewise.
7935 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7937 (aarch64_override_options_internal): Update to set
7938 PARAM_PREFETCH_DYNAMIC_STRIDES.
7939 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7940 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7941 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7942 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7943 prefetch-dynamic-strides setting.
7945 2018-05-07 Luis Machado <luis.machado@linaro.org>
7947 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7948 <minimum_stride>: New const int field.
7949 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7950 minimum_stride field.
7951 (exynosm1_prefetch_tune): Likewise.
7952 (thunderxt88_prefetch_tune): Likewise.
7953 (thunderx_prefetch_tune): Likewise.
7954 (thunderx2t99_prefetch_tune): Likewise.
7955 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7956 (aarch64_override_options_internal): Update to set
7957 PARAM_PREFETCH_MINIMUM_STRIDE.
7958 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7959 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7960 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7961 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7962 stride is constant and is below the minimum stride threshold.
7964 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7967 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7968 the type is addressable. Don't force op into register if it has
7971 2018-05-05 Roland McGrath <mcgrathr@google.com>
7974 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7975 any section for which we don't know a specific type it should have,
7976 regardless of name. Previously this was done only for the exact
7977 names ".init_array", ".fini_array", and ".preinit_array".
7978 (default_elf_asm_named_section): Add comment about
7979 relationship with default_section_type_flags and SECTION_NOTYPE.
7980 (get_section): Don't consider it a type conflict if one side has
7981 SECTION_NOTYPE and the other doesn't, as long as neither has the
7982 SECTION_BSS et al used in the default_section_type_flags logic.
7984 2018-05-05 Tom de Vries <tom@codesourcery.com>
7987 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7988 (workaround_barsyncs): New function.
7989 (nvptx_reorg): Use workaround_barsyncs.
7990 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7991 (define_expand "nvptx_membar_cta"): New define_expand.
7992 (define_insn "*nvptx_membar_cta"): New insn.
7994 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7996 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7997 To improve optimization opportunities.
7998 * builtin-types.def: The new needed builtin types for the above.
8000 2018-05-04 Richard Biener <rguenther@suse.de>
8002 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
8003 * gimple-ssa-store-merging.c
8004 (imm_store_chain_info::output_merged_store): Remove redundant create,
8005 release split_store vector contents on failure.
8006 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
8007 scalar stmt vector on cache hit.
8009 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
8011 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
8013 * config.gcc (powerpc-xilinx-eabi*): Remove.
8014 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
8016 (fusion_addis_mem_combo_load): Ditto.
8017 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8019 (rs6000_cpu_cpp_builtins): Ditto.
8020 * config/rs6000/rs6000-linux.c
8021 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8022 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8023 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8025 (rs6000_setup_reg_addr_masks): Ditto.
8026 (rs6000_init_hard_regno_mode_ok): Ditto.
8027 (rs6000_option_override_internal): Ditto.
8028 (legitimate_lo_sum_address_p): Ditto.
8029 (rs6000_legitimize_address): Ditto.
8030 (rs6000_legitimize_reload_address): Ditto.
8031 (rs6000_legitimate_address_p): Ditto.
8032 (abi_v4_pass_in_fpr): Ditto.
8033 (setup_incoming_varargs): Ditto.
8034 (rs6000_gimplify_va_arg): Ditto.
8035 (rs6000_split_multireg_move): Ditto.
8036 (rs6000_savres_strategy): Ditto.
8037 (rs6000_emit_prologue_components): Ditto.
8038 (rs6000_emit_epilogue_components): Ditto.
8039 (rs6000_emit_prologue): Ditto.
8040 (rs6000_emit_epilogue): Ditto.
8041 (rs6000_elf_file_end): Ditto.
8042 (rs6000_function_value): Ditto.
8043 (rs6000_libcall_value): Ditto.
8044 * config/rs6000/rs6000.h: Ditto.
8045 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8046 (TARGET_MINMAX): ... this. New.
8047 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8048 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8049 (*movsi_internal1_single): Delete.
8050 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8051 mfpu=, mxilinx-fpu): Delete.
8052 * config/rs6000/singlefp.h: Delete.
8053 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8054 * config/rs6000/t-rs6000: Ditto.
8055 * config/rs6000/t-xilinx: Delete.
8056 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
8057 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
8058 (VStype_simple): Delete.
8059 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8060 * config/rs6000/xfpu.h: Delete.
8061 * config/rs6000/xfpu.md: Delete.
8062 * config/rs6000/xilinx.h: Delete.
8063 * config/rs6000/xilinx.opt: Delete.
8064 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8065 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8067 2018-05-04 Tom de Vries <tom@codesourcery.com>
8070 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8073 2018-05-04 Richard Biener <rguenther@suse.de>
8076 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8077 (expand_complex_div_wide): Likewise.
8078 (expand_complex_operations_1): Likewise.
8079 (expand_complex_libcall): Preserve EH info of the original stmt.
8080 (tree_lower_complex): Handle removed blocks.
8081 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8082 on complex multiplication and division libcall builtins.
8084 2018-05-04 Richard Biener <rguenther@suse.de>
8087 * fold-const.c (negate_expr_p): Restrict negation of operand
8088 zero of a division to when we know that can happen without
8090 (fold_negate_expr_1): Likewise.
8092 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8095 * real.h (real_nextafter): Declare.
8096 * real.c (real_nextafter): New function.
8097 * fold-const-call.c (fold_const_nextafter): New function.
8098 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8099 CASE_CFN_NEXTTOWARD.
8100 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8101 even when arg1_mode is different from arg0_mode.
8103 2018-05-03 Nathan Sidwell <nathan@acm.org>
8105 * doc/extend.texi (Deprecated Features): Remove
8107 (Backwards Compatibility): Likewise.
8108 * doc/invoke.texi (C++ Language Options): Likewise.
8109 (C++ Dialect Options): Likewise.
8111 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8114 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8115 _mm512_mask_mullox_epi64): New intrinsics.
8117 2018-05-03 Tom de Vries <tom@codesourcery.com>
8120 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8121 dump files): Add offload-tree.
8123 2018-05-03 Richard Biener <rguenther@suse.de>
8125 PR tree-optimization/85615
8126 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8127 to loops not nested in BBs loop father to avoid creating multi-entry
8130 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8132 PR tree-optimization/70291
8133 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8134 arguments. Change return type to tree. Emit libcall as a new
8135 statement rather than replacing existing one when inplace_p is true.
8136 (expand_complex_multiplication_components): New function.
8137 (expand_complex_multiplication): Expand floating-point complex
8138 multiplication using the above.
8139 (expand_complex_division): Rename inner_type parameter to type.
8140 Update expand_complex_libcall call-site.
8141 (expand_complex_operations_1): Update expand_complex_multiplication
8142 and expand_complex_division call-sites.
8144 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8147 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8148 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8149 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8150 the highest significant bit of the shift count mask is clear. In
8151 check whether and[sq]i3 is needed verify that all significant bits
8152 of the shift count other than the highest are set.
8154 2018-05-02 Tom de Vries <tom@codesourcery.com>
8157 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8158 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8159 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8160 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8161 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8162 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8163 __builtin_goacc_parlevel_size.
8165 2018-05-02 Richard Biener <rguenther@suse.de>
8167 PR tree-optimization/85597
8168 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8169 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8172 2018-05-02 Tom de Vries <tom@codesourcery.com>
8175 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8176 dump files): Add ltrans-tree.
8178 2018-05-02 Tom de Vries <tom@codesourcery.com>
8181 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8182 dump files): Add wpa-ipa.
8184 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8186 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8187 powerpc*-*-linux*paired* target.
8188 * config/rs6000/750cl.h: Delete.
8189 * config/rs6000/paired.h: Delete.
8190 * config/rs6000/paired.md: Delete.
8191 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8193 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8194 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8195 comment. Remove paired float support.
8196 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8197 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8199 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8200 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8202 * config/rs6000/rs6000.c: Remove paired float support.
8203 (paired_expand_vector_init, paired_expand_vector_move,
8204 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8205 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8206 paired_expand_builtin, paired_expand_predicate_builtin,
8207 paired_init_builtins): Delete.
8208 * config/rs6000/rs6000.h: Remove paired float support.
8209 * config/rs6000/rs6000.md: Remove paired float support.
8210 (move_from_CR_ov_bit): Delete.
8211 * config/rs6000/rs6000.opt (mpaired): Delete.
8212 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8213 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8215 2018-05-02 Richard Biener <rguenther@suse.de>
8218 * gimplify.c (gimplify_save_expr): When in SSA form allow
8219 SAVE_EXPRs to compute to SSA vars.
8221 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8224 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8225 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8226 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8227 clobber operands[2], instead use a new pseudo. Formatting fixes.
8229 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8231 PR tree-optimization/85586
8232 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8233 exit early for statements in the same group if the accesses are
8236 2018-05-02 Tom de Vries <tom@codesourcery.com>
8239 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8242 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8244 PR tree-optimization/85143
8245 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8247 2018-05-01 Tom de Vries <tom@codesourcery.com>
8250 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8251 not found" error message.
8253 2018-05-01 Tom de Vries <tom@codesourcery.com>
8256 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8257 * vec.c (test_ordered_remove_if): New function.
8258 (vec_c_tests): Call test_ordered_remove_if.
8259 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8260 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8261 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8262 VEC_ORDERED_REMOVE_IF.
8264 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8266 PR tree-optimization/82665
8267 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8268 pointer subtraction where arguments come from a memchr call.
8270 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8272 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8273 --push-state --as-needed and --pop-state instead of --as-needed and
8274 --no-as-needed if ld supports it.
8275 * configure: Regenerated.
8278 * doc/install.texi2html: Replace _002d with - and _002a with * in
8279 generated html files using sed.
8281 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8284 * gcc-rich-location.c (blank_line_before_p): New function.
8285 (use_new_line): New function.
8286 (gcc_rich_location::add_fixit_insert_formatted): New function.
8287 * gcc-rich-location.h
8288 (gcc_rich_location::add_fixit_insert_formatted): New function.
8290 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8292 * selftest.c (assert_streq): Rename "expected" and "actual" to
8293 "val1" and "val2". Extend NULL-handling to cover both inputs
8294 symmetrically, while still requiring both to be non-NULL for a pass.
8295 * selftest.h (assert_streq): Rename "expected" and "actual" to
8297 (ASSERT_EQ): Likewise.
8298 (ASSERT_EQ_AT): Likewise.
8299 (ASSERT_KNOWN_EQ): Likewise.
8300 (ASSERT_KNOWN_EQ_AT): Likewise.
8301 (ASSERT_NE): Likewise.
8302 (ASSERT_MAYBE_NE): Likewise.
8303 (ASSERT_MAYBE_NE_AT): Likewise.
8304 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8305 the assertion to pass.
8306 (ASSERT_STREQ_AT): Likewise.
8308 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8310 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8311 interaction with -pie.
8313 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8315 * selftest.h: Fix alphabetization of per-source-file selftest
8318 2018-04-30 Jason Merrill <jason@redhat.com>
8320 PR c++/61982 - dead stores to destroyed objects.
8321 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8324 2018-04-30 Jason Merrill <jason@redhat.com>
8326 * tree.c (build_clobber): New.
8327 * tree.h: Declare it.
8328 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8330 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8332 * diagnostic-show-locus.c (layout::layout): Update for
8333 location_get_source_line returning a char_span.
8334 (struct char_span): Move to input.h.
8335 (struct correction): Update for fields in char_span becoming
8337 (struct source_line): Update for location_get_source_line
8338 returning a char_span.
8339 (layout::print_line): Likewise.
8340 * edit-context.c (edited_file::print_content): Likewise.
8341 (edited_file::print_diff_hunk): Likewise.
8342 (edited_file::print_run_of_changed_lines): Likewise.
8343 (edited_file::get_num_lines): Likewise.
8344 (edited_line::edited_line): Likewise.
8345 * final.c (asm_show_source): Likewise.
8346 * input.c (location_get_source_line): Convert return type
8347 from const char * to char_span, losing the final "line_len"
8349 (dump_location_info): Update for the above.
8350 (get_substring_ranges_for_loc): Likewise. Use a char_span
8351 when handling the literal within the line.
8352 (test_reading_source_line): Update for location_get_source_line
8353 returning a char_span.
8354 * input.h (class char_span): Move here from
8355 diagnostic-show-locus.c, converting from a struct to a class.
8356 Make data members private.
8357 (char_span::operator bool): New.
8358 (char_span::length): New.
8359 (char_span::get_buffer): New.
8360 (char_span::operator[]): New.
8361 (char_span::subspan): Make const.
8362 (char_span::xstrdup): New.
8363 (location_get_source_line): Convert return type from const char *
8364 to char_span, losing the final "line_size" param.
8366 2018-04-30 Jan Hubicka <jh@suse.cz>
8368 * lto-wrapper.c (ltrans_priorities): New static var.
8369 (cmp_priority): New.
8370 (run_gcc): Read priorities and if doing parallel build order
8371 the Makefile by them.
8373 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8375 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8377 2018-04-30 Richard Biener <rguenther@suse.de>
8379 * tree-cfg.c (verify_address): Remove base argument, add
8380 flag whether to check TREE_ADDRESSABLE and do that.
8381 (verify_expr): Remove.
8382 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8383 (verify_gimple_assign_single): Likewise.
8384 (verify_gimple_switch): Likewise.
8385 (verify_expr_location_1): Dereference tp once. Add (disabled)
8386 piece from verify_expr.
8387 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8389 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8391 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8393 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8395 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8396 (small_data_pattern): Likewise.
8397 (arc_rewrite_small_data): Likewise.
8398 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8399 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8400 (get_symbol_alignment): New function.
8401 (legitimate_small_data_address_p): Likewise.
8402 (legitimate_scaled_address): Update, call
8403 legitimate_small_data_address_p.
8404 (output_sdata): New static variable.
8405 (arc_print_operand): Update how we handle small data operands.
8406 (arc_print_operand_address): Likewise.
8407 (arc_legitimate_address_p): Update, use
8408 legitimate_small_data_address_p.
8409 (arc_rewrite_small_data_p): Remove.
8410 (arc_rewrite_small_data_1): Likewise.
8411 (arc_rewrite_small_data): Likewise.
8412 (small_data_pattern): Likewise.
8413 (compact_sda_memory_operand): Update to use
8414 legitimate_small_data_address_p and get_symbol_alignment.
8415 (prepare_move_operands): Don't rewite sdata pattern.
8416 (prepare_extend_operands): Remove.
8417 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8419 (zero_extendqisi2): Likewise.
8420 (zero_extendhisi2): Likewise.
8421 (extendqihi2): Likewise.
8422 (extendqisi2): Likewise.
8423 (extendhisi2): Likewise.
8427 * config/arc/constraints.md (Usd): Change it to memory constraint.
8429 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8431 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8432 as source of std instructions.
8433 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8434 allow 6-bit constants as source for store instructions.
8435 (movdi_insn): Update instruction pattern to allow 6-bit constants
8436 as source for store instructions.
8438 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8440 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8442 2018-04-30 Nathan Sidwell <nathan@acm.org>
8443 Sandra Loosemore <sandra@codesourcery.com>
8445 * dumpfile.c (dump_open): Allow '-' for stdout.
8446 * doc/invoke.texi (Developer Options): Document dump filename
8447 determination early. Document stdin/stdout selection.
8449 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8451 Microblaze Target: PIC data text relative
8453 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8454 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8456 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8457 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8458 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8459 New addressing mode for data-text relative position indepenedent code.
8460 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8461 'ADDRESS_SYMBOLIC_TXT_REL'.
8462 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8463 (microblaze_legitimate_pic_operand): Exclude function calls from
8464 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8465 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8467 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8468 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8469 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8470 for 'address + offset'.
8471 (microblaze_expand_prologue): Add new function prologue call for
8473 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8474 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8475 table in case of TARGET_PIC_DATA_TEXT_REL.
8476 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8477 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8478 Add new macros 'UNSPEC_TEXT',
8479 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8480 + exclude function calls from 'UNSPEC_PLT' in case of data text
8482 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8483 new target hook for generating address diff vector tables in case of
8485 * doc/tm.texi : Regenerate.
8486 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8487 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8488 of addr diff vector generation.
8489 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8490 target hook definition.
8491 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8492 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8493 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8494 option for data text relative.
8496 2018-04-30 Richard Biener <rguenther@suse.de>
8498 * tree-chrec.h (evolution_function_is_constant_p): Remove
8500 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8502 2018-04-30 Richard Biener <rguenther@suse.de>
8505 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8507 2018-04-30 Richard Biener <rguenther@suse.de>
8509 PR tree-optimization/28364
8510 PR tree-optimization/85275
8511 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8512 copying first exit test.
8514 2018-04-28 Mark Wielaard <mark@klomp.org>
8516 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8518 (dwarf_AT): Handle DW_AT_addr_base.
8519 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8521 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8524 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8525 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8526 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8527 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8529 2018-04-28 Richard Biener <rguenther@suse.de>
8531 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8532 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8533 to reflect use. Only add interesting stmts.
8535 2018-04-27 Martin Jambor <mjambor@suse.cz>
8538 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8539 the jump function allows for passing through aggregate values.
8541 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8543 * input.h (in_system_header_at): Convert from macro to inline
8545 (from_macro_expansion_at): Likewise.
8546 (from_macro_definition_at): Likewise.
8548 2018-04-27 Jeff Law <law@redhat.com>
8550 * config.gcc: Mark tile* targets as deprecated/obsolete.
8552 2018-04-27 Richard Biener <rguenther@suse.de>
8554 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8557 2018-04-27 Richard Biener <rguenther@suse.de>
8559 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8561 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8563 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8564 with Yd constraint. Set "preferred_for_speed" attribute from
8565 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8567 (*movdi_internal): Ditto.
8568 (movti_interunit splitters): Remove
8569 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8570 (movdi_interunit splitters): Ditto.
8571 * config/i386/constraints.md (Ye): Remove.
8572 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8574 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8577 * config/aarch64/constraints.md (Usg): Limit to 31.
8580 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8582 PR tree-optimization/85529
8583 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8584 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8585 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8586 zero extension or masking of the MSB bit.
8587 (optimize_range_tests): Add FIRST_BB argument, pass it through
8588 to optimize_range_tests_var_bound.
8589 (maybe_optimize_range_tests, reassociate_bb): Adjust
8590 optimize_range_tests callers.
8592 2018-04-26 Richard Biener <rguenther@suse.de>
8593 Jakub Jelinek <jakub@redhat.com>
8595 * cgraph.h (symbol_table): Just declare debug method here.
8596 * symtab.c (symbol_table::debug): Define.
8598 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8600 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8602 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8604 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8605 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8606 (*movdi_internal): Substitute Yi and Yj constraint with x
8607 and Ym and Yn constraint with y constraint. Update "isa"
8608 attribute and set "preferred_for_speed" attribute from
8609 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8610 (*movsi_internal): Ditto.
8611 (*movdf_internal): Ditto.
8612 (*movsf_internal): Ditto.
8613 (*zero_extendsidi2): Ditto.
8614 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8615 (sse2_loadld): Ditto.
8616 (*vec_extract<ssevecmodelower>_0): Ditto.
8617 (*vec_extractv4si_0_zext_sse4): Ditto.
8618 (vec_concatv2di): Ditto.
8619 (*vec_dup<mode>): Ditto.
8620 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8621 * config/i386/constraints.md (Yi): Remove.
8626 2018-04-26 Nathan Sidwell <nathan@acm.org>
8628 * dumpfile.c (dump_open): New.
8629 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8630 (dump_finish): Detect stdio/stderr by value not name.
8632 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8634 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8636 2018-04-26 Tom de Vries <tom@codesourcery.com>
8639 * config/nvptx/nvptx.c (verify_neutering_jumps)
8640 (verify_neutering_labels): New function
8641 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8643 2018-04-26 Tom de Vries <tom@codesourcery.com>
8646 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8647 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8650 2018-04-26 Richard Biener <rguenther@suse.de>
8651 Tom de Vries <tom@codesourcery.com>
8654 * lto-streamer-out.c (output_function): Fixup loops if required to match
8655 discovery done in the reader.
8657 2018-04-26 Richard Biener <rguenther@suse.de>
8659 PR tree-optimization/85116
8660 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8661 have a loop exit from the single latch predecessor. Remove
8662 case of header with just condition.
8663 (ch_base::copy_headers): Exclude infinite loops from any
8665 (pass_ch::execute): Record exits.
8667 2018-04-26 Richard Biener <rguenther@suse.de>
8669 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8670 prologue cost vector and pass it to vect_get_load_cost.
8671 (vect_get_peeling_costs_all_drs): Likewise.
8672 (vect_peeling_hash_get_lowest_cost): Likewise.
8673 (vect_enhance_data_refs_alignment): Likewise.
8675 2018-04-26 Richard Biener <rguenther@suse.de>
8678 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8679 checking of integer<->pointer conversions.
8680 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8681 sign-/zero-extending pointer types.
8682 (expand_omp_for_static_chunk): Likewise.
8684 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8685 Jean Lee <xiaoyur347@gmail.com>
8687 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8688 (TARGET_ASAN_SHADOW_OFFSET): Define.
8689 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8690 true for -fsanitize=address.
8692 2018-04-25 Mark Wielaard <mark@klomp.org>
8694 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8697 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8699 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8700 than "alu", remove explicit "memory" and "imm_disp" attributes.
8701 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8704 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8705 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8706 gen_lowpart_no_emit.
8708 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8711 * config/i386/i386.c (ix86_expand_builtin): Change memory
8712 operand to XI, extend p0 to Pmode.
8713 * config/i386/i386.md: Change unspec volatile and operand
8714 1 mode to XI, change operand 0 mode to P.
8716 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8718 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8719 GET_MODE_MASK before any checking.
8720 (nds32_can_use_bset_p): Likewise.
8721 (nds32_can_use_btgl_p): Likewise.
8723 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8725 * config/nds32/nds32-doubleword.md: New define_split pattern for
8726 illegal register number.
8728 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8730 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8732 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8734 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8736 2018-04-25 Richard Biener <rguenther@suse.de>
8738 * lto-streamer.h (LTO_major_version): Bump to 8.
8740 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8742 * BASE-VER: Set to 9.0.0.
8744 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8746 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8747 in __abskf2 and __powikf2.
8749 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8752 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8753 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8754 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8755 Use the above on operand 2. Reindent.
8756 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8758 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8761 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8763 * config/i386/i386.opt (mcet): Removed.
8764 * doc/install.texi: Remove -mcet documentation.
8765 * doc/invoke.texi: Likewise.
8767 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8770 * doc/install.texi: Remove -mcet from bootstrap-cet.
8772 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8775 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8776 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8780 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8781 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8782 containing a CONST_VECTOR.
8784 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8786 * doc/install.texi: Update newlib dependency for nvptx.
8788 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8791 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8792 instead of INTVAL when shifting x left.
8794 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8796 PR tree-optimization/85478
8797 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8798 vect_grouped_store_supported for single element vectors.
8800 2018-04-24 Richard Biener <rguenther@suse.de>
8803 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8804 load cost increase to the case of non-constant step.
8806 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8809 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8810 destination if any_malformed_asm.
8812 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8815 * expr.c (store_field): In the bitfield case, if the value comes from
8816 a function call and is returned in registers by means of a PARALLEL,
8817 do not change the mode of the temporary unless BLKmode and VOIDmode.
8819 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8821 PR rtl-optimization/85423
8822 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8823 dependencies to debug insns when the previous insn is non-debug.
8825 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8827 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8828 enums into a single definition.
8829 (fls): Fix predicates and printing.
8832 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8834 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8835 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8836 and short u6 immediate.
8837 (check_if_valid_sleep_operand): Remove.
8838 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8840 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8842 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8843 flag_always_save_lp condition.
8844 * config/nds32/nds32.opt (malways-save-lp): New option.
8846 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8848 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8849 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8850 * config/nds32/nds32.h
8851 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8852 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8854 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8856 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8857 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8859 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8860 Chung-Ju Wu <jasonwucj@gmail.com>
8862 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8864 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8865 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8867 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8869 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8871 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8873 * config/nds32/nds32-protos.h (nds32_data_alignment,
8874 nds32_local_alignment): Declare.
8875 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8876 nds32_local_alignment): New functions.
8877 (TARGET_CONSTANT_ALIGNMENT): Define.
8878 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8880 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8882 * config/nds32/nds32.c
8883 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8884 (TARGET_MODES_TIEABLE_P): Likewise.
8886 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8888 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8891 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8892 Chung-Ju Wu <jasonwucj@gmail.com>
8894 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8895 for unaligned access.
8896 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8898 * config/nds32/nds32-intrinsic.md: Likewise.
8899 * config/nds32/nds32_intrinsic.h: Likewise.
8900 * config/nds32/nds32.h (nds32_builtins): Likewise.
8901 * config/nds32/nds32.opt (munaligned-access): New option.
8902 * config/nds32/nds32.c (nds32_asm_file_start): Display
8903 flag_unaligned_access status.
8905 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8907 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8908 -mno-relax is present.
8909 * config/riscv/linux.h (LINK_SPEC): Ditto.
8911 2018-04-20 Martin Sebor <msebor@redhat.com>
8914 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8916 (gimple_fold_builtin_stxcpy_chk): Same.
8917 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8919 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8922 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8923 __powikf2 when long double is IEEE 128-bit.
8925 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8927 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8928 step to make sure stack always aligned.
8930 2018-04-20 Carl Love <cel@us.ibm.com>
8933 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8934 size check for arg0.
8936 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8937 Tom de Vries <tom@codesourcery.com>
8940 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8941 Emit insns for calls too.
8942 (nvptx_find_par): Always look for worker-level predecessor insn.
8943 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8945 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8946 (nvptx_process_pars): Propagate frames for calls.
8948 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8951 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8953 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8954 (ix86_handle_option): Don't handle OPT_mibt.
8955 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8957 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8959 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8960 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8961 (ix86_target_macros): Define __CET__ with flag_cf_protection
8962 for -fcf-protection.
8963 * config/i386/i386.c (isa2_opts): Remove -mibt.
8964 * config/i386/i386.h (TARGET_IBT): Removed.
8965 (TARGET_IBT_P): Likewise.
8966 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8967 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8968 * config/i386/i386.opt (mcet): Update help message.
8971 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8972 -mcet as an alias for -mshstk.
8974 2018-04-20 Richard Biener <rguenther@suse.de>
8977 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8978 complexity by forcing a single use of the multiply operand.
8980 2018-04-20 Martin Jambor <mjambor@suse.cz>
8983 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8984 recursion dependency to only apply to non-clones.
8986 2018-04-20 Martin Jambor <mjambor@suse.cz>
8989 * ipa-cp.c (create_specialized_node): Check that clones of
8990 self-recursive edges exist during IPA-CP.
8992 2018-04-19 Toon Moene <toon@moene.org>
8994 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8997 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8999 PR tree-optimization/85467
9000 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
9001 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
9002 VECTOR_CST element to type.
9004 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9007 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
9008 * config/i386/i386.md (builtin_setjmp_setup): Removed.
9009 (builtin_longjmp): Likewise.
9010 (save_stack_nonlocal): New pattern.
9011 (restore_stack_nonlocal): Likewise.
9013 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9016 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9017 Replace ASM_OUTPUT_LABEL with fprintf.
9019 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9022 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9023 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9024 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9025 define __IBT__ and __SHSTK__ for -fcf-protection.
9026 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9028 (ix86_trampoline_init): Likewise.
9029 (x86_output_mi_thunk): Likewise.
9030 (ix86_notrack_prefixed_insn_p): Likewise.
9031 (ix86_option_override_internal): Don't disallow -fcf-protection.
9032 * config/i386/i386.md (rdssp<mode>): Also enable for
9034 (incssp<mode>): Likewise.
9035 (nop_endbr): Likewise.
9036 * config/i386/i386.opt (mcet): Change help message to built-in
9040 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9041 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9042 enable CET built-in functions.
9044 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9046 * common/config/i386/i386-common.c
9047 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9048 OPTION_MASK_ISA_MOVDIRI_UNSET,
9049 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9050 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9051 * config.gcc (movdirintrin.h): New header.
9052 * config/i386/cpuid.h (bit_MOVDIRI,
9053 bit_MOVDIR64B): New bits.
9054 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9056 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9057 (VOID, PVOID, PCVOID)): New function types.
9058 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9059 __builtin_ia32_directstoreu_u64,
9060 __builtin_ia32_movdir64b): New builtins.
9061 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9062 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9064 (ix86_valid_target_attribute_inner_p): Ditto.
9065 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9066 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9067 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9068 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9069 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9070 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9071 (movdiri<mode>, movdir64b_<mode>): New.
9072 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9073 * config/i386/immintrin.h: Include movdirintrin.h.
9074 * config/i386/movdirintrin.h: New file.
9075 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9077 2018-04-19 Richard Biener <rguenther@suse.de>
9080 * cfg.c (clear_bb_flags): When loop state says we have
9081 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9083 2018-04-19 Richard Biener <rguenther@suse.de>
9085 PR tree-optimization/84737
9086 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9087 copying restrict info.
9088 (vect_setup_realignment): Use it.
9089 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9090 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9091 the first DR to all generated stores.
9092 (vectorizable_load): Likewise for loads.
9094 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9096 PR tree-optimization/85446
9097 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9098 the integral and pointer types to have the same precision.
9100 * doc/install.texi: Document --disable-cet being the default and
9103 2018-04-18 Martin Liska <mliska@suse.cz>
9105 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9108 2018-04-18 Martin Liska <mliska@suse.cz>
9111 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9114 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9115 arguments if they are comparable.
9117 2018-04-18 Martin Liska <mliska@suse.cz>
9120 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9123 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9126 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9129 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9130 ENDBR after calling __morestack.
9132 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9135 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9136 by using gcc_base_ver to generate a gcc_driver_version, and use
9137 it when generating GCC_DRIVER_NAME.
9138 * configure: Regenerate.
9140 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9143 * config.gcc: Obsolete powerpc*-*-*spe*.
9145 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9148 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9149 (stabstr_D): Change type of unum from unsigned int to
9150 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9153 2018-04-17 Jim Wilson <jimw@sifive.com>
9156 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9157 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9158 Set arg_pointer_offset after using pretend_args_size.
9160 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9162 PR rtl-optimization/85431
9163 * dse.c (record_store): Ignore zero width stores.
9166 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9167 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9168 __builtin_stack_restore rather than after it.
9169 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9170 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9171 argument instead of virtual_dynamic_stack_rtx.
9173 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9175 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9177 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9178 Add note to error message to explain internal mapping of overloaded
9179 built-in function name to non-overloaded built-in function name.
9180 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9183 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9186 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9187 where the inputs overlap with the output.
9189 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9191 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9192 (=v, v) alternative and explicit "memory" attribute.
9193 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9194 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9196 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9197 "sselog1" type instead of "sselog".
9198 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9199 "sselog". Remove explicit "memory" attribute.
9200 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9201 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9203 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9204 "sselog1" type instead of "sselog". Remove explicit "memory"
9206 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9207 use "sselog1" type instead of "sselog". Remove explicit "memory"
9209 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9210 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9212 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9213 "sselog1" type instead of "sselog". Remove explicit "memory"
9215 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9216 use "sselog1" type instead of "sselog". Remove explicit "memory"
9220 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9223 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9226 2018-04-17 Martin Jambor <mjambor@suse.cz>
9229 * ipa-cp.c (create_specialized_node): Call
9230 expand_all_artificial_thunks if necessary.
9232 2018-04-17 Martin Liska <mliska@suse.cz>
9235 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9236 in message, remote space in between '_G' and '('.
9238 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9241 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9242 avx512f_vmcmp<mode>3<round_saeonly_name>,
9243 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9244 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9245 avx512f_rndscale<mode><round_saeonly_name>,
9246 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9247 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9248 Use %<iptr>2 instead of %2 for -masm=intel.
9249 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9250 avx512f_vcvttss2usi<round_saeonly_name>,
9251 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9253 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9254 avx512f_vcvttsd2usi<round_saeonly_name>,
9255 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9256 Use %q1 instead of %1 for -masm=intel.
9257 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9258 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9259 of %3 for -masm=intel.
9260 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9262 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9264 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9266 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9267 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9269 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9271 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9272 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9273 %g1 and one with %0 and %1.
9274 (avx512er_vmrcp28<mode><round_saeonly_name>,
9275 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9277 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9278 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9279 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9280 of %0 and %{%4%} for -masm=intel.
9281 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9282 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9283 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9284 order of %0 and %{%5%}%{z%} for -masm=intel.
9286 2018-04-17 Jan Hubicka <jh@suse.cz>
9289 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9291 2018-04-17 Martin Liska <mliska@suse.cz>
9294 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9295 for target_clone error message. Make default implementation
9296 clone to be a local declaration.
9297 (separate_attrs): Add new argument and check for an empty
9299 (expand_target_clones): Handle it.
9300 (ipa_target_clone): Make redirection just for target_clones
9303 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9304 Tom de Vries <tom@codesourcery.com>
9307 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9308 tiled basic blocks without omp continue statements.
9310 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9313 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9314 vec_extract expression as having side effects to make sure it gets
9317 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9320 * config/i386/i386.c (get_builtin_code_for_version): Check
9323 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9326 * gcc/config.gcc: Support "skylake".
9327 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9329 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9330 (processor_target_table): Add "skylake".
9331 (ix86_option_override_internal): Add "skylake".
9332 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9333 PROCESSOR_CANNONLAKE.
9334 (get_builtin_code_for_version): Fix priority for
9335 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9336 PROCESSOR_SKYLAKE-AVX512.
9337 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9338 (processor_type): Add PROCESSOR_SKYLAKE.
9340 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9341 Jason Merrill <jason@redhat.com>
9344 * convert.c (convert_to_integer_1): Use direct recursion for
9345 enumeral types and types with a precision less than the number
9346 of bits in their mode.
9348 2018-04-16 Julia Koval <julia.koval@intel.com>
9351 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9352 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9354 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9357 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9358 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9359 and -mno-direct-move.
9361 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9364 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9365 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9366 Ensure negative shifts result in {0}.
9368 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9370 PR rtl-optimization/79916
9371 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9372 regs (if any) to define how to gnerate SD moves when LRA is in
9375 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9377 PR rtl-optimization/85393
9378 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9379 * except.c (expand_dw2_landing_pad_for_region): Make static.
9380 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9381 a label and unconditional jump to old_bb, rather than
9382 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9385 PR rtl-optimization/85376
9386 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9387 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9388 instead of a specific value.
9390 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9391 Bin Cheng <bin.cheng@arm.com>
9393 PR tree-optimization/82965
9394 PR tree-optimization/83991
9395 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9396 by_profile_only parameter.
9397 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9398 information if the loop was predicted to iterate too many times.
9399 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9401 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9404 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9407 2018-04-13 Martin Liska <mliska@suse.cz>
9408 Jakub Jelinek <jakub@redhat.com>
9411 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9412 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9413 * builtins.c (expand_builtin_memory_copy_args): Use
9414 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9415 handle dest_addr == pc_rtx.
9417 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9420 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9421 asked to not generate direct moves.
9422 (fix_trunc<mode>si2_stfiwx): Similar.
9423 (fix_trunc<mode>si2_internal): Similar.
9425 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9428 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9429 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9430 lookup if dest in some wider mode is known to be const0_rtx and
9431 if so, record permanent equivalence for it to be ZERO_EXTEND of
9432 the narrower mode destination.
9434 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9436 * lto-streamer-out.c (output_function): Revert 259346.
9437 * omp-expand.c (expand_oacc_for): Likewise.
9439 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9441 PR rtl-optimization/85354
9442 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9443 * sel-sched.c (sel_global_init): ... here.
9445 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9448 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9449 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9450 mode for PE-COFF targets.
9451 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9452 (i386_pe_asm_lto_end): Likewise.
9453 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9454 (TARGET_ASM_LTO_END): Likewise.
9455 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9456 (i386_pe_asm_lto_start): New function.
9457 (i386_pe_asm_lto_end): Likewise.
9459 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9460 Richard Biener <rguenther@suse.de>
9463 * lto-streamer-out.c (output_function): Fix CFG loop state before
9465 * omp-expand.c (expand_oacc_for): Handle calls to internal
9466 functions like regular functions.
9468 2018-04-12 Richard Biener <rguenther@suse.de>
9471 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9472 for the early LTO debug to properly generate references to it
9473 during DIE emission. Do not re-use that for the skeleton for
9475 (dwarf2out_early_finish): Likewise.
9477 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9480 * config/i386/sse.md
9481 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9482 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9483 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9484 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9485 and output is a reg, avoid creating invalid lowpart subreg, but
9486 instead split into a 512-bit move. Don't split if not AVX512VL,
9487 input is xmm16+ reg and output is a mem.
9488 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9489 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9490 xmm16+ reg and output is a mem.
9492 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9494 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9495 also for flag_dwarf2_cfi_asm.
9497 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9499 PR rtl-optimization/85342
9500 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9501 a bool scalar var inside of the loop instead. Don't try to update
9502 recog_data.operand after failed apply_change_group.
9504 2018-04-12 Tom de Vries <tom@codesourcery.com>
9507 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9508 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9509 array with flexible array member as array without given dimension.
9510 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9511 argument for undefined param to true.
9513 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9516 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9517 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9518 from PowerPC section.
9519 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9520 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9521 help text that is too long.
9522 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9523 help text that is too long.
9524 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9525 help text that is too long.
9527 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9529 * config/alpha/alpha.md (stack_probe_internal): Rename
9530 from "probe_stack". Update all callers.
9532 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9534 PR rtl-optimization/84566
9535 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9536 sched_macro_fuse_insns.
9538 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9541 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9542 (compute_block_dependences): ... from here.
9544 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9546 PR tree-optimization/85331
9547 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9548 from int to HOST_WIDE_INT.
9550 2018-04-11 Martin Jambor <mjambor@suse.cz>
9553 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9554 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9555 not the same as the source val.
9556 (cgraph_edge_brings_value_p): New parameter.
9557 (gather_edges_for_value): Pass destination value to
9558 cgraph_edge_brings_value_p.
9559 (perhaps_add_new_callers): Likewise.
9560 (get_info_about_necessary_edges): Likewise and exclude values brought
9561 only by self-recursive edges.
9562 (create_specialized_node): Redirect only clones of self-calling edges.
9563 (+self_recursive_pass_through_p): New function.
9564 (find_more_scalar_values_for_callers_subset): Use it.
9565 (find_aggregate_values_for_callers_subset): Likewise.
9566 (known_aggs_to_agg_replacement_list): Removed.
9567 (decide_whether_version_node): Re-calculate known constants for all
9568 remaining context clones.
9570 2018-04-11 Richard Biener <rguenther@suse.de>
9573 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9574 from early DWARF output.
9575 (dwarf2out_early_finish): Output line info unconditionally into
9576 early DWARF and add reference to it.
9578 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9581 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9582 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9583 other than V2DFmode using iptr mode attribute.
9584 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9586 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9588 PR rtl-optimization/84659
9589 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9591 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9594 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9596 (output_loc_list): Pass address of a dummy size variable even in the
9597 locview handling loop.
9598 (index_location_lists): Add comment on why skip_loc_list_entry can't
9601 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9604 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9607 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9610 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9611 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9612 and -mstring-compare-inline-limit.
9614 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9617 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9618 for stack clash protection in a register whenever we need it to be in
9621 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9623 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9624 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9626 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9629 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9631 (mlong-double-): Ditto.
9632 * config/rs6000/sysv4.opt (msdata=): Ditto.
9633 (mtls-size=): Ditto.
9635 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9637 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9638 erroneous entries for
9639 "vector int vec_ldl (int, long int *)", and
9640 "vector unsigned int vec_ldl (int, unsigned long int *)".
9641 Add comments and entries for
9642 "vector bool char vec_ldl (int, bool char *)",
9643 "vector bool short vec_ldl (int, bool short *)",
9644 "vector bool int vec_ldl (int, bool int *)",
9645 "vector bool long long vec_ldl (int, bool long long *)",
9646 "vector pixel vec_ldl (int, pixel *)",
9647 "vector long long vec_ldl (int, long long *)",
9648 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9649 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9650 type tree bool_long_long_type_node and correct definition of
9651 bool_V2DI_type_node to make reference to this new type tree.
9652 (rs6000_mangle_type): Replace erroneous reference to
9653 bool_long_type_node with bool_long_long_type_node.
9654 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9655 comments to emphasize sign distinctions for char and int types and
9656 replace RS6000_BTI_bool_long constant with
9657 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9658 use of RS6000_BTI_pixel.
9659 (bool_long_type_node): Remove this macro definition.
9660 (bool_long_long_type_node): New macro definition
9662 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9664 PR rtl-optimization/85300
9665 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9666 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9667 simplify_unary_operation fails.
9669 2018-04-10 Martin Liska <mliska@suse.cz>
9671 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9672 cgraph_edge and ipa_ref.
9674 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9678 * config/i386/sse.md
9679 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9680 computation of the VEC_MERGE selector from mask.
9681 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9682 Fix decoding of the VEC_MERGE selector into mask.
9684 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9686 PR tree-optimization/85286
9687 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9689 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9691 * final.c (final_1): Set insn_last_address as well as
9692 insn_current_address.
9694 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9697 * explow.c (emit_stack_probe): Call validize_mem on memory location
9698 before passing it to gen_probe_stack. Create address operand and
9699 legitimize it for the probe_stack_address case.
9701 2018-04-09 Jan Hubicka <jh@suse.cz>
9704 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9705 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9706 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9708 (free_lang_data): Rebuild type inheritance graph.
9710 2018-04-09 Martin Sebor <msebor@redhat.com>
9712 * invoke.texi (-finline-small-functions): Mention other optimization
9714 (-findirect-inlining, -fpartial-inlining): Same.
9715 (-finline-functions-called-once): Same.
9716 (-freorder-blocks-and-partition): Same.
9718 2018-04-09 Jan Hubicka <jh@suse.cz>
9721 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9722 jumps; choose last target that matches the criteria (i.e.
9723 no partition changes for non-crossing jumps).
9724 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9725 support for redirecting crossing jumps to non-crossing.
9727 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9729 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9730 also for naked functions.
9732 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9734 * config/arc/arc.md (add_shift): New pattern.
9735 (add_shift2): Likewise.
9736 (sub_shift): Likewise.
9737 (sub_shift_cmp0_noout): Likewise.
9738 (compare_si_ashiftsi): Likewise.
9739 (xbfu_cmp0_noout): New combine pattern.
9740 (xbfu_cmp0"): Likewise.
9741 (movsi_set_cc_insn): Place the predicable variant first.
9742 (commutative_binary_cmp0_noout): Remove clobber.
9743 (commutative_binary_cmp0): New pattern.
9744 (noncommutative_binary_cmp0): Likewise.
9745 (noncommutative_binary_cmp0_noout): Likewise.
9746 (noncommutative_binary_comparison_result_used): Removed.
9747 (rsub_cmp0): New pattern.
9748 (rsub_cmp0_noout): Likewise.
9749 (extzvsi): Changed, keep only meaningful variants.
9750 (SQH, SEZ): New iterators.
9751 (SQH_postfix): New mode attribute.
9752 (SEZ_prefix): New code attribute.
9753 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9754 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9755 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9757 (noncommutative_operator): Check the availability of barrel
9760 2018-04-09 Richard Biener <rguenther@suse.de>
9762 PR tree-optimization/85284
9763 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9764 Only use the niter constraining form of simple_iv when the exit
9767 2018-04-09 Tom de Vries <tom@codesourcery.com>
9770 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9771 (define_expand "*memory_barrier"): New define_expand.
9772 (define_insn "memory_barrier"): New insn.
9774 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9776 PR rtl-optimization/80463
9777 PR rtl-optimization/83972
9778 PR rtl-optimization/83480
9780 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9781 correct producer for the insn.
9782 (tidy_control_flow): Fixup seqnos in case of debug insns.
9784 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9786 PR rtl-optimization/83913
9788 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9789 different sched-times when merging exprs.
9791 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9793 PR rtl-optimization/83962
9795 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9796 tidy_fallthru_edge and tidy_control_flow.
9798 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9800 PR rtl-optimization/83530
9802 * sel-sched.c (force_next_insn): New global variable.
9803 (remove_insn_for_debug): When force_next_insn is true, also leave only
9804 next insn in the ready list.
9805 (sel_sched_region): When the region wasn't scheduled, make another pass
9806 over it with force_next_insn set to 1.
9808 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9810 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9812 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9813 for interrupt control.
9814 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9815 functions for interrupt control.
9816 * config/nds32/nds32-intrinsic.md: Likewise.
9817 * config/nds32/nds32_intrinsic.h: Likewise.
9818 * config/nds32/nds32.h (nds32_builtins): Likewise.
9820 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9822 * config/nds32/nds32.c (nds32_init_machine_status,
9823 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9824 strict_aligned_p field.
9825 (nds32_expand_to_rtl_hook): New function.
9826 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9827 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9829 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9830 Chung-Ju Wu <jasonwucj@gmail.com>
9832 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9833 * config/nds32/nds32-n7.md: New file.
9834 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9835 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9837 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9838 * config/nds32/nds32.md (pipeline_model): Add n7.
9839 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9840 * config/nds32/pipelines.md: Include n7 settings.
9842 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9843 Chung-Ju Wu <jasonwucj@gmail.com>
9845 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9846 * config/nds32/nds32-e8.md: New file.
9847 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9848 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9850 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9851 * config/nds32/nds32.md (pipeline_model): Add e8.
9852 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9853 * config/nds32/pipelines.md: Include e8 settings.
9855 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9856 Chung-Ju Wu <jasonwucj@gmail.com>
9858 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9859 * config/nds32/nds32-n8.md: New file.
9860 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9861 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9863 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9864 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9865 * config/nds32/nds32.md (pipeline_model): Add n8.
9866 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9867 * config/nds32/pipelines.md: Include n8 settings.
9869 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9870 Chung-Ju Wu <jasonwucj@gmail.com>
9872 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9873 * config/nds32/nds32-n9-2r1w.md: New file.
9874 * config/nds32/nds32-n9-3r2w.md: New file.
9875 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9876 nds32_register_ports): New or modify for cpu n9.
9877 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9879 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9880 * config/nds32/nds32-utils.c: New file.
9881 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9882 TARGET_MUL_SLOW): Define.
9883 * config/nds32/nds32.md (pipeline_model): New attribute.
9884 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9885 New options that support cpu n9.
9886 * config/nds32/pipelines.md: Include n9 settings.
9887 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9889 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9891 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9892 information if necessary.
9893 (output_cond_branch_compare_zero): Likewise.
9894 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9895 (nds32_target_alignment): Refine for alignment.
9896 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9897 (FUNCTION_BOUNDARY): Modify.
9898 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9900 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9902 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9904 * config/nds32/constants.md (unspec_volatile_element): Add values for
9905 TLB operation and data prefetch.
9906 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9907 functions for TLB operation and data prefetch.
9908 * config/nds32/nds32-intrinsic.md: Likewise.
9909 * config/nds32/nds32_intrinsic.h: Likewise.
9910 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9911 (nds32_print_operand): Likewise.
9912 * config/nds32/nds32.h (nds32_builtins): Likewise.
9914 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9915 Andrew Pinski <pinsika@gcc.gnu.org>
9918 * match.pd: Use constant_boolean_node of correct type instead of
9919 boolean_true_node or boolean_false_node for simplifying
9920 pointer comparisons to zero.
9922 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9924 PR tree-optimization/80021
9925 * tree.c (verify_type_variant): Make error call in verify_variant_match
9926 translatable and remove final full stop.
9928 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9930 * config/nds32/constants.md (unspec_volatile_element): Add
9931 UNSPEC_VOLATILE_EH_RETURN.
9932 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9933 nds32_output_stack_pop): Support dwarf exception handling process.
9934 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9935 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9936 exception handling process.
9937 (nds32_compute_stack_frame): Likewise.
9938 (nds32_return_addr_rtx): Likewise.
9939 (nds32_initial_elimination_offset): Likewise.
9940 (nds32_expand_prologue): Likewise.
9941 (nds32_expand_epilogue): Likewise.
9942 (nds32_dynamic_chain_address): New function.
9943 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9945 (DYNAMIC_CHAIN_ADDRESS): Define.
9946 (EH_RETURN_DATA_REGNO): Define.
9947 (EH_RETURN_STACKADJ_RTX): Define.
9948 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9949 patterns for dwarf exception handling.
9951 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9953 * config/nds32/nds32.h: Clean up obsolete macros.
9955 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9957 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9958 Add enum values for particular instructions.
9959 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9960 particular intrinsic functions.
9961 * config/nds32/nds32-intrinsic.md: Likewise.
9962 * config/nds32/nds32_intrinsic.h: Likewise.
9963 * config/nds32/nds32.h (nds32_builtins): Likewise.
9964 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9965 (btst, ave): New patterns for particular instructions.
9967 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9969 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9970 Add enum values for atomic load/store and memory sync.
9971 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9973 * config/nds32/nds32-intrinsic.md: Likewise.
9974 * config/nds32/nds32_intrinsic.h: Likewise.
9975 * config/nds32/nds32.h (nds32_builtins): Likewise.
9977 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9979 PR tree-optimization/85257
9980 * fold-const.c (native_encode_vector): If not all elts could fit
9981 and off is -1, return 0 rather than offset.
9982 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9983 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9984 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9985 adjust buffer in native_interpret_expr call.
9987 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9989 * config/nds32/constants.md (unspec_volatile_element): Add cache
9990 control enum values.
9991 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9992 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9993 * config/nds32/nds32.c (nds32_cctl_names): New.
9994 (nds32_print_operand): Handle cache control register names.
9995 * config/nds32/nds32.h (nds32_builtins): New enum values.
9996 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9998 * config/nds32/nds32.md (type): Add mmu.
9999 * config/nds32/pipelines.md (simple_insn): Add mmu.
10001 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10003 * config/nds32/nds32.md (type): Remove call.
10004 * config/nds32/pipelines.md (simple_insn): Likewise.
10006 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10008 * config/nds32/constants.md (unspec_volatile_element): Add
10009 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
10010 UNSPEC_VOLATILE_FMFCFG.
10011 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
10012 description for fmfcfg and fmfcsr.
10013 (bdesc_1arg): Add fmtcsr.
10014 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
10015 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
10016 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
10017 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
10018 unspec_fmfcfg): New patterns.
10019 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10020 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10021 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10022 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10023 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10024 __nds32__fmfcfg): Define.
10026 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10028 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10029 intrinsic register names.
10030 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10031 intrinsic register enum values and macros.
10033 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10035 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10036 for load/store addressing form.
10037 (nds32_print_operand_address): Likewise.
10039 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10042 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10043 based on LABEL_REF. Remove useless assertion.
10044 (pic_address_needs_scratch): Fix formatting.
10045 (sparc_legitimize_pic_address): Minor tweaks.
10046 (sparc_delegitimize_address): Adjust assertion accordingly.
10047 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10048 into symbolic_operand.
10049 (movsi_high_pic_label_ref): Likewise.
10050 (movsi_lo_sum_pic_label_ref): Likewise.
10051 (movdi_pic_label_ref): Likewise.
10052 (movdi_high_pic_label_ref): Likewise.
10053 (movdi_lo_sum_pic_label_ref): Likewise.
10055 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10057 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10058 custom LIB_SPEC setup.
10060 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10061 Kito Cheng <kito.cheng@gmail.com>
10063 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10064 * config/riscv/freebsd.h: New.
10066 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10068 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10069 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10072 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10073 Kito Cheng <kito.cheng@gmail.com>
10075 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10076 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10077 * config/nds32/nds32-protos.h (nds32_output_call,
10078 nds32_output_return): Declare.
10079 * config/nds32/nds32.md: Refine all the call and return patterns.
10081 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10084 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10085 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10087 PR rtl-optimization/84872
10088 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10089 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10090 EDGE_CROSSING edge.
10092 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10094 * expr.c (copy_blkmode_to_reg): Revert 254862.
10095 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10097 2018-04-06 Richard Biener <rguenther@suse.de>
10099 PR middle-end/85244
10100 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10101 after seeing a component reference with an adjacent field. Treat
10102 refs to arrays at struct end of external decls similar to
10103 refs to unconstrained commons.
10105 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10108 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10109 look through SAVE_EXPRs with non-side-effects argument. Adjust
10111 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10114 2018-04-06 Richard Biener <rguenther@suse.de>
10116 PR middle-end/85180
10117 * alias.c (find_base_term): New wrapper around find_base_term
10118 unwinding CSELIB_VAL_PTR changes.
10119 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10122 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10124 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10126 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10127 constant definitions.
10128 ("nop"): lr 0,0 -> nopr r0
10129 ("nop_lr0", "nop_lr1"): New insn definitions.
10131 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10133 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10134 NDS32_V3PUSH_AVAILABLE_P macro.
10136 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10137 Chung-Ju Wu <jasonwucj@gmail.com>
10139 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10140 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10141 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10142 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10143 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10144 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10145 * config/nds32/constraints.md: New constraints and checking for hard
10146 float configuration.
10147 * config/nds32/iterators.md: New mode iterator and attribute for hard
10148 float configuration.
10149 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10151 * config/nds32/nds32-fpu.md: New file.
10152 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10153 deal with hard float code generation.
10154 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10156 (abi_type, float_reg_number): New enum type.
10157 * config/nds32/nds32-predicates.c: New predicates for hard float.
10158 * config/nds32/nds32-protos.h: Declare functions for hard float.
10159 * config/nds32/nds32.c: Implementation for hard float configuration.
10160 * config/nds32/nds32.h: Definitions for hard float configuration.
10161 * config/nds32/nds32.md: Include hard float machine description and
10162 modify patterns for hard float configuration.
10163 * config/nds32/nds32.opt: New options for hard float configuration.
10164 * config/nds32/predicates.md: New predicates for hard float
10167 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10169 * common/config/nds32/nds32-common.c
10170 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10172 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10174 PR middle-end/85195
10175 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10176 CONSTRUCTOR_ELT (ctor, ...)->value.
10178 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10181 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10183 2018-04-05 Tom de Vries <tom@codesourcery.com>
10186 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10189 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10190 Kito Cheng <kito.cheng@gmail.com>
10192 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10193 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10194 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10195 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10197 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10198 Kito Cheng <kito.cheng@gmail.com>
10200 * config/nds32/constraints.md (Ufe): New memory constraint.
10201 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10202 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10203 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10205 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10206 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10208 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10210 * config/nds32/nds32.md: Use optimize_size in the condition for
10211 alu-shift instructions.
10213 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10215 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10217 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10219 * config/nds32/nds32.md (negsi2): Refine pattern.
10221 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10222 Chung-Ju Wu <jasonwucj@gmail.com>
10224 * config/nds32/iterators.md (shift_rotate): New code iterator.
10225 (shift): New code attribute.
10226 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10227 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10228 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10229 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10230 bit-wise operations.
10231 (andsi3, *andsi3): Ditto.
10232 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10233 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10234 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10235 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10236 nds32_ior_operand, nds32_xor_operand): New predicates.
10238 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10240 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10241 (addsi3, subsi3): ... this.
10243 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10245 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10247 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10249 * config/nds32/nds32.md: Adjust indention.
10251 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10253 * config/nds32/nds32.md (feature): New attribute.
10255 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10257 * config/nds32/nds32.md (subtype): New attribute.
10259 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10262 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10263 expansion to perform a bitwise AND of the argument followed by a
10264 boolean negation of the result.
10266 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10268 PR rtl-optimization/84878
10269 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10270 the basic block. Assert the use reference is not artificial and that
10271 it has an associated insn.
10273 2018-04-04 Michael Matz <matz@suse.de>
10275 * builtins.c (compute_objsize): Pass correct operand
10276 to array_at_struct_end_p.
10278 2018-04-04 Richard Biener <rguenther@suse.de>
10281 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10282 from contexts for DINFO_LEVEL_TERSE and below.
10284 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10286 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10287 resiter_operand condition.
10288 * config/nds32/nds32.md (*move<mode>): Ditto.
10290 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10291 Monk Chiang <sh.chiang04@gmail.com>
10293 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10295 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10297 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10299 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10300 Kito Cheng <kito.cheng@gmail.com>
10302 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10303 nds32_cond_code_str, output_cond_branch,
10304 output_cond_branch_compare_zero, nds32_expand_cbranch,
10305 nds32_expand_cstore, nds32_expand_movcc,
10306 nds32_output_cbranchsi4_equality_zero,
10307 nds32_output_cbranchsi4_equality_reg,
10308 nds32_output_cbranchsi4_equality_reg_or_const_int,
10309 nds32_output_cbranchsi4_greater_less_zero: New functions.
10310 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10311 nds32_expand_cstore, nds32_expand_movcc,
10312 nds32_output_cbranchsi4_equality_zero,
10313 nds32_output_cbranchsi4_equality_reg,
10314 nds32_output_cbranchsi4_equality_reg_or_const_int,
10315 nds32_output_cbranchsi4_greater_less_zero): Declare.
10316 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10317 nds32_rimm11s_operand): New predicates.
10318 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10319 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10322 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10324 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10325 * config/nds32/nds32.md: Ditto.
10326 * config/nds32/pipelines.md: Ditto.
10328 2018-04-04 Richard Biener <rguenther@suse.de>
10330 PR tree-optimization/85168
10331 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10332 propagating abnormals.
10334 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10336 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10338 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10339 Kito Cheng <kito.cheng@gmail.com>
10341 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10342 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10343 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10344 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10345 * config/nds32/nds32.md (sibcall_internal): New.
10346 (sibcall_register): Remove.
10347 (sibcall_immediate): Remove.
10348 (sibcall_value_internal): New.
10349 (sibcall_value_register): Remove.
10350 (sibcall_value_immediate): Remove.
10351 * config/nds32/predicates.md (nds32_general_register_operand): New.
10352 (nds32_call_address_operand): New.
10354 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10356 PR rtl-optimization/85167
10357 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10358 bb_defs if *split_p, instead preinitialize it to NULL.
10360 PR tree-optimization/85156
10361 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10362 evaluating the argument multiple times.
10364 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10366 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10368 (_mm_cvtpd_ps): Likewise.
10369 (_mm_cvttpd_epi32): Likewise.
10370 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10371 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10372 vector, pixel, and bool following altivec.h include.
10374 2018-04-03 Martin Sebor <msebor@redhat.com>
10376 * doc/extend.texi (Common Function Attributes): Clarify.
10377 (const attribute): Likewise.
10378 (pure attribute): Likewise.
10380 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10383 * config/i386/i386.c (ix86_expand_vector_set): Use
10384 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10386 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10388 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10389 instructions when changing rounding bits to preserve precision bits
10390 in the x87 control word.
10392 2018-04-03 Martin Liska <mliska@suse.cz>
10394 PR tree-optimization/82491
10395 * rtl.h (strip_offset_and_add): Replace += suboffset with
10396 poly_uint64 () + suboffset.
10398 2018-03-29 Martin Liska <mliska@suse.cz>
10399 Martin Jambor <mjambor@suse.cz>
10402 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10403 param_type is not an integral or pointer type.
10405 2018-04-03 Richard Biener <rguenther@suse.de>
10407 * sese.h (recompute_all_dominators): Remove.
10409 2018-04-02 Martin Sebor <msebor@redhat.com>
10411 * doc/invoke.texi (-Wrestrict): Fix typos.
10413 2018-04-02 Jim Wilson <jimw@sifive.com>
10415 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10416 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10417 (<optab>di3, <optab>si3_extend): Likewise.
10418 (<optab>si3_mask, <optab>si3_mask_1): New.
10419 (<optab>di3_mask, <optab>di3_mask_1): New.
10420 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10421 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10422 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10424 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10426 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10429 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10431 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10432 (nds32_canonicalize_comparison): New function.
10434 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10435 Kito Cheng <kito.cheng@gmail.com>
10436 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10438 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10439 * config/nds32/constants.md (unspec_volatile_element): Add
10440 UNSPEC_VOLATILE_RELAX_GROUP.
10441 * config/nds32/nds32-relax-opt.c: New file.
10442 * config/nds32/nds32-predicates.c
10443 (nds32_symbol_load_store_p): New function.
10444 * config/nds32/nds32-protos.h
10445 (nds32_symbol_load_store_p): Declare function.
10446 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10447 * config/nds32/nds32.c
10448 (nds32_register_pass): New function to register pass.
10449 (nds32_register_passes): New function to register passes.
10450 * config/nds32/nds32.md (relax_group): New pattern.
10451 * config/nds32/nds32.opt (mrelax-hint): New option.
10452 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10454 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10456 * config/nds32/t-nds32: Modify files dependency.
10458 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10460 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10461 (PROFILE_HOOK): Define its implementation.
10463 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10465 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10466 type and 32-bit size.
10468 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10470 PR middle-end/85090
10471 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10472 (V_128_256): New mode iterator.
10473 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10474 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10475 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10477 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10478 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10480 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10483 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10484 NaN inputs correctly.
10486 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10489 * config/rs6000/vsx.md (??r): New mode attribute.
10490 (*vsx_mov<mode>_64bit): Use it.
10491 (*vsx_mov<mode>_32bit): Likewise.
10493 2018-03-30 Martin Sebor <msebor@redhat.com>
10495 PR tree-optimization/84818
10496 * builtins.c (check_access): Use warning_n.
10498 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10501 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10503 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10506 2018-03-30 Julia Koval <julia.koval@intel.com>
10509 * x86-tune.def (movx, partial_reg_dependency): Enable for
10512 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10514 PR inline-asm/84985
10515 * lra-constraints.c (process_alt_operands): Move setting
10516 this_alternative_matches below.
10518 2018-03-29 Martin Liska <mliska@suse.cz>
10521 * doc/invoke.texi: Document how LTO works with debug info.
10522 Describe auto-load support of binutils. Mention 'x86-64'
10523 as valid option value of -march option.
10525 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10527 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10530 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10531 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10532 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10535 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10538 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10539 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10540 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10541 for RS6000_BTM_POWERPC64.
10542 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10543 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10544 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10547 (DIVDEU): Likewise.
10549 2018-03-28 Carl Love <cel@us.ibm.com>
10552 2017-09-27 Carl Love <cel@us.ibm.com>
10554 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10555 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10556 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10559 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10561 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10562 instead of __vector bool.
10563 (_mm_max_pu8): Likewise.
10564 (_mm_min_pi16): Likewise.
10566 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10569 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10570 (DIVWEUO): Likewise.
10571 (DIVDEO): Likewise.
10572 (DIVDEUO): Likewise.
10573 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10574 DIVWEUO and DIVDEUO.
10575 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10576 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10577 (div_extend): Likewise.
10578 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10580 (__builtin_divweuo): Likewise.
10581 (__builtin_divdeo): Likewise.
10582 (__builtin_divdeuo): Likewise.
10584 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10587 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10588 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10590 PR tree-optimization/82004
10591 * gimple-match-head.c (optimize_pow_to_exp): New function.
10592 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10593 Don't fold to exp if optimize_pow_to_exp is false.
10595 2018-03-28 Martin Liska <mliska@suse.cz>
10598 * calls.c (initialize_argument_information): Fix trailing space.
10599 * common.opt: Fix typo and provide better explanation for
10600 -fsanitize-coverage option.
10601 * config/i386/i386.opt: Fix typo.
10603 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10604 Martin Liska <mliska@suse.cz>
10607 * gimplify.c (asan_poison_variable): Don't do the check for
10608 gimplify_omp_ctxp here.
10609 (gimplify_decl_expr): Do it here.
10610 (gimplify_target_expr): Likewise.
10612 2018-03-28 Martin Liska <mliska@suse.cz>
10615 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10616 chkp_type_bounds_count if MPX is not enabled.
10618 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10620 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10622 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10625 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10626 function to create the function decl for complex long double
10627 multiply and divide for -mabi=ieeelongdouble.
10628 (init_float128_ieee): Call it.
10630 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10633 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10634 -fcf-protection=branch -mibt.
10635 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10637 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10640 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10641 arm_word_relocations.
10643 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10646 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10647 extern array declarations.
10649 2018-03-27 Richard Biener <rguenther@suse.de>
10651 PR middle-end/84067
10652 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10653 explicit single_use checks.
10655 2018-03-27 Richard Biener <rguenther@suse.de>
10657 PR tree-optimization/85082
10658 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10661 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10663 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10664 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10665 Turn on fasynchronous-unwind-tables and funwind-tables.
10667 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10670 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10671 (*bmi_blsr_<mode>_ccz): Ditto.
10673 2018-03-26 Tom de Vries <tom@codesourcery.com>
10675 PR tree-optimization/85063
10676 * omp-general.c (offloading_function_p): New function. Factor out
10678 * omp-offload.c (pass_omp_target_link::gate): ... here.
10679 * omp-general.h (offloading_function_p): Declare.
10680 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10681 with attribute omp declare target for offloading functions.
10683 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10685 PR tree-optimization/84005
10686 * tree-data-ref.h (get_base_for_alignment): Declare.
10687 * tree-data-ref.c (get_base_for_alignment_1): New function.
10688 (get_base_for_alignment): Likewise.
10689 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10690 get_base_for_alignment to find a suitable base object, instead
10691 of always using drb->base_address.
10693 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10695 PR inline-asm/85022
10696 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10697 known size by default.
10699 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10701 PR inline-asm/85030
10702 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10703 and non BLKmode operands.
10705 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10708 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10709 Clean up attributes.
10711 2018-03-23 Richard Biener <rguenther@suse.de>
10714 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10715 we are going to emit early debug for LTO.
10717 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10719 PR inline-asm/85034
10720 * function.c (match_asm_constraints_1): Don't optimize if input
10721 doesn't satisfy general_operand predicate for output's mode.
10723 PR inline-asm/85022
10724 * alias.c (write_dependence_p): Don't require for x_canonicalized
10725 non-VOIDmode if x has VOIDmode.
10728 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10729 just don't try to optimize it rather than assert it never happens.
10731 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10733 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10734 macro expansions for definition of ST_INTERNAL_<mode> and
10735 LD_INTERNAL_<mode> builtins.
10736 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10738 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10740 (altivec_expand_st_builtin): Likewise.
10741 (altivec_expand_builtin): Remove calls to deleted functions.
10742 (rs6000_address_for_altivec): Delete this function.
10743 * config/rs6000/vector.md: Remove expands for
10744 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10746 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10749 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10750 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10751 re-computing once computed.
10752 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10753 (arm_init_machine_status): Initialize
10754 machine->static_chain_stack_bytes.
10756 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10759 * doc/extend.texi: Add four new prototypes for vec_ld.
10760 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10761 definitions for more logical presentation.
10762 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10763 entries for V1TI variants of __builtin_altivec_ld builtin.
10764 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10765 handling of V1TI variant of LVX icode pattern.
10766 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10767 (rs6000_gimple_fold_builtin): Likewise.
10768 (altivec_init_builtins): Add code to define
10769 __builtin_altivec_lvx_v1ti function.
10771 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10773 PR inline-asm/84941
10774 * function.c (match_asm_constraints_1): Don't do the optimization
10775 if input isn't a REG, SUBREG, MEM or constant.
10777 2018-03-22 Tom de Vries <tom@codesourcery.com>
10779 PR tree-optimization/84956
10780 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10781 bb_has_abnormal_pred.
10783 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10786 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10787 DECL_INITIAL (decl) to decl at the end.
10788 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10789 adjust the comment.
10791 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10793 * doc/extend.texi (__builtin_tgmath): Document when complex
10794 integer types are treated as _Complex _Float64.
10796 2018-03-21 Tom de Vries <tom@codesourcery.com>
10798 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10800 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10802 PR tree-optimization/84960
10803 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10804 if it is ENTRY block, move them into single succ of ENTRY in that case.
10806 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10808 PR tree-optimization/84811
10809 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10810 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10811 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10812 (known_subrange_p): Update accordingly. Cast each value involved
10813 in the size comparison, rather than casting the result of the
10816 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10818 PR tree-optimization/84982
10819 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10820 by flipping the least significant bit rather than all bits from
10821 bitpos to bitpos + bitsize - 1.
10823 2018-03-21 Nathan Sidwell <nathan@acm.org>
10825 * doc/extend.texi (Deprecated Features): Remove mention of
10826 long-deleted deprecations.
10828 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10831 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10832 * configure: Regenerate.
10834 2018-03-21 Tom de Vries <tom@codesourcery.com>
10836 PR tree-optimization/83126
10837 * tree-parloops.c (num_phis): New function.
10838 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10840 2018-03-21 Nathan Sidwell <nathan@acm.org>
10842 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10843 mention anon-struct/union members and trailing attributes.
10845 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10847 PR tree-optimization/84969
10848 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10849 builtin memset partitions if they set different rhs values.
10851 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10853 PR rtl-optimization/84989
10854 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10855 VEC_DUPLICATE with scalar result mode.
10857 2018-03-21 Martin Liska <mliska@suse.cz>
10860 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10861 not intended return statement.
10863 2018-03-21 Martin Liska <mliska@suse.cz>
10866 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10867 (chkp_find_bound_slots_1): Limit number of iterations.
10869 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10872 * Minor grammar fixes for x86 options.
10874 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10877 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10878 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10880 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10883 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10884 (altivec_lvx_<mode>_1op): Likewise.
10885 (altivec_stvx_<mode>_2op): Likewise.
10886 (altivec_stvx_<mode>_1op): Likewise.
10887 (altivec_lvx_<VM2:mode>): New define_expand.
10888 (altivec_stvx_<VM2:mode>): Likewise.
10889 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10890 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10891 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10892 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10893 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10894 (rs6000_gen_lvx): Likewise.
10895 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10896 (altivec_expand_stv_builtin): Likewise.
10897 (altivec_expand_builtin): Likewise.
10898 * config/rs6000/vector.md: Likewise.
10900 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10903 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10906 2018-03-20 Richard Biener <rguenther@suse.de>
10909 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10910 sign-conversions as zero, fall back to standard scalar_stmt
10913 2018-03-20 Martin Liska <mliska@suse.cz>
10916 * predict.c (rebuild_frequencies): Handle case when we have
10917 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10919 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10922 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10923 flag_section_anchors.
10924 * varasm.c (use_blocks_for_decl_p): Remove hack for
10925 dw2_force_const_mem.
10928 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10930 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10931 be created, use lowpart_subreg of operands[0] rather than operands[0]
10933 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10934 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10935 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10936 and n constraint instead of aarch64_shift_imm_di and Usd.
10937 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10938 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10940 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10943 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10944 to favor GPR over NEON registers.
10945 (<shift>di3_neon): Likewise.
10947 2018-03-20 Tom de Vries <tom@codesourcery.com>
10950 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10951 (nvptx_process_pars): Emit bar.sync asap and alap.
10953 2018-03-20 Tom de Vries <tom@codesourcery.com>
10956 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10957 seen_label if seen_label is already set.
10959 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10962 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10963 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10964 Use 1U instead of 1. Formatting fixes.
10967 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10968 instead of TREE_TYPE (s1) for the return value.
10970 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10972 PR tree-optimization/84946
10973 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10974 bitsize + bitsize in poly_uint64 rather than poly_int64.
10977 * dwarf2asm.c: Include fold-const.c.
10978 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10979 of decl rather than decl itself.
10981 PR rtl-optimization/84643
10982 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10984 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10987 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10988 calling assemble_variable.
10990 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10993 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10994 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10996 2018-03-19 Jim Wilson <jimw@sifive.com>
10999 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
11000 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
11001 (riscv_first_stack_step): Likewise.
11002 (riscv_option_override): Use STACK_BOUNDARY instead of
11003 MIN_STACK_BOUNDARY.
11004 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
11005 MIN_STACK_BOUNDARY.
11006 (BIGGEST_ALIGNMENT): Set to 128.
11007 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
11008 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
11011 2018-03-19 Richard Biener <rguenther@suse.de>
11013 PR tree-optimization/84933
11014 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
11015 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
11017 2018-03-19 Richard Biener <rguenther@suse.de>
11019 PR tree-optimization/84859
11020 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11021 (cond_if_else_store_replacement): Perform sinking operation on
11022 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11023 Generalize what a BB with a single eligible store is.
11025 2018-03-19 Richard Biener <rguenther@suse.de>
11027 PR tree-optimization/84929
11028 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11029 chrec_is_positive against non-chrec arg.
11031 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11034 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11036 2018-03-18 Martin Liska <mliska@suse.cz>
11038 PR rtl-optimization/84635
11039 * regrename.c (build_def_use): Use matches_mode only when
11042 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11044 PR tree-optimization/84913
11045 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11046 vectorize chains of COND_EXPRs.
11048 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11050 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11052 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11054 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11056 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11058 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11060 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11061 Kito Cheng <kito.cheng@gmail.com>
11063 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11064 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11065 (nds32_adjust_reg_alloc_order): New function.
11066 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11068 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11070 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11071 nds32_print_operand, nds32_print_operand_address): Use
11072 HOST_WIDE_INT_PRINT_DEC instead.
11074 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11076 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11078 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11081 * config/i386/i386.c (initial_ix86_tune_features,
11082 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11083 unsigned long long.
11084 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11085 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11086 rather than 1u << ix86_tune. Formatting fix.
11087 (ix86_option_override_internal): Change ix86_arch_mask from
11088 unsigned int to unsigned HOST_WIDE_INT, initialize to
11089 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11090 (ix86_function_specific_restore): Likewise.
11092 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11095 * postreload.c (reload_combine_recognize_pattern): Perform
11096 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11097 truncate_int_for_mode the result for the destination's mode.
11100 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11101 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11103 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11106 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11109 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11110 regno_allocno_class_array and sorted_pseudos.
11111 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11112 insns where regno is used.
11114 2018-03-16 Martin Liska <mliska@suse.cz>
11117 * multiple_target.c (create_dispatcher_calls): Redirect
11118 reference in the symbol table.
11120 2018-03-16 Martin Liska <mliska@suse.cz>
11123 * multiple_target.c (create_dispatcher_calls): Redirect also
11126 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11130 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11131 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11132 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11134 2018-03-16 Julia Koval <julia.koval@intel.com>
11136 * doc/invoke.texi (Skylake Server): Add CLWB.
11137 Cannonlake): Remove CLWB.
11139 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11141 PR tree-optimization/84841
11142 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11144 (FLOAT_ONE_CONST_TYPE): Define.
11145 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11146 (sort_by_operand_rank): Put entries with higher constant_type last
11147 rather than first to match comments.
11149 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11151 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11154 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11157 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11158 check and use address of target always.
11160 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11163 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11164 (indirect_thunk_bnd_needed): Likewise.
11165 (indirect_thunks_used): Likewise.
11166 (indirect_thunks_bnd_used): Likewise.
11167 (indirect_return_needed): New.
11168 (indirect_return_bnd_needed): Likewise.
11169 (output_indirect_thunk_function): Add a bool argument for
11171 (output_indirect_thunk_function): Don't generate alias for
11172 function return thunk.
11173 (ix86_code_end): Call output_indirect_thunk_function to generate
11174 function return thunks.
11175 (ix86_output_function_return): Set indirect_return_bnd_needed
11176 and indirect_return_needed instead of indirect_thunk_bnd_needed
11177 and indirect_thunk_needed.
11179 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11181 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11182 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11183 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11185 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11186 Paul Hua <paul.hua.gm@gmail.com>
11189 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11191 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11193 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11194 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11195 resp. SFmode cases.
11197 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11200 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11201 instead of GET_MODE_SIZE when comparing Units.
11203 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11206 * varasm.c (hash_section): Return an unchangeble hash value
11207 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11208 Return !aarch64_can_use_per_function_literal_pools_p ().
11210 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11213 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11214 rather than address of cmode as last argument to prepare_cmp_insn.
11216 2018-03-15 Julia Koval <julia.koval@intel.com>
11218 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11219 F_AVX512VNNI, F_AVX512BITALG): New.
11221 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11224 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11225 insn for floating-point loads and stores.
11227 2018-03-14 Carl Love <cel@us.ibm.com>
11229 * config/rs6000/rs6000-c.c: Add macro definitions for
11230 ALTIVEC_BUILTIN_VEC_PERMXOR.
11231 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11232 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11233 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11234 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11236 * config/doc/extend.texi: Add prototypes for vec_permxor.
11238 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11241 * diagnostic-show-locus.c (class layout_point): Convert m_line
11242 from int to linenum_type.
11243 (line_span::comparator): Use linenum "compare" function when
11244 comparing line numbers.
11245 (test_line_span): New function.
11246 (layout_range::contains_point): Convert param "row" from int to
11248 (layout_range::intersects_line_p): Likewise.
11249 (layout::will_show_line_p): Likewise.
11250 (layout::print_source_line): Likewise.
11251 (layout::should_print_annotation_line_p): Likewise.
11252 (layout::print_annotation_line): Likewise.
11253 (layout::print_leading_fixits): Likewise.
11254 (layout::annotation_line_showed_range_p): Likewise.
11255 (struct line_corrections): Likewise for field m_row.
11256 (line_corrections::line_corrections): Likewise for param "row".
11257 (layout::print_trailing_fixits): Likewise.
11258 (layout::get_state_at_point): Likewise.
11259 (layout::get_x_bound_for_row): Likewise.
11260 (layout::print_line): Likewise.
11261 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11262 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11263 * input.c (selftest::test_linenum_comparisons): New function.
11264 (selftest::input_c_tests): Call it.
11265 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11266 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11267 * selftest.h (ASSERT_GT): New macro.
11268 (ASSERT_GT_AT): New macro.
11269 (ASSERT_LT): New macro.
11270 (ASSERT_LT_AT): New macro.
11272 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11274 PR rtl-optimization/84780
11275 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11277 2018-03-14 Martin Liska <mliska@suse.cz>
11279 * tree.c (record_node_allocation_statistics): Use
11280 get_stats_node_kind.
11281 (get_stats_node_kind): New function extracted from
11282 record_node_allocation_statistics.
11283 (free_node): Use get_stats_node_kind.
11285 2018-03-14 Richard Biener <rguenther@suse.de>
11287 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11288 that the value-set of ANTIC_IN doesn't grow.
11291 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11293 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11294 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11296 2018-03-14 Julia Koval <julia.koval@intel.com>
11298 * config.gcc (icelake-client, icelake-server): New.
11300 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11301 (initial_ix86_arch_features): Ditto.
11302 (PTA_SKYLAKE): Add SGX.
11303 (PTA_ICELAKE): Remove.
11304 (PTA_ICELAKE_CLIENT): New.
11305 (PTA_ICELAKE_SERVER): New.
11306 (ix86_option_override_internal): Split up icelake on icelake client and
11308 (get_builtin_code_for_version): Ditto.
11309 (fold_builtin_cpu): Ditto.
11310 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11311 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11312 * config/i386/i386.h (processor_type): Ditto.
11313 * doc/invoke.texi: Ditto.
11315 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11318 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11319 INTEGER_CST offset, add it together with bitpos / 8 and
11320 sign extend based on POINTER_SIZE.
11324 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11327 * config/i386/constraints.md (Yc): New register constraint.
11328 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11329 Use Yc constraint for alternative 2 of operand 0. Remove
11330 preferred_for_speed attribute.
11332 2018-03-14 Richard Biener <rguenther@suse.de>
11334 PR tree-optimization/84830
11335 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11336 with the old one to avoid oscillations.
11338 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11341 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11343 (assign_by_spills): Return a flag of reload assignment failure.
11344 Do not process the reload assignment failures. Do not spill other
11345 reload pseudos if they has the same reg class. Update n if
11347 (lra_assign): Add a return arg. Set up from the result of
11348 assign_by_spills call.
11349 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11350 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11351 usage_insns if it is not NULL.
11352 (spill_hard_reg_in_range): New function.
11353 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11354 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11355 function prototypes.
11356 (lra_assign): Change prototype.
11357 * lra.c (lra): Add code to deal with fails by splitting hard reg
11360 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11362 * config/riscv/riscv.opt (mrelax): New option.
11363 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11364 "norelax" when riscv_mrelax is disabled.
11365 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11367 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11370 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11371 reassociation for int modes.
11373 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11375 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11376 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11378 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11379 * config/aarch64/aarch64-sve.md
11380 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11381 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11382 (*extend<mode><Vwide>2): Rename to...
11383 (aarch64_sve_extend<mode><Vwide>2): ...this.
11384 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11385 renaming the old pattern to...
11386 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11388 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11389 define_expand, renaming the old pattern to...
11390 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11391 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11392 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11393 account when deciding which SVE instruction the optab should use.
11394 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11396 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11398 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11399 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11400 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11401 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11402 (tlsdesc_small_<mode>): Turn a define_expand and use
11403 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11404 (tlsdesc_small_advsimd_<mode>): ...this.
11405 (tlsdesc_small_sve_<mode>): New pattern.
11407 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11409 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11410 (UNSPEC_UMUL_HIGHPART): New constants.
11411 (MUL_HIGHPART): New int iteraor.
11412 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11413 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11415 (*<su>mul<mode>3_highpart): New define_insn.
11417 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11420 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11423 2018-03-13 Martin Liska <mliska@suse.cz>
11426 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11428 (sem_item_optimizer::~sem_item_optimizer): Release it.
11429 (sem_item_optimizer::merge_classes): Register variable aliases.
11430 (sem_item_optimizer::fixup_pt_set): New function.
11431 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11432 * ipa-icf.h: Declare new variables and functions.
11434 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11436 PR middle-end/84834
11437 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11438 integer_pow2p@2 and test integer_pow2p in condition.
11439 (A < 0 ? C : 0): Similarly for @1.
11441 PR middle-end/84831
11442 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11443 characters starting at p contain '\0' character, don't look beyond
11447 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11448 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11451 * reg-stack.c (change_stack): Change update_end var from int to
11452 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11453 also call set_block_for_insn on the newly added insns and rescan.
11456 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11457 on the last operand.
11460 * tree.c (stabilize_reference_1): Return save_expr (e) for
11461 STATEMENT_LIST even if it doesn't have side-effects.
11463 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11465 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11467 2018-03-12 Renlin Li <renlin.li@arm.com>
11469 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11470 aarch64_output_scalar_simd_mov_immediate.
11472 2018-03-12 Martin Sebor <msebor@redhat.com>
11474 PR tree-optimization/83456
11475 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11476 for perfectly overlapping calls to memcpy.
11477 (gimple_fold_builtin_memory_chk): Same.
11478 (gimple_fold_builtin_strcpy): Handle no-warning.
11479 (gimple_fold_builtin_stxcpy_chk): Same.
11480 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11482 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11484 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11485 parameter. Use it for SFmode.
11486 (rs6000_function_arg_advance_1): Adjust.
11487 (rs6000_function_arg): Adjust.
11488 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11490 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11492 PR rtl-optimization/84169
11493 PR rtl-optimization/84780
11494 * combine.c (can_combine_p): Check for a 2-insn combination whether
11495 the destination register is used between the two insns, too.
11497 2018-03-12 Richard Biener <rguenther@suse.de>
11499 PR tree-optimization/84803
11500 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11501 for refs DR analysis didn't process.
11503 2018-03-12 Richard Biener <rguenther@suse.de>
11505 PR tree-optimization/84777
11506 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11507 force-vectorize loops ignore whether we are optimizing for size.
11509 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11511 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11512 (TARGET_MD_ASM_ADJUST): Define.
11514 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11515 Kito Cheng <kito.cheng@gmail.com>
11516 Chung-Ju Wu <jasonwucj@gmail.com>
11518 * config/nds32/nds32.c (nds32_compute_stack_frame,
11519 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11520 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11521 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11522 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11523 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11524 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11525 * config/nds32/nds32.md (prologue, epilogue): Use macro
11526 NDS32_V3PUSH_AVAILABLE_P to do checking.
11528 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11531 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11532 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11533 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11534 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11535 addition of most attributes on !orig_type_die or the attribute not
11536 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11538 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11539 Chung-Ju Wu <jasonwucj@gmail.com>
11541 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11542 __NDS32_VH__ macro.
11543 * config/nds32/nds32.opt (mvh): New option.
11545 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11546 Chung-Ju Wu <jasonwucj@gmail.com>
11548 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11550 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11551 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11554 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11555 Chung-Ju Wu <jasonwucj@gmail.com>
11557 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11559 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11560 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11562 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11563 Kito Cheng <kito.cheng@gmail.com>
11564 Chung-Ju Wu <jasonwucj@gmail.com>
11566 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11567 UNSPEC_FFMISM and UNSPEC_FLMISM.
11568 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11569 for ffb, ffmism and flmism.
11570 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11571 (unspec_ffmism): Ditto.
11572 (unspec_flmism): Ditto.
11573 (nds32_expand_builtin_impl): Check if string extension is available.
11574 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11575 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11577 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11580 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11583 * lra-assigns.c (assign_by_spills): Return a flag of reload
11584 assignment failure. Do not process the reload assignment
11585 failures. Do not spill other reload pseudos if they has the same
11587 (lra_assign): Add a return arg. Set up from the result of
11588 assign_by_spills call.
11589 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11590 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11591 usage_insns if it is not NULL.
11592 (spill_hard_reg_in_range): New function.
11593 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11594 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11595 function prototypes.
11596 (lra_assign): Change prototype.
11597 * lra.c (lra): Add code to deal with fails by splitting hard reg
11600 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11603 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11605 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11608 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11609 (dw_val_node): Add val_symbolic_view.
11610 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11611 (symview_upper_bound): New.
11612 (new_line_info_table): Initialize symviews_since_reset.
11613 (dwarf2out_source_line): Count symviews_since_reset and set
11614 symview_upper_bound.
11615 (dw_val_equal_p): Handle symview.
11616 (add_AT_symview): New.
11617 (print_dw_val): Handle symview.
11618 (attr_checksum, attr_checksum_ordered): Likewise.
11619 (same_dw_val_p, size_of_die): Likewise.
11620 (value_format, output_die): Likewise.
11621 (add_high_low_attributes): Use add_AT_symview for entry_view.
11622 (dwarf2out_finish): Reset symview_upper_bound, clear
11625 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11628 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11629 Add strict argument and use it.
11630 (rs6000_split_multireg_move): Update for new strict argument.
11631 (mem_operand_gpr): Disallow all non-offsettable addresses.
11632 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11634 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11637 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11638 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11639 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11642 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11643 decl, use remap_type if we want to use the type.
11645 2018-03-09 Martin Sebor <msebor@redhat.com>
11647 PR tree-optimization/84526
11648 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11650 (builtin_access::generic_overlap): Be prepared to handle non-array
11653 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11655 PR rtl-optimization/84682
11656 * lra-constraints.c (process_address_1): Check is_address flag
11657 for address constraints.
11658 (process_alt_operands): Likewise.
11659 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11660 preprocess_constraints.
11661 * recog.h (preprocess_constraints): Add oploc parameter.
11663 * recog.c (preprocess_constraints): Test address_operand for
11664 CT_ADDRESS constraints.
11666 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11669 * lra-assigns.c (assign_by_spills): Return a flag of reload
11670 assignment failure. Do not process the reload assignment
11671 failures. Do not spill other reload pseudos if they has the same
11673 (lra_assign): Add a return arg. Set up from the result of
11674 assign_by_spills call.
11675 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11676 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11677 usage_insns if it is not NULL.
11678 (spill_hard_reg_in_range): New function.
11679 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11680 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11681 function prototypes.
11682 (lra_assign): Change prototype.
11683 * lra.c (lra): Add code to deal with fails by splitting hard reg
11686 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11689 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11690 Accept complain bool parameter. Only emit errors if it is true.
11691 (arm_parse_cpu_option_name): Likewise.
11692 (arm_target_thumb_only): Adjust callers of the above.
11693 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11694 prototype to take a default true bool parameter.
11695 (arm_parse_arch_option_name): Likewise.
11697 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11698 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11702 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11703 * configure: Regenerate.
11704 * configure.ac ("linker --version-script option"): New.
11705 ("linker soname option"): New.
11707 2018-03-09 Richard Biener <rguenther@suse.de>
11709 PR tree-optimization/84775
11710 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11711 immediate uses of predicate stmts and mark them modified.
11714 PR tree-optimization/84178
11715 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11717 (version_loop_for_if_conversion): Delay update_ssa call.
11718 (tree_if_conversion): Delay update_ssa until after predicate
11721 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11724 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11725 when the function accesses prior frames.
11727 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11730 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11731 gen_llsym, otherwise call maybe_gen_llsym.
11733 PR inline-asm/84742
11734 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11735 has ',' character inside of it.
11737 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11740 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11741 as clobbering CC_REGNUM.
11743 2018-03-08 Richard Biener <rguenther@suse.de>
11745 PR middle-end/84552
11746 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11747 (follow_copies_to_constant): Do not follow SSA names registered
11750 2018-03-08 Richard Biener <rguenther@suse.de>
11752 PR tree-optimization/84178
11753 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11755 (version_loop_for_if_conversion): Delay update_ssa call.
11756 (tree_if_conversion): Delay update_ssa until after predicate
11759 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11761 PR tree-optimization/84178
11762 * tree-if-conv.c (release_bb_predicate): Remove the
11763 the assertion that the stmts have NULL use_ops.
11764 Discard the statements, asserting that they haven't
11765 yet been added to a BB.
11767 2018-03-08 Richard Biener <rguenther@suse.de>
11769 PR tree-optimization/84746
11770 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11771 (phi_translate): Pass in destination ANTIC_OUT set.
11772 (phi_translate_1): Likewise. For a simplified result lookup
11773 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11774 (phi_translate_set): Adjust.
11775 (do_pre_regular_insertion): Likewise.
11776 (do_pre_partial_partial_insertion): Likewise.
11778 2018-03-08 Martin Liska <mliska@suse.cz>
11780 PR gcov-profile/84735
11781 * doc/gcov.texi: Document usage of profile files.
11782 * gcov-io.h: Document changes in the format.
11784 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11788 * dwarf2out.c (struct dw_line_info_table): Update comments for
11790 (FORCE_RESET_NEXT_VIEW): New.
11791 (FORCE_RESETTING_VIEW_P): New.
11792 (RESETTING_VIEW_P): Check for -1 too.
11793 (ZERO_VIEW_P): Likewise.
11794 (new_line_info_table): Force-reset next view.
11795 (dwarf2out_begin_function): Likewise.
11796 (dwarf2out_source_line): Simplify zero_view_p initialization.
11797 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11798 view directly. Omit view when omitting .loc at line 0.
11800 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11802 PR tree-optimization/84740
11803 * tree-switch-conversion.c (process_switch): Call build_constructors
11804 only if info.phi_count is non-zero.
11806 PR tree-optimization/84739
11807 * tree-tailcall.c (find_tail_calls): Check call arguments against
11808 DECL_ARGUMENTS (current_function_decl) rather than
11809 DECL_ARGUMENTS (func) when checking for tail recursion.
11811 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11813 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11814 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11815 Volker Reichelt's entry and add entries for people that perform
11816 GCC fuzzy testing and report numerous bugs.
11818 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11821 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11822 readonly data in sdata, if that is disabled.
11823 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11824 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11825 -mreadonly-in-sdata option.
11827 2018-03-07 Martin Sebor <msebor@redhat.com>
11829 PR tree-optimization/84468
11830 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11831 basic block when looking for nul assignment.
11833 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11836 * except.h (output_function_exception_table): Adjust prototype.
11837 * except.c (output_function_exception_table): Remove FNNAME parameter
11838 and add SECTION parameter. Ouput one part of the table at a time.
11839 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11840 the first part of the exception table and emit unwind directives.
11841 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11842 (i386_pe_seh_cold_init): Likewise.
11843 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11844 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11845 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11846 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11847 * config/i386/winnt.c: Include except.h.
11848 (struct seh_frame_state): Add reg_offset, after_prologue and
11849 in_cold_section fields.
11850 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11851 (i386_pe_seh_cold_init): New function.
11852 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11853 to seh->in_cold_section.
11854 (seh_emit_push): Record the offset of the push.
11855 (seh_emit_save): Record the offet of the save.
11856 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11857 Test seh->after_prologue to disregard the epilogue.
11858 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11859 (i386_pe_end_cold_function): New function.
11861 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11864 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11865 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11868 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11870 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11871 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11873 PR middle-end/84723
11874 * multiple_target.c: Include tree-inline.h and intl.h.
11875 (expand_target_clones): Diagnose and fail if node->definition and
11876 !tree_versionable_function_p (node->decl).
11878 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11880 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11882 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11883 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11884 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11886 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11889 * combine.c (try_combine): Use reg_or_subregno instead of handling
11890 just paradoxical SUBREGs and REGs.
11892 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11894 * config/arc/arc.c (arc_finalize_pic): Remove function.
11895 (arc_must_save_register): We use single base PIC register, remove
11896 checks to save/restore the PIC register.
11897 (arc_expand_prologue): Likewise.
11898 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11900 (arc_verify_short): Likewise.
11901 (arc_attr_type): Likewise.
11902 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11903 (walk_stores): Likewise.
11904 (arc_address_cost): Make it static.
11905 (arc_verify_short): Likewise.
11906 (branch_dest): Likewise.
11907 (arc_attr_type): Likewise.
11908 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11909 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11910 (arc_final_prescan_insn): Remove inserting the nops due to
11911 hardware hazards. It is done in reorg step.
11912 (insn_length_variant_t): Remove.
11913 (insn_length_parameters_t): Likewise.
11914 (arc_insn_length_parameters): Likewise.
11915 (arc_get_insn_variants): Likewise.
11916 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11918 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11920 PR inline-asm/84683
11921 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11924 PR tree-optimization/84687
11925 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11927 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11929 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11931 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11932 Rename to ppc_speculation_barrier.
11933 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11934 __builtin_ppc_speculation_barrier.
11936 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11939 * combine.c (combine_simplify_rtx): Don't try to simplify if
11940 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11943 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11945 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11946 to 32 bytes when compiling for POWER9.
11948 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11951 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11952 regparm >= 3 with no arg reg available also for calls with
11953 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11956 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11958 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11960 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11963 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11965 2018-03-05 Richard Biener <rguenther@suse.de>
11967 PR tree-optimization/84486
11968 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11969 When inserting a __builtin_assume_aligned call set the LHS
11970 SSA name alignment info accordingly.
11972 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11974 PR tree-optimization/84114
11975 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11976 Avoid reassociation of FLOAT_MODE addition.
11978 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11980 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11981 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11982 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11983 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11984 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11985 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11986 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11988 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11989 __builtin_ia32_wbinvd): New builtins.
11990 (SPECIAL_ARGS2): New.
11991 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11992 (SPECIAL_ARGS2): New.
11993 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11994 (ix86_valid_target_attribute_inner_p): Ditto.
11995 (ix86_init_mmx_sse_builtins): Add special_args2.
11996 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11997 TARGET_WBNOINVD_P): New.
11998 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11999 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
12000 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
12001 * config/i386/immintrin.h (_wbinvd): New intrinsic.
12002 * config/i386/pconfigintrin.h: New file.
12003 * config/i386/wbnoinvdintrin.h: Ditto.
12004 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
12006 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
12008 2018-03-05 Richard Biener <rguenther@suse.de>
12010 PR tree-optimization/84670
12011 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
12013 (BB_VISITED_WITH_VISITED_SUCCS): New define.
12014 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
12015 (compute_antic_aux): Only assert the number of values in ANTIC_IN
12016 doesn't grow if all successors (recursively) were visited at least
12019 2018-03-05 Richard Biener <rguenther@suse.de>
12021 PR tree-optimization/84650
12022 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12023 if executed in the loop pipeline.
12025 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12027 * doc/configfiles.texi (Configuration Files): Move info about
12028 conditionalizing $target-protos.h to...
12029 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12030 differs from $target-protos.h.
12032 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12033 Chung-Ju Wu <jasonwucj@gmail.com>
12035 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12036 * config/nds32/nds32-multiple.md (setmemsi): Define.
12037 * config/nds32/nds32-memory-manipulation.c
12038 (nds32_gen_dup_4_byte_to_word_value): New.
12039 (emit_setmem_word_loop): New.
12040 (emit_setmem_byte_loop): New.
12041 (nds32_expand_setmem_loop): New.
12042 (nds32_expand_setmem_loop_v3m): New.
12043 (nds32_expand_setmem_unroll): New.
12044 (nds32_expand_setmem): New.
12046 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12047 Chung-Ju Wu <jasonwucj@gmail.com>
12049 * config/nds32/nds32-memory-manipulation.c
12050 (nds32_emit_load_store): New.
12051 (nds32_emit_post_inc_load_store): New.
12052 (nds32_emit_mem_move): New.
12053 (nds32_emit_mem_move_block): New.
12054 (nds32_expand_movmemsi_loop_unknown_size): New.
12055 (nds32_expand_movmemsi_loop_known_size): New.
12056 (nds32_expand_movmemsi_loop): New.
12057 (nds32_expand_movmemsi_unroll): New.
12058 (nds32_expand_movmemqi): Rename ...
12059 (nds32_expand_movmemsi): ... to this.
12060 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12061 (movmemsi): ... to this.
12062 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12063 (nds32_expand_movmemsi): ... to this.
12065 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12066 Monk Chiang <sh.chiang04@gmail.com>
12067 Chung-Ju Wu <jasonwucj@gmail.com>
12069 * config/nds32/nds32-protos.h
12070 (nds32_expand_load_multiple): New arguments.
12071 (nds32_expand_store_multiple): Ditto.
12072 (nds32_valid_multiple_load_store): Rename ...
12073 (nds32_valid_multiple_load_store_p): ... to this.
12074 * config/nds32/nds32-memory-manipulation.c
12075 (nds32_expand_load_multiple): Refine implementation.
12076 (nds32_expand_store_multiple): Ditto.
12077 * config/nds32/nds32-multiple.md
12078 (load_multiple): Update nds32_expand_load_multiple interface.
12079 (store_multiple): Update nds32_expand_store_multiple interface.
12080 * config/nds32/nds32-predicates.c
12081 (nds32_valid_multiple_load_store): Rename ...
12082 (nds32_valid_multiple_load_store_p): ... to this and refine
12084 * config/nds32/predicates.md
12085 (nds32_load_multiple_and_update_address_operation): New predicate.
12086 (nds32_store_multiple_and_update_address_operation): New predicate.
12088 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12089 Chung-Ju Wu <jasonwucj@gmail.com>
12091 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12092 (combo): New attribute.
12093 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12095 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12097 * config/nds32/nds32.opt: Change -mcmodel= default value.
12099 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12100 Monk Chiang <sh.chiang04@gmail.com>
12101 Chung-Ju Wu <jasonwucj@gmail.com>
12103 * config/nds32/constants.md (unspec_element): New enum.
12104 * config/nds32/constraints.md (Umw): New constraint.
12105 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12106 * config/nds32/nds32-intrinsic.md: Likewise.
12107 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12108 (nds32_valid_smw_lwm_base_p): New.
12109 (nds32_output_smw_single_word): New.
12110 (nds32_output_lmw_single_word): New.
12111 (nds32_expand_unaligned_load): New.
12112 (nds32_expand_unaligned_store): New.
12113 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12114 (nds32_output_smw_single_word): Declare.
12115 (nds32_output_lmw_single_word): Declare.
12116 (nds32_expand_unaligned_load): Declare.
12117 (nds32_expand_unaligned_store): Declare.
12118 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12119 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12120 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12121 NDS32_BUILTIN_UASTORE_DW.
12122 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12125 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12126 Kito Cheng <kito.cheng@gmail.com>
12127 Chung-Ju Wu <jasonwucj@gmail.com>
12129 * config/nds32/nds32-intrinsic.c
12130 (nds32_expand_builtin_null_ftype_reg): Delete.
12131 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12132 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12133 (nds32_read_argument): New.
12134 (nds32_legitimize_target): Ditto.
12135 (nds32_legitimize_argument): Ditto.
12136 (nds32_check_constant_argument): Ditto.
12137 (nds32_expand_unop_builtin): Ditto.
12138 (nds32_expand_unopimm_builtin): Ditto.
12139 (nds32_expand_binop_builtin): Ditto.
12140 (nds32_builtin_decl_impl): Ditto.
12141 (builtin_description): Ditto.
12142 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12143 (nds32_init_builtins_impl): Ditto.
12144 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12145 (nds32_builtin_decl): New.
12146 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12147 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12149 2018-03-02 Jeff Law <law@redhat.com>
12151 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12152 (redundant_insn, fill_simple_delay_slots): Likewise.
12153 (fill_slots_from_thread): Likewise.
12154 * resource.c (mark_referenced_resources): Likewise.
12155 (mark_set_resources, find_dead_or_set_registers): Likewise.
12157 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12159 * substring-locations.h (format_warning_va): Formatting fix for
12160 ATTRIBUTE_GCC_DIAG.
12161 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12163 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12164 * substring-locations.c: Include intl.h.
12165 (format_warning_va): Turned into small wrapper around
12166 format_warning_n_va, renamed to ...
12167 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12168 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12170 (format_warning_at_substring_n): New function.
12171 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12172 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12173 format_warning_at_substring with just a shorter name instead of
12174 const function pointer.
12175 (fmtwarn_n): New function.
12176 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12177 appropriate, get rid of all the fmtstr temporaries, move conditionals
12178 with G_() wrapped string literals directly into fmtwarn arguments,
12179 cast dir.len to (int), formatting fixes.
12181 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12183 * doc/invoke.texi: Remove "Cilk Plus" references.
12185 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12186 Richard Biener <rguenther@suse.de>
12189 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12190 for error or warning attributes if CALL_FROM_THUNK_P is set.
12193 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12196 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12197 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12200 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12201 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12203 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12204 instead of -1U in last predictors element's probability member.
12206 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12209 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12210 arguments if they are comparable.
12212 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12214 PR tree-optimization/84634
12215 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12216 masks and masked_loop_p with a single loop_masks, making sure it's
12217 null for bb vectorization.
12219 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12221 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12222 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12223 loop->force_vectorize to check whether there is no alias.
12225 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12228 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12230 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12232 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12233 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12234 instead of a loop around prev_real_insn.
12235 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12238 PR inline-asm/84625
12239 * config/i386/i386.c (ix86_print_operand): Use conditional
12240 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12243 2018-03-02 Richard Biener <rguenther@suse.de>
12245 PR tree-optimization/84427
12246 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12247 (bitmap_set_subtract_values): Rewrite to handle multiple
12250 (prune_clobbered_mems): Likewise.
12251 (phi_translate): Take edge instead of pred/phiblock.
12252 (phi_translate_1): Likewise.
12253 (phi_translate_set): Likewise. Insert all translated
12254 exprs for a value into the set, keeping possibly multiple
12255 expressions per value.
12256 (compute_antic_aux): Adjust for phi_translate changes.
12257 When intersecting union the expressions and prune those
12258 not in the final value set, keeping possibly multiple
12259 expressions per value. Do not use value-insertion
12260 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12261 all expressions. Add verification that the value-sets
12262 only shrink during iteration.
12263 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12264 (do_pre_regular_insertion): Likewise.
12265 (do_pre_partial_partial_insertion): Likewise.
12267 2018-03-02 Richard Biener <rguenther@suse.de>
12270 * config/darwin.c (saved_debug_info_level): New static global.
12271 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12272 (darwin_asm_lto_end): Restore debug info generation settings.
12274 2018-03-01 Martin Liska <mliska@suse.cz>
12277 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12278 volatile arguments.
12280 2018-03-01 Richard Biener <rguenther@suse.de>
12283 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12284 type annotation with LTO when debug was disabled at compile-time.
12286 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12288 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12290 (mips_final_postscan_insn): Likewise.
12292 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12294 PR rtl-optimization/84528
12295 * alias.c (init_alias_target): Add commentary.
12296 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12297 a unique base value if the frame pointer is not eliminated
12298 to the stack pointer.
12300 2018-03-01 Tom de Vries <tom@codesourcery.com>
12302 PR rtl-optimization/83327
12303 * lra-int.h (hard_regs_spilled_into): Declare.
12304 * lra.c (hard_regs_spilled_into): Define.
12305 (init_reg_info): Init hard_regs_spilled_into.
12306 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12307 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12308 (process_bb_lives): Handle hard_regs_spilled_into.
12309 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12310 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12312 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12314 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12315 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12316 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12317 * config/rs6000/aix72.h: New file.
12319 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12321 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12322 instead of warning_at with conditional singular and plural messages
12326 * stor-layout.c (update_alignment_for_field): For
12327 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12328 && !DECL_PACKED (field), do the alignment update, just use
12329 only desired_align instead of MAX (type_align, desired_align)
12331 (place_field): Don't do known_align < desired_align handling
12332 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12333 is non-NULL, instead do it after rli->prev_field handling and
12334 only if not within a bitfield word. For DECL_PACKED (field)
12335 use type_align of BITS_PER_UNIT.
12337 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12339 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12340 superfluous parentheses and trailing spaces.
12342 2018-02-28 Richard Biener <rguenther@suse.de>
12344 PR tree-optimization/84584
12345 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12346 SCoPs with fake exit edge.
12348 2018-02-28 Martin Liska <mliska@suse.cz>
12351 * timevar.c (timer::print): Fix format to properly print 100%
12354 2018-02-28 Richard Biener <rguenther@suse.de>
12356 PR middle-end/84607
12357 * genmatch.c (capture_info::walk_match): Do not mark
12358 captured expressions without operands as expr_p given
12359 they act more like predicates and should be subject to
12360 "lost tail" side-effect preserving.
12362 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12364 PR rtl-optimization/81611
12365 * auto-inc-dec.c (attempt_change): Move dead note from
12366 mem_insn if it's the next use of regno
12367 (find_address): Take address use of reg holding
12368 non-incremented value. Add parm to limit search to the named
12370 (merge_in_block): Attempt to use a mem insn that is the next
12371 use of the original regno.
12373 2018-02-27 Martin Sebor <msebor@redhat.com>
12376 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12377 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12379 2018-02-27 Martin Sebor <msebor@redhat.com>
12381 PR translation/84207
12382 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12383 n argument to unsigned HOST_WIDE_INT.
12384 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12385 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12386 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12387 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12389 2018-02-27 Richard Biener <rguenther@suse.de>
12391 PR tree-optimization/84512
12392 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12393 Do not use the estimate returned from record_stmt_cost for
12394 the scalar iteration cost but sum properly using add_stmt_cost.
12396 2018-02-27 Richard Biener <rguenther@suse.de>
12398 PR tree-optimization/84466
12399 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12400 Adjust last change to less strictly validate use operands.
12402 2018-02-27 Martin Liska <mliska@suse.cz>
12404 PR gcov-profile/84548
12405 * gcov.c (process_file): Allow partial overlap and consider it
12406 also as group functions.
12407 (output_lines): Properly calculate range of lines for a group.
12409 2018-02-27 Martin Liska <mliska@suse.cz>
12411 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12412 'ggc' suffixes. Change first column width.
12413 (timer::print): Fix formatting of the column.
12415 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12417 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12418 preserve inline entry blocks for the sake of debug inline
12419 entry point markers alone.
12420 (remove_unused_locals): Suggest in comments a better place to
12421 force the preservation of inline entry blocks that are
12422 otherwise unused, but do not preserve them.
12424 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12426 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12428 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12431 * config/i386/constraints.md (Bs): Replace
12432 ix86_indirect_branch_register with
12433 TARGET_INDIRECT_BRANCH_REGISTER.
12435 * config/i386/i386.md (indirect_jump): Likewise.
12436 (tablejump): Likewise.
12437 (*sibcall_memory): Likewise.
12438 (*sibcall_value_memory): Likewise.
12439 Peepholes of indirect call and jump via memory: Likewise.
12440 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12441 (*sibcall_value_GOT_32): Likewise.
12442 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12443 (GOT_memory_operand): Likewise.
12444 (call_insn_operand): Likewise.
12445 (sibcall_insn_operand): Likewise.
12446 (GOT32_symbol_operand): Likewise.
12447 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12449 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12451 PR rtl-optimization/83496
12452 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12453 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12454 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12455 redundant insn, if any.
12456 (relax_delay_slots): Likewise.
12457 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12459 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12461 PR tree-optimization/83965
12462 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12463 that grouped statements are part of a reduction chain. Return
12464 true if the statement is not marked as a reduction itself but
12465 is part of a group.
12466 (vect_recog_dot_prod_pattern): Don't check whether the statement
12467 is part of a group here.
12468 (vect_recog_sad_pattern): Likewise.
12469 (vect_recog_widen_sum_pattern): Likewise.
12471 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12474 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12476 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12479 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12481 (ix86_output_indirect_function_return): New prototype.
12482 (ix86_split_simple_return_pop_internal): Likewise.
12483 * config/i386/i386.c (indirect_return_via_cx): New.
12484 (indirect_return_via_cx_bnd): Likewise.
12485 (indirect_thunk_name): Handle return va CX_REG.
12486 (output_indirect_thunk_function): Create alias for
12487 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12488 (ix86_output_indirect_jmp): Remove the bool argument.
12489 (ix86_output_indirect_function_return): New function.
12490 (ix86_split_simple_return_pop_internal): Likewise.
12491 * config/i386/i386.md (*indirect_jump): Don't pass false
12492 to ix86_output_indirect_jmp.
12493 (*tablejump_1): Likewise.
12494 (simple_return_pop_internal): Change it to define_insn_and_split.
12495 Call ix86_split_simple_return_pop_internal to split it for
12496 -mfunction-return=.
12497 (simple_return_indirect_internal): Call
12498 ix86_output_indirect_function_return instead of
12499 ix86_output_indirect_jmp.
12501 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12504 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12505 memset and value initialization afterwards.
12507 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12509 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12511 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12514 * common/config/aarch64/aarch64-common.c
12515 (aarch_option_optimization_table[]): Switch
12516 off fomit-frame-pointer
12518 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12519 Chung-Ju Wu <jasonwucj@gmail.com>
12521 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12523 (store_multiple): Ditto.
12525 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12527 * config.gcc: Add --with-cpu support for nds32 target.
12528 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12529 * config/nds32/nds32.opt: Add -mcpu= option.
12531 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12533 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12534 isel=yes): Warn for these deprecated options.
12536 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12538 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12539 ISA_2_5_MASKS_EMBEDDED.
12541 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12543 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12544 p->max as pointers rather than using iterative_hash_expr.
12546 2018-02-23 Carl Love <cel@us.ibm.com>
12548 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12549 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12551 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12552 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12553 P8V_BUILTIN_VEC_VUNSIGNED2.
12555 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12558 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12559 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12561 * lra-constraints.c (curr_insn_transform): Set up
12562 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12564 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12565 * lra-eliminations.c (spill_pseudos): Ditto.
12566 (process_insn_for_elimination): Ditto.
12567 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12568 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12569 LRA_NON_CLOBBERED_ALT.
12571 2018-02-22 Martin Sebor <msebor@redhat.com>
12573 PR tree-optimization/84480
12574 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12575 to maybe_diag_stxncpy_trunc. Call it.
12576 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12577 from gimple_fold_builtin_strcpy. Print inlining stack.
12578 (handle_builtin_stxncpy): Print inlining stack.
12579 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12581 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12584 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12585 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12586 and -fcheck-pointer-bounds are used together.
12587 (indirect_thunk_prefix): New enum.
12588 (indirect_thunk_need_prefix): New function.
12589 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12590 "_nt" instead of "_bnd" for NOTRACK prefix.
12591 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12592 (output_indirect_thunk_function): Likewise.
12594 (ix86_code_end): Update output_indirect_thunk_function calls.
12595 (ix86_output_indirect_branch_via_reg): Replace
12596 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12597 (ix86_output_indirect_branch_via_push): Likewise.
12598 (ix86_output_function_return): Likewise.
12599 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12600 incompatible with -fcf-protection=branch and
12601 -fcheck-pointer-bounds.
12603 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12606 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12607 Change gcc_assert call to output_operand_lossage.
12609 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12611 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12613 2018-02-22 DJ Delorie <dj@redhat.com>
12614 Sebastian Perta <sebastian.perta@renesas.com>
12615 Oleg Endo <olegendo@gcc.gnu.org>
12617 * config/rx/rx.c (rx_rtx_costs): New function.
12618 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12620 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12622 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12624 2018-02-22 Martin Liska <mliska@suse.cz>
12627 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12628 Add "native" as a possible value.
12630 2018-02-22 Martin Liska <mliska@suse.cz>
12633 * config/i386/i386.c (ix86_option_override_internal):
12634 Add "native" as a possible value for -march and -mtune.
12636 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12639 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12640 to all type variants.
12642 PR tree-optimization/84503
12643 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12644 width as info->bitpos + info->bitsize - start.
12645 (merged_store_group::merge_overlapping): Simplify width computation.
12646 (check_no_overlap): New function.
12647 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12648 start + width and last_order of the group, fail if check_no_overlap
12650 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12651 to group if check_no_overlap fails.
12653 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12655 * config/rs6000/altivec.md: Delete contraint arguments to
12656 define_expand, define_split, and define_peephole2, and in
12657 define_insn_and_split if always unused.
12658 * config/rs6000/darwin.md: Ditto.
12659 * config/rs6000/dfp.md: Ditto.
12660 * config/rs6000/rs6000.md: Ditto.
12661 * config/rs6000/sync.md: Ditto.
12662 * config/rs6000/vector.md: Ditto.
12663 * config/rs6000/vsx.md: Ditto.
12665 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12667 * config/rs6000/altivec.md: Write output control strings as braced
12668 blocks instead of double-quoted strings.
12669 * config/rs6000/darwin.md: Ditto.
12670 * config/rs6000/rs6000.md: Ditto.
12671 * config/rs6000/vector.md: Ditto.
12672 * config/rs6000/vsx.md: Ditto.
12674 2018-02-21 Jason Merrill <jason@redhat.com>
12676 PR c++/84314 - ICE with templates and fastcall attribute.
12677 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12679 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12681 * ipa-cp.c (determine_versionability): Fix comment typos.
12683 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12686 * ipa-cp.c (determine_versionability): Do not version functions caling
12689 2018-02-21 Martin Liska <mliska@suse.cz>
12692 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12693 Add "native" as a possible value.
12694 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12695 the macro when native cpu detection is available.
12697 2018-02-21 Martin Liska <mliska@suse.cz>
12700 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12701 Add "native" as a possible value.
12702 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12703 when native cpu detection is available.
12705 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12706 Martin Sebor <msebor@redhat.com>
12708 PR tree-optimization/84478
12709 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12711 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12712 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12713 support which is conservatively correct, for 2 only stay conservative
12714 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12715 argument to the 2 argument get_range_strlen, adjust 6 arg
12716 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12718 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12719 (gimple_fold_builtin_strlen): Pass true as last argument to
12722 2018-02-20 Martin Sebor <msebor@redhat.com>
12724 PR middle-end/84095
12725 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12726 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12727 (builtin_memref::builtin_memref): Factor out parts into
12728 set_base_and_offset and call it.
12730 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12732 PR middle-end/84406
12733 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12734 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12735 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12736 search at the associated MODE_INT.
12738 2018-02-20 Jeff Law <law@redhat.com>
12740 PR middle-end/82123
12741 PR tree-optimization/81592
12742 PR middle-end/79257
12743 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12744 for range data rather than using global data.
12745 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12746 range data rather than using global data.
12747 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12748 pass it to children as needed.
12749 (struct directive::fmtresult): Similarly.
12750 (struct directive::set_width): Similarly.
12751 (struct directive::set_precision): Similarly.
12752 (format_integer, format_directive, parse_directive): Similarly.
12753 (format_none): Accept unnamed vr_values parameter.
12754 (format_percent, format_floating, format_character): Similarly.
12755 (format_string, format_plain): Similarly.
12756 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12757 the EVRP range analyzer for range data rather than using global data.
12758 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12759 gimple-ssa-evrp-analyze.h
12760 (class sprintf_dom_walker): Add after_dom_children member function.
12761 Add evrp_range_analyzer member.
12762 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12763 range analyzer as needed.
12764 (sprintf_dom_walker::after_dom_children): New member function.
12765 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12767 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12768 (evrp_range_analyzer::pop_to_marker): Likewise.
12770 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12772 PR tree-optimization/84419
12773 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12774 with the required type if its current type is compatible but
12777 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12779 PR middle-end/82004
12780 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12781 after vectorization.
12783 2018-02-20 Martin Liska <mliska@suse.cz>
12786 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12787 possible values if we don't have a hint.
12789 2018-02-20 Martin Liska <mliska@suse.cz>
12793 * final.c (shorten_branches): Build align_tab array with one
12795 * opts.c (finish_options): Add alignment option limit check.
12796 (MAX_CODE_ALIGN): Likewise.
12797 (MAX_CODE_ALIGN_VALUE): Likewise.
12798 * doc/invoke.texi: Document maximum allowed option value for
12799 all -falign-* options.
12801 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12804 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12805 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12806 * var-tracking.c (emit_note_insn_var_location): Remove all references
12807 to NOTE_INSN_CALL_ARG_LOCATION.
12808 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12809 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12810 Use copy_rtx_if_shared.
12811 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12812 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12813 (dwarf2out_var_location): Remove handling of
12814 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12816 * final.c (final_scan_insn): Remove all references to
12817 NOTE_INSN_CALL_ARG_LOCATION.
12818 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12819 before dumping final insns.
12820 * except.c (emit_note_eh_region_end): Remove all references to
12821 NOTE_INSN_CALL_ARG_LOCATION.
12822 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12823 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12824 * config/arc/arc.c (hwloop_optimize): Likewise.
12825 * config/arm/arm.c (create_fix_barrier): Likewise.
12826 * config/s390/s390.c (s390_chunkify_start): Likewise.
12827 * config/sh/sh.c (find_barrier): Likewise.
12828 * config/i386/i386.c (rest_of_insert_endbranch,
12829 ix86_seh_fixup_eh_fallthru): Likewise.
12830 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12831 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12832 * config/frv/frv.c (frv_function_prologue): Likewise.
12833 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12835 (note_outside_basic_block_p): Remove all references to
12836 NOTE_INSN_CALL_ARG_LOCATION.
12837 * gengtype.c (adjust_field_rtx_def): Likewise.
12838 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12840 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12841 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12844 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12847 PR tree-optimization/84452
12848 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12849 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12852 2018-02-19 Martin Liska <mliska@suse.cz>
12855 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12857 2018-02-19 Martin Liska <mliska@suse.cz>
12858 Richard Sandiford <richard.sandiford@linaro.org>
12860 PR tree-optimization/82491
12861 * gimple-fold.c (get_base_constructor): Make earlier bail out
12864 2018-02-19 Carl Love <cel@us.ibm.com>
12866 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12867 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12869 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12870 P8V_BUILTIN_VEC_NEG.
12872 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12874 * config/rl78/rl78.md (movdf): New define expand.
12876 2018-02-19 Martin Liska <mliska@suse.cz>
12879 * doc/invoke.texi: Fix typo.
12880 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12882 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12884 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12885 handle rs6000_single_float and rs6000_double_float specially for
12888 2018-02-16 Jeff Law <law@redhat.com>
12890 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12891 (pop_regs): Use it.
12893 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12896 * ipa-inline.c (inline_small_functions): Fix a typo.
12898 2018-02-16 Nathan Sidwell <nathan@acm.org>
12900 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12902 2018-02-16 Carl Love <cel@us.ibm.com>
12904 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12905 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12906 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12907 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12908 expansion to P8V_BUILTIN_VEC_FLOAT2.
12910 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12912 PR rtl-optimization/70023
12913 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12914 src_regno into account.
12916 2018-02-16 Carl Love <cel@us.ibm.com>
12918 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12919 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12920 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12921 * config/rs6000/rs6000.c: Remove case statements for
12922 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12923 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12924 and P9V_BUILTIN_VEC_VINSERT4B.
12925 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12926 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12927 * config/rs6000/vsx.md:
12928 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12931 2018-02-16 Carl Love <cel@us.ibm.com>
12933 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12935 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12937 * config/rs6000/rs6000-c.c: Add the definitions for
12938 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12939 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12940 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12941 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12942 definition for insert4b and define insn *insert3b_internal.
12943 * doc/extend.texi: Add documentation for vec_extract4b.
12945 2018-02-16 Nathan Sidwell <nathan@acm.org>
12947 * doc/extend.texi (Backwards Compatibility): Mention friend
12948 injection. Note for-scope is deprecated.
12949 * doc/invoke.texi (-ffriend-injection): Deprecate.
12951 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12953 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12954 that moved to I2, also allow destinations that are a paradoxical
12955 subreg (instead of a normal reg).
12957 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12960 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12963 2018-02-16 Richard Biener <rguenther@suse.de>
12965 PR tree-optimization/84037
12966 PR tree-optimization/84016
12968 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12969 Adjust vec_construct for the fact we need additional higher latency
12970 128bit inserts for AVX256 and AVX512 vector builds.
12971 (ix86_add_stmt_cost): Scale vector construction cost for
12974 2018-02-16 Richard Biener <rguenther@suse.de>
12976 PR tree-optimization/84417
12977 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12978 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12979 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12981 2018-02-16 Martin Liska <mliska@suse.cz>
12984 * internal-fn.def (ASAN_CHECK): Set proper flags.
12985 (ASAN_MARK): Likewise.
12987 2018-02-16 Julia Koval <julia.koval@intel.com>
12989 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12990 from PTA_CANNONLAKE.
12992 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12995 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12996 Use ++iter rather than iter++ for std::list iterators.
12997 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12998 defer deleting them until all nodes in the forest are processed. Do
12999 free even leaf nodes. Change to_process into auto_vec.
13002 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
13003 * vec.h (vec_default_construct): Use memset instead of placement new
13004 if BROKEN_VALUE_INITIALIZATION is defined.
13005 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
13006 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
13009 PR rtl-optimization/83723
13010 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
13011 * lra.c (lra_substitute_pseudo): Likewise. If true, use
13012 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
13014 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
13016 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
13018 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13020 PR rtl-optimization/81443
13021 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13022 from inner REGs to paradoxical SUBREGs.
13024 2018-02-16 Richard Biener <rguenther@suse.de>
13026 PR tree-optimization/84399
13027 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13028 For operands we can analyze at their definition make sure we can
13029 analyze them at each use as well.
13031 2018-02-16 Richard Biener <rguenther@suse.de>
13033 PR tree-optimization/84190
13034 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13035 volatile accesses if the decl isn't volatile.
13037 2018-02-15 Jason Merrill <jason@redhat.com>
13039 PR c++/84314 - ICE with templates and fastcall attribute.
13040 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13041 TYPE_CANONICAL on an existing type.
13043 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13045 PR tree-optimization/84383
13046 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13047 dstoff nor call operand_equal_p if dstbase is NULL.
13049 PR tree-optimization/84334
13050 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13051 also a CONSTANT_CLASS_P, punt.
13053 2018-02-14 Jim Wilson <jimw@sifive.com>
13055 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13056 first SMALL_OPERAND check. New local min_second_step. Move assert
13057 to where locals are set. Add TARGET_RVC support.
13058 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13060 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13062 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13064 2018-02-14 Martin Sebor <msebor@redhat.com>
13066 PR tree-optimization/83698
13067 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13068 arrays constrain the offset range to their bounds.
13069 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13070 (builtin_access::overlap): Avoid setting the size of overlap if it's
13072 (maybe_diag_overlap): Also consider arrays when deciding what values
13073 of offsets to include in diagnostics.
13075 2018-02-14 Martin Sebor <msebor@redhat.com>
13078 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13079 that correspond to the kind of a declaration.
13081 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13084 * config/pa/pa.md: Load address of PIC label using the linkage table
13085 if the label is nonlocal.
13087 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13089 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13090 warning message if user requests -maltivec=be.
13091 * doc/invoke.texi: Document deprecation of -maltivec=be.
13093 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13096 * config/rs6000/rs6000-c.c: Update definitions for
13097 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13098 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13100 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13103 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13104 add _get_ssp intrinsics. Remove argument from
13105 __builtin_ia32_rdssp[d|q].
13106 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13107 * config/i386/i386-builtin.def: Remove argument from
13108 __builtin_ia32_rdssp[d|q].
13109 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13110 ix86_expand_special_args_builtin for _rdssp[d|q].
13111 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13112 Clear register before usage.
13113 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13114 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13116 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13118 PR tree-optimization/84357
13119 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13120 operand 1 of an ARRAY_REF too.
13122 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13125 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13126 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13128 (set_of_reg): New struct.
13129 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13130 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13131 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13133 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13134 Split into bitclr, bitset, bitinvert patterns if appropriate.
13135 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13136 use rx_fuse_in_memory_bitop.
13137 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13138 to named insn, correct maximum insn length.
13140 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13143 * machmode.def: Define a complex mode for PARTIAL_INT.
13144 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13146 * doc/rtl.texi: Document CSPImode.
13147 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13149 (msp430_hard_regno_nregs_with_padding): Likewise.
13151 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13154 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13156 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13158 PR rtl-optimization/84169
13159 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13160 we generated a parallel as new i3 and we split that to new i2 and i3
13161 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13162 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13163 those to i2, not i1. Partially rewrite this scan code.
13165 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13168 * stor-layout.c (place_field): For variable length fields, adjust
13169 offset_align afterwards not just based on the field's alignment,
13170 but also on the size.
13172 PR middle-end/84309
13173 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13174 of exps and logs in the use_exp2 case.
13176 2018-02-13 Jeff Law <law@redhat.com>
13178 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13179 entry for "vector".
13181 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13184 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13188 * common.opt (gas-loc-support, gas-locview-support): New.
13189 (ginline-points, ginternal-reset-location-views): New.
13190 * doc/invoke.texi: Document them. Use @itemx where intended.
13191 (gvariable-location-views): Adjust.
13192 * target.def (reset_location_view): New.
13193 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13194 (TARGET_RESET_LOCATION_VIEW): New.
13195 * doc/tm.texi: Rebuilt.
13196 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13197 (dwarf2out_default_as_locview_support): New.
13198 (output_asm_line_debug_info): Use option variables.
13199 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13200 (output_loc_list): Likewise.
13201 (add_high_low_attributes): Check option variables.
13202 Don't output entry view attribute in strict mode.
13203 (gen_inlined_subroutine_die): Check option variables.
13204 (dwarf2out_inline_entry): Likewise.
13205 (init_sections_and_labels): Likewise.
13206 (dwarf2out_early_finish): Likewise.
13207 (maybe_reset_location_view): New, from...
13208 (dwarf2out_var_location): ... here. Call it.
13209 * debug.h (dwarf2out_default_as_loc_support): Declare.
13210 (dwarf2out_default_as_locview_support): Declare.
13211 * hooks.c (hook_int_rtx_insn_0): New.
13212 * hooks.h (hook_int_rtx_insn_0): Declare.
13213 * toplev.c (process_options): Take -gas-loc-support and
13214 -gas-locview-support from dwarf2out. Enable
13215 -gvariable-location-views by default only with locview
13216 assembler support. Enable -ginternal-reset-location-views by
13217 default only if the target defines the corresponding hook.
13218 Enable -ginline-points by default if location views are
13219 enabled; force it disabled if statement frontiers are
13221 * tree-inline.c (expand_call_inline): Check option variables.
13222 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13224 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13226 PR tree-optimization/84321
13227 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13228 handling. Also check whether the anti-range contains any values
13229 that satisfy the mask; switch to a VR_RANGE if not.
13231 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13234 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13236 2018-02-13 Martin Jambor <mjambor@suse.cz>
13239 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13240 of call statements, also set location of a load to a temporary.
13242 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13244 * config/rl78/rl78.c (add_vector_labels): New function.
13245 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13246 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13247 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13248 which checks that no arguments are passed.
13249 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13250 * doc/extend.texi: Documentation for the new attribute.
13252 2018-02-13 Andreas Schwab <schwab@suse.de>
13254 * config/riscv/linux.h (CPP_SPEC): Define.
13256 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13259 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13260 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13261 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13262 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13263 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13264 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13265 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13266 temporarily for AES and PCLMUL builtins.
13268 PR tree-optimization/84339
13269 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13270 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13273 PR middle-end/84309
13274 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13275 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13276 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13278 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13280 * omp-simd-clone.h: New file.
13281 * omp-simd-clone.c: Include omp-simd-clone.h.
13282 (expand_simd_clones): No longer static.
13283 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13284 cgraph.h and omp-simd-clone.h.
13285 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13286 (vect_recog_widen_shift_pattern): Formatting fix.
13287 (vect_pattern_recog_1): Don't check optab for calls.
13290 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13291 operands[2] into a REG before using gen_lowpart on it.
13293 2018-02-12 Jeff Law <law@redhat.com>
13296 * config/sh/sh.c (find_barrier): Consider a sibling call
13299 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13300 successfully back substituting a reg.
13302 2018-02-12 Richard Biener <rguenther@suse.de>
13304 PR tree-optimization/84037
13305 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13306 parameter, move visited init to caller.
13307 (vect_slp_analyze_operations): Separate cost from validity
13308 check, initialize visited once for all instances.
13309 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13311 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13313 (vect_model_promotion_demotion_cost): Likewise.
13314 (vectorizable_bswap): Guard cost modeling with !slp_node
13315 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13317 (vectorizable_call): Likewise.
13318 (vectorizable_conversion): Likewise.
13319 (vectorizable_assignment): Likewise.
13320 (vectorizable_shift): Likewise.
13321 (vectorizable_operation): Likewise.
13322 (vectorizable_store): Likewise.
13323 (vectorizable_load): Likewise.
13324 (vectorizable_condition): Likewise.
13325 (vectorizable_comparison): Likewise.
13327 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13330 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13331 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13332 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13334 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13336 PR middle-end/83665
13337 * params.def (inline-min-speedup): Increase from 8 to 15.
13338 (max-inline-insns-auto): Decrease from 40 to 30.
13339 * ipa-split.c (consider_split): Add some buffer for function to
13340 be considered inlining candidate.
13341 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13344 2018-02-12 Richard Biener <rguenther@suse.de>
13346 PR tree-optimization/84037
13347 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13348 matched stmts if we cannot swap the non-matched ones.
13350 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13352 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13353 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13354 _mm_maskz_scalef_round_ss): New intrinsics.
13355 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13356 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13357 __builtin_ia32_scalefss_round): Remove.
13358 (__builtin_ia32_scalefsd_mask_round,
13359 __builtin_ia32_scalefss_mask_round): New intrinsics.
13360 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13361 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13362 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13363 "<round_constraint>")): Changed to ...
13364 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13365 "<round_scalar_constraint>")): ... this.
13366 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13367 %0, %1, %2<round_op3>}"): Changed to ...
13368 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13369 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13370 %2<round_scalar_mask_op3>}"): ... this.
13371 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13373 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13375 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13376 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13377 (_mm_maskz_sqrt_round_ss): New intrinsics.
13378 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13379 (__builtin_ia32_sqrtsd_mask_round)
13380 (__builtin_ia32_sqrtss_mask_round): New builtins.
13381 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13382 (__builtin_ia32_sqrtss_round): Remove.
13383 (__builtin_ia32_sqrtsd_mask_round)
13384 (__builtin_ia32_sqrtss_mask_round): New builtins.
13385 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13386 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13387 ((match_operand:VF_128 1 "vector_operand"
13388 "xBm,<round_constraint>")): Changed to ...
13389 ((match_operand:VF_128 1 "vector_operand"
13390 "xBm,<round_scalar_constraint>")): ... this.
13391 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13392 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13393 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13394 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13395 %<iptr>1<round_scalar_mask_op3>}): ... this.
13396 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13397 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13399 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13402 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13403 Cast vec_cmpeq result to correct type.
13404 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13405 Cast vec_cmpgt result to correct type.
13407 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13409 * final.c (final_scan_insn_1): Renamed from...
13410 (final_scan_insn): ... this. New wrapper, to recover
13411 seen from the outermost call in recursive ones.
13412 * config/sparc/sparc.c (output_return): Drop seen from call.
13413 (output_sibcall): Likewise.
13414 * config/visium/visium.c (output_branch): Likewise.
13416 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13418 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13421 2018-02-10 Alan Modra <amodra@gmail.com>
13424 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13425 Specify LR as an input.
13427 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13430 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13431 remove_member_access_dummy_vars): New functions.
13432 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13433 lower_omp_1, execute_lower_omp): Use them.
13435 PR rtl-optimization/84308
13436 * shrink-wrap.c (spread_components): Release todo vector.
13438 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13440 PR rtl-optimization/57193
13441 * ira-color.c (struct allocno_color_data): Add member
13442 conflict_allocno_hard_prefs.
13443 (update_conflict_allocno_hard_prefs): New.
13444 (bucket_allocno_compare_func): Add a preference based on
13445 conflict_allocno_hard_prefs.
13446 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13447 (color_allocnos): Remove a dead code. Initiate
13448 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13450 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13453 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13454 constraint from =wa to wa. Avoid a subreg on the output operand,
13455 instead use a pseudo and subreg it in a move.
13456 (p9_xxbrd_<mode>): Changed to ...
13457 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13458 (p9_xxbrd_v2df): New expander.
13459 (p9_xxbrw_<mode>): Changed to ...
13460 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13461 (p9_xxbrw_v4sf): New expander.
13463 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13465 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13467 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13470 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13471 multiply in 32-bit mode.
13472 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13473 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13476 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13478 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13479 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13480 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13481 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13483 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13486 * dwarf2out.c (is_trivial_indirect_ref): New function.
13487 (dwarf2out_late_global_decl): Do not generate a location
13488 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13489 and that are not defined in the current unit.
13491 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13493 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13494 instead of a libcall for UNORDERED.
13496 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13499 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13500 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13502 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13505 * config/s390/s390.c (s390_set_current_function): Invoke
13506 s390_indirect_branch_settings also if fndecl didn't change.
13508 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13510 * config/rs6000/rs6000.md (blockage): Set length to zero.
13512 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13514 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13516 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13519 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13520 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13524 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13525 PARALLEL incoming that failed vt_get_decl_and_offset check.
13527 PR middle-end/84237
13528 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13529 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13531 (get_variable_section): For decls in named .bss* sections pass true as
13532 second argument to bss_initializer_p.
13534 2018-02-09 Marek Polacek <polacek@redhat.com>
13535 Jakub Jelinek <jakub@redhat.com>
13538 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13539 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13540 Sync some changes from cxx_fold_indirect_ref.
13542 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13544 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13546 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13547 (BLOCK_INLINE_ENTRY_LABEL): New.
13548 (dwarf2out_var_location): Disregard inline entry markers.
13549 (inline_entry_data): New struct.
13550 (inline_entry_data_hasher): New hashtable type.
13551 (inline_entry_data_hasher::hash): New.
13552 (inline_entry_data_hasher::equal): New.
13553 (inline_entry_data_table): New variable.
13554 (add_high_low_attributes): Add DW_AT_entry_pc and
13555 DW_AT_GNU_entry_view attributes if a pending entry is found
13556 in inline_entry_data_table. Add old entry_pc attribute only
13557 if debug nonbinding markers are disabled.
13558 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13559 markers are enabled.
13560 (block_within_block_p, dwarf2out_inline_entry): New.
13561 (dwarf2out_finish): Check that no entries remained in
13562 inline_entry_data_table.
13563 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13564 (final_scan_insn, notice_source_line): Likewise.
13565 (rest_of_clean_state): Skip inline entry markers.
13566 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13568 * gimple.c (gimple_build_debug_inline_entry): New.
13569 * gimple.h (enum gimple_debug_subcode): Add
13570 GIMPLE_DEBUG_INLINE_ENTRY.
13571 (gimple_build_debug_inline_entry): Declare.
13572 (gimple_debug_inline_entry_p): New.
13573 (gimple_debug_nonbind_marker_p): Adjust.
13574 * insn-notes.def (INLINE_ENTRY): New.
13575 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13576 inline entry marker notes.
13577 (print_insn): Likewise.
13578 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13579 (INSN_DEBUG_MARKER_KIND): Likewise.
13580 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13581 * tree-inline.c (expand_call_inline): Build and insert
13582 debug_inline_entry stmt.
13583 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13584 inline entry blocks early, if nonbind markers are enabled.
13585 (dump_scope_block): Dump fragment info.
13586 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13587 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13588 (gimple_build_debug_inline_entry): New.
13589 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13590 Enable/disable inline entry points too.
13591 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13592 (DEBUG_INSN): Describe inline entry markers.
13594 * common.opt (gvariable-location-views): New.
13595 (gvariable-location-views=incompat5): New.
13596 * config.in: Rebuilt.
13597 * configure: Rebuilt.
13598 * configure.ac: Test assembler for view support.
13599 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13600 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13601 * dwarf2out.c (var_loc_view): New typedef.
13602 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13603 (dwarf2out_locviews_in_attribute): New.
13604 (dwarf2out_locviews_in_loclist): New.
13605 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13606 (enum dw_line_info_opcode): Add LI_adv_address.
13607 (struct dw_line_info_table): Add view.
13608 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13609 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13610 (zero_view_p): New variable.
13611 (ZERO_VIEW_P): New macro.
13612 (output_asm_line_debug_info): New.
13613 (struct var_loc_node): Add view.
13614 (add_AT_view_list, AT_loc_list): New.
13615 (add_var_loc_to_decl): Add view param. Test it against last.
13616 (new_loc_list): Add view params. Record them.
13617 (AT_loc_list_ptr): Handle loc and view lists.
13618 (view_list_to_loc_list_val_node): New.
13619 (print_dw_val): Handle dw_val_class_view_list.
13620 (size_of_die): Likewise.
13621 (value_format): Likewise.
13622 (loc_list_has_views): New.
13623 (gen_llsym): Set vl_symbol too.
13624 (maybe_gen_llsym, skip_loc_list_entry): New.
13625 (dwarf2out_maybe_output_loclist_view_pair): New.
13626 (output_loc_list): Output view list or entries too.
13627 (output_view_list_offset): New.
13628 (output_die): Handle dw_val_class_view_list.
13629 (output_dwarf_version): New.
13630 (output_compilation_unit_header): Use it.
13631 (output_skeleton_debug_sections): Likewise.
13632 (output_rnglists, output_line_info): Likewise.
13633 (output_pubnames, output_aranges): Update version comments.
13634 (output_one_line_info_table): Output view numbers in asm comments.
13635 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13636 Call maybe_gen_llsym.
13637 (loc_list_from_tree_1): Adjust.
13638 (add_AT_location_description): Create view list attribute if
13639 needed, check it's absent otherwise.
13640 (convert_cfa_to_fb_loc_list): Adjust.
13641 (maybe_emit_file): Call output_asm_line_debug_info for test.
13642 (dwarf2out_var_location): Reset views as needed. Precompute
13643 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13644 attribute. Set view.
13645 (new_line_info_table): Reset next view.
13646 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13647 (dwarf2out_source_line): Likewise. Output view resets and labels to
13648 the assembler, or select appropriate line info opcodes.
13649 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13650 (optimize_string_length): Catch it. Adjust.
13651 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13652 dw_val_class_view_list, and remove it if no longer needed.
13653 (hash_loc_list): Hash view numbers.
13654 (loc_list_hasher::equal): Compare them.
13655 (optimize_location_lists): Check whether a view list symbol is
13656 needed, and whether the locview attribute is present, and
13657 whether they match. Remove the locview attribute if no longer
13659 (index_location_lists): Call skip_loc_list_entry for test.
13660 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13661 Use output_dwarf_version.
13662 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13663 (struct dw_val_node): Add val_view_list.
13664 * final.c (SEEN_NEXT_VIEW): New.
13665 (set_next_view_needed): New.
13666 (clear_next_view_needed): New.
13667 (maybe_output_next_view): New.
13668 (final_start_function): Rename to...
13669 (final_start_function_1): ... this. Take pointer to FIRST,
13670 add SEEN parameter. Emit param bindings in the initial view.
13671 (final_start_function): Reintroduce SEEN-less interface.
13672 (final): Rename to...
13673 (final_1): ... this. Take SEEN parameter. Output final pending
13674 next view at the end.
13675 (final): Reintroduce seen-less interface.
13676 (final_scan_insn): Output pending next view before switching
13677 sections or ending a block. Mark the next view as needed when
13678 outputting variable locations. Notify debug backend of section
13679 changes, and of location view changes.
13680 (rest_of_handle_final): Adjust.
13681 * toplev.c (process_options): Autodetect value for debug variable
13682 location views option. Warn on incompat5 without -gdwarf-5.
13683 * doc/invoke.texi (gvariable-location-views): New.
13684 (gvariable-location-views=incompat5): New.
13685 (gno-variable-location-views): New.
13687 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13689 PR tree-optimization/84136
13690 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13691 that the result of find_edge is non-NULL.
13693 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13696 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13697 storing integer register in SImode. Fix cost of 256 and 512
13698 byte aligned SSE register store.
13700 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13702 * config/i386/i386.c (ix86_multiplication_cost): Fix
13703 multiplication cost for TARGET_AVX512DQ.
13705 2018-02-08 Marek Polacek <polacek@redhat.com>
13707 PR tree-optimization/84238
13708 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13711 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13713 PR tree-optimization/84265
13714 * tree-vect-stmts.c (vectorizable_store): Don't treat
13715 VMAT_CONTIGUOUS accesses as grouped.
13716 (vectorizable_load): Likewise.
13718 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13720 PR tree-optimization/81635
13721 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13722 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13723 (test_round_for_mask): New functions.
13724 (wide_int_cc_tests): Call test_round_for_mask.
13725 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13726 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13727 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13728 range returned by get_range_info.
13730 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13733 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13734 * symtab.c: Include builtins.h
13735 (symtab_node::output_to_lto_symbol_table_p): Move here
13736 from lto-streamer-out.c:output_symbol_p.
13737 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13738 (output_symbol_p): Move all logic to symtab.c
13739 (produce_symtab): Update.
13741 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13743 * config/s390/s390-opts.h (enum indirect_branch): Define.
13744 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13745 (s390_indirect_branch_via_thunk)
13746 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13747 (enum s390_indirect_branch_type): Define.
13748 * config/s390/s390.c (struct s390_frame_layout, struct
13749 machine_function): Remove.
13750 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13751 (indirect_branch_table_label_no, indirect_branch_table_name):
13753 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13754 (enum s390_indirect_branch_option): Define.
13755 (s390_return_addr_from_memory): New function.
13756 (s390_handle_string_attribute): New function.
13757 (s390_attribute_table): Add new attribute handler.
13758 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13759 (s390_indirect_branch_via_thunk): New function.
13760 (s390_indirect_branch_via_inline_thunk): New function.
13761 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13762 sibling call optimization for non z10 compiles.
13763 (s390_emit_call): Force indirect branch target to be a single
13764 register. Add r1 clobber for non-z10 compiles.
13765 (s390_emit_epilogue): Emit return jump via return_use expander.
13766 (s390_reorg): Handle JUMP_INSNs as execute targets.
13767 (s390_option_override_internal): Perform validity checks for the
13768 new command line options.
13769 (s390_indirect_branch_attrvalue): New function.
13770 (s390_indirect_branch_settings): New function.
13771 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13772 (s390_output_indirect_thunk_function): New function.
13773 (s390_code_end): Implement target hook.
13774 (s390_case_values_threshold): Implement target hook.
13775 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13777 * config/s390/s390.h (struct s390_frame_layout)
13778 (struct machine_function): Move here from s390.c.
13779 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13780 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13781 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13782 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13783 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13784 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13785 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13786 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13787 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13788 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13789 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13790 (mnemonic attribute): Add values which aren't recognized
13792 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13793 pattern for branch conversion. Fix mnemonic attribute.
13794 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13795 indirect branch via thunk if requested.
13796 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13797 ("*indirect_jump"): Disable for branch conversion using out of
13799 ("indirect_jump_via_thunk<mode>_z10")
13800 ("indirect_jump_via_thunk<mode>")
13801 ("indirect_jump_via_inlinethunk<mode>_z10")
13802 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13803 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13804 ("casesi_jump_via_inlinethunk<mode>_z10")
13805 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13806 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13807 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13808 ("*indirect2_jump"): Disable for branch conversion.
13809 ("casesi_jump"): Turn into expander and expand patterns for branch
13811 ("return_use"): New expander.
13812 ("*return"): Emit return via thunk and rename it to ...
13813 ("*return<mode>"): ... this one.
13814 * config/s390/s390.opt: Add new options and and enum for the
13817 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13819 * lra-constraints.c (match_reload): Unconditionally use
13820 gen_lowpart_SUBREG, rather than selecting between that
13821 and equivalent gen_rtx_SUBREG code.
13823 2018-02-08 Richard Biener <rguenther@suse.de>
13825 PR tree-optimization/84233
13826 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13827 changed flag instead of boguously re-using phi_inserted.
13829 2018-02-08 Martin Jambor <mjambor@suse.cz>
13831 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13832 static local variables.
13834 2018-02-08 Richard Biener <rguenther@suse.de>
13836 PR tree-optimization/84278
13837 * tree-vect-stmts.c (vectorizable_store): When looking for
13838 smaller vector types to perform grouped strided loads/stores
13839 make sure the mode is supported by the target.
13840 (vectorizable_load): Likewise.
13842 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13844 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13845 Increase LDP/STP opportunities by adding adjacent callee-saves.
13847 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13849 PR rtl-optimization/84068
13850 PR rtl-optimization/83459
13851 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13853 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13855 PR tree-optimization/84224
13856 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13857 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13858 non-zero arguments.
13860 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13863 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13864 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13865 count, remove one USE.
13867 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13869 * doc/install.texi (Configuration): Document the
13870 --with-long-double-format={ibm,ieee} PowerPC configuration
13874 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13875 Convert from define_expand to be define_insn_and_split. Rework
13876 float/double/_Float128 conversions to QI/HI/SImode to work with
13877 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13878 conversions to QI/HImode types did a store and then a load to
13879 truncate the value. For conversions to VSX registers, don't split
13880 the insn, instead emit the code directly. Use the code iterator
13881 any_fix to combine signed and unsigned conversions.
13882 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13883 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13884 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13885 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13886 (fix_<mode>di2_hw): Likewise.
13887 (fixuns_<mode>di2_hw): Likewise.
13888 (fix_<mode>si2_hw): Likewise.
13889 (fixuns_<mode>si2_hw): Likewise.
13890 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13891 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13892 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13893 fix<uns>_trunc<SFDF:mode>si2_p8.
13894 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13895 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13896 (fix<uns>_<mode>_mem): Likewise.
13897 (fctiw<u>z_<mode>_mem): Likewise.
13898 (fix<uns>_<mode>_mem): Likewise.
13899 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13900 the register allocator from doing a direct move to the GPRs to do
13901 a store, and instead use the ISA 3.0 store byte/half-word from
13902 vector register instruction. For IEEE 128-bit floating point,
13903 also optimize stores of 32-bit ints.
13904 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13906 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13908 * genextract.c (push_pathstr_operand): New function to support
13910 (walk_rtx): Call push_pathstr_operand.
13911 (print_path): Support [a-zA-Z].
13913 2018-02-07 Richard Biener <rguenther@suse.de>
13915 PR tree-optimization/84037
13916 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13917 (cse_and_gimplify_to_preheader): Declare.
13918 (vect_get_place_in_interleaving_chain): Likewise.
13919 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13921 (_loop_vec_info::~_loop_vec_info): Delete it.
13922 (cse_and_gimplify_to_preheader): New function.
13923 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13924 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13925 (vectorizable_load): Likewise. For grouped stores always base
13926 the IV on the first element.
13927 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13928 condition before gimplifying.
13930 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13932 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13933 *DIV_EXPR and *MOD_EXPR.
13935 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13938 * config/i386/i386.c (ix86_option_override_internal): Mask out
13939 the CF_SET bit when checking -fcf-protection.
13941 2018-02-07 Tom de Vries <tom@codesourcery.com>
13944 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13947 2018-02-07 Richard Biener <rguenther@suse.de>
13949 PR tree-optimization/84204
13950 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13953 PR tree-optimization/84205
13954 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13955 special-case isl_ast_op_zdiv_r.
13957 PR tree-optimization/84223
13958 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13959 Only add conditions from within the region.
13960 (gather_bbs::after_dom_children): Adjust.
13962 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13965 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13966 * config/avr/avr.md: Only post-reload split REG-REG moves if
13967 either register is GENERAL_REG_P.
13969 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13971 PR tree-optimization/84235
13972 * tree-ssa-scopedtables.c
13973 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13974 if the subtraction is performed in floating point type where NaNs are
13975 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13976 build 1. Formatting fix.
13978 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13981 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13982 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13983 and skip it regardless of bb boundaries. Use CALL_P macro,
13984 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13985 unnecessarily, formatting fix.
13987 2018-02-06 Michael Collison <michael.collison@arm.com>
13989 * config/arm/thumb2.md:
13990 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13991 (*thumb_mov_notscc): Ditto.
13993 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13996 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13997 unsigned_fix, not "s".
13999 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14001 * configure.ac (gcc_fn_eh_frame_ro): New function.
14002 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
14003 correct .eh_frame permissions.
14004 * configure: Regenerate.
14006 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
14008 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
14009 irrelevant options.
14011 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14013 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14014 Display warning message for -mno-speculate-indirect-jumps.
14016 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
14018 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14020 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14022 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14024 PR tree-optimization/84225
14025 * tree-eh.c (find_trapping_overflow): Only call
14026 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14028 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14031 * config/i386/i386.c: Reimplement the check of possible options
14032 -mibt/-mshstk conbination. Change error messages.
14033 * doc/invoke.texi: Fix a typo: remove extra '='.
14035 2018-02-06 Marek Polacek <polacek@redhat.com>
14037 PR tree-optimization/84228
14038 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14040 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14043 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14044 emitted arch directives.
14045 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14046 __ARM_FEATURE_COPROC before changing architectures.
14048 2018-02-06 Richard Biener <rguenther@suse.de>
14050 * config/i386/i386.c (print_reg): Fix typo.
14051 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14053 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14055 * configure: Regenerate.
14057 2018-02-05 Martin Sebor <msebor@redhat.com>
14059 PR tree-optimization/83369
14060 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14063 2018-02-05 Martin Liska <mliska@suse.cz>
14065 * doc/invoke.texi: Cherry-pick upstream r323995.
14067 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14069 * ira.c (ira_init_register_move_cost): Adjust comment.
14071 2018-02-05 Martin Liska <mliska@suse.cz>
14073 PR gcov-profile/84137
14074 * doc/gcov.texi: Fix typo in documentation.
14076 2018-02-05 Martin Liska <mliska@suse.cz>
14078 PR gcov-profile/83879
14079 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14080 using dlopen functionality.
14082 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14084 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14085 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14086 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14087 _mm_maskz_range_round_ss): New intrinsics.
14088 (__builtin_ia32_rangesd128_round)
14089 (__builtin_ia32_rangess128_round): Remove.
14090 (__builtin_ia32_rangesd128_mask_round,
14091 __builtin_ia32_rangess128_mask_round): New builtins.
14092 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14093 __builtin_ia32_rangess128_round): Remove.
14094 (__builtin_ia32_rangesd128_mask_round,
14095 __builtin_ia32_rangess128_mask_round): New builtins.
14096 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14097 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14098 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14099 "<round_saeonly_constraint>")): Changed to ...
14100 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14101 "<round_saeonly_scalar_constraint>")): ... this.
14102 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14103 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14104 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14105 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14106 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14108 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14110 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14112 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14113 Remove all values except native, 8540 and 8548.
14115 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14117 * config/i386/i386.c (ix86_output_function_return): Pass
14118 INVALID_REGNUM, instead of -1, as invalid register number to
14119 indirect_thunk_name and output_indirect_thunk.
14121 2018-02-02 Julia Koval <julia.koval@intel.com>
14123 * config.gcc: Add -march=icelake.
14124 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14125 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14126 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14127 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14128 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14129 (processor_target_table): Add icelake.
14130 (ix86_option_override_internal): Handle new PTAs.
14131 (get_builtin_code_for_version): Handle icelake.
14132 (M_INTEL_COREI7_ICELAKE): New.
14133 (fold_builtin_cpu): Handle icelake.
14134 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14135 * doc/invoke.texi: Add -march=icelake.
14137 2018-02-02 Julia Koval <julia.koval@intel.com>
14139 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14140 to wide_int_bitmask.
14141 * wide-int-bitmask.h: New.
14143 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14146 * config/i386/i386.md: Replace Pmode with word_mode in
14147 builtin_setjmp_setup and builtin_longjmp to support x32.
14149 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14153 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14155 (rs6000_supported_cpu_names): New static variable.
14156 (linux_cpu_translation_table): Likewise.
14157 (elf_platform) <cpu>: Define new static variable and use it.
14158 Translate kernel AT_PLATFORM name to canonical name if needed.
14159 Error if platform name is unknown.
14161 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14164 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14166 2018-02-01 Jeff Law <law@redhat.com>
14169 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14170 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14171 the scratch if RELEASE_VIA_POP is false.
14172 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14173 If we have to save a temporary register, decrement SIZE appropriately.
14174 Pass new arguments to release_scratch_register_on_entry.
14175 (ix86_adjust_stack_and_probe): Likewise.
14176 (ix86_emit_probe_stack_range): Pass new arguments to
14177 release_scratch_register_on_entry.
14179 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14181 PR rtl-optimization/84157
14182 * combine.c (change_zero_ext): Use REG_P predicate in
14183 front of HARD_REGISTER_P predicate.
14185 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14187 * config/avr/avr.c (avr_option_override): Move disabling of
14188 -fdelete-null-pointer-checks to...
14189 * common/config/avr/avr-common.c (avr_option_optimization_table):
14192 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14194 PR tree-optimization/81635
14195 * tree-data-ref.c (split_constant_offset_1): For types that
14196 wrap on overflow, try to use range info to prove that wrapping
14199 2018-02-01 Renlin Li <renlin.li@arm.com>
14202 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14203 TAILCALL_ADDR_REGS.
14204 (aarch64_register_move_cost): Likewise.
14205 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14206 TAILCALL_ADDR_REGS.
14207 (REG_CLASS_NAMES): Likewise.
14208 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14209 TAILCALL_ADDR_REGS. Remove IP registers.
14210 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14212 2018-02-01 Richard Biener <rguenther@suse.de>
14214 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14215 for specifying RPO order and allow NULL for that.
14216 * domwalk.c (dom_walker::dom_walker): Likewise.
14217 (dom_walker::walk): Handle NULL RPO order.
14218 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14220 (rewrite_update_dom_walker): Likewise.
14221 (mark_def_dom_walker): Likewise.
14223 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14225 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14226 (aarch64_maybe_expand_sve_subreg_move): Declare.
14227 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14228 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14230 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14231 that are semantically a reverse operation.
14232 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14233 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14234 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14236 (aarch64_can_change_mode_class): For big-endian, forbid changes
14237 between two SVE modes if they have different element sizes.
14239 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14241 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14242 the TImode handling for big-endian targets.
14244 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14246 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14247 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14249 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14250 Remove BSWAP handing for big-endian targets and use the form of
14251 LD1RQ appropariate for the mode.
14253 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14255 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14256 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14257 duplicated element.
14259 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14262 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14263 check for operands that need to go through aarch64_sve_reload_be.
14265 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14267 PR tree-optimization/81661
14268 PR tree-optimization/84117
14269 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14270 * tree-eh.c: Include gimplify.h.
14271 (find_trapping_overflow, replace_trapping_overflow,
14272 rewrite_to_non_trapping_overflow): New functions.
14273 * tree-vect-loop.c: Include tree-eh.h.
14274 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14275 * tree-data-ref.c: Include tree-eh.h.
14276 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14278 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14280 PR rtl-optimization/84123
14281 * combine.c (change_zero_ext): Check if hard register satisfies
14282 can_change_dest_mode before calling gen_lowpart_SUBREG.
14284 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14287 * ira.c (ira_init_register_move_cost): Remove assert.
14289 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14291 PR rtl-optimization/84071
14292 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14293 * doc/tm.texi: Regenerate.
14295 2018-01-31 Richard Biener <rguenther@suse.de>
14297 PR tree-optimization/84132
14298 * tree-data-ref.c (analyze_miv_subscript): Properly
14299 check whether evolution_function_is_affine_multivariate_p
14300 before calling gcd_of_steps_may_divide_p.
14302 2018-01-31 Julia Koval <julia.koval@intel.com>
14305 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14306 * config/i386/i386.md (rdpid_rex64) New.
14307 (rdpid): Make 32bit only.
14309 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14312 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14313 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14315 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14318 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14320 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14322 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14324 PR rtl-optimization/84071
14325 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14326 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14328 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14330 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14331 (arc_attribute_table): Add 'aux' attribute.
14332 (arc_in_small_data_p): Consider aux like variables.
14333 (arc_is_aux_reg_p): New function.
14334 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14335 (arc_get_aux_arg): New function.
14336 (prepare_move_operands): Handle aux-register access.
14337 (arc_handle_aux_attribute): New function.
14338 * doc/extend.texi (ARC Variable attributes): Add subsection.
14340 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14342 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14343 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14344 (arc_attribute_table): Add 'uncached' attribute.
14345 (arc_print_operand): Print '.di' flag for uncached memory
14347 (arc_in_small_data_p): Do not consider for small data the uncached
14349 (arc_is_uncached_mem_p): New function.
14350 * config/arc/predicates.md (compact_store_memory_operand): Check
14351 for uncached memory accesses.
14352 (nonvol_nonimm_operand): Likewise.
14353 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14355 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14358 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14359 falign-loops=): Add Optimization flag.
14361 2018-01-30 Jeff Law <law@redhat.com>
14364 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14365 INT_REGISTERS_SAVED. Check it prior to calling
14366 get_scratch_register_on_entry.
14367 (ix86_adjust_stack_and_probe): Similarly.
14368 (ix86_emit_probe_stack_range): Similarly.
14369 (ix86_expand_prologue): Corresponding changes.
14371 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14374 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14375 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14377 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14380 * lra-constraints.c (curr_insn_transform): Process AND in the
14383 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14385 PR rtl-optimization/83986
14386 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14387 dependence against last_pending_memory_flush in addition to
14388 pending_jump_insns.
14390 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14392 PR tree-optimization/81611
14393 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14396 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14399 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14402 2018-01-30 Richard Biener <rguenther@suse.de>
14403 Jakub Jelinek <jakub@redhat.com>
14405 PR tree-optimization/84111
14406 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14407 inner loops added during recursion, as they don't have up-to-date
14410 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14413 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14414 (can_inline_edge_by_limits_p): ... here.
14415 (can_early_inline_edge_p, check_callers,
14416 update_caller_keys, update_callee_keys, recursive_inlining,
14417 add_new_edges_to_heap, speculation_useful_p,
14418 inline_small_functions,
14419 inline_small_functions, flatten_function,
14420 inline_to_all_callers_1): Update.
14422 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14424 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14427 2018-01-30 Richard Biener <rguenther@suse.de>
14429 PR tree-optimization/83008
14430 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14431 invariant and constant vector uses in stmts when they need
14432 more than one stmt.
14434 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14437 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14438 * configure: Regenerate.
14440 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14442 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14444 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14445 Use gen_rtx_REG rather than gen_lowpart.
14447 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14449 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14450 rather than 0 when creating partial subregs.
14452 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14454 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14457 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14460 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14461 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14462 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14463 flags. This restores the settings used before the 2017-07-24.
14464 Turning off pre increment/decrement/modify allows IVOPTS to
14465 optimize DF/SF loops where the index is an int.
14467 2018-01-29 Richard Biener <rguenther@suse.de>
14468 Kelvin Nilsen <kelvin@gcc.gnu.org>
14471 * tree-vect-stmts.c (vectorizable_call): Don't call
14472 targetm.vectorize_builtin_md_vectorized_function if callee is
14475 2018-01-22 Carl Love <cel@us.ibm.com>
14477 * doc/extend.tex: Fix typo in second arg in
14478 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14480 2018-01-29 Richard Biener <rguenther@suse.de>
14482 PR tree-optimization/84086
14483 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14484 (flush_ssaname_freelist): When SSA names were released reset
14485 the SCEV hash table.
14487 2018-01-29 Richard Biener <rguenther@suse.de>
14489 PR tree-optimization/84057
14490 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14491 removed paths when removing edges.
14493 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14495 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14496 -mfunction-return=@var{choice}.
14498 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14500 PR diagnostic/84034
14501 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14502 Handle CR like TAB.
14503 (layout::print_source_line): Likewise.
14504 (test_get_line_width_without_trailing_whitespace): Add test cases.
14506 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14508 PR middle-end/84040
14509 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14512 2018-01-26 Jim Wilson <jimw@sifive.com>
14514 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14516 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14519 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14521 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14522 and CMP + SUB-immediate -> SUBS.
14524 2018-01-26 Martin Sebor <msebor@redhat.com>
14526 PR tree-optimization/83896
14527 * tree-ssa-strlen.c (get_string_len): Rename...
14528 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14529 Avoid assuming length is constant.
14530 (handle_char_store): Use HOST_WIDE_INT for string length.
14532 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14535 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14536 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14538 2018-01-26 Richard Biener <rguenther@suse.de>
14540 PR rtl-optimization/84003
14541 * dse.c (record_store): Only record redundant stores when
14542 the earlier store aliases at least all accesses the later one does.
14544 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14546 PR rtl-optimization/83985
14547 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14548 REG_CFA_RESTORE insns.
14549 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14550 REG_CFA_RESTORE insns here.
14553 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14554 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14556 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14558 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14559 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14560 (arc_init): Likewise.
14561 (arc_override_options): Likewise.
14562 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14564 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14566 * config/arc/arc.h (TARGET_DBNZ): Define.
14567 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14568 properly set the tune attribute.
14569 (dbnz): Use TARGET_DBNZ guard.
14570 * config/arc/arc.opt (mtune): Add core3 option.
14572 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14574 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14575 recognize new pic like addresses.
14576 (arc_delegitimize_address): Clean up.
14578 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14580 * config/arc/arc-arches.def: Option mrf16 valid for all
14582 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14583 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14584 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14585 * config/arc/arc-tables.opt: Regenerate.
14586 * config/arc/arc.c (arc_conditional_register_usage): Handle
14587 reduced register file case.
14588 (arc_file_start): Set must have build attributes.
14589 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14590 mrf16 option value.
14591 * config/arc/arc.opt (mrf16): Add new option.
14592 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14593 * config/arc/genmultilib.awk: Handle new mrf16 option.
14594 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14595 * config/arc/t-multilib: Regenerate.
14596 * doc/invoke.texi (ARC Options): Document mrf16 option.
14598 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14600 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14601 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14602 (arc_attribute_table): Add 'secure_call' attribute.
14603 (arc_print_operand): Print secure call operand.
14604 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14606 (arc_is_secure_call_p): New function. * config/arc/arc.md
14607 (call_i): Add support for sjli instruction.
14608 (call_value_i): Likewise.
14609 * config/arc/constraints.md (Csc): New constraint.
14611 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14612 John Eric Martin <John.Martin@emmicro-us.com>
14614 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14615 * config/arc/arc.c (_arc_jli_section): New struct.
14616 (arc_jli_section): New type.
14617 (rc_jli_sections): New static variable.
14618 (arc_handle_jli_attribute): New function.
14619 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14620 (arc_file_end): New function.
14621 (TARGET_ASM_FILE_END): Define.
14622 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14623 (arc_add_jli_section): New function.
14624 (jli_call_scan): Likewise.
14625 (arc_reorg): Call jli_call_scan.
14626 (arc_output_addsi): Remove 'S' from printing asm operand.
14627 (arc_is_jli_call_p): New function.
14628 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14630 (movhi_insn): Likewise.
14631 (movsi_insn): Likewise.
14632 (movsi_set_cc_insn): Likewise.
14633 (loadqi_update): Likewise.
14634 (load_zeroextendqisi_update): Likewise.
14635 (load_signextendqisi_update): Likewise.
14636 (loadhi_update): Likewise.
14637 (load_zeroextendhisi_update): Likewise.
14638 (load_signextendhisi_update): Likewise.
14639 (loadsi_update): Likewise.
14640 (loadsf_update): Likewise.
14641 (movsicc_insn): Likewise.
14642 (bset_insn): Likewise.
14643 (bxor_insn): Likewise.
14644 (bclr_insn): Likewise.
14645 (bmsk_insn): Likewise.
14646 (bicsi3_insn): Likewise.
14647 (cmpsi_cc_c_insn): Likewise.
14648 (movsi_ne): Likewise.
14649 (movsi_cond_exec): Likewise.
14650 (clrsbsi2): Likewise.
14651 (norm_f): Likewise.
14661 (call_i): Remove 'S' asm letter, add jli instruction.
14662 (call_value_i): Likewise.
14663 * config/arc/arc.op (mjli-always): New option.
14664 * config/arc/constraints.md (Cji): New constraint.
14665 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14667 (subsf3_fpx): Likewise.
14668 (mulsf3_fpx): Likewise.
14669 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14671 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14672 function attrbutes.
14673 * doc/invoke.texi (ARC): Document mjli-always option.
14675 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14677 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14678 avoid addition with 0 and use incw and decw where possible.
14680 2018-01-26 Richard Biener <rguenther@suse.de>
14682 PR tree-optimization/81082
14683 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14684 association if it requires casting to unsigned.
14685 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14686 from fold_plusminus_mult_expr to catch important cases late when
14687 range info is available.
14689 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14691 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14692 * configure.ac (hidden_linkonce): New test.
14693 * configure: Regenerate.
14694 * config.in: Regenerate.
14696 2018-01-26 Julia Koval <julia.koval@intel.com>
14698 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14699 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14700 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14701 _mm_mask_bitshuffle_epi64_mask): Fix type.
14702 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14703 USI_FTYPE_V4DI_V4DI_USI): Remove.
14704 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14705 __builtin_ia32_vpshufbitqmb256_mask,
14706 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14707 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14708 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14710 2018-01-26 Alan Modra <amodra@gmail.com>
14713 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14714 UNSPEC_VBPERMQ. Sort other unspecs.
14716 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14718 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14720 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14722 PR middle-end/83055
14723 * predict.c (drop_profile): Do not push/pop cfun; update also
14725 (handle_missing_profiles): Fix logic looking for zero profiles.
14727 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14729 PR middle-end/83977
14730 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14731 on functions with #pragma omp declare simd or functions with simd
14733 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14734 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14735 Remove trailing \n from warning_at calls.
14737 2018-01-25 Tom de Vries <tom@codesourcery.com>
14740 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14741 for neutered workers.
14743 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14746 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14747 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14749 2018-01-24 Jeff Law <law@redhat.com>
14752 * i386.c (get_probe_interval): Move to earlier point.
14753 (ix86_compute_frame_layout): If -fstack-clash-protection and
14754 the frame is larger than the probe interval, then use pushes
14755 to save registers rather than reg->mem moves.
14756 (ix86_expand_prologue): Remove conditional for int_registers_saved
14759 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14762 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14763 min/max for never referenced object.
14765 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14767 PR middle-end/83977
14768 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14770 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14771 attributes from DECL_ATTRIBUTES (decl) without affecting
14772 DECL_ATTRIBUTES (current_function_decl).
14773 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14774 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14776 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14778 PR tree-optimization/83979
14779 * fold-const.c (fold_comparison): Use constant_boolean_node
14780 instead of boolean_{true,false}_node.
14782 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14784 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14787 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14789 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14790 Simplify the clause that sets the length attribute.
14791 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14792 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14793 clause that sets the length attribute.
14794 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14796 2018-01-24 Tom de Vries <tom@codesourcery.com>
14799 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14800 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14801 Add strict parameter.
14802 (prevent_branch_around_nothing): Insert dummy insn between branch to
14803 label and label with no ptx insn inbetween.
14804 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14806 2018-01-24 Tom de Vries <tom@codesourcery.com>
14809 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14810 for neutered threads in warp.
14811 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14813 2018-01-24 Richard Biener <rguenther@suse.de>
14815 PR tree-optimization/83176
14816 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14819 2018-01-24 Richard Biener <rguenther@suse.de>
14821 PR tree-optimization/82819
14822 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14823 code generating pluses that are no-ops in the target precision.
14825 2018-01-24 Richard Biener <rguenther@suse.de>
14827 PR middle-end/84000
14828 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14830 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14832 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14833 to merge probabilities.
14834 * predict.c (probably_never_executed): Also mark as cold functions
14835 with global 0 profile and guessed local profile.
14836 * profile-count.c (profile_probability::combine_with_count): New
14838 * profile-count.h (profile_probability::operator*,
14839 profile_probability::operator*=, profile_probability::operator/,
14840 profile_probability::operator/=): Reduce precision to adjusted
14841 and set value to guessed on contradictory divisions.
14842 (profile_probability::combine_with_freq): Remove.
14843 (profile_probability::combine_wiht_count): Declare.
14844 (profile_count::force_nonzero):: Set to adjusted.
14845 (profile_count::probability_in):: Set quality to adjusted.
14846 * tree-ssa-tail-merge.c (replace_block_by): Use
14847 combine_with_count.
14849 2018-01-23 Andrew Waterman <andrew@sifive.com>
14850 Jim Wilson <jimw@sifive.com>
14852 * config/riscv/riscv.c (riscv_stack_boundary): New.
14853 (riscv_option_override): Set riscv_stack_boundary. Handle
14854 riscv_preferred_stack_boundary_arg.
14855 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14856 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14857 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14858 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14859 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14860 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14862 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14865 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14866 of struct ix86_frame.
14867 (ix86_expand_epilogue): Likewise. Add a local variable for
14868 the reg_save_offset field in struct ix86_frame.
14870 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14872 PR tree-optimization/82604
14873 * tree-loop-distribution.c (enum partition_kind): New enum item
14874 PKIND_PARTIAL_MEMSET.
14875 (partition_builtin_p): Support above new enum item.
14876 (generate_code_for_partition): Ditto.
14877 (compute_access_range): Differentiate cases that equality can be
14878 proven at all loops, the innermost loops or no loops.
14879 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14880 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14881 (finalize_partitions, distribute_loop): Don't fuse partition of
14882 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14883 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14884 parloop is enabled.
14886 2018-01-23 Martin Liska <mliska@suse.cz>
14888 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14889 order to ignore the predictor.
14890 (PRED_POLYMORPHIC_CALL): Likewise.
14891 (PRED_RECURSIVE_CALL): Likewise.
14893 2018-01-23 Martin Liska <mliska@suse.cz>
14895 * tree-profile.c (tree_profiling): Print function header to
14896 aware reader which function we are working on.
14897 * value-prof.c (gimple_find_values_to_profile): Do not print
14898 not interesting value histograms.
14900 2018-01-23 Martin Liska <mliska@suse.cz>
14902 * profile-count.h (enum profile_quality): Add
14903 profile_uninitialized as the first value. Do not number values
14904 as they are zero based.
14905 (profile_count::verify): Update sanity check.
14906 (profile_probability::verify): Likewise.
14908 2018-01-23 Nathan Sidwell <nathan@acm.org>
14910 * doc/invoke.texi (ffor-scope): Deprecate.
14912 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14914 PR tree-optimization/83510
14915 * domwalk.c (set_all_edges_as_executable): New function.
14916 (dom_walker::dom_walker): Convert bool param
14917 "skip_unreachable_blocks" to enum reachability. Move setup of
14918 edge flags to set_all_edges_as_executable and only do it when
14919 reachability is REACHABLE_BLOCKS.
14920 * domwalk.h (enum dom_walker::reachability): New enum.
14921 (dom_walker::dom_walker): Convert bool param
14922 "skip_unreachable_blocks" to enum reachability.
14923 (set_all_edges_as_executable): New decl.
14924 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14925 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14927 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14928 but converting true to REACHABLE_BLOCKS.
14929 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14931 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14932 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14933 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14935 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14936 if check_all_array_refs will be called.
14938 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14940 * tree.c (selftest::test_location_wrappers): Add more test
14943 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14945 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14946 (selftest::test_bit_in_range): Likewise.
14948 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14951 * doc/sourcebuild.texi (vect_float): Say that the selector
14952 only describes the situation when -funsafe-math-optimizations is on.
14953 (vect_float_strict): Document.
14955 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14957 PR tree-optimization/83965
14958 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14959 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14960 instead of checking only for a reduction.
14961 (vect_recog_widen_sum_pattern): Likewise.
14963 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14965 * predict.c (probably_never_executed): Only use precise profile info.
14966 (compute_function_frequency): Skip after inlining hack since we now
14967 have quality checking.
14969 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14971 * profile-count.h (profile_probability::very_unlikely,
14972 profile_probability::unlikely, profile_probability::even): Set
14973 precision to guessed.
14975 2018-01-23 Richard Biener <rguenther@suse.de>
14977 PR tree-optimization/83963
14978 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14979 Properly terminate dominator walk when crossing the exit edge not
14980 when visiting its source block.
14982 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14985 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14986 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14988 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14990 PR tree-optimization/83957
14991 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14992 semicolon after for body surrounded by braces.
14994 PR tree-optimization/83081
14995 * profile-count.h (profile_probability::split): New method.
14996 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14997 Use profile_probability::split.
14998 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14999 when splitting a single conditional jump into 2.
15001 2018-01-22 David Malcolm <dmalcolm@redhat.com>
15003 PR tree-optimization/69452
15004 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
15007 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15009 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
15010 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
15011 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
15013 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15015 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
15017 * config/rl78/rl78.md (movdi): New define_expand.
15018 * config/rl78/rl78.c (rl78_split_movdi): New function.
15020 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15023 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15025 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15026 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15027 128-bit to produce an UNSPEC move to get the double word with the
15028 signbit and then a shift directly to do signbit.
15029 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15030 implementation with a new version that just does either a direct
15031 move or a regular move. Move memory interface to separate insns.
15032 Move insns so they are next to the expander.
15033 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15034 with signbit move. Split big and little endian case.
15035 (signbit<mode>2_dm_mem_le): Likewise.
15036 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15037 (signbit<mode>2_dm2): Likewise.
15039 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15041 * config/rl78/rl78.md (anddi3): New define_expand.
15043 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15045 * config/rl78/rl78.md (umindi3): New define_expand.
15047 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15049 * config/rl78/rl78.md (smindi3): New define_expand.
15051 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15053 * config/rl78/rl78.md (smaxdi3): New define_expand.
15055 2018-01-22 Carl Love <cel@us.ibm.com>
15057 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15058 LVX_V1TI): Add macro expansion.
15059 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15060 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15061 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15062 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15063 Change check to determine if the instruction is a byte reversing
15064 entry. Fix typo in comment.
15065 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15066 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15067 Add def_builtin calls for new builtins.
15068 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15069 Add define_insn expansion.
15071 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15073 * config/rl78/rl78.md (umaxdi3): New define_expand.
15075 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15077 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15078 for non-QImode registers.
15080 2018-01-22 Richard Biener <rguenther@suse.de>
15082 PR tree-optimization/83963
15083 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15084 including the loop exit block.
15085 (scop_detection::merge_sese): Likewise.
15086 (scop_detection::add_scop): Do it here instead.
15088 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15090 * doc/sourcebuild.texi (arm_softfloat): Document.
15092 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15095 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15096 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15097 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15099 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15100 David Edelsohn <dje.gcc@gmail.com>
15103 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15104 Change "crset eq" to "crset 2".
15105 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15106 (*call_indirect_aix<mode>_nospec): Likewise.
15107 (*call_value_indirect_aix<mode>_nospec): Likewise.
15108 (*call_indirect_elfv2<mode>_nospec): Likewise.
15109 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15110 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15111 change assembly output from . to $.
15112 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15113 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15114 (*tablejump<mode>_internal1_nospec): Likewise.
15116 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15119 * config/sh/sh_optimize_sett_clrt.cc:
15120 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15122 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15124 PR tree-optimization/83940
15125 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15126 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15127 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15128 use it to pass back the definition type.
15129 (vect_check_store_rhs): Likewise rhs_dt_out.
15130 (vect_build_gather_load_calls): Add a mask_dt argument and use
15131 it instead of a call to vect_is_simple_use.
15132 (vectorizable_store): Update calls to vect_check_load_store_mask
15133 and vect_check_store_rhs. Use the dt returned by the latter instead
15134 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15135 instead of calls to vect_is_simple_use. Pass the scalar rather
15136 than the vector operand to vect_is_simple_use when handling
15137 second and subsequent copies of an rhs value.
15138 (vectorizable_load): Update calls to vect_check_load_store_mask
15139 and vect_build_gather_load_calls. Use the cached mask_dt and
15140 gs_info.offset_dt instead of calls to vect_is_simple_use.
15142 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15144 PR middle-end/83945
15145 * tree-emutls.c: Include gimplify.h.
15146 (lower_emutls_2): New function.
15147 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15148 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15149 it before further processing.
15152 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15153 UINTVAL (trueop1) instead of INTVAL (op1).
15155 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15159 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15160 INCOMING_FRAME_SP_OFFSET if not defined.
15161 (scan_trace): Add ENTRY argument. If true and
15162 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15163 emit a note to adjust the CFA offset.
15164 (create_cfi_notes): Adjust scan_trace callers.
15165 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15166 INCOMING_FRAME_SP_OFFSET in the CIE.
15167 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15168 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15170 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15171 * doc/tm.texi: Regenerated.
15173 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15175 PR rtl-optimization/83147
15176 * lra-constraints.c (remove_inheritance_pseudos): Use
15177 lra_substitute_pseudo_within_insn.
15179 2018-01-19 Tom de Vries <tom@codesourcery.com>
15180 Cesar Philippidis <cesar@codesourcery.com>
15183 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15185 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15188 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15189 spaces for function labels.
15191 2018-01-19 Martin Liska <mliska@suse.cz>
15193 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15194 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15195 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15196 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15197 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15198 (PRED_CONST_RETURN): Change from 69 to 65.
15199 (PRED_NULL_RETURN): Change from 91 to 71.
15200 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15201 (PRED_LOOP_GUARD): Change from 66 to 73.
15203 2018-01-19 Martin Liska <mliska@suse.cz>
15205 * predict.c (predict_insn_def): Add new assert.
15206 (struct branch_predictor): Change type to signed integer.
15207 (test_prediction_value_range): Amend test to cover
15208 PROB_UNINITIALIZED.
15209 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15210 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15211 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15212 (PRED_LOOP_IV_COMPARE): Likewise.
15213 * predict.h (PROB_UNINITIALIZED): Define new constant.
15215 2018-01-19 Martin Liska <mliska@suse.cz>
15217 * predict.c (dump_prediction): Add new format for
15218 analyze_brprob.py script which is enabled with -details
15220 * profile-count.h (precise_p): New function.
15222 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15224 PR tree-optimization/83922
15225 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15226 there are no statements that need masking.
15227 (vect_active_double_reduction_p): New function.
15228 (vect_analyze_loop_operations): Use it when handling phis that
15229 are not in the loop header.
15231 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15233 PR tree-optimization/83914
15234 * tree-vect-loop.c (vectorizable_induction): Don't convert
15235 init_expr or apply the peeling adjustment for inductions
15236 that are nested within the vectorized loop.
15238 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15240 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15243 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15247 * function.h (gimplify_parameters): Add gimple_seq * argument.
15248 * function.c: Include gimple.h and options.h.
15249 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15250 for the added local temporaries if needed.
15251 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15252 if there are any parameter cleanups, wrap whole body into a
15253 try/finally with the cleanups.
15255 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15258 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15259 Use GET_MODE_CLASS for scalar floating point.
15261 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15265 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15266 Fix call of call_cgraph_insertion_hooks.
15268 2018-01-18 Martin Sebor <msebor@redhat.com>
15270 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15272 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15275 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15278 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15281 * common.opt: (-ffile-prefix-map): New option.
15282 * opts.c (common_handle_option): Defer it.
15283 * opts-global.c (handle_common_deferred_options): Handle it.
15284 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15285 * file-prefix-map.h: New file.
15286 (remap_debug_filename, add_debug_prefix_map): ...here.
15287 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15288 * final.c (debug_prefix_map, add_debug_prefix_map
15289 remap_debug_filename): Move to...
15290 * file-prefix-map.c: New file.
15291 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15292 generalize, get rid of alloca(), use strrchr() instead of strchr().
15293 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15294 Implement in terms of add_prefix_map().
15295 (remap_macro_filename, remap_debug_filename): Implement in term of
15297 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15298 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15299 * dbxout.c: Include file-prefix-map.h.
15300 * varasm.c: Likewise.
15301 * vmsdbgout.c: Likewise.
15302 * xcoffout.c: Likewise.
15303 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15304 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15305 * doc/invoke.texi (-ffile-prefix-map): Document.
15306 (-fdebug-prefix-map): Update description.
15308 2018-01-18 Martin Liska <mliska@suse.cz>
15310 * config/i386/i386.c (indirect_thunk_name): Document that also
15312 (output_indirect_thunk): Document why both instructions
15313 (pause and lfence) are generated.
15315 2018-01-18 Richard Biener <rguenther@suse.de>
15317 PR tree-optimization/83887
15318 * graphite-scop-detection.c
15319 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15320 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15321 (scop_detection::merge_sese): Re-implement with a flood-fill
15322 algorithm that properly finds a SESE region if it exists.
15324 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15327 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15328 pointer_diff optimizations use view_convert instead of convert.
15330 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15332 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15333 Generate different code for -mno-speculate-indirect-jumps.
15334 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15335 (*call_indirect_aix<mode>): Disable for
15336 -mno-speculate-indirect-jumps.
15337 (*call_indirect_aix<mode>_nospec): New define_insn.
15338 (*call_value_indirect_aix<mode>): Disable for
15339 -mno-speculate-indirect-jumps.
15340 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15341 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15342 -mno-speculate-indirect-jumps.
15343 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15345 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15347 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15348 long double type, set the flags for noting the default long double
15349 type, even if we don't pass or return a long double type.
15351 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15354 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15357 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15359 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15360 support for merge[hl].
15361 (fold_mergehl_helper): New helper function.
15362 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15363 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15364 (altivec_vmrglw_direct): Add xxmrglw insn.
15366 2018-01-17 Andrew Waterman <andrew@sifive.com>
15368 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15369 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15371 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15374 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15375 call the lto_location_cache before reading the
15376 DECL_SOURCE_LOCATION of the types.
15378 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15379 Richard Sandiford <richard.sandiford@linaro.org>
15381 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15382 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15383 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15384 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15385 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15387 * config/aarch64/constraints.md (aarch64_movti_operand):
15389 * config/aarch64/predicates.md (Uti): Add new constraint.
15391 2018-01-17 Carl Love <cel@us.ibm.com>
15393 * config/rs6000/vsx.md (define_expand xl_len_r,
15394 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15395 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15397 (define_expand, define_insn): Move the shift left from the
15398 define_insn to the define_expand for lxvl and stxvl instructions.
15399 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15400 and XL_LEN_R definitions to PURE.
15402 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15404 * config/i386/i386.c (indirect_thunk_name): Declare regno
15405 as unsigned int. Compare regno with INVALID_REGNUM.
15406 (output_indirect_thunk): Ditto.
15407 (output_indirect_thunk_function): Ditto.
15408 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15409 in the call to output_indirect_thunk_function.
15411 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15413 PR middle-end/83884
15414 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15415 rather than the size of inner_type to determine the stack slot size
15416 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15418 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15421 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15424 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15426 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15427 endian Linux systems to optionally enable multilibs for selecting
15428 the long double type if the user configured an explicit type.
15429 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15430 have no long double multilibs if not defined.
15431 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15432 warn if the user used -mabi={ieee,ibm}longdouble and we built
15433 multilibs for long double.
15434 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15435 appropriate multilib option.
15436 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15438 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15439 for building long double multilibs.
15440 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15442 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15444 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15447 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15449 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15452 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15455 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15458 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15460 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15461 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15463 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15465 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15466 different rtl trees depending on TARGET_64BIT.
15467 (rs6000_gen_lvx): Likewise.
15469 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15471 * config/visium/visium.md (nop): Tweak comment.
15472 (hazard_nop): Likewise.
15474 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15476 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15477 -mspeculate-indirect-jumps.
15478 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15479 for -mno-speculate-indirect-jumps.
15480 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15481 (*call_value_indirect_elfv2<mode>): Disable for
15482 -mno-speculate-indirect-jumps.
15483 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15484 (indirect_jump): Emit different RTL for
15485 -mno-speculate-indirect-jumps.
15486 (*indirect_jump<mode>): Disable for
15487 -mno-speculate-indirect-jumps.
15488 (*indirect_jump<mode>_nospec): New define_insn.
15489 (tablejump): Emit different RTL for
15490 -mno-speculate-indirect-jumps.
15491 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15492 (tablejumpsi_nospec): New define_expand.
15493 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15494 (tablejumpdi_nospec): New define_expand.
15495 (*tablejump<mode>_internal1): Disable for
15496 -mno-speculate-indirect-jumps.
15497 (*tablejump<mode>_internal1_nospec): New define_insn.
15498 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15501 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15503 * caller-save.c (insert_save): Drop unnecessary parameter. All
15506 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15507 Richard Biener <rguenth@suse.de>
15510 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15511 return early, inline manually is_gimple_sizepos. Make sure if we
15512 call gimplify_expr we don't end up with a gimple constant.
15513 * tree.c (variably_modified_type_p): Don't return true for
15514 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15515 * gimplify.h (is_gimple_sizepos): Remove.
15517 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15519 PR tree-optimization/83857
15520 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15521 vectorizable_live_operation for pure SLP statements.
15522 (vectorizable_live_operation): Handle PHIs.
15524 2018-01-16 Richard Biener <rguenther@suse.de>
15526 PR tree-optimization/83867
15527 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15528 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15530 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15533 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15534 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15535 If off is not INTEGER_CST, issue a may not be aligned warning
15536 rather than isn't aligned. Use isn%'t rather than isn't.
15537 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15539 <case MULT_EXPR>: Improve the case when bottom and one of the
15540 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15541 operand, in that case check if the other operand is multiple of
15542 bottom divided by the INTEGER_CST operand.
15544 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15547 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15548 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15549 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15550 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15551 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15552 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15553 (pa_function_arg_size): New function.
15555 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15557 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15558 in a separate statement.
15560 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15562 PR tree-optimization/83847
15563 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15564 group gathers and scatters.
15566 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15568 PR rtl-optimization/86620
15569 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15571 PR rtl-optimization/83213
15572 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15573 to last if both are JUMP_INSNs.
15575 PR tree-optimization/83843
15576 * gimple-ssa-store-merging.c
15577 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15578 store_immediate_info for bswap/nop orig_stores.
15580 2018-01-15 Andrew Waterman <andrew@sifive.com>
15582 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15584 <UDIV>: Increase cost if !TARGET_DIV.
15586 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15588 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15589 (define_attr "cr_logical_3op"): New.
15590 (cceq_ior_compare): Adjust.
15591 (cceq_ior_compare_complement): Adjust.
15592 (*cceq_rev_compare): Adjust.
15593 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15594 (is_cracked_insn): Adjust.
15595 (insn_must_be_first_in_group): Adjust.
15596 * config/rs6000/40x.md: Adjust.
15597 * config/rs6000/440.md: Adjust.
15598 * config/rs6000/476.md: Adjust.
15599 * config/rs6000/601.md: Adjust.
15600 * config/rs6000/603.md: Adjust.
15601 * config/rs6000/6xx.md: Adjust.
15602 * config/rs6000/7450.md: Adjust.
15603 * config/rs6000/7xx.md: Adjust.
15604 * config/rs6000/8540.md: Adjust.
15605 * config/rs6000/cell.md: Adjust.
15606 * config/rs6000/e300c2c3.md: Adjust.
15607 * config/rs6000/e500mc.md: Adjust.
15608 * config/rs6000/e500mc64.md: Adjust.
15609 * config/rs6000/e5500.md: Adjust.
15610 * config/rs6000/e6500.md: Adjust.
15611 * config/rs6000/mpc.md: Adjust.
15612 * config/rs6000/power4.md: Adjust.
15613 * config/rs6000/power5.md: Adjust.
15614 * config/rs6000/power6.md: Adjust.
15615 * config/rs6000/power7.md: Adjust.
15616 * config/rs6000/power8.md: Adjust.
15617 * config/rs6000/power9.md: Adjust.
15618 * config/rs6000/rs64.md: Adjust.
15619 * config/rs6000/titan.md: Adjust.
15621 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15623 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15624 ix86_indirect_branch_register logic.
15626 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15628 * config/i386/constraints.md (Bs): Update
15629 ix86_indirect_branch_register check. Don't check
15630 ix86_indirect_branch_register with GOT_memory_operand.
15632 * config/i386/predicates.md (GOT_memory_operand): Don't check
15633 ix86_indirect_branch_register here.
15634 (GOT32_symbol_operand): Likewise.
15636 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15638 * config/i386/predicates.md (constant_call_address_operand):
15639 Rewrite ix86_indirect_branch_register logic.
15640 (sibcall_insn_operand): Likewise.
15642 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15644 * config/i386/constraints.md (Bs): Replace
15645 ix86_indirect_branch_thunk_register with
15646 ix86_indirect_branch_register.
15648 * config/i386/i386.md (indirect_jump): Likewise.
15649 (tablejump): Likewise.
15650 (*sibcall_memory): Likewise.
15651 (*sibcall_value_memory): Likewise.
15652 Peepholes of indirect call and jump via memory: Likewise.
15653 * config/i386/i386.opt: Likewise.
15654 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15655 (GOT_memory_operand): Likewise.
15656 (call_insn_operand): Likewise.
15657 (sibcall_insn_operand): Likewise.
15658 (GOT32_symbol_operand): Likewise.
15660 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15662 PR middle-end/83837
15663 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15664 type rather than type addr's type points to.
15665 (expand_omp_atomic_mutex): Likewise.
15666 (expand_omp_atomic): Likewise.
15668 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15671 * config/i386/i386.c (output_indirect_thunk_function): Use
15672 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15673 for __x86_return_thunk.
15675 2018-01-15 Richard Biener <rguenther@suse.de>
15677 PR middle-end/83850
15678 * expmed.c (extract_bit_field_1): Fix typo.
15680 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15683 * config/arm/iterators.md (VF): New mode iterator.
15684 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15685 Remove integer-related logic from pattern.
15686 (neon_vabd<mode>_3): Likewise.
15688 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15690 PR middle-end/82694
15691 * common.opt (fstrict-overflow): No longer an alias.
15692 (fwrapv-pointer): New option.
15693 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15694 also for pointer types based on flag_wrapv_pointer.
15695 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15696 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15697 opts->x_flag_wrapv got set.
15698 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15699 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15700 POINTER_TYPE_OVERFLOW_UNDEFINED.
15701 * match.pd: Likewise in address comparison pattern.
15702 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15704 2018-01-15 Richard Biener <rguenther@suse.de>
15707 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15708 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15709 Reset type names to their identifier if their TYPE_DECL doesn't
15710 have linkage (and thus is used for ODR and devirt).
15711 (save_debug_info_for_decl): Remove.
15712 (save_debug_info_for_type): Likewise.
15713 (add_tree_to_fld_list): Adjust.
15714 * tree-pretty-print.c (dump_generic_node): Make dumping of
15715 type names more robust.
15717 2018-01-15 Richard Biener <rguenther@suse.de>
15719 * BASE-VER: Bump to 8.0.1.
15721 2018-01-14 Martin Sebor <msebor@redhat.com>
15724 * builtins.c (check_access): Avoid warning when the no-warning bit
15727 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15729 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15730 * ira-color (allocno_hard_regs_compare): Likewise.
15732 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15735 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15736 Use .pushsection/.popsection.
15738 2018-01-14 Martin Sebor <msebor@redhat.com>
15741 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15743 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15745 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15746 entry from extra_headers.
15747 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15748 extra_headers, make the list bitwise identical to the i?86-*-* one.
15750 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15752 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15753 -mcmodel=large with -mindirect-branch=thunk,
15754 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15755 -mfunction-return=thunk-extern.
15756 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15757 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15758 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15760 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15762 * config/i386/i386.c (print_reg): Print the name of the full
15763 integer register without '%'.
15764 (ix86_print_operand): Handle 'V'.
15765 * doc/extend.texi: Document 'V' modifier.
15767 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15769 * config/i386/constraints.md (Bs): Disallow memory operand for
15770 -mindirect-branch-register.
15772 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15773 (GOT_memory_operand): Likewise.
15774 (call_insn_operand): Likewise.
15775 (sibcall_insn_operand): Likewise.
15776 (GOT32_symbol_operand): Likewise.
15777 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15778 for -mindirect-branch-register.
15779 (tablejump): Likewise.
15780 (*sibcall_memory): Likewise.
15781 (*sibcall_value_memory): Likewise.
15782 Disallow peepholes of indirect call and jump via memory for
15783 -mindirect-branch-register.
15784 (*call_pop): Replace m with Bw.
15785 (*call_value_pop): Likewise.
15786 (*sibcall_pop_memory): Replace m with Bs.
15787 * config/i386/i386.opt (mindirect-branch-register): New option.
15788 * doc/invoke.texi: Document -mindirect-branch-register option.
15790 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15792 * config/i386/i386-protos.h (ix86_output_function_return): New.
15793 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15794 set function_return_type.
15795 (indirect_thunk_name): Add ret_p to indicate thunk for function
15797 (output_indirect_thunk_function): Pass false to
15798 indirect_thunk_name.
15799 (ix86_output_indirect_branch_via_reg): Likewise.
15800 (ix86_output_indirect_branch_via_push): Likewise.
15801 (output_indirect_thunk_function): Create alias for function
15802 return thunk if regno < 0.
15803 (ix86_output_function_return): New function.
15804 (ix86_handle_fndecl_attribute): Handle function_return.
15805 (ix86_attribute_table): Add function_return.
15806 * config/i386/i386.h (machine_function): Add
15807 function_return_type.
15808 * config/i386/i386.md (simple_return_internal): Use
15809 ix86_output_function_return.
15810 (simple_return_internal_long): Likewise.
15811 * config/i386/i386.opt (mfunction-return=): New option.
15812 (indirect_branch): Mention -mfunction-return=.
15813 * doc/extend.texi: Document function_return function attribute.
15814 * doc/invoke.texi: Document -mfunction-return= option.
15816 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15818 * config/i386/i386-opts.h (indirect_branch): New.
15819 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15820 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15821 with local indirect jump when converting indirect call and jump.
15822 (ix86_set_indirect_branch_type): New.
15823 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15824 (indirectlabelno): New.
15825 (indirect_thunk_needed): Likewise.
15826 (indirect_thunk_bnd_needed): Likewise.
15827 (indirect_thunks_used): Likewise.
15828 (indirect_thunks_bnd_used): Likewise.
15829 (INDIRECT_LABEL): Likewise.
15830 (indirect_thunk_name): Likewise.
15831 (output_indirect_thunk): Likewise.
15832 (output_indirect_thunk_function): Likewise.
15833 (ix86_output_indirect_branch_via_reg): Likewise.
15834 (ix86_output_indirect_branch_via_push): Likewise.
15835 (ix86_output_indirect_branch): Likewise.
15836 (ix86_output_indirect_jmp): Likewise.
15837 (ix86_code_end): Call output_indirect_thunk_function if needed.
15838 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15840 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15841 (ix86_attribute_table): Add indirect_branch.
15842 * config/i386/i386.h (machine_function): Add indirect_branch_type
15843 and has_local_indirect_jump.
15844 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15846 (tablejump): Likewise.
15847 (*indirect_jump): Use ix86_output_indirect_jmp.
15848 (*tablejump_1): Likewise.
15849 (simple_return_indirect_internal): Likewise.
15850 * config/i386/i386.opt (mindirect-branch=): New option.
15851 (indirect_branch): New.
15854 (thunk-inline): Likewise.
15855 (thunk-extern): Likewise.
15856 * doc/extend.texi: Document indirect_branch function attribute.
15857 * doc/invoke.texi: Document -mindirect-branch= option.
15859 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15862 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15864 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15866 * ipa-inline.c (want_inline_small_function_p): Return false if
15867 inlining has already failed with CIF_FINAL_ERROR.
15868 (update_caller_keys): Call want_inline_small_function_p before
15870 (update_callee_keys): Likewise.
15872 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15874 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15876 (rs6000_quadword_masked_address_p): Likewise.
15877 (quad_aligned_load_p): Likewise.
15878 (quad_aligned_store_p): Likewise.
15879 (const_load_sequence_p): Add comment to describe the outer-most loop.
15880 (mimic_memory_attributes_and_flags): New function.
15881 (rs6000_gen_stvx): Likewise.
15882 (replace_swapped_aligned_store): Likewise.
15883 (rs6000_gen_lvx): Likewise.
15884 (replace_swapped_aligned_load): Likewise.
15885 (replace_swapped_load_constant): Capitalize argument name in
15886 comment describing this function.
15887 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15888 and stores that access quad-word aligned addresses and replace
15889 with stvx or lvx instructions when appropriate.
15890 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15891 New function prototype.
15892 (rs6000_quadword_masked_address_p): Likewise.
15893 (rs6000_gen_lvx): Likewise.
15894 (rs6000_gen_stvx): Likewise.
15895 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15896 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15897 when memory address is aligned.
15898 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15899 this split to select lvx instruction when memory address is aligned.
15900 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15901 instruction when memory address is aligned.
15902 (*vsx_le_perm_load_v16qi): Likewise.
15903 (four unnamed splitters): Modify to select the stvx instruction
15904 when memory is aligned.
15906 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15908 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15909 which appears in the queue multiple times.
15911 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15912 Alan Hayward <alan.hayward@arm.com>
15913 David Sherwood <david.sherwood@arm.com>
15915 * tree-vectorizer.h (vec_lower_bound): New structure.
15916 (_loop_vec_info): Add check_nonzero and lower_bounds.
15917 (LOOP_VINFO_CHECK_NONZERO): New macro.
15918 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15919 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15920 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15921 fields. Make seg_len the distance travelled, not including the
15923 (dr_direction_indicator): Declare.
15924 (dr_zero_step_indicator): Likewise.
15925 (dr_known_forward_stride_p): Likewise.
15926 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15928 (runtime_alias_check_p): Allow runtime alias checks with
15930 (operator ==): Compare access_size and align.
15931 (prune_runtime_alias_test_list): Rework for new distinction between
15932 the access_size and seg_len.
15933 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15935 (get_segment_min_max): New function.
15936 (create_intersect_range_checks): Use it.
15937 (dr_step_indicator): New function.
15938 (dr_direction_indicator): Likewise.
15939 (dr_zero_step_indicator): Likewise.
15940 (dr_known_forward_stride_p): Likewise.
15941 * tree-loop-distribution.c (data_ref_segment_size): Return
15942 DR_STEP * (niters - 1).
15943 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15945 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15946 (vect_preserves_scalar_order_p): New function, split out from...
15947 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15948 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15949 (vect_vfa_access_size): New function.
15950 (vect_vfa_align): Likewise.
15951 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15952 (dump_lower_bound): New function.
15953 (vect_check_lower_bound): Likewise.
15954 (vect_small_gap_p): Likewise.
15955 (vectorizable_with_step_bound_p): Likewise.
15956 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15957 depencies if the vectorization factor is 1. Convert the checks
15958 for nonzero steps into checks on the bounds of DR_STEP. Try using
15959 a bunds check for variable steps if the minimum required step is
15960 relatively small. Update calls to the dr_with_seg_len
15961 constructor and to vect_compile_time_alias.
15962 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15964 (vect_loop_versioning): Call it.
15965 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15967 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15969 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15970 Alan Hayward <alan.hayward@arm.com>
15971 David Sherwood <david.sherwood@arm.com>
15973 * doc/sourcebuild.texi (vect_scatter_store): Document.
15974 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15976 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15978 * genopinit.c (main): Add supports_vec_scatter_store and
15979 supports_vec_scatter_store_cached to target_optabs.
15980 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15981 IFN_MASK_SCATTER_STORE.
15982 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15984 * internal-fn.h (internal_store_fn_p): Declare.
15985 (internal_fn_stored_value_index): Likewise.
15986 * internal-fn.c (scatter_store_direct): New macro.
15987 (expand_scatter_store_optab_fn): New function.
15988 (direct_scatter_store_optab_supported_p): New macro.
15989 (internal_store_fn_p): New function.
15990 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15991 IFN_MASK_SCATTER_STORE.
15992 (internal_fn_mask_index): Likewise.
15993 (internal_fn_stored_value_index): New function.
15994 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15995 for scatter stores.
15996 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15997 * optabs-query.c (supports_vec_scatter_store_p): New function.
15998 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15999 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
16000 true for scatter stores.
16001 (vect_gather_scatter_fn_p): Handle scatter stores too.
16002 (vect_check_gather_scatter): Consider using scatter stores if
16003 supports_vec_scatter_store_p.
16004 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
16005 scatter stores too.
16006 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16007 internal_fn_stored_value_index.
16008 (check_load_store_masking): Handle scatter stores too.
16009 (vect_get_store_rhs): Make public.
16010 (vectorizable_call): Use internal_store_fn_p.
16011 (vectorizable_store): Handle scatter store internal functions.
16012 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
16013 when deciding whether the end of the group has been reached.
16014 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
16015 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
16016 (mask_scatter_store<mode>): New insns.
16018 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16019 Alan Hayward <alan.hayward@arm.com>
16020 David Sherwood <david.sherwood@arm.com>
16022 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16023 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16024 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16026 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16027 Use vect_truncate_gather_scatter_offset if we can't treat the
16028 operation as a normal gather load or scatter store.
16029 (get_group_load_store_type): Take the gather_scatter_info
16030 as argument. Try using a gather load or scatter store for
16031 single-element groups.
16032 (get_load_store_type): Update calls to get_group_load_store_type
16033 and vect_use_strided_gather_scatters_p.
16035 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16036 Alan Hayward <alan.hayward@arm.com>
16037 David Sherwood <david.sherwood@arm.com>
16039 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16040 optional tree argument.
16041 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16043 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16044 but continue to use the current value as a fallback.
16045 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16046 to compare the updates.
16047 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16048 (get_load_store_type): Use it when handling a strided access.
16049 (vect_get_strided_load_store_ops): New function.
16050 (vect_get_data_ptr_increment): Likewise.
16051 (vectorizable_load): Handle strided gather loads. Always pass
16052 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16054 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16055 Alan Hayward <alan.hayward@arm.com>
16056 David Sherwood <david.sherwood@arm.com>
16058 * doc/md.texi (gather_load@var{m}): Document.
16059 (mask_gather_load@var{m}): Likewise.
16060 * genopinit.c (main): Add supports_vec_gather_load and
16061 supports_vec_gather_load_cached to target_optabs.
16062 * optabs-tree.c (init_tree_optimization_optabs): Use
16063 ggc_cleared_alloc to allocate target_optabs.
16064 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16065 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16067 * internal-fn.h (internal_load_fn_p): Declare.
16068 (internal_gather_scatter_fn_p): Likewise.
16069 (internal_fn_mask_index): Likewise.
16070 (internal_gather_scatter_fn_supported_p): Likewise.
16071 * internal-fn.c (gather_load_direct): New macro.
16072 (expand_gather_load_optab_fn): New function.
16073 (direct_gather_load_optab_supported_p): New macro.
16074 (direct_internal_fn_optab): New function.
16075 (internal_load_fn_p): Likewise.
16076 (internal_gather_scatter_fn_p): Likewise.
16077 (internal_fn_mask_index): Likewise.
16078 (internal_gather_scatter_fn_supported_p): Likewise.
16079 * optabs-query.c (supports_at_least_one_mode_p): New function.
16080 (supports_vec_gather_load_p): Likewise.
16081 * optabs-query.h (supports_vec_gather_load_p): Declare.
16082 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16083 and memory_type field.
16084 (NUM_PATTERNS): Bump to 15.
16085 * tree-vect-data-refs.c: Include internal-fn.h.
16086 (vect_gather_scatter_fn_p): New function.
16087 (vect_describe_gather_scatter_call): Likewise.
16088 (vect_check_gather_scatter): Try using internal functions for
16089 gather loads. Recognize existing calls to a gather load function.
16090 (vect_analyze_data_refs): Consider using gather loads if
16091 supports_vec_gather_load_p.
16092 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16093 (vect_get_gather_scatter_offset_type): Likewise.
16094 (vect_convert_mask_for_vectype): Likewise.
16095 (vect_add_conversion_to_patterm): Likewise.
16096 (vect_try_gather_scatter_pattern): Likewise.
16097 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16098 (vect_vect_recog_func_ptrs): Add it.
16099 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16100 internal_fn_mask_index and internal_gather_scatter_fn_p.
16101 (check_load_store_masking): Take the gather_scatter_info as an
16102 argument and handle gather loads.
16103 (vect_get_gather_scatter_ops): New function.
16104 (vectorizable_call): Check internal_load_fn_p.
16105 (vectorizable_load): Likewise. Handle gather load internal
16107 (vectorizable_store): Update call to check_load_store_masking.
16108 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16109 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16110 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16111 (aarch64_gather_scale_operand_d): New predicates.
16112 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16113 (mask_gather_load<mode>): New insns.
16115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16116 Alan Hayward <alan.hayward@arm.com>
16117 David Sherwood <david.sherwood@arm.com>
16119 * optabs.def (fold_left_plus_optab): New optab.
16120 * doc/md.texi (fold_left_plus_@var{m}): Document.
16121 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16122 * internal-fn.c (fold_left_direct): Define.
16123 (expand_fold_left_optab_fn): Likewise.
16124 (direct_fold_left_optab_supported_p): Likewise.
16125 * fold-const-call.c (fold_const_fold_left): New function.
16126 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16127 * tree-parloops.c (valid_reduction_p): New function.
16128 (gather_scalar_reductions): Use it.
16129 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16130 (vect_finish_replace_stmt): Declare.
16131 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16132 (needs_fold_left_reduction_p): New function, split out from...
16133 (vect_is_simple_reduction): ...here. Accept reductions that
16134 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16135 (vect_force_simple_reduction): Also store the reduction type in
16136 the assignment's STMT_VINFO_REDUC_TYPE.
16137 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16138 (merge_with_identity): New function.
16139 (vect_expand_fold_left): Likewise.
16140 (vectorize_fold_left_reduction): Likewise.
16141 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16142 scalar phi in place for it. Check for target support and reject
16143 cases that would reassociate the operation. Defer the transform
16144 phase to vectorize_fold_left_reduction.
16145 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16146 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16147 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16149 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16151 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16152 call to ifc_temp_var.
16154 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16155 Alan Hayward <alan.hayward@arm.com>
16156 David Sherwood <david.sherwood@arm.com>
16158 * target.def (legitimize_address_displacement): Take the original
16159 offset as a poly_int.
16160 * targhooks.h (default_legitimize_address_displacement): Update
16162 * targhooks.c (default_legitimize_address_displacement): Likewise.
16163 * doc/tm.texi: Regenerate.
16164 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16165 as an argument, moving assert of ad->disp == ad->disp_term to...
16166 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16167 Try calling targetm.legitimize_address_displacement before expanding
16168 the address rather than afterwards, and adjust for the new interface.
16169 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16170 Match the new hook interface. Handle SVE addresses.
16171 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16172 new hook interface.
16174 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16176 * Makefile.in (OBJS): Add early-remat.o.
16177 * target.def (select_early_remat_modes): New hook.
16178 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16179 * doc/tm.texi: Regenerate.
16180 * targhooks.h (default_select_early_remat_modes): Declare.
16181 * targhooks.c (default_select_early_remat_modes): New function.
16182 * timevar.def (TV_EARLY_REMAT): New timevar.
16183 * passes.def (pass_early_remat): New pass.
16184 * tree-pass.h (make_pass_early_remat): Declare.
16185 * early-remat.c: New file.
16186 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16188 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16191 Alan Hayward <alan.hayward@arm.com>
16192 David Sherwood <david.sherwood@arm.com>
16194 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16195 vfm1 with a bound_epilog parameter.
16196 (vect_do_peeling): Update calls accordingly, and move the prologue
16197 call earlier in the function. Treat the base bound_epilog as 0 for
16198 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16199 this base when peeling for gaps.
16200 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16201 with fully-masked loops.
16202 (vect_estimate_min_profitable_iters): Handle the single peeled
16203 iteration in that case.
16205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16206 Alan Hayward <alan.hayward@arm.com>
16207 David Sherwood <david.sherwood@arm.com>
16209 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16210 single-element interleaving even if the size is not a power of 2.
16211 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16212 accesses for single-element interleaving if the group size is
16215 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16216 Alan Hayward <alan.hayward@arm.com>
16217 David Sherwood <david.sherwood@arm.com>
16219 * doc/md.texi (fold_extract_last_@var{m}): Document.
16220 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16221 * optabs.def (fold_extract_last_optab): New optab.
16222 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16223 * internal-fn.c (fold_extract_direct): New macro.
16224 (expand_fold_extract_optab_fn): Likewise.
16225 (direct_fold_extract_optab_supported_p): Likewise.
16226 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16227 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16228 EXTRACT_LAST_REDUCTION.
16229 (get_initial_def_for_reduction): Do not create an initial vector
16230 for EXTRACT_LAST_REDUCTION reductions.
16231 (vectorizable_reduction): Leave the scalar phi in place for
16232 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16233 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16234 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16235 transform phase to vectorizable_condition.
16236 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16238 (vect_finish_stmt_generation): ...here.
16239 (vect_finish_replace_stmt): New function.
16240 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16241 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16243 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16245 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16246 Alan Hayward <alan.hayward@arm.com>
16247 David Sherwood <david.sherwood@arm.com>
16249 * doc/md.texi (extract_last_@var{m}): Document.
16250 * optabs.def (extract_last_optab): New optab.
16251 * internal-fn.def (EXTRACT_LAST): New internal function.
16252 * internal-fn.c (cond_unary_direct): New macro.
16253 (expand_cond_unary_optab_fn): Likewise.
16254 (direct_cond_unary_optab_supported_p): Likewise.
16255 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16256 loops using EXTRACT_LAST.
16257 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16258 (extract_last_<mode>): ...this optab.
16259 (vec_extract<mode><Vel>): Update accordingly.
16261 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16262 Alan Hayward <alan.hayward@arm.com>
16263 David Sherwood <david.sherwood@arm.com>
16265 * target.def (empty_mask_is_expensive): New hook.
16266 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16267 * doc/tm.texi: Regenerate.
16268 * targhooks.h (default_empty_mask_is_expensive): Declare.
16269 * targhooks.c (default_empty_mask_is_expensive): New function.
16270 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16271 if the target says that empty masks are expensive.
16272 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16274 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16276 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16277 Alan Hayward <alan.hayward@arm.com>
16278 David Sherwood <david.sherwood@arm.com>
16280 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16281 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16282 (vect_use_loop_mask_for_alignment_p): New function.
16283 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16284 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16285 niters_skip argument. Make sure that the first niters_skip elements
16286 of the first iteration are inactive.
16287 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16288 Update call to vect_set_loop_masks_directly.
16289 (get_misalign_in_elems): New function, split out from...
16290 (vect_gen_prolog_loop_niters): ...here.
16291 (vect_update_init_of_dr): Take a code argument that specifies whether
16292 the adjustment should be added or subtracted.
16293 (vect_update_init_of_drs): Likewise.
16294 (vect_prepare_for_masked_peels): New function.
16295 (vect_do_peeling): Skip prologue peeling if we're using a mask
16296 instead. Update call to vect_update_inits_of_drs.
16297 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16299 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16300 alignment. Do not include the number of peeled iterations in
16301 the minimum threshold in that case.
16302 (vectorizable_induction): Adjust the start value down by
16303 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16304 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16305 Take the number of skipped iterations into account when calculating
16307 * tree-vect-stmts.c (vect_gen_while_not): New function.
16309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16310 Alan Hayward <alan.hayward@arm.com>
16311 David Sherwood <david.sherwood@arm.com>
16313 * doc/sourcebuild.texi (vect_fully_masked): Document.
16314 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16315 default value to 0.
16316 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16318 (vect_analyze_loop_2): ...here. Don't check the vectorization
16319 factor against the number of loop iterations if the loop is
16322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16323 Alan Hayward <alan.hayward@arm.com>
16324 David Sherwood <david.sherwood@arm.com>
16326 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16327 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16328 (dump_groups): Update accordingly.
16329 (iv_use::mem_type): New member variable.
16330 (address_p): New function.
16331 (record_use): Add a mem_type argument and initialize the new
16333 (record_group_use): Add a mem_type argument. Use address_p.
16334 Remove obsolete null checks of base_object. Update call to record_use.
16335 (find_interesting_uses_op): Update call to record_group_use.
16336 (find_interesting_uses_cond): Likewise.
16337 (find_interesting_uses_address): Likewise.
16338 (get_mem_type_for_internal_fn): New function.
16339 (find_address_like_use): Likewise.
16340 (find_interesting_uses_stmt): Try find_address_like_use before
16341 calling find_interesting_uses_op.
16342 (addr_offset_valid_p): Use the iv mem_type field as the type
16343 of the addressed memory.
16344 (add_autoinc_candidates): Likewise.
16345 (get_address_cost): Likewise.
16346 (split_small_address_groups_p): Use address_p.
16347 (split_address_groups): Likewise.
16348 (add_iv_candidate_for_use): Likewise.
16349 (autoinc_possible_for_pair): Likewise.
16350 (rewrite_groups): Likewise.
16351 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16352 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16353 (get_alias_ptr_type_for_ptr_address): New function.
16354 (rewrite_use_address): Rewrite address uses in calls that were
16355 identified by find_address_like_use.
16357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16358 Alan Hayward <alan.hayward@arm.com>
16359 David Sherwood <david.sherwood@arm.com>
16361 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16363 * gimple-expr.h (is_gimple_addressable: Likewise.
16364 * gimple-expr.c (is_gimple_address): Likewise.
16365 * internal-fn.c (expand_call_mem_ref): New function.
16366 (expand_mask_load_optab_fn): Use it.
16367 (expand_mask_store_optab_fn): Likewise.
16369 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16370 Alan Hayward <alan.hayward@arm.com>
16371 David Sherwood <david.sherwood@arm.com>
16373 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16374 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16375 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16376 (cond_umax@var{mode}): Document.
16377 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16378 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16379 (cond_umin_optab, cond_umax_optab): New optabs.
16380 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16381 (COND_IOR, COND_XOR): New internal functions.
16382 * internal-fn.h (get_conditional_internal_fn): Declare.
16383 * internal-fn.c (cond_binary_direct): New macro.
16384 (expand_cond_binary_optab_fn): Likewise.
16385 (direct_cond_binary_optab_supported_p): Likewise.
16386 (get_conditional_internal_fn): New function.
16387 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16388 Cope with reduction statements that are vectorized as calls rather
16390 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16391 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16392 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16393 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16394 (UNSPEC_COND_EOR): New unspecs.
16395 (optab): Add mappings for them.
16396 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16397 (sve_int_op, sve_fp_op): New int attributes.
16399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16400 Alan Hayward <alan.hayward@arm.com>
16401 David Sherwood <david.sherwood@arm.com>
16403 * optabs.def (while_ult_optab): New optab.
16404 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16405 * internal-fn.def (WHILE_ULT): New internal function.
16406 * internal-fn.h (direct_internal_fn_supported_p): New override
16407 that takes two types as argument.
16408 * internal-fn.c (while_direct): New macro.
16409 (expand_while_optab_fn): New function.
16410 (convert_optab_supported_p): Likewise.
16411 (direct_while_optab_supported_p): New macro.
16412 * wide-int.h (wi::udiv_ceil): New function.
16413 * tree-vectorizer.h (rgroup_masks): New structure.
16414 (vec_loop_masks): New typedef.
16415 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16416 and fully_masked_p.
16417 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16418 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16419 (vect_max_vf): New function.
16420 (slpeel_make_loop_iterate_ntimes): Delete.
16421 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16422 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16423 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16424 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16425 internal-fn.h, stor-layout.h and optabs-query.h.
16426 (vect_set_loop_mask): New function.
16427 (add_preheader_seq): Likewise.
16428 (add_header_seq): Likewise.
16429 (interleave_supported_p): Likewise.
16430 (vect_maybe_permute_loop_masks): Likewise.
16431 (vect_set_loop_masks_directly): Likewise.
16432 (vect_set_loop_condition_masked): Likewise.
16433 (vect_set_loop_condition_unmasked): New function, split out from
16434 slpeel_make_loop_iterate_ntimes.
16435 (slpeel_make_loop_iterate_ntimes): Rename to..
16436 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16437 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16438 (vect_do_peeling): Update call accordingly.
16439 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16441 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16442 mask_compare_type, can_fully_mask_p and fully_masked_p.
16443 (release_vec_loop_masks): New function.
16444 (_loop_vec_info): Use it to free the loop masks.
16445 (can_produce_all_loop_masks_p): New function.
16446 (vect_get_max_nscalars_per_iter): Likewise.
16447 (vect_verify_full_masking): Likewise.
16448 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16449 retries, and free the mask rgroups before retrying. Check loop-wide
16450 reasons for disallowing fully-masked loops. Make the final decision
16451 about whether use a fully-masked loop or not.
16452 (vect_estimate_min_profitable_iters): Do not assume that peeling
16453 for the number of iterations will be needed for fully-masked loops.
16454 (vectorizable_reduction): Disable fully-masked loops.
16455 (vectorizable_live_operation): Likewise.
16456 (vect_halve_mask_nunits): New function.
16457 (vect_double_mask_nunits): Likewise.
16458 (vect_record_loop_mask): Likewise.
16459 (vect_get_loop_mask): Likewise.
16460 (vect_transform_loop): Handle the case in which the final loop
16461 iteration might handle a partial vector. Call vect_set_loop_condition
16462 instead of slpeel_make_loop_iterate_ntimes.
16463 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16464 (check_load_store_masking): New function.
16465 (prepare_load_store_mask): Likewise.
16466 (vectorizable_store): Handle fully-masked loops.
16467 (vectorizable_load): Likewise.
16468 (supportable_widening_operation): Use vect_halve_mask_nunits for
16470 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16471 (vect_gen_while): New function.
16472 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16473 (aarch64_uqdec<mode>): New insn.
16475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16476 Alan Hayward <alan.hayward@arm.com>
16477 David Sherwood <david.sherwood@arm.com>
16479 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16480 (reduc_xor_scal_optab): New optabs.
16481 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16482 (reduc_xor_scal_@var{m}): Document.
16483 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16484 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16485 internal functions.
16486 * fold-const-call.c (fold_const_call): Handle them.
16487 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16488 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16489 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16490 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16491 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16492 (UNSPEC_XORV): New unspecs.
16493 (optab): Add entries for them.
16494 (BITWISEV): New int iterator.
16495 (bit_reduc_op): New int attributes.
16497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16498 Alan Hayward <alan.hayward@arm.com>
16499 David Sherwood <david.sherwood@arm.com>
16501 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16502 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16503 * optabs.def (vec_shl_insert_optab): New optab.
16504 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16505 (duplicate_and_interleave): Likewise.
16506 * tree-vect-loop.c: Include internal-fn.h.
16507 (neutral_op_for_slp_reduction): New function, split out from
16508 get_initial_defs_for_reduction.
16509 (get_initial_def_for_reduction): Handle option 2 for variable-length
16510 vectors by loading the neutral value into a vector and then shifting
16511 the initial value into element 0.
16512 (get_initial_defs_for_reduction): Replace the code argument with
16513 the neutral value calculated by neutral_op_for_slp_reduction.
16514 Use gimple_build_vector for constant-length vectors.
16515 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16516 but the first group_size elements have a neutral value.
16517 Use duplicate_and_interleave otherwise.
16518 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16519 Update call to get_initial_defs_for_reduction. Handle SLP
16520 reductions for variable-length vectors by creating one vector
16521 result for each scalar result, with the elements associated
16522 with other scalar results stubbed out with the neutral value.
16523 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16524 Require IFN_VEC_SHL_INSERT for double reductions on
16525 variable-length vectors, or SLP reductions that have
16526 a neutral value. Require can_duplicate_and_interleave_p
16527 support for variable-length unchained SLP reductions if there
16528 is no neutral value, such as for MIN/MAX reductions. Also require
16529 the number of vector elements to be a multiple of the number of
16530 SLP statements when doing variable-length unchained SLP reductions.
16531 Update call to vect_create_epilog_for_reduction.
16532 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16533 and remove initial values.
16534 (duplicate_and_interleave): Make public.
16535 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16536 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16539 Alan Hayward <alan.hayward@arm.com>
16540 David Sherwood <david.sherwood@arm.com>
16542 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16543 (can_duplicate_and_interleave_p): New function.
16544 (vect_get_and_check_slp_defs): Take the vector of statements
16545 rather than just the current one. Remove excess parentheses.
16546 Restriction rejectinon of vect_constant_def and vect_external_def
16547 for variable-length vectors to boolean types, or types for which
16548 can_duplicate_and_interleave_p is false.
16549 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16550 (duplicate_and_interleave): New function.
16551 (vect_get_constant_vectors): Use gimple_build_vector for
16552 constant-length vectors and suitable variable-length constant
16553 vectors. Use duplicate_and_interleave for other variable-length
16554 vectors. Don't defer the update when inserting new statements.
16556 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16557 Alan Hayward <alan.hayward@arm.com>
16558 David Sherwood <david.sherwood@arm.com>
16560 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16561 min_profitable_iters doesn't go negative.
16563 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16564 Alan Hayward <alan.hayward@arm.com>
16565 David Sherwood <david.sherwood@arm.com>
16567 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16568 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16569 * optabs.def (vec_mask_load_lanes_optab): New optab.
16570 (vec_mask_store_lanes_optab): Likewise.
16571 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16572 (MASK_STORE_LANES): Likewise.
16573 * internal-fn.c (mask_load_lanes_direct): New macro.
16574 (mask_store_lanes_direct): Likewise.
16575 (expand_mask_load_optab_fn): Handle masked operations.
16576 (expand_mask_load_lanes_optab_fn): New macro.
16577 (expand_mask_store_optab_fn): Handle masked operations.
16578 (expand_mask_store_lanes_optab_fn): New macro.
16579 (direct_mask_load_lanes_optab_supported_p): Likewise.
16580 (direct_mask_store_lanes_optab_supported_p): Likewise.
16581 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16583 (vect_load_lanes_supported): Likewise.
16584 * tree-vect-data-refs.c (strip_conversion): New function.
16585 (can_group_stmts_p): Likewise.
16586 (vect_analyze_data_ref_accesses): Use it instead of checking
16587 for a pair of assignments.
16588 (vect_store_lanes_supported): Take a masked_p parameter.
16589 (vect_load_lanes_supported): Likewise.
16590 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16591 vect_store_lanes_supported and vect_load_lanes_supported.
16592 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16593 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16594 parameter. Don't allow gaps for masked accesses.
16595 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16596 and vect_load_lanes_supported.
16597 (get_load_store_type): Take a masked_p parameter and update
16598 call to get_group_load_store_type.
16599 (vectorizable_store): Update call to get_load_store_type.
16600 Handle IFN_MASK_STORE_LANES.
16601 (vectorizable_load): Update call to get_load_store_type.
16602 Handle IFN_MASK_LOAD_LANES.
16604 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16605 Alan Hayward <alan.hayward@arm.com>
16606 David Sherwood <david.sherwood@arm.com>
16608 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16610 * config/aarch64/aarch64-protos.h
16611 (aarch64_sve_struct_memory_operand_p): Declare.
16612 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16613 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16614 (VPRED, vpred): Handle SVE structure modes.
16615 * config/aarch64/constraints.md (Utx): New constraint.
16616 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16617 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16618 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16619 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16620 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16621 structure modes. Split into pieces after RA.
16622 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16623 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16625 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16626 SVE structure modes.
16627 (aarch64_classify_address): Likewise.
16628 (sizetochar): Move earlier in file.
16629 (aarch64_print_operand): Handle SVE register lists.
16630 (aarch64_array_mode): New function.
16631 (aarch64_sve_struct_memory_operand_p): Likewise.
16632 (TARGET_ARRAY_MODE): Redefine.
16634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16635 Alan Hayward <alan.hayward@arm.com>
16636 David Sherwood <david.sherwood@arm.com>
16638 * target.def (array_mode): New target hook.
16639 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16640 * doc/tm.texi: Regenerate.
16641 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16642 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16643 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16644 targetm.array_mode.
16645 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16649 Alan Hayward <alan.hayward@arm.com>
16650 David Sherwood <david.sherwood@arm.com>
16652 * fold-const.c (fold_binary_loc): Check the argument types
16653 rather than the result type when testing for a vector operation.
16655 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16657 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16658 * doc/tm.texi: Regenerate.
16660 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16661 Alan Hayward <alan.hayward@arm.com>
16662 David Sherwood <david.sherwood@arm.com>
16664 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16665 (sve): Document new AArch64 extension.
16666 * doc/md.texi (w): Extend the description of the AArch64
16667 constraint to include SVE vectors.
16668 (Upl, Upa): Document new AArch64 predicate constraints.
16669 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16671 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16672 (msve-vector-bits=): New option.
16673 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16674 SVE when these are disabled.
16675 (sve): New extension.
16676 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16677 modes. Adjust their number of units based on aarch64_sve_vg.
16678 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16679 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16680 aarch64_addr_query_type.
16681 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16682 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16683 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16684 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16685 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16686 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16687 (aarch64_simd_imm_zero_p): Delete.
16688 (aarch64_check_zero_based_sve_index_immediate): Declare.
16689 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16690 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16691 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16692 (aarch64_sve_float_mul_immediate_p): Likewise.
16693 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16694 rather than an rtx.
16695 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16696 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16697 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16698 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16699 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16700 (aarch64_regmode_natural_size): Likewise.
16701 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16702 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16704 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16705 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16706 for VG and the SVE predicate registers.
16707 (V_ALIASES): Add a "z"-prefixed alias.
16708 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16709 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16710 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16711 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16712 (REG_CLASS_NAMES): Add entries for them.
16713 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16714 and the predicate registers.
16715 (aarch64_sve_vg): Declare.
16716 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16717 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16718 (REGMODE_NATURAL_SIZE): Define.
16719 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16721 * config/aarch64/aarch64.c: Include cfgrtl.h.
16722 (simd_immediate_info): Add a constructor for series vectors,
16723 and an associated step field.
16724 (aarch64_sve_vg): New variable.
16725 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16726 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16727 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16728 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16729 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16730 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16731 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16732 (aarch64_get_mask_mode): New functions.
16733 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16734 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16735 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16736 predicate modes and predicate registers. Explicitly restrict
16737 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16738 to store a vector mode if it is recognized by
16739 aarch64_classify_vector_mode.
16740 (aarch64_regmode_natural_size): New function.
16741 (aarch64_hard_regno_caller_save_mode): Return the original mode
16743 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16744 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16745 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16746 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16748 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16749 does not overlap dest if the function is frame-related. Handle
16751 (aarch64_split_add_offset): New function.
16752 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16753 them aarch64_add_offset.
16754 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16755 and update call to aarch64_sub_sp.
16756 (aarch64_add_cfa_expression): New function.
16757 (aarch64_expand_prologue): Pass extra temporary registers to the
16758 functions above. Handle the case in which we need to emit new
16759 DW_CFA_expressions for registers that were originally saved
16760 relative to the stack pointer, but now have to be expressed
16761 relative to the frame pointer.
16762 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16764 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16765 IP0 and IP1 values for SVE frames.
16766 (aarch64_expand_vec_series): New function.
16767 (aarch64_expand_sve_widened_duplicate): Likewise.
16768 (aarch64_expand_sve_const_vector): Likewise.
16769 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16770 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16771 into the register, rather than emitting a SET directly.
16772 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16773 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16774 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16775 (offset_9bit_signed_scaled_p): New functions.
16776 (aarch64_replicate_bitmask_imm): New function.
16777 (aarch64_bitmask_imm): Use it.
16778 (aarch64_cannot_force_const_mem): Reject expressions involving
16779 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16780 (aarch64_classify_index): Handle SVE indices, by requiring
16781 a plain register index with a scale that matches the element size.
16782 (aarch64_classify_address): Handle SVE addresses. Assert that
16783 the mode of the address is VOIDmode or an integer mode.
16784 Update call to aarch64_classify_symbol.
16785 (aarch64_classify_symbolic_expression): Update call to
16786 aarch64_classify_symbol.
16787 (aarch64_const_vec_all_in_range_p): New function.
16788 (aarch64_print_vector_float_operand): Likewise.
16789 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16790 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16791 and the FP immediates 1.0 and 0.5.
16792 (aarch64_print_address_internal): Handle SVE addresses.
16793 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16794 (aarch64_regno_regclass): Handle predicate registers.
16795 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16797 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16798 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16799 (aarch64_convert_sve_vector_bits): New function.
16800 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16801 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16802 rather than an rtx.
16803 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16804 Handle SVE vector and predicate modes. Accept VL-based constants
16805 that need only one temporary register, and VL offsets that require
16806 no temporary registers.
16807 (aarch64_conditional_register_usage): Mark the predicate registers
16808 as fixed if SVE isn't available.
16809 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16810 Return true for SVE vector and predicate modes.
16811 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16812 rather than an unsigned int. Handle SVE modes.
16813 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16815 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16817 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16818 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16819 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16820 (aarch64_sve_float_mul_immediate_p): New functions.
16821 (aarch64_sve_valid_immediate): New function.
16822 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16823 Explicitly reject structure modes. Check for INDEX constants.
16824 Handle PTRUE and PFALSE constants.
16825 (aarch64_check_zero_based_sve_index_immediate): New function.
16826 (aarch64_simd_imm_zero_p): Delete.
16827 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16828 vector modes. Accept constants in the range of CNT[BHWD].
16829 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16830 ask for an Advanced SIMD mode.
16831 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16832 (aarch64_simd_vector_alignment): Handle SVE predicates.
16833 (aarch64_vectorize_preferred_vector_alignment): New function.
16834 (aarch64_simd_vector_alignment_reachable): Use it instead of
16836 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16837 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16839 (MAX_VECT_LEN): Delete.
16840 (expand_vec_perm_d): Add a vec_flags field.
16841 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16842 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16843 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16845 (aarch64_evpc_rev): Rename to...
16846 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16847 (aarch64_evpc_rev_global): New function.
16848 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16849 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16851 (aarch64_evpc_sve_tbl): New function.
16852 (aarch64_expand_vec_perm_const_1): Update after rename of
16853 aarch64_evpc_rev. Handle SVE permutes too, trying
16854 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16855 than aarch64_evpc_tbl.
16856 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16857 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16858 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16859 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16860 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16861 (aarch64_expand_sve_vcond): New functions.
16862 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16863 of aarch64_vector_mode_p.
16864 (aarch64_dwarf_poly_indeterminate_value): New function.
16865 (aarch64_compute_pressure_classes): Likewise.
16866 (aarch64_can_change_mode_class): Likewise.
16867 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16868 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16869 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16870 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16871 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16872 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16873 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16874 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16876 (Dn, Dl, Dr): Accept const as well as const_vector.
16877 (Dz): Likewise. Compare against CONST0_RTX.
16878 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16879 of "vector" where appropriate.
16880 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16881 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16882 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16883 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16884 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16885 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16886 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16887 (v_int_equiv): Extend to SVE modes.
16888 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16890 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16891 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16892 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16893 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16894 (SVE_COND_FP_CMP): New int iterators.
16895 (perm_hilo): Handle the new unpack unspecs.
16896 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16898 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16899 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16900 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16901 (aarch64_equality_operator, aarch64_constant_vector_operand)
16902 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16903 (aarch64_sve_nonimmediate_operand): Likewise.
16904 (aarch64_sve_general_operand): Likewise.
16905 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16906 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16907 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16908 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16909 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16910 (aarch64_sve_float_arith_immediate): Likewise.
16911 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16912 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16913 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16914 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16915 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16916 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16917 (aarch64_sve_float_arith_operand): Likewise.
16918 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16919 (aarch64_sve_float_mul_operand): Likewise.
16920 (aarch64_sve_vec_perm_operand): Likewise.
16921 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16922 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16923 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16924 as well as const_vector.
16925 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16926 in file. Use CONST0_RTX and CONSTM1_RTX.
16927 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16928 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16929 Use aarch64_simd_imm_zero.
16930 * config/aarch64/aarch64-sve.md: New file.
16931 * config/aarch64/aarch64.md: Include it.
16932 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16933 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16934 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16935 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16936 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16937 (sve): New attribute.
16938 (enabled): Disable instructions with the sve attribute unless
16940 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16941 aarch64_expand_mov_immediate.
16942 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16943 CNT[BHSD] immediates.
16944 (movti): Split CONST_POLY_INT moves into two halves.
16945 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16946 Split additions that need a temporary here if the destination
16947 is the stack pointer.
16948 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16949 (*add<mode>3_poly_1): New instruction.
16950 (set_clobber_cc): New expander.
16952 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16954 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16955 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16956 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16957 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16958 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16959 Change innermode from fixed_mode_size to machine_mode.
16960 (simplify_subreg): Update call accordingly. Handle a constant-sized
16961 subreg of a variable-length CONST_VECTOR.
16963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16964 Alan Hayward <alan.hayward@arm.com>
16965 David Sherwood <david.sherwood@arm.com>
16967 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16968 (add_offset_to_base): New function, split out from...
16969 (create_mem_ref): ...here. When handling a scale other than 1,
16970 check first whether the address is valid without the offset.
16971 Add it into the base if so, leaving the index and scale as-is.
16973 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16976 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16977 fold_for_warn before checking if arg2 is INTEGER_CST.
16979 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16981 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16982 (store_multiple_operation): Delete.
16983 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16984 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16985 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16986 guarded by TARGET_STRING.
16987 (rs6000_output_load_multiple): Delete.
16988 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16989 OPTION_MASK_STRING / TARGET_STRING handling.
16990 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16991 (const rs6000_opt_masks) <"string">: Change mask to 0.
16992 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16993 (MASK_STRING): Delete.
16994 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16996 (load_multiple): Delete.
17003 (store_multiple): Delete.
17010 (movmemsi_8reg): Delete.
17011 (corresponding unnamed define_insn): Delete.
17012 (movmemsi_6reg): Delete.
17013 (corresponding unnamed define_insn): Delete.
17014 (movmemsi_4reg): Delete.
17015 (corresponding unnamed define_insn): Delete.
17016 (movmemsi_2reg): Delete.
17017 (corresponding unnamed define_insn): Delete.
17018 (movmemsi_1reg): Delete.
17019 (corresponding unnamed define_insn): Delete.
17020 * config/rs6000/rs6000.opt (mno-string): New.
17021 (mstring): Replace by deprecation warning stub.
17022 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17024 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17026 * regrename.c (regrename_do_replace): If replacing the same
17027 reg multiple times, try to reuse last created gen_raw_REG.
17030 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17031 main to workaround a bug in GDB.
17033 2018-01-12 Tom de Vries <tom@codesourcery.com>
17036 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17038 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17040 PR rtl-optimization/80481
17041 * ira-color.c (get_cap_member): New function.
17042 (allocnos_conflict_by_live_ranges_p): Use it.
17043 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17044 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17046 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17049 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17050 (*saddl_se_1): Ditto.
17051 (*ssubsi_1): Ditto.
17052 (*ssubl_se_1): Ditto.
17054 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17056 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17057 rather than wi::to_widest for DR_INITs.
17058 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17059 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17060 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17062 (vect_analyze_group_access_1): Note that here.
17064 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17066 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17067 polynomial type sizes.
17069 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17071 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17072 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17073 (gimple_add_tmp_var): Likewise.
17075 2018-01-12 Martin Liska <mliska@suse.cz>
17077 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17078 (gimple_alloc_sizes): Likewise.
17079 (dump_gimple_statistics): Use PRIu64 in printf format.
17080 * gimple.h: Change uint64_t to int.
17082 2018-01-12 Martin Liska <mliska@suse.cz>
17084 * tree-core.h: Use uint64_t instead of int.
17085 * tree.c (tree_node_counts): Likewise.
17086 (tree_node_sizes): Likewise.
17087 (dump_tree_statistics): Use PRIu64 in printf format.
17089 2018-01-12 Martin Liska <mliska@suse.cz>
17091 * Makefile.in: As qsort_chk is implemented in vec.c, add
17092 vec.o to linkage of gencfn-macros.
17093 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17094 passing the info to record_node_allocation_statistics.
17095 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17097 * ggc-common.c (struct ggc_usage): Add operator== and use
17098 it in operator< and compare function.
17099 * mem-stats.h (struct mem_usage): Likewise.
17100 * vec.c (struct vec_usage): Remove operator< and compare
17101 function. Can be simply inherited.
17103 2018-01-12 Martin Jambor <mjambor@suse.cz>
17106 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17107 * tree-ssa-math-opts.c: Include domwalk.h.
17108 (convert_mult_to_fma_1): New function.
17109 (fma_transformation_info): New type.
17110 (fma_deferring_state): Likewise.
17111 (cancel_fma_deferring): New function.
17112 (result_of_phi): Likewise.
17113 (last_fma_candidate_feeds_initial_phi): Likewise.
17114 (convert_mult_to_fma): Added deferring logic, split actual
17115 transformation to convert_mult_to_fma_1.
17116 (math_opts_dom_walker): New type.
17117 (math_opts_dom_walker::after_dom_children): New method, body moved
17118 here from pass_optimize_widening_mul::execute, added deferring logic
17120 (pass_optimize_widening_mul::execute): Moved most of code to
17121 math_opts_dom_walker::after_dom_children.
17122 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17123 * config/i386/i386.c (ix86_option_override_internal): Added
17124 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17126 2018-01-12 Richard Biener <rguenther@suse.de>
17129 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17130 inline instance vars.
17132 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17135 * config/rx/rx.c (rx_is_restricted_memory_address):
17136 Handle SUBREG case.
17138 2018-01-12 Richard Biener <rguenther@suse.de>
17140 PR tree-optimization/80846
17141 * target.def (split_reduction): New target hook.
17142 * targhooks.c (default_split_reduction): New function.
17143 * targhooks.h (default_split_reduction): Declare.
17144 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17145 target requests first reduce vectors by combining low and high
17147 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17148 (get_vectype_for_scalar_type_and_size): Export.
17149 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17150 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17151 * doc/tm.texi: Regenerate.
17152 * config/i386/i386.c (ix86_split_reduction): Implement
17153 TARGET_VECTORIZE_SPLIT_REDUCTION.
17155 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17158 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17159 in PIC mode except for TARGET_VXWORKS_RTP.
17160 * config/sparc/sparc.c: Include cfgrtl.h.
17161 (TARGET_INIT_PIC_REG): Define.
17162 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17163 (sparc_pic_register_p): New predicate.
17164 (sparc_legitimate_address_p): Use it.
17165 (sparc_legitimize_pic_address): Likewise.
17166 (sparc_delegitimize_address): Likewise.
17167 (sparc_mode_dependent_address_p): Likewise.
17168 (gen_load_pcrel_sym): Remove 4th parameter.
17169 (load_got_register): Adjust call to above. Remove obsolete stuff.
17170 (sparc_expand_prologue): Do not call load_got_register here.
17171 (sparc_flat_expand_prologue): Likewise.
17172 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17173 (sparc_use_pseudo_pic_reg): New function.
17174 (sparc_init_pic_reg): Likewise.
17175 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17176 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17178 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17180 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17181 Add item for branch_cost.
17183 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17185 PR rtl-optimization/83565
17186 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17187 not extend the result to a larger mode for rotate operations.
17188 (num_sign_bit_copies1): Likewise.
17190 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17193 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17195 Use values-Xc.o for -pedantic.
17196 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17198 2018-01-12 Martin Liska <mliska@suse.cz>
17201 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17203 (possible_polymorphic_call_targets): Use it.
17204 (ipa_devirt): Likewise.
17206 2018-01-12 Martin Liska <mliska@suse.cz>
17208 * profile-count.h (enum profile_quality): Use 0 as invalid
17209 enum value of profile_quality.
17211 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17213 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17214 -mext-string options.
17216 2018-01-12 Richard Biener <rguenther@suse.de>
17218 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17219 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17220 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17222 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17224 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17226 * configure.ac (--with-long-double-format): Add support for the
17227 configuration option to change the default long double format on
17229 * config.gcc (powerpc*-linux*-*): Likewise.
17230 * configure: Regenerate.
17231 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17232 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17233 used without modification.
17235 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17237 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17238 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17239 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17240 MISC_BUILTIN_SPEC_BARRIER.
17241 (rs6000_init_builtins): Likewise.
17242 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17244 (speculation_barrier): New define_insn.
17245 * doc/extend.texi: Document __builtin_speculation_barrier.
17247 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17250 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17251 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17252 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17254 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17255 integral modes instead of "ss" and "sd".
17256 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17257 vectors with 32-bit and 64-bit elements.
17258 (vecdupssescalarmodesuffix): New mode attribute.
17259 (vec_dup<mode>): Use it.
17261 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17264 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17265 frame if argument is passed on stack.
17267 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17270 * ree.c (combine_reaching_defs): Optimize also
17271 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17272 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17274 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17276 PR middle-end/83189
17277 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17279 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17281 PR middle-end/83718
17282 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17283 after they are computed.
17285 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17287 PR tree-optimization/83695
17288 * gimple-loop-linterchange.cc
17289 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17290 reset cached scev information after interchange.
17291 (pass_linterchange::execute): Remove call to scev_reset_htab.
17293 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17295 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17296 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17297 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17298 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17299 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17300 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17301 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17302 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17303 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17304 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17305 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17306 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17307 (V_lane_reg): Likewise.
17308 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17310 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17311 (vfmal_lane_low<mode>_intrinsic,
17312 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17313 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17314 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17315 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17316 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17317 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17319 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17321 * config/arm/arm-cpus.in (fp16fml): New feature.
17322 (ALL_SIMD): Add fp16fml.
17323 (armv8.2-a): Add fp16fml as an option.
17324 (armv8.3-a): Likewise.
17325 (armv8.4-a): Add fp16fml as part of fp16.
17326 * config/arm/arm.h (TARGET_FP16FML): Define.
17327 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17329 * config/arm/arm-modes.def (V2HF): Define.
17330 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17331 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17332 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17333 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17334 vfmsl_low, vfmsl_high): New set of builtins.
17335 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17336 (vfml_op): New code attribute.
17337 (VFMLHALVES): New int iterator.
17338 (VFML, VFMLSEL): New mode attributes.
17339 (V_reg): Define mapping for V2HF.
17340 (V_hi, V_lo): New mode attributes.
17341 (VF_constraint): Likewise.
17342 (vfml_half, vfml_half_selector): New int attributes.
17343 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17345 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17346 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17348 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17349 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17350 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17351 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17353 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17354 Document new effective target and option set.
17356 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17358 * config/arm/arm-cpus.in (armv8_4): New feature.
17359 (ARMv8_4a): New fgroup.
17360 (armv8.4-a): New arch.
17361 * config/arm/arm-tables.opt: Regenerate.
17362 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17363 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17364 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17365 Add matching rules for -march=armv8.4-a and extensions.
17366 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17368 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17371 * config/rx/rx.md (BW): New mode attribute.
17372 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17374 2018-01-11 Richard Biener <rguenther@suse.de>
17376 PR tree-optimization/83435
17377 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17378 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17379 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17381 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17382 Alan Hayward <alan.hayward@arm.com>
17383 David Sherwood <david.sherwood@arm.com>
17385 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17387 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17388 (aarch64_print_address_internal): Use it to check for a zero offset.
17390 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17391 Alan Hayward <alan.hayward@arm.com>
17392 David Sherwood <david.sherwood@arm.com>
17394 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17395 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17396 Return a poly_int64 rather than a HOST_WIDE_INT.
17397 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17398 rather than a HOST_WIDE_INT.
17399 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17400 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17401 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17402 final_offset from HOST_WIDE_INT to poly_int64.
17403 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17404 to_constant when getting the number of units in an Advanced SIMD
17406 (aarch64_builtin_vectorized_function): Check for a constant number
17408 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17410 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17411 attribute instead of GET_MODE_NUNITS.
17412 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17413 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17414 GET_MODE_SIZE for fixed-size registers.
17415 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17416 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17417 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17418 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17419 (aarch64_print_operand, aarch64_print_address_internal)
17420 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17421 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17422 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17423 Handle polynomial GET_MODE_SIZE.
17424 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17425 wider than SImode without modification.
17426 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17427 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17428 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17429 passing and returning SVE modes.
17430 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17431 rather than GEN_INT.
17432 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17433 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17434 (aarch64_allocate_and_probe_stack_space): Likewise.
17435 (aarch64_layout_frame): Cope with polynomial offsets.
17436 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17437 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17438 polynomial offsets.
17439 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17440 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17441 poly_int64 rather than a HOST_WIDE_INT.
17442 (aarch64_get_separate_components, aarch64_process_components)
17443 (aarch64_expand_prologue, aarch64_expand_epilogue)
17444 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17445 (aarch64_anchor_offset): New function, split out from...
17446 (aarch64_legitimize_address): ...here.
17447 (aarch64_builtin_vectorization_cost): Handle polynomial
17448 TYPE_VECTOR_SUBPARTS.
17449 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17451 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17452 number of elements from the PARALLEL rather than the mode.
17453 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17454 rather than GET_MODE_BITSIZE.
17455 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17456 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17457 (aarch64_expand_vec_perm_const_1): Handle polynomial
17458 d->perm.length () and d->perm elements.
17459 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17460 Apply to_constant to d->perm elements.
17461 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17462 polynomial CONST_VECTOR_NUNITS.
17463 (aarch64_move_pointer): Take amount as a poly_int64 rather
17465 (aarch64_progress_pointer): Avoid temporary variable.
17466 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17467 the mode attribute instead of GET_MODE.
17469 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17470 Alan Hayward <alan.hayward@arm.com>
17471 David Sherwood <david.sherwood@arm.com>
17473 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17474 x exists before using it.
17475 (aarch64_add_constant_internal): Rename to...
17476 (aarch64_add_offset_1): ...this. Replace regnum with separate
17477 src and dest rtxes. Handle the case in which they're different,
17478 including when the offset is zero. Replace scratchreg with an rtx.
17479 Use 2 additions if there is no spare register into which we can
17480 move a 16-bit constant.
17481 (aarch64_add_constant): Delete.
17482 (aarch64_add_offset): Replace reg with separate src and dest
17483 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17484 Use aarch64_add_offset_1.
17485 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17486 an rtx rather than an int. Take the delta as a poly_int64
17487 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17488 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17489 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17490 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17491 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17492 and aarch64_add_sp.
17493 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17494 aarch64_add_constant.
17496 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17498 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17499 Use scalar_float_mode.
17501 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17503 * config/aarch64/aarch64-simd.md
17504 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17505 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17506 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17507 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17508 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17509 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17510 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17511 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17512 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17513 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17515 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17518 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17519 targ_options->x_arm_arch_string is non NULL.
17521 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17523 * config/aarch64/aarch64.h
17524 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17526 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17529 * expmed.c (emit_store_flag_force): Swap if const op0
17530 and change VOIDmode to mode of op0.
17532 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17534 PR rtl-optimization/83761
17535 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17536 than bytes to mode_for_size.
17538 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17540 PR middle-end/83189
17541 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17542 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17545 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17547 PR middle-end/83575
17548 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17549 when in layout mode.
17550 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17551 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17554 2018-01-10 Michael Collison <michael.collison@arm.com>
17556 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17557 * config/aarch64/aarch64-option-extension.def: Add
17558 AARCH64_OPT_EXTENSION of 'fp16fml'.
17559 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17560 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17561 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17562 * config/aarch64/constraints.md (Ui7): New constraint.
17563 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17564 (VFMLA_SEL_W): Ditto.
17567 (VFMLA16_LOW): New int iterator.
17568 (VFMLA16_HIGH): Ditto.
17569 (UNSPEC_FMLAL): New unspec.
17570 (UNSPEC_FMLSL): Ditto.
17571 (UNSPEC_FMLAL2): Ditto.
17572 (UNSPEC_FMLSL2): Ditto.
17573 (f16mac): New code attribute.
17574 * config/aarch64/aarch64-simd-builtins.def
17575 (aarch64_fmlal_lowv2sf): Ditto.
17576 (aarch64_fmlsl_lowv2sf): Ditto.
17577 (aarch64_fmlalq_lowv4sf): Ditto.
17578 (aarch64_fmlslq_lowv4sf): Ditto.
17579 (aarch64_fmlal_highv2sf): Ditto.
17580 (aarch64_fmlsl_highv2sf): Ditto.
17581 (aarch64_fmlalq_highv4sf): Ditto.
17582 (aarch64_fmlslq_highv4sf): Ditto.
17583 (aarch64_fmlal_lane_lowv2sf): Ditto.
17584 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17585 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17586 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17587 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17588 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17589 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17590 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17591 (aarch64_fmlal_lane_highv2sf): Ditto.
17592 (aarch64_fmlsl_lane_highv2sf): Ditto.
17593 (aarch64_fmlal_laneq_highv2sf): Ditto.
17594 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17595 (aarch64_fmlalq_lane_highv4sf): Ditto.
17596 (aarch64_fmlsl_lane_highv4sf): Ditto.
17597 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17598 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17599 * config/aarch64/aarch64-simd.md:
17600 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17601 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17602 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17603 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17604 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17605 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17606 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17607 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17608 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17609 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17610 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17611 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17612 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17613 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17614 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17615 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17616 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17617 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17618 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17619 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17620 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17621 (vfmlsl_low_u32): Ditto.
17622 (vfmlalq_low_u32): Ditto.
17623 (vfmlslq_low_u32): Ditto.
17624 (vfmlal_high_u32): Ditto.
17625 (vfmlsl_high_u32): Ditto.
17626 (vfmlalq_high_u32): Ditto.
17627 (vfmlslq_high_u32): Ditto.
17628 (vfmlal_lane_low_u32): Ditto.
17629 (vfmlsl_lane_low_u32): Ditto.
17630 (vfmlal_laneq_low_u32): Ditto.
17631 (vfmlsl_laneq_low_u32): Ditto.
17632 (vfmlalq_lane_low_u32): Ditto.
17633 (vfmlslq_lane_low_u32): Ditto.
17634 (vfmlalq_laneq_low_u32): Ditto.
17635 (vfmlslq_laneq_low_u32): Ditto.
17636 (vfmlal_lane_high_u32): Ditto.
17637 (vfmlsl_lane_high_u32): Ditto.
17638 (vfmlal_laneq_high_u32): Ditto.
17639 (vfmlsl_laneq_high_u32): Ditto.
17640 (vfmlalq_lane_high_u32): Ditto.
17641 (vfmlslq_lane_high_u32): Ditto.
17642 (vfmlalq_laneq_high_u32): Ditto.
17643 (vfmlslq_laneq_high_u32): Ditto.
17644 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17645 (AARCH64_FL_FOR_ARCH8_4): New.
17646 (AARCH64_ISA_F16FML): New ISA flag.
17647 (TARGET_F16FML): New feature flag for fp16fml.
17648 (doc/invoke.texi): Document new fp16fml option.
17650 2018-01-10 Michael Collison <michael.collison@arm.com>
17652 * config/aarch64/aarch64-builtins.c:
17653 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17654 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17655 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17656 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17657 (AARCH64_ISA_SHA3): New ISA flag.
17658 (TARGET_SHA3): New feature flag for sha3.
17659 * config/aarch64/iterators.md (sha512_op): New int attribute.
17660 (CRYPTO_SHA512): New int iterator.
17661 (UNSPEC_SHA512H): New unspec.
17662 (UNSPEC_SHA512H2): Ditto.
17663 (UNSPEC_SHA512SU0): Ditto.
17664 (UNSPEC_SHA512SU1): Ditto.
17665 * config/aarch64/aarch64-simd-builtins.def
17666 (aarch64_crypto_sha512hqv2di): New builtin.
17667 (aarch64_crypto_sha512h2qv2di): Ditto.
17668 (aarch64_crypto_sha512su0qv2di): Ditto.
17669 (aarch64_crypto_sha512su1qv2di): Ditto.
17670 (aarch64_eor3qv8hi): Ditto.
17671 (aarch64_rax1qv2di): Ditto.
17672 (aarch64_xarqv2di): Ditto.
17673 (aarch64_bcaxqv8hi): Ditto.
17674 * config/aarch64/aarch64-simd.md:
17675 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17676 (aarch64_crypto_sha512su0qv2di): Ditto.
17677 (aarch64_crypto_sha512su1qv2di): Ditto.
17678 (aarch64_eor3qv8hi): Ditto.
17679 (aarch64_rax1qv2di): Ditto.
17680 (aarch64_xarqv2di): Ditto.
17681 (aarch64_bcaxqv8hi): Ditto.
17682 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17683 (vsha512h2q_u64): Ditto.
17684 (vsha512su0q_u64): Ditto.
17685 (vsha512su1q_u64): Ditto.
17686 (veor3q_u16): Ditto.
17687 (vrax1q_u64): Ditto.
17688 (vxarq_u64): Ditto.
17689 (vbcaxq_u16): Ditto.
17690 * config/arm/types.md (crypto_sha512): New type attribute.
17691 (crypto_sha3): Ditto.
17692 (doc/invoke.texi): Document new sha3 option.
17694 2018-01-10 Michael Collison <michael.collison@arm.com>
17696 * config/aarch64/aarch64-builtins.c:
17697 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17698 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17699 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17700 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17701 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17702 (AARCH64_ISA_SM4): New ISA flag.
17703 (TARGET_SM4): New feature flag for sm4.
17704 * config/aarch64/aarch64-simd-builtins.def
17705 (aarch64_sm3ss1qv4si): Ditto.
17706 (aarch64_sm3tt1aq4si): Ditto.
17707 (aarch64_sm3tt1bq4si): Ditto.
17708 (aarch64_sm3tt2aq4si): Ditto.
17709 (aarch64_sm3tt2bq4si): Ditto.
17710 (aarch64_sm3partw1qv4si): Ditto.
17711 (aarch64_sm3partw2qv4si): Ditto.
17712 (aarch64_sm4eqv4si): Ditto.
17713 (aarch64_sm4ekeyqv4si): Ditto.
17714 * config/aarch64/aarch64-simd.md:
17715 (aarch64_sm3ss1qv4si): Ditto.
17716 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17717 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17718 (aarch64_sm4eqv4si): Ditto.
17719 (aarch64_sm4ekeyqv4si): Ditto.
17720 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17721 (sm3part_op): Ditto.
17722 (CRYPTO_SM3TT): Ditto.
17723 (CRYPTO_SM3PART): Ditto.
17724 (UNSPEC_SM3SS1): New unspec.
17725 (UNSPEC_SM3TT1A): Ditto.
17726 (UNSPEC_SM3TT1B): Ditto.
17727 (UNSPEC_SM3TT2A): Ditto.
17728 (UNSPEC_SM3TT2B): Ditto.
17729 (UNSPEC_SM3PARTW1): Ditto.
17730 (UNSPEC_SM3PARTW2): Ditto.
17731 (UNSPEC_SM4E): Ditto.
17732 (UNSPEC_SM4EKEY): Ditto.
17733 * config/aarch64/constraints.md (Ui2): New constraint.
17734 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17735 * config/arm/types.md (crypto_sm3): New type attribute.
17736 (crypto_sm4): Ditto.
17737 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17738 (vsm3tt1aq_u32): Ditto.
17739 (vsm3tt1bq_u32): Ditto.
17740 (vsm3tt2aq_u32): Ditto.
17741 (vsm3tt2bq_u32): Ditto.
17742 (vsm3partw1q_u32): Ditto.
17743 (vsm3partw2q_u32): Ditto.
17744 (vsm4eq_u32): Ditto.
17745 (vsm4ekeyq_u32): Ditto.
17746 (doc/invoke.texi): Document new sm4 option.
17748 2018-01-10 Michael Collison <michael.collison@arm.com>
17750 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17751 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17752 (AARCH64_FL_FOR_ARCH8_4): New.
17753 (AARCH64_FL_V8_4): New flag.
17754 (doc/invoke.texi): Document new armv8.4-a option.
17756 2018-01-10 Michael Collison <michael.collison@arm.com>
17758 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17759 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17760 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17761 * config/aarch64/aarch64-option-extension.def: Add
17762 AARCH64_OPT_EXTENSION of 'sha2'.
17763 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17764 (crypto): Disable sha2 and aes if crypto disabled.
17765 (crypto): Enable aes and sha2 if enabled.
17766 (simd): Disable sha2 and aes if simd disabled.
17767 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17769 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17770 (TARGET_SHA2): New feature flag for sha2.
17771 (TARGET_AES): New feature flag for aes.
17772 * config/aarch64/aarch64-simd.md:
17773 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17774 conditional on TARGET_AES.
17775 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17776 (aarch64_crypto_sha1hsi): Make pattern conditional
17778 (aarch64_crypto_sha1hv4si): Ditto.
17779 (aarch64_be_crypto_sha1hv4si): Ditto.
17780 (aarch64_crypto_sha1su1v4si): Ditto.
17781 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17782 (aarch64_crypto_sha1su0v4si): Ditto.
17783 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17784 (aarch64_crypto_sha256su0v4si): Ditto.
17785 (aarch64_crypto_sha256su1v4si): Ditto.
17786 (doc/invoke.texi): Document new aes and sha2 options.
17788 2018-01-10 Martin Sebor <msebor@redhat.com>
17790 PR tree-optimization/83781
17791 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17794 2018-01-11 Martin Sebor <msebor@gmail.com>
17795 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17797 PR tree-optimization/83501
17798 PR tree-optimization/81703
17800 * tree-ssa-strlen.c (get_string_cst): Rename...
17801 (get_string_len): ...to this. Handle global constants.
17802 (handle_char_store): Adjust.
17804 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17805 Jim Wilson <jimw@sifive.com>
17807 * config/riscv/riscv-protos.h (riscv_output_return): New.
17808 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17809 (riscv_attribute_table, riscv_output_return),
17810 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17811 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17812 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17814 (riscv_expand_prologue): Add early return for naked function.
17815 (riscv_expand_epilogue): Likewise.
17816 (riscv_function_ok_for_sibcall): Return false for naked function.
17817 (riscv_set_current_function): New.
17818 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17819 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17820 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17821 * doc/extend.texi (RISC-V Function Attributes): New.
17823 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17825 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17826 check for 128-bit long double before checking TCmode.
17827 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17828 128-bit long doubles before checking TFmode or TCmode.
17829 (FLOAT128_IBM_P): Likewise.
17831 2018-01-10 Martin Sebor <msebor@redhat.com>
17833 PR tree-optimization/83671
17834 * builtins.c (c_strlen): Unconditionally return zero for the empty
17836 Use -Warray-bounds for warnings.
17837 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17838 for non-constant array indices with COMPONENT_REF, arrays of
17839 arrays, and pointers to arrays.
17840 (gimple_fold_builtin_strlen): Determine and set length range for
17841 non-constant character arrays.
17843 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17845 PR middle-end/81897
17846 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17849 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17851 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17853 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17856 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17857 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17858 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17859 indexed_or_indirect_operand predicate.
17860 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17861 (*vsx_le_perm_load_v8hi): Likewise.
17862 (*vsx_le_perm_load_v16qi): Likewise.
17863 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17864 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17865 (*vsx_le_perm_store_v8hi): Likewise.
17866 (*vsx_le_perm_store_v16qi): Likewise.
17867 (eight unnamed splitters): Likewise.
17869 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17871 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17872 * config/rs6000/emmintrin.h: Likewise.
17873 * config/rs6000/mmintrin.h: Likewise.
17874 * config/rs6000/xmmintrin.h: Likewise.
17876 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17879 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17881 * tree.c (tree_nop_conversion): Return true for location wrapper
17883 (maybe_wrap_with_location): New function.
17884 (selftest::check_strip_nops): New function.
17885 (selftest::test_location_wrappers): New function.
17886 (selftest::tree_c_tests): Call it.
17887 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17888 (maybe_wrap_with_location): New decl.
17889 (EXPR_LOCATION_WRAPPER_P): New macro.
17890 (location_wrapper_p): New inline function.
17891 (tree_strip_any_location_wrapper): New inline function.
17893 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17896 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17897 stack_realign_offset for the largest alignment of stack slot
17899 (ix86_find_max_used_stack_alignment): New function.
17900 (ix86_finalize_stack_frame_flags): Use it. Set
17901 max_used_stack_alignment if we don't realign stack.
17902 * config/i386/i386.h (machine_function): Add
17903 max_used_stack_alignment.
17905 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17907 * config/arm/arm.opt (-mbranch-cost): New option.
17908 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17911 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17914 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17915 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17917 2018-01-10 Richard Biener <rguenther@suse.de>
17920 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17921 early out so it also covers the case where we have a non-NULL
17924 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17926 PR tree-optimization/83753
17927 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17928 for non-strided grouped accesses if the number of elements is 1.
17930 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17933 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17934 * i386.h (TARGET_USE_GATHER): Define.
17935 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17937 2018-01-10 Martin Liska <mliska@suse.cz>
17940 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17941 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17943 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17944 CLEANUP_NO_PARTITIONING is not set.
17946 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17948 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17949 for vectors, as a partial revert of r254296.
17950 * rtl.h (const_vec_p): Delete.
17951 (const_vec_duplicate_p): Don't test for vector CONSTs.
17952 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17953 * expmed.c (make_tree): Likewise.
17956 * common.md (E, F): Use CONSTANT_P instead of checking for
17958 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17959 checking for CONST_VECTOR.
17961 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17963 PR middle-end/83575
17964 * predict.c (force_edge_cold): Handle in more sane way edges
17965 with no prediction.
17967 2018-01-09 Carl Love <cel@us.ibm.com>
17969 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17971 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17972 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17973 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17974 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17975 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17976 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17977 * config/rs6000/rs6000-protos.h: Add extern defition for
17978 rs6000_generate_float2_double_code.
17979 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17981 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17982 (float2_v2df): Add define_expand.
17984 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17987 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17988 op_mode in the force_to_mode call.
17990 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17992 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17993 instead of checking each element individually.
17994 (aarch64_evpc_uzp): Likewise.
17995 (aarch64_evpc_zip): Likewise.
17996 (aarch64_evpc_ext): Likewise.
17997 (aarch64_evpc_rev): Likewise.
17998 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17999 instead of checking each element individually. Return true without
18001 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
18002 whether all selected elements come from the same input, instead of
18003 checking each element individually. Remove calls to gen_rtx_REG,
18004 start_sequence and end_sequence and instead assert that no rtl is
18007 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18009 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
18010 order of HIGH and CONST checks.
18012 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
18014 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
18015 if the destination isn't an SSA_NAME.
18017 2018-01-09 Richard Biener <rguenther@suse.de>
18019 PR tree-optimization/83668
18020 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18022 (canonicalize_loop_form): ... here, renamed from ...
18023 (canonicalize_loop_closed_ssa_form): ... this and amended to
18024 swap successor edges for loop exit blocks to make us use
18025 the RPO order we need for initial schedule generation.
18027 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18029 PR tree-optimization/64811
18030 * match.pd: When optimizing comparisons with Inf, avoid
18031 introducing or losing exceptions from comparisons with NaN.
18033 2018-01-09 Martin Liska <mliska@suse.cz>
18036 * asan.c (shadow_mem_size): Add gcc_assert.
18038 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18040 Don't save registers in main().
18043 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18044 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18045 * config/avr/avr.c (avr_set_current_function): Don't error if
18046 naked, OS_task or OS_main are specified at the same time.
18047 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18049 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18051 * common/config/avr/avr-common.c (avr_option_optimization_table):
18052 Switch on -mmain-is-OS_task for optimizing compilations.
18054 2018-01-09 Richard Biener <rguenther@suse.de>
18056 PR tree-optimization/83572
18057 * graphite.c: Include cfganal.h.
18058 (graphite_transform_loops): Connect infinite loops to exit
18059 and remove fake edges at the end.
18061 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18063 * ipa-inline.c (edge_badness): Revert accidental checkin.
18065 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18068 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18069 symbols; not inline clones.
18071 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18074 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18075 hard registers. Formatting fixes.
18077 PR preprocessor/83722
18078 * gcc.c (try_generate_repro): Pass
18079 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18080 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18083 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18084 Kito Cheng <kito.cheng@gmail.com>
18086 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18087 (riscv_leaf_function_p): Delete.
18088 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18090 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18092 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18094 (do_ifelse): New function.
18095 (do_isel): New function.
18096 (do_sub3): New function.
18097 (do_add3): New function.
18098 (do_load_mask_compare): New function.
18099 (do_overlap_load_compare): New function.
18100 (expand_compare_loop): New function.
18101 (expand_block_compare): Call expand_compare_loop() when appropriate.
18102 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18103 option description.
18104 (-mblock-compare-inline-loop-limit): New option.
18106 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18109 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18110 Reverse order of second and third operands in first alternative.
18111 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18112 of first and second elements in UNSPEC_VPERMR vector.
18113 (altivec_expand_vec_perm_le): Likewise.
18115 2018-01-08 Jeff Law <law@redhat.com>
18117 PR rtl-optimizatin/81308
18118 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18119 (process_switch): If group_case_labels makes a change, then set
18121 (pass_convert_switch::execute): If a switch is converted, then
18122 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18124 PR rtl-optimization/81308
18125 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18128 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18130 PR target/83663 - Revert r255946
18131 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18132 generation for cases where splatting a value is not useful.
18133 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18134 across a vec_duplicate and a paradoxical subreg forming a vector
18135 mode to a vec_concat.
18137 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18139 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18140 -march=armv8.3-a variants.
18141 * config/arm/t-multilib: Likewise.
18142 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18144 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18146 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18148 (cceq_ior_compare_complement): Give it a name so I can use it, and
18149 change boolean_or_operator predicate to boolean_operator so it can
18150 be used to generate a crand.
18151 (eqne): New code iterator.
18152 (bd/bd_neg): New code_attrs.
18153 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18154 a single define_insn.
18155 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18156 decrement (bdnzt/bdnzf/bdzt/bdzf).
18157 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18158 with the new names of the branch decrement patterns, and added the
18159 names of the branch decrement conditional patterns.
18161 2018-01-08 Richard Biener <rguenther@suse.de>
18163 PR tree-optimization/83563
18164 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18167 2018-01-08 Richard Biener <rguenther@suse.de>
18169 PR middle-end/83713
18170 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18172 2018-01-08 Richard Biener <rguenther@suse.de>
18174 PR tree-optimization/83685
18175 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18176 references to abnormals.
18178 2018-01-08 Richard Biener <rguenther@suse.de>
18181 * dwarf2out.c (output_indirect_strings): Handle empty
18182 skeleton_debug_str_hash.
18183 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18185 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18187 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18188 (emit_store_direct): Likewise.
18189 (arc_trampoline_adjust_address): Likewise.
18190 (arc_asm_trampoline_template): New function.
18191 (arc_initialize_trampoline): Use asm_trampoline_template.
18192 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18193 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18194 * config/arc/arc.md (flush_icache): Delete pattern.
18196 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18198 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18199 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18202 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18205 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18206 by not USED_FOR_TARGET.
18207 (make_pass_resolve_sw_modes): Likewise.
18209 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18211 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18214 2018-01-08 Richard Biener <rguenther@suse.de>
18216 PR middle-end/83580
18217 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18219 2018-01-08 Richard Biener <rguenther@suse.de>
18221 PR middle-end/83517
18222 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18224 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18226 PR middle-end/81897
18227 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18228 basic blocks with a small number of successors.
18229 (convert_control_dep_chain_into_preds): Improve handling of
18231 (dump_predicates): Split apart into...
18232 (dump_pred_chain): ...here...
18233 (dump_pred_info): ...and here.
18234 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18235 (can_chain_union_be_invalidated_p): Improve check for invalidation
18237 (uninit_uses_cannot_happen): Avoid unnecessary if
18238 convert_control_dep_chain_into_preds yielded nothing.
18240 2018-01-06 Martin Sebor <msebor@redhat.com>
18242 PR tree-optimization/83640
18243 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18244 subtracting negative offset from size.
18245 (builtin_access::overlap): Adjust offset bounds of the access to fall
18246 within the size of the object if possible.
18248 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18250 PR rtl-optimization/83699
18251 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18252 extract_bit_field_as_subreg to cases in which the extracted
18253 value is also a vector.
18255 * lra-constraints.c (process_alt_operands): Test for the equivalence
18256 substitutions when detecting a possible reload cycle.
18258 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18261 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18262 by default if flag_selective_schedling{,2}. Formatting fixes.
18264 PR rtl-optimization/83682
18265 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18266 if it has non-VECTOR_MODE element mode.
18267 (vec_duplicate_p): Likewise.
18269 PR middle-end/83694
18270 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18271 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18273 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18276 * config/i386/i386-builtin.def
18277 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18278 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18279 Require also OPTION_MASK_ISA_AVX512F in addition to
18280 OPTION_MASK_ISA_GFNI.
18281 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18282 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18283 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18284 to OPTION_MASK_ISA_GFNI.
18285 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18286 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18287 OPTION_MASK_ISA_AVX512BW.
18288 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18289 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18290 addition to OPTION_MASK_ISA_GFNI.
18291 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18292 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18293 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18294 to OPTION_MASK_ISA_GFNI.
18295 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18296 a requirement for all ISAs rather than any of them with a few
18298 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18300 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18301 bitmasks to be enabled with 3 exceptions, instead of requiring any
18302 enabled ISA with lots of exceptions.
18303 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18304 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18305 Change avx512bw in isa attribute to avx512f.
18306 * config/i386/sgxintrin.h: Add license boilerplate.
18307 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18308 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18309 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18310 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18312 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18313 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18314 temporarily sse2 rather than sse if not enabled already.
18317 * config/i386/sse.md (VI248_VLBW): Rename to ...
18318 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18319 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18320 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18321 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18322 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18323 mode iterator instead of VI248_VLBW.
18325 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18327 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18328 (record_modified): Skip clobbers; add debug output.
18329 (param_change_prob): Use sreal frequencies.
18331 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18333 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18334 punt for user-aligned variables.
18336 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18338 * tree-chrec.c (chrec_contains_symbols): Return true for
18341 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18344 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18345 of (x|y) == x for BICS pattern.
18347 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18349 PR tree-optimization/83605
18350 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18351 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18354 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18356 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18357 * config/epiphany/rtems.h: New file.
18359 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18360 Uros Bizjak <ubizjak@gmail.com>
18363 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18364 QIreg_operand instead of register_operand predicate.
18365 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18366 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18367 comments instead of -fmitigate[-_]rop.
18369 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18372 * cgraphunit.c (symbol_table::compile): Switch to text_section
18373 before calling assembly_start debug hook.
18374 * run-rtl-passes.c (run_rtl_passes): Likewise.
18377 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18379 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18380 range_int_cst_p rather than !symbolic_range_p before calling
18381 extract_range_from_multiplicative_op_1.
18383 2018-01-04 Jeff Law <law@redhat.com>
18385 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18386 redundant test in assertion.
18388 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18390 * doc/rtl.texi: Document machine_mode wrapper classes.
18392 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18394 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18395 using tree_to_uhwi.
18397 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18399 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18400 the VEC_PERM_EXPR fold to fail.
18402 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18405 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18406 to switched_sections.
18408 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18411 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18412 test for d.testing.
18414 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18417 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18418 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18420 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18423 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18424 is BLKmode and bitpos not zero or mode change is needed.
18426 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18429 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18432 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18435 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18436 instead of MULT rtx. Update all corresponding splitters.
18437 (*saddl_se): Ditto.
18438 (*ssub<modesuffix>): Ditto.
18439 (*ssubl_se): Ditto.
18440 (*cmp_sadd_di): Update split patterns.
18441 (*cmp_sadd_si): Ditto.
18442 (*cmp_sadd_sidi): Ditto.
18443 (*cmp_ssub_di): Ditto.
18444 (*cmp_ssub_si): Ditto.
18445 (*cmp_ssub_sidi): Ditto.
18446 * config/alpha/predicates.md (const23_operand): New predicate.
18447 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18448 Look for ASHIFT, not MULT inner operand.
18449 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18451 2018-01-04 Martin Liska <mliska@suse.cz>
18453 PR gcov-profile/83669
18454 * gcov.c (output_intermediate_file): Add version to intermediate
18456 * doc/gcov.texi: Document new field 'version' in intermediate
18457 file format. Fix location of '-k' option of gcov command.
18459 2018-01-04 Martin Liska <mliska@suse.cz>
18462 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18464 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18466 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18468 2018-01-03 Martin Sebor <msebor@redhat.com>
18470 PR tree-optimization/83655
18471 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18472 checking calls with invalid arguments.
18474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18476 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18477 (vectorizable_mask_load_store): Delete.
18478 (vectorizable_call): Return false for masked loads and stores.
18479 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18480 instead of gimple_assign_rhs1.
18481 (vectorizable_load): Handle IFN_MASK_LOAD.
18482 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18486 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18488 (vectorizable_mask_load_store): ...here.
18489 (vectorizable_load): ...and here.
18491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18493 * tree-vect-stmts.c (vect_build_all_ones_mask)
18494 (vect_build_zero_merge_argument): New functions, split out from...
18495 (vectorizable_load): ...here.
18497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18499 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18501 (vectorizable_mask_load_store): ...here.
18502 (vectorizable_store): ...and here.
18504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18506 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18508 (vectorizable_mask_load_store): ...here.
18510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18512 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18513 (vect_model_store_cost): Take a vec_load_store_type instead of a
18515 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18516 (vect_model_store_cost): Take a vec_load_store_type instead of a
18518 (vectorizable_mask_load_store): Update accordingly.
18519 (vectorizable_store): Likewise.
18520 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18524 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18525 IFN_MASK_LOAD calls here rather than...
18526 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18528 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18529 Alan Hayward <alan.hayward@arm.com>
18530 David Sherwood <david.sherwood@arm.com>
18532 * expmed.c (extract_bit_field_1): For vector extracts,
18533 fall back to extract_bit_field_as_subreg if vec_extract
18536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18537 Alan Hayward <alan.hayward@arm.com>
18538 David Sherwood <david.sherwood@arm.com>
18540 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18541 they are variable or constant sized.
18542 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18543 slots for constant-sized data.
18545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18546 Alan Hayward <alan.hayward@arm.com>
18547 David Sherwood <david.sherwood@arm.com>
18549 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18550 handling COND_EXPRs with boolean comparisons, try to find a better
18551 basis for the mask type than the boolean itself.
18553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18555 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18556 is calculated and how it can be overridden.
18557 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18558 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18560 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18564 Alan Hayward <alan.hayward@arm.com>
18565 David Sherwood <david.sherwood@arm.com>
18567 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18568 Remove the mode argument.
18569 (aarch64_simd_valid_immediate): Remove the mode and inverse
18571 * config/aarch64/iterators.md (bitsize): New iterator.
18572 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18573 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18574 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18575 aarch64_simd_valid_immediate.
18576 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18577 (aarch64_reg_or_bic_imm): Likewise.
18578 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18579 with an insn_type enum and msl with a modifier_type enum.
18580 Replace element_width with a scalar_mode. Change the shift
18581 to unsigned int. Add constructors for scalar_float_mode and
18582 scalar_int_mode elements.
18583 (aarch64_vect_float_const_representable_p): Delete.
18584 (aarch64_can_const_movi_rtx_p)
18585 (aarch64_simd_scalar_immediate_valid_for_move)
18586 (aarch64_simd_make_constant): Update call to
18587 aarch64_simd_valid_immediate.
18588 (aarch64_advsimd_valid_immediate_hs): New function.
18589 (aarch64_advsimd_valid_immediate): Likewise.
18590 (aarch64_simd_valid_immediate): Remove mode and inverse
18591 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18592 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18593 and aarch64_float_const_representable_p on the result.
18594 (aarch64_output_simd_mov_immediate): Remove mode argument.
18595 Update call to aarch64_simd_valid_immediate and use of
18596 simd_immediate_info.
18597 (aarch64_output_scalar_simd_mov_immediate): Update call
18600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18601 Alan Hayward <alan.hayward@arm.com>
18602 David Sherwood <david.sherwood@arm.com>
18604 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18605 (mode_nunits): Likewise CONST_MODE_NUNITS.
18606 * machmode.def (ADJUST_NUNITS): Document.
18607 * genmodes.c (mode_data::need_nunits_adj): New field.
18608 (blank_mode): Update accordingly.
18609 (adj_nunits): New variable.
18610 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18612 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18613 listed in adj_nunits.
18614 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18615 listed in adj_nunits. Don't emit case statements for such modes.
18616 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18617 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18618 nothing if adj_nunits is nonnull.
18619 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18620 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18621 (emit_mode_fbit): Update use of print_maybe_const_decl.
18622 (emit_move_size): Likewise. Treat the array as non-const
18624 (emit_mode_adjustments): Handle adj_nunits.
18626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18628 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18629 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18630 (VECTOR_MODES): Use it.
18631 (make_vector_modes): Take the prefix as an argument.
18633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18634 Alan Hayward <alan.hayward@arm.com>
18635 David Sherwood <david.sherwood@arm.com>
18637 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18638 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18639 for MODE_VECTOR_BOOL.
18640 * machmode.def (VECTOR_BOOL_MODE): Document.
18641 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18642 (make_vector_bool_mode): New function.
18643 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18645 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18646 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18648 * stor-layout.c (int_mode_for_mode): Likewise.
18649 * tree.c (build_vector_type_for_mode): Likewise.
18650 * varasm.c (output_constant_pool_2): Likewise.
18651 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18652 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18653 for MODE_VECTOR_BOOL.
18654 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18655 of mode class checks.
18656 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18657 instead of a list of mode class checks.
18658 (expand_vector_scalar_condition): Likewise.
18659 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18662 Alan Hayward <alan.hayward@arm.com>
18663 David Sherwood <david.sherwood@arm.com>
18665 * machmode.h (mode_size): Change from unsigned short to
18667 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18668 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18669 or if measurement_type is not polynomial.
18670 (fixed_size_mode::includes_p): Check for constant-sized modes.
18671 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18672 return a poly_uint16 rather than an unsigned short.
18673 (emit_mode_size): Change the type of mode_size from unsigned short
18674 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18675 (emit_mode_adjustments): Cope with polynomial vector sizes.
18676 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18678 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18680 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18681 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18682 * caller-save.c (setup_save_areas): Likewise.
18683 (replace_reg_with_saved_mem): Likewise.
18684 * calls.c (emit_library_call_value_1): Likewise.
18685 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18686 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18687 (gen_lowpart_for_combine): Likewise.
18688 * convert.c (convert_to_integer_1): Likewise.
18689 * cse.c (equiv_constant, cse_insn): Likewise.
18690 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18691 (cselib_subst_to_values): Likewise.
18692 * dce.c (word_dce_process_block): Likewise.
18693 * df-problems.c (df_word_lr_mark_ref): Likewise.
18694 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18695 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18696 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18697 (rtl_for_decl_location): Likewise.
18698 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18699 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18700 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18701 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18702 (expand_expr_real_1): Likewise.
18703 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18704 (pad_below): Likewise.
18705 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18706 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18707 * ira.c (get_subreg_tracking_sizes): Likewise.
18708 * ira-build.c (ira_create_allocno_objects): Likewise.
18709 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18710 (ira_sort_regnos_for_alter_reg): Likewise.
18711 * ira-costs.c (record_operand_costs): Likewise.
18712 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18713 (resolve_simple_move): Likewise.
18714 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18715 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18716 (lra_constraints): Likewise.
18717 (CONST_POOL_OK_P): Reject variable-sized modes.
18718 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18719 (add_pseudo_to_slot, lra_spill): Likewise.
18720 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18721 * optabs-query.c (get_best_extraction_insn): Likewise.
18722 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18723 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18724 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18725 * recog.c (offsettable_address_addr_space_p): Likewise.
18726 * regcprop.c (maybe_mode_change): Likewise.
18727 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18728 * regrename.c (build_def_use): Likewise.
18729 * regstat.c (dump_reg_info): Likewise.
18730 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18731 (find_reloads, find_reloads_subreg_address): Likewise.
18732 * reload1.c (eliminate_regs_1): Likewise.
18733 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18734 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18735 (simplify_binary_operation_1, simplify_subreg): Likewise.
18736 * targhooks.c (default_function_arg_padding): Likewise.
18737 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18738 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18739 (verify_gimple_assign_ternary): Likewise.
18740 * tree-inline.c (estimate_move_cost): Likewise.
18741 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18742 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18743 (get_address_cost_ainc): Likewise.
18744 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18745 (vect_supportable_dr_alignment): Likewise.
18746 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18747 (vectorizable_reduction): Likewise.
18748 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18749 (vectorizable_operation, vectorizable_load): Likewise.
18750 * tree.c (build_same_sized_truth_vector_type): Likewise.
18751 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18752 * var-tracking.c (emit_note_insn_var_location): Likewise.
18753 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18754 (ADDR_VEC_ALIGN): Likewise.
18756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18757 Alan Hayward <alan.hayward@arm.com>
18758 David Sherwood <david.sherwood@arm.com>
18760 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18762 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18763 or if measurement_type is polynomial.
18764 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18765 * combine.c (make_extraction): Likewise.
18766 * dse.c (find_shift_sequence): Likewise.
18767 * dwarf2out.c (mem_loc_descriptor): Likewise.
18768 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18769 (extract_bit_field, extract_low_bits): Likewise.
18770 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18771 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18772 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18773 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18774 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18775 * reload.c (find_reloads): Likewise.
18776 * reload1.c (alter_reg): Likewise.
18777 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18778 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18779 * tree-if-conv.c (predicate_mem_writes): Likewise.
18780 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18781 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18782 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18783 * valtrack.c (dead_debug_insert_temp): Likewise.
18784 * varasm.c (mergeable_constant_section): Likewise.
18785 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18787 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18788 Alan Hayward <alan.hayward@arm.com>
18789 David Sherwood <david.sherwood@arm.com>
18791 * expr.c (expand_assignment): Cope with polynomial mode sizes
18792 when assigning to a CONCAT.
18794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18795 Alan Hayward <alan.hayward@arm.com>
18796 David Sherwood <david.sherwood@arm.com>
18798 * machmode.h (mode_precision): Change from unsigned short to
18800 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18802 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18803 or if measurement_type is not polynomial.
18804 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18805 in which the mode is already known to be a scalar_int_mode.
18806 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18807 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18809 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18810 for GET_MODE_PRECISION.
18811 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18812 for GET_MODE_PRECISION.
18813 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18815 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18816 (expand_field_assignment, make_extraction): Likewise.
18817 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18818 (get_last_value): Likewise.
18819 * convert.c (convert_to_integer_1): Likewise.
18820 * cse.c (cse_insn): Likewise.
18821 * expr.c (expand_expr_real_1): Likewise.
18822 * lra-constraints.c (simplify_operand_subreg): Likewise.
18823 * optabs-query.c (can_atomic_load_p): Likewise.
18824 * optabs.c (expand_atomic_load): Likewise.
18825 (expand_atomic_store): Likewise.
18826 * ree.c (combine_reaching_defs): Likewise.
18827 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18828 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18829 * tree.h (type_has_mode_precision_p): Likewise.
18830 * ubsan.c (instrument_si_overflow): Likewise.
18832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18833 Alan Hayward <alan.hayward@arm.com>
18834 David Sherwood <david.sherwood@arm.com>
18836 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18837 polynomial numbers of units.
18838 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18839 (valid_vector_subparts_p): New function.
18840 (build_vector_type): Remove temporary shim and take the number
18841 of units as a poly_uint64 rather than an int.
18842 (build_opaque_vector_type): Take the number of units as a
18843 poly_uint64 rather than an int.
18844 * tree.c (build_vector_from_ctor): Handle polynomial
18845 TYPE_VECTOR_SUBPARTS.
18846 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18847 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18848 (build_vector_from_val): If the number of units is variable,
18849 use build_vec_duplicate_cst for constant operands and
18850 VEC_DUPLICATE_EXPR otherwise.
18851 (make_vector_type): Remove temporary is_constant ().
18852 (build_vector_type, build_opaque_vector_type): Take the number of
18853 units as a poly_uint64 rather than an int.
18854 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18856 * cfgexpand.c (expand_debug_expr): Likewise.
18857 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18858 (store_constructor, expand_expr_real_1): Likewise.
18859 (const_scalar_mask_from_tree): Likewise.
18860 * fold-const-call.c (fold_const_reduction): Likewise.
18861 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18862 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18863 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18864 (fold_relational_const): Likewise.
18865 (native_interpret_vector): Likewise. Change the size from an
18866 int to an unsigned int.
18867 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18868 TYPE_VECTOR_SUBPARTS.
18869 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18870 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18871 duplicating a non-constant operand into a variable-length vector.
18872 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18873 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18874 * ipa-icf.c (sem_variable::equals): Likewise.
18875 * match.pd: Likewise.
18876 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18877 * print-tree.c (print_node): Likewise.
18878 * stor-layout.c (layout_type): Likewise.
18879 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18880 * tree-cfg.c (verify_gimple_comparison): Likewise.
18881 (verify_gimple_assign_binary): Likewise.
18882 (verify_gimple_assign_ternary): Likewise.
18883 (verify_gimple_assign_single): Likewise.
18884 * tree-pretty-print.c (dump_generic_node): Likewise.
18885 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18886 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18887 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18888 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18889 (vect_shift_permute_load_chain): Likewise.
18890 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18891 (expand_vector_condition, optimize_vector_constructor): Likewise.
18892 (lower_vec_perm, get_compute_type): Likewise.
18893 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18894 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18895 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18896 (vect_recog_mask_conversion_pattern): Likewise.
18897 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18898 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18899 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18900 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18901 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18902 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18903 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18904 (supportable_widening_operation): Likewise.
18905 (supportable_narrowing_operation): Likewise.
18906 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18908 * varasm.c (output_constant): Likewise.
18910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18911 Alan Hayward <alan.hayward@arm.com>
18912 David Sherwood <david.sherwood@arm.com>
18914 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18915 so that both the length == 3 and length != 3 cases set up their
18916 own permute vectors. Add comments explaining why we know the
18917 number of elements is constant.
18918 (vect_permute_load_chain): Likewise.
18920 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18921 Alan Hayward <alan.hayward@arm.com>
18922 David Sherwood <david.sherwood@arm.com>
18924 * machmode.h (mode_nunits): Change from unsigned char to
18926 (ONLY_FIXED_SIZE_MODES): New macro.
18927 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18928 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18929 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18931 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18932 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18933 or if measurement_type is not polynomial.
18934 * genmodes.c (ZERO_COEFFS): New macro.
18935 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18937 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18938 Use ZERO_COEFFS when emitting initializers.
18939 * data-streamer.h (bp_pack_poly_value): New function.
18940 (bp_unpack_poly_value): Likewise.
18941 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18942 for GET_MODE_NUNITS.
18943 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18944 for GET_MODE_NUNITS.
18945 * tree.c (make_vector_type): Remove temporary shim and make
18946 the real function take the number of units as a poly_uint64
18947 rather than an int.
18948 (build_vector_type_for_mode): Handle polynomial nunits.
18949 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18950 * emit-rtl.c (const_vec_series_p_1): Likewise.
18951 (gen_rtx_CONST_VECTOR): Likewise.
18952 * fold-const.c (test_vec_duplicate_folding): Likewise.
18953 * genrecog.c (validate_pattern): Likewise.
18954 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18955 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18956 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18957 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18958 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18959 * rtlanal.c (subreg_get_info): Likewise.
18960 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18961 (vect_grouped_load_supported): Likewise.
18962 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18963 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18964 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18965 (simplify_const_unary_operation, simplify_binary_operation_1)
18966 (simplify_const_binary_operation, simplify_ternary_operation)
18967 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18968 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18969 instead of CONST_VECTOR_NUNITS.
18970 * varasm.c (output_constant_pool_2): Likewise.
18971 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18972 explicit-encoded elements in the XVEC for variable-length vectors.
18974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18976 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18979 Alan Hayward <alan.hayward@arm.com>
18980 David Sherwood <david.sherwood@arm.com>
18982 * coretypes.h (fixed_size_mode): Declare.
18983 (fixed_size_mode_pod): New typedef.
18984 * builtins.h (target_builtins::x_apply_args_mode)
18985 (target_builtins::x_apply_result_mode): Change type to
18986 fixed_size_mode_pod.
18987 * builtins.c (apply_args_size, apply_result_size, result_vector)
18988 (expand_builtin_apply_args_1, expand_builtin_apply)
18989 (expand_builtin_return): Update accordingly.
18991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18993 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18994 * cselib.c (cselib_hash_rtx): Likewise.
18995 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18996 CONST_VECTOR encoding.
18998 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18999 Jeff Law <law@redhat.com>
19002 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
19003 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
19004 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
19005 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
19008 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
19009 explicitly probe *sp in a noreturn function if there were any callee
19010 register saves or frame pointer is needed.
19012 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19015 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
19016 BLKmode for ternary, binary or unary expressions.
19019 * var-tracking.c (delete_vta_debug_insn): New inline function.
19020 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19021 insns from get_insns () to NULL instead of each bb separately.
19022 Use delete_vta_debug_insn. No longer static.
19023 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19024 delete_vta_debug_insns callers.
19025 * rtl.h (delete_vta_debug_insns): Declare.
19026 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19027 instead of variable_tracking_main.
19029 2018-01-03 Martin Sebor <msebor@redhat.com>
19031 PR tree-optimization/83603
19032 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19033 arguments past the endof the argument list in functions declared
19034 without a prototype.
19035 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19036 Avoid checking when arguments are null.
19038 2018-01-03 Martin Sebor <msebor@redhat.com>
19041 * doc/extend.texi (attribute const): Fix a typo.
19042 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19043 issuing -Wsuggest-attribute for void functions.
19045 2018-01-03 Martin Sebor <msebor@redhat.com>
19047 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19048 offset_int::from instead of wide_int::to_shwi.
19049 (maybe_diag_overlap): Remove assertion.
19050 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19051 * gimple-ssa-sprintf.c (format_directive): Same.
19052 (parse_directive): Same.
19053 (sprintf_dom_walker::compute_format_length): Same.
19054 (try_substitute_return_value): Same.
19056 2018-01-03 Jeff Law <law@redhat.com>
19058 PR middle-end/83654
19059 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19060 non-constant residual for zero at runtime and avoid probing in
19061 that case. Reorganize code for trailing problem to mirror handling
19064 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19066 PR tree-optimization/83501
19067 * tree-ssa-strlen.c (get_string_cst): New.
19068 (handle_char_store): Call get_string_cst.
19070 2018-01-03 Martin Liska <mliska@suse.cz>
19072 PR tree-optimization/83593
19073 * tree-ssa-strlen.c: Include tree-cfg.h.
19074 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19075 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19076 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19078 (strlen_dom_walker::before_dom_children): Call
19079 gimple_purge_dead_eh_edges. Dump tranformation with details
19081 (strlen_dom_walker::before_dom_children): Update call by adding
19082 new argument cleanup_eh.
19083 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19085 2018-01-03 Martin Liska <mliska@suse.cz>
19088 * cif-code.def (VARIADIC_THUNK): New enum value.
19089 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19092 2018-01-03 Jan Beulich <jbeulich@suse.com>
19094 * sse.md (mov<mode>_internal): Tighten condition for when to use
19095 vmovdqu<ssescalarsize> for TI and OI modes.
19097 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19099 Update copyright years.
19101 2018-01-03 Martin Liska <mliska@suse.cz>
19104 * ipa-visibility.c (function_and_variable_visibility): Skip
19105 functions with noipa attribure.
19107 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19109 * gcc.c (process_command): Update copyright notice dates.
19110 * gcov-dump.c (print_version): Ditto.
19111 * gcov.c (print_version): Ditto.
19112 * gcov-tool.c (print_version): Ditto.
19113 * gengtype.c (create_file): Ditto.
19114 * doc/cpp.texi: Bump @copying's copyright year.
19115 * doc/cppinternals.texi: Ditto.
19116 * doc/gcc.texi: Ditto.
19117 * doc/gccint.texi: Ditto.
19118 * doc/gcov.texi: Ditto.
19119 * doc/install.texi: Ditto.
19120 * doc/invoke.texi: Ditto.
19122 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19124 * vector-builder.h (vector_builder::m_full_nelts): Change from
19125 unsigned int to poly_uint64.
19126 (vector_builder::full_nelts): Update prototype accordingly.
19127 (vector_builder::new_vector): Likewise.
19128 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19129 (vector_builder::operator ==): Likewise.
19130 (vector_builder::finalize): Likewise.
19131 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19132 Take the number of elements as a poly_uint64 rather than an
19134 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19135 from unsigned int to poly_uint64.
19136 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19137 (vec_perm_indices::new_vector): Likewise.
19138 (vec_perm_indices::length): Likewise.
19139 (vec_perm_indices::nelts_per_input): Likewise.
19140 (vec_perm_indices::input_nelts): Likewise.
19141 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19142 number of elements per input as a poly_uint64 rather than an
19143 unsigned int. Use the original encoding for variable-length
19144 vectors, rather than clamping each individual element.
19145 For the second and subsequent elements in each pattern,
19146 clamp the step and base before clamping their sum.
19147 (vec_perm_indices::series_p): Handle polynomial element counts.
19148 (vec_perm_indices::all_in_range_p): Likewise.
19149 (vec_perm_indices_to_tree): Likewise.
19150 (vec_perm_indices_to_rtx): Likewise.
19151 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19152 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19153 (tree_vector_builder::new_binary_operation): Handle polynomial
19154 element counts. Return false if we need to know the number
19155 of elements at compile time.
19156 * fold-const.c (fold_vec_perm): Punt if the number of elements
19157 isn't known at compile time.
19159 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19161 * vec-perm-indices.h (vec_perm_builder): Change element type
19162 from HOST_WIDE_INT to poly_int64.
19163 (vec_perm_indices::element_type): Update accordingly.
19164 (vec_perm_indices::clamp): Handle polynomial element_types.
19165 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19166 (vec_perm_indices::all_in_range_p): Likewise.
19167 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19169 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19170 polynomial vec_perm_indices element types.
19171 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19172 * fold-const.c (fold_vec_perm): Likewise.
19173 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19174 * tree-vect-generic.c (lower_vec_perm): Likewise.
19175 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19176 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19177 element type to HOST_WIDE_INT.
19179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19180 Alan Hayward <alan.hayward@arm.com>
19181 David Sherwood <david.sherwood@arm.com>
19183 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19184 rather than an int. Use plus_constant.
19185 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19186 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19189 Alan Hayward <alan.hayward@arm.com>
19190 David Sherwood <david.sherwood@arm.com>
19192 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19193 a HOST_WIDE_INT to a poly_int64.
19195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19196 Alan Hayward <alan.hayward@arm.com>
19197 David Sherwood <david.sherwood@arm.com>
19199 * calls.c (load_register_parameters): Cope with polynomial
19200 mode sizes. Require a constant size for BLKmode parameters
19201 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19202 forces a parameter to be padded at the lsb end in order to
19203 fill a complete number of words, require the parameter size
19204 to be ordered wrt UNITS_PER_WORD.
19206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19207 Alan Hayward <alan.hayward@arm.com>
19208 David Sherwood <david.sherwood@arm.com>
19210 * reload1.c (spill_stack_slot_width): Change element type
19211 from unsigned int to poly_uint64_pod.
19212 (alter_reg): Treat mode sizes as polynomial.
19214 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19215 Alan Hayward <alan.hayward@arm.com>
19216 David Sherwood <david.sherwood@arm.com>
19218 * reload.c (complex_word_subreg_p): New function.
19219 (reload_inner_reg_of_subreg, push_reload): Use it.
19221 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19222 Alan Hayward <alan.hayward@arm.com>
19223 David Sherwood <david.sherwood@arm.com>
19225 * lra-constraints.c (process_alt_operands): Reject matched
19226 operands whose sizes aren't ordered.
19227 (match_reload): Refer to this check here.
19229 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19230 Alan Hayward <alan.hayward@arm.com>
19231 David Sherwood <david.sherwood@arm.com>
19233 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19234 that the mode size is in the set {1, 2, 4, 8, 16}.
19236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19237 Alan Hayward <alan.hayward@arm.com>
19238 David Sherwood <david.sherwood@arm.com>
19240 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19241 Use plus_constant instead of gen_rtx_PLUS.
19243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19244 Alan Hayward <alan.hayward@arm.com>
19245 David Sherwood <david.sherwood@arm.com>
19247 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19248 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19249 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19250 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19251 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19252 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19253 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19254 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19255 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19256 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19258 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19259 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19260 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19261 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19262 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19263 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19264 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19265 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19266 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19267 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19269 * expr.c (emit_move_resolve_push): Treat the input and result
19270 of PUSH_ROUNDING as a poly_int64.
19271 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19272 (emit_push_insn): Likewise.
19273 * lra-eliminations.c (mark_not_eliminable): Likewise.
19274 * recog.c (push_operand): Likewise.
19275 * reload1.c (elimination_effects): Likewise.
19276 * rtlanal.c (nonzero_bits1): Likewise.
19277 * calls.c (store_one_arg): Likewise. Require the padding to be
19278 known at compile time.
19280 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19281 Alan Hayward <alan.hayward@arm.com>
19282 David Sherwood <david.sherwood@arm.com>
19284 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19285 Use plus_constant instead of gen_rtx_PLUS.
19287 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19288 Alan Hayward <alan.hayward@arm.com>
19289 David Sherwood <david.sherwood@arm.com>
19291 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19292 rather than an int.
19294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19295 Alan Hayward <alan.hayward@arm.com>
19296 David Sherwood <david.sherwood@arm.com>
19298 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19299 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19300 via stack temporaries. Treat the mode size as polynomial too.
19302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19303 Alan Hayward <alan.hayward@arm.com>
19304 David Sherwood <david.sherwood@arm.com>
19306 * expr.c (expand_expr_real_2): When handling conversions involving
19307 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19308 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19309 as a poly_uint64 too.
19311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19312 Alan Hayward <alan.hayward@arm.com>
19313 David Sherwood <david.sherwood@arm.com>
19315 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19318 Alan Hayward <alan.hayward@arm.com>
19319 David Sherwood <david.sherwood@arm.com>
19321 * combine.c (can_change_dest_mode): Handle polynomial
19322 REGMODE_NATURAL_SIZE.
19323 * expmed.c (store_bit_field_1): Likewise.
19324 * expr.c (store_constructor): Likewise.
19325 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19326 and polynomial REGMODE_NATURAL_SIZE.
19327 (gen_lowpart_common): Likewise.
19328 * reginfo.c (record_subregs_of_mode): Likewise.
19329 * rtlanal.c (read_modify_subreg_p): Likewise.
19331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19332 Alan Hayward <alan.hayward@arm.com>
19333 David Sherwood <david.sherwood@arm.com>
19335 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19336 numbers of elements.
19338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19339 Alan Hayward <alan.hayward@arm.com>
19340 David Sherwood <david.sherwood@arm.com>
19342 * match.pd: Cope with polynomial numbers of vector elements.
19344 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19345 Alan Hayward <alan.hayward@arm.com>
19346 David Sherwood <david.sherwood@arm.com>
19348 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19349 in a POINTER_PLUS_EXPR.
19351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19352 Alan Hayward <alan.hayward@arm.com>
19353 David Sherwood <david.sherwood@arm.com>
19355 * omp-simd-clone.c (simd_clone_subparts): New function.
19356 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19357 (ipa_simd_modify_function_body): Likewise.
19359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19360 Alan Hayward <alan.hayward@arm.com>
19361 David Sherwood <david.sherwood@arm.com>
19363 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19364 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19365 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19366 (expand_vector_condition, vector_element): Likewise.
19367 (subparts_gt): New function.
19368 (get_compute_type): Use subparts_gt.
19369 (count_type_subparts): Delete.
19370 (expand_vector_operations_1): Use subparts_gt instead of
19371 count_type_subparts.
19373 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19374 Alan Hayward <alan.hayward@arm.com>
19375 David Sherwood <david.sherwood@arm.com>
19377 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19378 (vect_compile_time_alias): ...this new function. Do the calculation
19379 on poly_ints rather than trees.
19380 (vect_prune_runtime_alias_test_list): Update call accordingly.
19382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19383 Alan Hayward <alan.hayward@arm.com>
19384 David Sherwood <david.sherwood@arm.com>
19386 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19388 (vect_schedule_slp_instance): Likewise.
19390 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19391 Alan Hayward <alan.hayward@arm.com>
19392 David Sherwood <david.sherwood@arm.com>
19394 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19395 constant and extern definitions for variable-length vectors.
19396 (vect_get_constant_vectors): Note that the number of units
19397 is known to be constant.
19399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19400 Alan Hayward <alan.hayward@arm.com>
19401 David Sherwood <david.sherwood@arm.com>
19403 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19404 of units as polynomial. Choose between WIDE and NARROW based
19407 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19408 Alan Hayward <alan.hayward@arm.com>
19409 David Sherwood <david.sherwood@arm.com>
19411 * tree-vect-stmts.c (simd_clone_subparts): New function.
19412 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19415 Alan Hayward <alan.hayward@arm.com>
19416 David Sherwood <david.sherwood@arm.com>
19418 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19419 vectors as polynomial. Use build_index_vector for
19420 IFN_GOMP_SIMD_LANE.
19422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19423 Alan Hayward <alan.hayward@arm.com>
19424 David Sherwood <david.sherwood@arm.com>
19426 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19427 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19428 for variable-length vectors.
19429 (vectorizable_mask_load_store): Treat the number of units as
19430 polynomial, asserting that it is constant if the condition has
19431 already been enforced.
19432 (vectorizable_store, vectorizable_load): Likewise.
19434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19435 Alan Hayward <alan.hayward@arm.com>
19436 David Sherwood <david.sherwood@arm.com>
19438 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19439 of units as polynomial. Punt if we can't tell at compile time
19440 which vector contains the final result.
19442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19443 Alan Hayward <alan.hayward@arm.com>
19444 David Sherwood <david.sherwood@arm.com>
19446 * tree-vect-loop.c (vectorizable_induction): Treat the number
19447 of units as polynomial. Punt on SLP inductions. Use an integer
19448 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19449 cast of such a series for variable-length floating-point
19452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19453 Alan Hayward <alan.hayward@arm.com>
19454 David Sherwood <david.sherwood@arm.com>
19456 * tree.h (build_index_vector): Declare.
19457 * tree.c (build_index_vector): New function.
19458 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19459 of units as polynomial, forcibly converting it to a constant if
19460 vectorizable_reduction has already enforced the condition.
19461 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19462 to create a {1,2,3,...} vector.
19463 (vectorizable_reduction): Treat the number of units as polynomial.
19464 Choose vectype_in based on the largest scalar element size rather
19465 than the smallest number of units. Enforce the restrictions
19468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19469 Alan Hayward <alan.hayward@arm.com>
19470 David Sherwood <david.sherwood@arm.com>
19472 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19473 number of units as polynomial.
19475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19476 Alan Hayward <alan.hayward@arm.com>
19477 David Sherwood <david.sherwood@arm.com>
19479 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19480 * target.def (autovectorize_vector_sizes): Return the vector sizes
19481 by pointer, using vector_sizes rather than a bitmask.
19482 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19483 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19484 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19486 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19487 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19488 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19489 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19490 * omp-general.c (omp_max_vf): Likewise.
19491 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19492 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19493 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19494 * tree-vect-slp.c (vect_slp_bb): Likewise.
19495 * doc/tm.texi: Regenerate.
19496 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19498 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19499 the vector size as a poly_uint64 rather than an unsigned int.
19500 (current_vector_size): Change from an unsigned int to a poly_uint64.
19501 (get_vectype_for_scalar_type): Update accordingly.
19502 * tree.h (build_truth_vector_type): Take the size and number of
19503 units as a poly_uint64 rather than an unsigned int.
19504 (build_vector_type): Add a temporary overload that takes
19505 the number of units as a poly_uint64 rather than an unsigned int.
19506 * tree.c (make_vector_type): Likewise.
19507 (build_truth_vector_type): Take the number of units as a poly_uint64
19508 rather than an unsigned int.
19510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19511 Alan Hayward <alan.hayward@arm.com>
19512 David Sherwood <david.sherwood@arm.com>
19514 * target.def (get_mask_mode): Take the number of units and length
19515 as poly_uint64s rather than unsigned ints.
19516 * targhooks.h (default_get_mask_mode): Update accordingly.
19517 * targhooks.c (default_get_mask_mode): Likewise.
19518 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19519 * doc/tm.texi: Regenerate.
19521 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19522 Alan Hayward <alan.hayward@arm.com>
19523 David Sherwood <david.sherwood@arm.com>
19525 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19526 * omp-general.c (omp_max_vf): Likewise.
19527 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19528 (expand_omp_simd): Handle polynomial safelen.
19529 * omp-low.c (omplow_simd_context): Add a default constructor.
19530 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19531 (lower_rec_simd_input_clauses): Update accordingly.
19532 (lower_rec_input_clauses): Likewise.
19534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19535 Alan Hayward <alan.hayward@arm.com>
19536 David Sherwood <david.sherwood@arm.com>
19538 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19539 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19540 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19541 (vect_analyze_slp_cost): Likewise.
19542 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19543 (vect_model_load_cost): Likewise.
19545 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19546 Alan Hayward <alan.hayward@arm.com>
19547 David Sherwood <david.sherwood@arm.com>
19549 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19550 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19551 from an unsigned int * to a poly_uint64_pod *.
19552 (calculate_unrolling_factor): New function.
19553 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19556 Alan Hayward <alan.hayward@arm.com>
19557 David Sherwood <david.sherwood@arm.com>
19559 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19560 from an unsigned int to a poly_uint64.
19561 (_loop_vec_info::slp_unrolling_factor): Likewise.
19562 (_loop_vec_info::vectorization_factor): Change from an int
19564 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19565 (vect_get_num_vectors): New function.
19566 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19567 (vect_get_num_copies): Use vect_get_num_vectors.
19568 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19569 to an unsigned int *.
19570 (vect_analyze_data_refs): Change min_vf from an int * to a
19572 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19573 than an unsigned HOST_WIDE_INT.
19574 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19575 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19576 to an unsigned int *.
19577 (vect_analyze_data_ref_dependences): Likewise.
19578 (vect_compute_data_ref_alignment): Handle polynomial vf.
19579 (vect_enhance_data_refs_alignment): Likewise.
19580 (vect_prune_runtime_alias_test_list): Likewise.
19581 (vect_shift_permute_load_chain): Likewise.
19582 (vect_supportable_dr_alignment): Likewise.
19583 (dependence_distance_ge_vf): Take the vectorization factor as a
19584 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19585 (vect_analyze_data_refs): Change min_vf from an int * to a
19587 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19588 vfm1 as a poly_uint64 rather than an int. Make the same change
19589 for the returned bound_scalar.
19590 (vect_gen_vector_loop_niters): Handle polynomial vf.
19591 (vect_do_peeling): Likewise. Update call to
19592 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19593 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19595 * tree-vect-loop.c (vect_determine_vectorization_factor)
19596 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19597 (vect_get_known_peeling_cost): Likewise.
19598 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19599 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19600 (vect_transform_loop): Likewise. Use the lowest possible VF when
19601 updating the upper bounds of the loop.
19602 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19603 rather than an int.
19604 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19605 polynomial unroll factors.
19606 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19607 (vect_make_slp_decision): Likewise.
19608 (vect_supported_load_permutation_p): Likewise, and polynomial
19610 (vect_analyze_slp_cost): Handle polynomial vf.
19611 (vect_slp_analyze_node_operations): Likewise.
19612 (vect_slp_analyze_bb_1): Likewise.
19613 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19614 than an unsigned HOST_WIDE_INT.
19615 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19616 (vectorizable_load): Handle polynomial vf.
19617 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19619 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19621 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19622 Alan Hayward <alan.hayward@arm.com>
19623 David Sherwood <david.sherwood@arm.com>
19625 * match.pd: Handle bit operations involving three constants
19626 and try to fold one pair.
19628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19630 * tree-vect-loop-manip.c: Include gimple-fold.h.
19631 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19632 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19633 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19634 Add a path that uses a step of VF instead of 1, but disable it
19636 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19637 and niters_no_overflow parameters. Update calls to
19638 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19639 Create a new SSA name if the latter choses to use a ste other
19640 than zero, and return it via niters_vector_mult_vf_var.
19641 * tree-vect-loop.c (vect_transform_loop): Update calls to
19642 vect_do_peeling, vect_gen_vector_loop_niters and
19643 slpeel_make_loop_iterate_ntimes.
19644 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19645 (vect_gen_vector_loop_niters): Update declarations after above changes.
19647 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19649 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19650 128-bit round to integer instructions.
19651 (ceil<mode>2): Likewise.
19652 (btrunc<mode>2): Likewise.
19653 (round<mode>2): Likewise.
19655 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19657 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19658 unaligned VSX load/store on P8/P9.
19659 (expand_block_clear): Allow the use of unaligned VSX
19660 load/store on P8/P9.
19662 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19664 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19666 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19667 swap associated with both a load and a store.
19669 2018-01-02 Andrew Waterman <andrew@sifive.com>
19671 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19672 * config/riscv/riscv.md (clear_cache): Use it.
19674 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19676 * web.c: Remove out-of-date comment.
19678 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19680 * expr.c (fixup_args_size_notes): Check that any existing
19681 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19682 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19683 (emit_single_push_insn): ...here.
19685 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19687 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19688 (const_vector_encoded_nelts): New function.
19689 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19690 (const_vector_int_elt, const_vector_elt): Declare.
19691 * emit-rtl.c (const_vector_int_elt_1): New function.
19692 (const_vector_elt): Likewise.
19693 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19694 of CONST_VECTOR_ELT.
19696 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19698 * expr.c: Include rtx-vector-builder.h.
19699 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19700 directly on the tree encoding.
19701 (const_vector_from_tree): Likewise.
19702 * optabs.c: Include rtx-vector-builder.h.
19703 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19704 sequence of "u" values.
19705 * vec-perm-indices.c: Include rtx-vector-builder.h.
19706 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19707 directly on the vec_perm_indices encoding.
19709 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19711 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19712 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19713 * rtx-vector-builder.h: New file.
19714 * rtx-vector-builder.c: Likewise.
19715 * rtl.h (rtx_def::u2): Add a const_vector field.
19716 (CONST_VECTOR_NPATTERNS): New macro.
19717 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19718 (CONST_VECTOR_DUPLICATE_P): Likewise.
19719 (CONST_VECTOR_STEPPED_P): Likewise.
19720 (CONST_VECTOR_ENCODED_ELT): Likewise.
19721 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19722 (unwrap_const_vec_duplicate): Likewise.
19723 (const_vec_series_p): Check for a non-duplicated vector encoding.
19724 Say that the function only returns true for integer vectors.
19725 * emit-rtl.c: Include rtx-vector-builder.h.
19726 (gen_const_vec_duplicate_1): Delete.
19727 (gen_const_vector): Call gen_const_vec_duplicate instead of
19728 gen_const_vec_duplicate_1.
19729 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19730 (gen_const_vec_duplicate): Use rtx_vector_builder.
19731 (gen_const_vec_series): Likewise.
19732 (gen_rtx_CONST_VECTOR): Likewise.
19733 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19734 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19735 Build a new vector rather than modifying a CONST_VECTOR in-place.
19736 (handle_special_swappables): Update call accordingly.
19737 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19738 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19739 Build a new vector rather than modifying a CONST_VECTOR in-place.
19740 (handle_special_swappables): Update call accordingly.
19742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19744 * simplify-rtx.c (simplify_const_binary_operation): Use
19745 CONST_VECTOR_ELT instead of XVECEXP.
19747 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19749 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19750 the selector elements to be different from the data elements
19751 if the selector is a VECTOR_CST.
19752 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19753 ssizetype for the selector.
19755 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19757 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19758 before testing each element individually.
19759 * tree-vect-generic.c (lower_vec_perm): Likewise.
19761 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19763 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19764 * selftest-run-tests.c (selftest::run_tests): Call it.
19765 * vector-builder.h (vector_builder::operator ==): New function.
19766 (vector_builder::operator !=): Likewise.
19767 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19768 (vec_perm_indices::all_from_input_p): New function.
19769 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19770 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19771 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19772 instead of reading the VECTOR_CST directly. Detect whether both
19773 vector inputs are the same before constructing the vec_perm_indices,
19774 and update the number of inputs argument accordingly. Use the
19775 utility functions added above. Only construct sel2 if we need to.
19777 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19779 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19780 the broadcast of the low byte.
19781 (expand_mult_highpart): Use an explicit encoding for the permutes.
19782 * optabs-query.c (can_mult_highpart_p): Likewise.
19783 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19784 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19785 (vectorizable_bswap): Likewise.
19786 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19787 explicit encoding for the power-of-2 permutes.
19788 (vect_permute_store_chain): Likewise.
19789 (vect_grouped_load_supported): Likewise.
19790 (vect_permute_load_chain): Likewise.
19792 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19794 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19795 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19796 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19797 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19798 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19799 (vect_gen_perm_mask_any): Likewise.
19801 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19803 * int-vector-builder.h: New file.
19804 * vec-perm-indices.h: Include int-vector-builder.h.
19805 (vec_perm_indices): Redefine as an int_vector_builder.
19806 (auto_vec_perm_indices): Delete.
19807 (vec_perm_builder): Redefine as a stand-alone class.
19808 (vec_perm_indices::vec_perm_indices): New function.
19809 (vec_perm_indices::clamp): Likewise.
19810 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19811 (vec_perm_indices::new_vector): New function.
19812 (vec_perm_indices::new_expanded_vector): Update for new
19813 vec_perm_indices class.
19814 (vec_perm_indices::rotate_inputs): New function.
19815 (vec_perm_indices::all_in_range_p): Operate directly on the
19816 encoded form, without computing elided elements.
19817 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19818 encoding. Update for new vec_perm_indices class.
19819 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19820 the given vec_perm_builder.
19821 (expand_vec_perm_var): Update vec_perm_builder constructor.
19822 (expand_mult_highpart): Use vec_perm_builder instead of
19823 auto_vec_perm_indices.
19824 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19825 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19826 or double series encoding as appropriate.
19827 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19828 vec_perm_indices instead of auto_vec_perm_indices.
19829 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19830 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19831 (vect_permute_store_chain): Likewise.
19832 (vect_grouped_load_supported): Likewise.
19833 (vect_permute_load_chain): Likewise.
19834 (vect_shift_permute_load_chain): Likewise.
19835 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19836 (vect_transform_slp_perm_load): Likewise.
19837 (vect_schedule_slp_instance): Likewise.
19838 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19839 (vectorizable_mask_load_store): Likewise.
19840 (vectorizable_bswap): Likewise.
19841 (vectorizable_store): Likewise.
19842 (vectorizable_load): Likewise.
19843 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19844 vec_perm_indices instead of auto_vec_perm_indices. Use
19845 tree_to_vec_perm_builder to read the vector from a tree.
19846 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19847 vec_perm_builder instead of a vec_perm_indices.
19848 (have_whole_vector_shift): Use vec_perm_builder and
19849 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19850 truncation to calc_vec_perm_mask_for_shift.
19851 (vect_create_epilog_for_reduction): Likewise.
19852 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19853 from auto_vec_perm_indices to vec_perm_indices.
19854 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19855 instead of changing individual elements.
19856 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19857 the vector in d.perm.
19858 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19859 from auto_vec_perm_indices to vec_perm_indices.
19860 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19861 instead of changing individual elements.
19862 (arm_vectorize_vec_perm_const): Use new_vector to install
19863 the vector in d.perm.
19864 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19865 Update vec_perm_builder constructor.
19866 (rs6000_expand_interleave): Likewise.
19867 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19868 (rs6000_expand_interleave): Likewise.
19870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19872 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19873 to qimode could truncate the indices.
19874 * optabs.c (expand_vec_perm_var): Likewise.
19876 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19878 * Makefile.in (OBJS): Add vec-perm-indices.o.
19879 * vec-perm-indices.h: New file.
19880 * vec-perm-indices.c: Likewise.
19881 * target.h (vec_perm_indices): Replace with a forward class
19883 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19884 * optabs.h: Include vec-perm-indices.h.
19885 (expand_vec_perm): Delete.
19886 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19887 (expand_vec_perm_const): Declare.
19888 * target.def (vec_perm_const_ok): Replace with...
19889 (vec_perm_const): ...this new hook.
19890 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19891 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19892 * doc/tm.texi: Regenerate.
19893 * optabs.def (vec_perm_const): Delete.
19894 * doc/md.texi (vec_perm_const): Likewise.
19895 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19896 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19897 expand_vec_perm for constant permutation vectors. Assert that
19898 the mode of variable permutation vectors is the integer equivalent
19899 of the mode that is being permuted.
19900 * optabs-query.h (selector_fits_mode_p): Declare.
19901 * optabs-query.c: Include vec-perm-indices.h.
19902 (selector_fits_mode_p): New function.
19903 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19904 is defined, instead of checking whether the vec_perm_const_optab
19905 exists. Use targetm.vectorize.vec_perm_const instead of
19906 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19907 fit in the vector mode before using a variable permute.
19908 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19909 vec_perm_indices instead of an rtx.
19910 (expand_vec_perm): Replace with...
19911 (expand_vec_perm_const): ...this new function. Take the selector
19912 as a vec_perm_indices rather than an rtx. Also take the mode of
19913 the selector. Update call to shift_amt_for_vec_perm_mask.
19914 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19915 Use vec_perm_indices::new_expanded_vector to expand the original
19916 selector into bytes. Check whether the indices fit in the vector
19917 mode before using a variable permute.
19918 (expand_vec_perm_var): Make global.
19919 (expand_mult_highpart): Use expand_vec_perm_const.
19920 * fold-const.c: Includes vec-perm-indices.h.
19921 * tree-ssa-forwprop.c: Likewise.
19922 * tree-vect-data-refs.c: Likewise.
19923 * tree-vect-generic.c: Likewise.
19924 * tree-vect-loop.c: Likewise.
19925 * tree-vect-slp.c: Likewise.
19926 * tree-vect-stmts.c: Likewise.
19927 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19929 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19930 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19931 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19932 (aarch64_vectorize_vec_perm_const): ...this new function.
19933 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19934 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19935 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19936 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19937 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19938 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19939 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19941 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19942 check for NEON modes.
19943 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19944 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19945 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19946 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19948 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19949 the old VEC_PERM_CONST conditions.
19950 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19951 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19952 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19953 (ia64_vectorize_vec_perm_const_ok): Merge into...
19954 (ia64_vectorize_vec_perm_const): ...this new function.
19955 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19956 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19957 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19958 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19959 * config/mips/mips.c (mips_expand_vec_perm_const)
19960 (mips_vectorize_vec_perm_const_ok): Merge into...
19961 (mips_vectorize_vec_perm_const): ...this new function.
19962 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19963 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19964 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19965 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19966 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19967 (rs6000_expand_vec_perm_const): Delete.
19968 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19970 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19971 (altivec_expand_vec_perm_const_le): Take each operand individually.
19972 Operate on constant selectors rather than rtxes.
19973 (altivec_expand_vec_perm_const): Likewise. Update call to
19974 altivec_expand_vec_perm_const_le.
19975 (rs6000_expand_vec_perm_const): Delete.
19976 (rs6000_vectorize_vec_perm_const_ok): Delete.
19977 (rs6000_vectorize_vec_perm_const): New function.
19978 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19979 an element count and rtx array.
19980 (rs6000_expand_extract_even): Update call accordingly.
19981 (rs6000_expand_interleave): Likewise.
19982 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19983 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19984 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19985 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19986 (rs6000_expand_vec_perm_const): Delete.
19987 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19988 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19989 (altivec_expand_vec_perm_const_le): Take each operand individually.
19990 Operate on constant selectors rather than rtxes.
19991 (altivec_expand_vec_perm_const): Likewise. Update call to
19992 altivec_expand_vec_perm_const_le.
19993 (rs6000_expand_vec_perm_const): Delete.
19994 (rs6000_vectorize_vec_perm_const_ok): Delete.
19995 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19996 reference to the SPE evmerge intructions.
19997 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19998 an element count and rtx array.
19999 (rs6000_expand_extract_even): Update call accordingly.
20000 (rs6000_expand_interleave): Likewise.
20001 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
20002 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
20004 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
20006 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20008 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
20009 vector mode and that that mode matches the mode of the data
20011 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
20012 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
20013 directly using expand_vec_perm_1 when forcing selectors into
20015 (expand_vec_perm_var): New function, split out from expand_vec_perm.
20017 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20019 * optabs-query.h (can_vec_perm_p): Delete.
20020 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20021 * optabs-query.c (can_vec_perm_p): Split into...
20022 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20023 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20024 particular selector is valid.
20025 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20026 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20027 (vect_grouped_load_supported): Likewise.
20028 (vect_shift_permute_load_chain): Likewise.
20029 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20030 (vect_transform_slp_perm_load): Likewise.
20031 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20032 (vectorizable_bswap): Likewise.
20033 (vect_gen_perm_mask_checked): Likewise.
20034 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20035 implementations of variable permutation vectors into account
20036 when deciding which selector to use.
20037 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20038 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20039 with a false third argument.
20040 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20041 to test whether the constant selector is valid and can_vec_perm_var_p
20042 to test whether a variable selector is valid.
20044 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20046 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20047 * optabs-query.c (can_vec_perm_p): Likewise.
20048 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20049 instead of vec_perm_indices.
20050 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20051 (vect_gen_perm_mask_checked): Likewise,
20052 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20053 (vect_gen_perm_mask_checked): Likewise,
20055 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20057 * optabs-query.h (qimode_for_vec_perm): Declare.
20058 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20059 (qimode_for_vec_perm): ...this new function.
20060 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20062 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20064 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20065 does not have a conditional at the top.
20067 2018-01-02 Richard Biener <rguenther@suse.de>
20069 * ipa-inline.c (big_speedup_p): Fix expression.
20071 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20074 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20077 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20081 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20082 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20083 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20084 cond_taken_branch_cost 3->4.
20086 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20088 PR tree-optimization/83581
20089 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20090 TODO_cleanup_cfg if any changes have been made.
20092 PR middle-end/83608
20093 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20094 convert_modes if target mode has the right side, but different mode
20097 PR middle-end/83609
20098 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20099 last argument when extracting from CONCAT. If either from_real or
20100 from_imag is NULL, use expansion through memory. If result is not
20101 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20102 the parts directly to inner mode, if even that fails, use expansion
20105 PR middle-end/83623
20106 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20107 check for bswap in mode rather than HImode and use that in expand_unop
20110 Copyright (C) 2018 Free Software Foundation, Inc.
20112 Copying and distribution of this file, with or without modification,
20113 are permitted in any medium without royalty provided the copyright
20114 notice and this notice are preserved.