1 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
3 * config/arc/arc.c (arc_label_align): Use align_labels instead of
4 deprecated align_labels_log.
6 2018-07-17 Richard Biener <rguenther@suse.de>
9 * dwarf2out.c (init_sections_and_labels): Always generate
10 a debug_line_str_section for early LTO debug.
11 (dwarf2out_finish): Reset debug_line_str_hash output early.
12 Bump counter for extra dwarf5 .debug_loc labels to not conflict
14 (dwarf2out_early_finish): Output debug_line_str.
16 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
18 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
19 index register on z196 or later.
21 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
23 * config/s390/s390.c (s390_default_align): Set default function
25 (s390_override_options_after_change): Call s390_default align.
26 (s390_option_override_internal): Call s390_default_align.
27 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
29 2018-07-17 Jakub Jelinek <jakub@redhat.com>
32 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
33 to _looptemp_ clauses, other than the first two.
35 2018-07-17 Martin Liska <mliska@suse.cz>
37 * opts.c: Do not enable OPT_falign_* for -Os.
39 2018-07-17 Martin Liska <mliska@suse.cz>
41 * align.h (MAX_CODE_ALIGN): New.
42 (MAX_CODE_ALIGN_VALUE): New.
43 * common/config/i386/i386-common.c (ix86_handle_option):
44 (MAX_CODE_ALIGN): Moved to align.h.
45 * final.c (MAX_CODE_ALIGN): Likewise.
46 * opts.c (parse_and_check_align_values):
47 (MAX_CODE_ALIGN): Likewise.
48 (MAX_CODE_ALIGN_VALUE): Likewise.
50 2018-07-17 Martin Liska <mliska@suse.cz>
52 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
53 in order to fulfil coding style.
54 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
55 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
56 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
57 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
58 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
59 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
60 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
61 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
62 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
64 2018-07-17 Martin Liska <mliska@suse.cz>
67 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
68 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
69 align_flags of label_to_alignment.
70 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
72 * config/m68k/m68k.c: Do not use removed align_labels_value and
74 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
75 (LOOP_ALIGN): Likewise.
76 (LABEL_ALIGN): Likewise.
77 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
78 Remove not used macro.
79 (rs6000_loop_align): Change return type to align_flags.
80 (rs6000_loop_align_max_skip): Remove.
81 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
82 Change return type to align_flags.
83 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
84 Remove not used macro.
85 (rs6000_loop_align): Change return type to align_flags.
86 (rs6000_loop_align_max_skip): Remove.
87 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
88 * config/rx/rx-protos.h (rx_align_for_label): Make it
90 * config/rx/rx.c (rx_align_for_label): Change return type
92 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
94 into align_flags class.
95 (LABEL_ALIGN): Likewise.
96 (LOOP_ALIGN): Likewise.
97 * config/s390/s390.c (s390_label_align): Use align_flags
99 (s390_asm_output_function_label): Likewise.
100 * config/sh/sh.c (sh_override_options_after_change):
101 Use align_flags class directly without macros.
102 (find_barrier): Likewise.
103 (barrier_align): Likewise.
104 (sh_loop_align): Likewise.
105 * config/spu/spu.c (spu_option_override):
106 Use align_flags_tuple::get_value instead of removed macros.
107 (spu_sched_init): Likewise.
108 * config/spu/spu.h (GTY): Likewise.
109 * config/visium/visium.c (visium_option_override):
110 Set "8" as default secondary alignment.
111 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
112 in order to guarantee secondary alignment of 8.
113 * coretypes.h: Include align.h header file.
114 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
115 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
116 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
117 * doc/tm.texi.in: Likewise.
118 * final.c (struct label_alignment): Remove not used structure.
119 (LABEL_ALIGN): Change type to align_flags.
120 (LOOP_ALIGN): Likewise.
121 (JUMP_ALIGN): Likewise.
122 (default_loop_align_max_skip): Remove.
123 (default_label_align_max_skip): Likewise.
124 (default_jump_align_max_skip): Likewise.
125 (default_label_align_after_barrier_max_skip):
126 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
127 (LABEL_TO_MAX_SKIP): Remove.
128 (label_to_alignment): Return align_flags type instead of integer.
129 (label_to_max_skip): Remove.
130 (align_fuzz): Use align_flags type.
131 (compute_alignments): Use align_flags type and use align_flags::max
132 to combine multiple alignments.
133 (grow_label_align): Grow vec instead of C array.
134 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
135 (shorten_branches): Use align_flags type and use align_flags::max
136 to combine multiple alignments.
137 (final_scan_insn_1): Remove usage of secondary alignment that comes
138 from label alignment, but instead use proper secondary alignment
139 which is computed in grow_label_align.
140 * flags.h (struct align_flags_tuple): Move to align.h.
141 (struct align_flags): Likewise.
142 (state_align_loops): Rename to align_loops.
143 (state_align_jumps): Rename to align_jumps.
144 (state_align_labels): Rename to align_labels.
145 (state_align_functions): Rename to align_functions.
146 (align_loops_log): Remove.
147 (align_jumps_log): Remove.
148 (align_labels_log): Remove.
149 (align_functions_log): Remove.
150 (align_loops_max_skip): Remove.
151 (align_jumps_max_skip): Remove.
152 (align_labels_max_skip): Remove.
153 (align_functions_max_skip): Remove.
154 (align_loops_value): Remove.
155 (align_jumps_value): Remove.
156 (align_labels_value): Remove.
157 (align_functions_value): Remove.
158 * output.h (label_to_alignment): Change return type to align_flags.
159 (label_to_max_skip): Remove.
160 * target.def: Remove loop_align_max_skip, label_align_max_skip,
161 jump_align_max_skip macros.
162 * targhooks.h (default_loop_align_max_skip): Remove.
163 (default_label_align_max_skip): Likewise.
164 (default_jump_align_max_skip): Likewise.
165 (default_label_align_after_barrier_max_skip): Remove.
166 * toplev.c (read_log_maxskip): Use ::normalize function.
167 (parse_N_M): Remove not used argument and also call ::normalize.
168 (parse_alignment_opts): Do not pass unused arguments.
169 * varasm.c (assemble_start_function): Use directly align_functions
170 instead of removed macros.
171 * system.h: Do not poison removed macros.
173 2018-07-17 Jakub Jelinek <jakub@redhat.com>
176 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
177 and cond temporaries don't have reference type if iterator has
178 pointer type. For init use &for_pre_body instead of pre_p if
179 for_pre_body is non-empty.
181 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
183 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
184 double-double modes to SFmode directly directly.
185 (trunc<mode>sf2_fprs): Delete.
187 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
189 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
190 for conversions between IFmode and the decimal floating point modes.
191 (init_float128_ieee): Use the correct names for conversions between
192 KFmode and the decimal floating point modes.
194 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
196 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
197 for the conversions between TDmode and IFmode.
198 (init_float128_ieee): Use more correct names for the conversions
199 between TDmode and KFmode.
201 2018-07-16 Jakub Jelinek <jakub@redhat.com>
203 PR tree-optimization/86526
204 * builtins.c (expand_builtin_memcmp): Formatting fixes.
205 (inline_expand_builtin_string_cmp): Likewise.
206 (inline_string_cmp): Likewise. Use c_readstr instead of
207 builtin_memcpy_read_str. Add unit_mode temporary.
209 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
212 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
213 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
215 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
217 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
218 Alphabetize prototypes of built-in functions, separating out
219 built-in functions that are listed in this section but should be
222 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
225 * expmed.c (emit_store_flag): Do not emit setcc followed by a
226 conditional move when trapping comparison was split to a
227 non-trapping one (and vice versa).
229 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
231 * config/s390/s390.c (s390_function_profiler): Generate nops
232 instead of profiler call sequences.
233 * config/s390/s390.opt: Add the new option.
235 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
237 * config/s390/s390.c (s390_function_profiler): Generate
238 __mcount_loc section.
239 * config/s390/s390.opt: Add the new option.
241 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
243 * common.opt: Add the new warning.
244 * config/s390/s390.c (s390_function_profiler): Emit "brasl
245 %r0,__fentry__" when -mfentry is specified.
246 (s390_option_override_internal): Disallow -mfentry for 31-bit
248 * config/s390/s390.opt: Add the new option.
250 2018-07-16 Richard Biener <rguenther@suse.de>
253 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
254 for function-local FUNCTION_DECL and RESULT_DECL immediately.
256 2018-07-16 Martin Liska <mliska@suse.cz>
259 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
262 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
264 * config/arc/arcHS.md: Update ARCHS scheduling rules.
266 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
268 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
270 * config/arc/arc-cpus.def (hs4x): New cpu.
272 * config/arc/arc-tables.opt: Regenerate.
273 * config/arc/arc.c (arc_sched_issue_rate): New function.
274 (TARGET_SCHED_ISSUE_RATE): Define.
275 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
276 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
278 (attr tune): Add ARCHS4x tune values.
279 (attr tune_dspmpy): Define.
280 (*tst): Correct instruction type.
281 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
282 * config/arc/arcHS4x.md: New file.
283 * config/arc/fpu.md: Update instruction type attributes.
284 * config/arc/t-multilib: Regenerate.
286 2018-07-16 Tom de Vries <tdevries@suse.de>
289 * var-tracking.c (vt_initialize): Fix pre_dec handling.
291 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
293 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
296 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
298 PR tree-optimization/86514
299 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
300 conversion to a boolean type from a type with greater precision.
302 2018-07-16 Tom de Vries <tdevries@suse.de>
304 * var-tracking.c (vt_initialize): Print adjusted insn slim if
305 dump_flags request TDF_SLIM.
307 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
309 * fold-const.c (int_const_binop_1): Abstract...
310 (wide_int_binop): ...wide int code here.
311 (poly_int_binop): ...poly int code here.
312 Abstract the rest of int_const_binop_1 into int_const_binop.
313 * fold-const.h (wide_int_binop): New.
314 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
315 Remove useless PLUS/MINUS_EXPR case.
316 (zero_nonzero_bits_from_vr): Move wide int code...
317 (zero_nonzero_bits_from_bounds): ...here.
318 (extract_range_from_binary_expr_1): Move mask optimization code...
319 (range_easy_mask_min_max): ...here.
320 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
321 (range_easy_mask_min_max): New.
323 2018-07-15 Jeff Law <law@redhat.com>
326 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
329 2018-07-14 Jim Wilson <jimw@sifive.com>
331 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
333 2018-07-14 Paul Koning <ni1d@arrl.net>
335 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
337 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
339 * lto-streamer-out.c (copy_function_or_variable): Dump info about
342 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
343 Steve Munroe <munroesj52@gmail.com>
345 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
346 (_mm_andnot_si128): Likewise.
347 (_mm_or_si128): Likewise.
348 (_mm_xor_si128): Likewise.
350 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
353 * builtins.c (expand_builtin_memcmp): Inline the calls first
354 when result_eq is false.
355 (expand_builtin_strcmp): Inline the calls first.
356 (expand_builtin_strncmp): Likewise.
357 (inline_string_cmp): New routine. Expand a string compare
358 call by using a sequence of char comparison.
359 (inline_expand_builtin_string_cmp): New routine. Inline expansion
360 a call to str(n)cmp/memcmp.
361 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
363 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
365 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
367 * config/arm/driver-arm.c: Include arm-native.h.
368 (host_detect_local_cpu): Use auto-generated data tables.
369 (vendors, arm_cpu_table): Delete. Move part information to ...
370 * config/arm/arm-cpus.in: ... here.
371 * config/arm/parsecpu.awk (gen_native): New function.
372 (vendor, part): New CPU fields.
373 (END): Add support for building the native CPU detection tables.
374 * config/arm/t-arm (arm-native.h): Add build rule.
375 (driver-arm.o): Add dependency on arm-native.h.
377 2018-07-13 Richard Biener <rguenther@suse.de>
380 * match.pd (addr1 - addr2): Allow either of the operand to
383 2018-07-13 Tom de Vries <tdevries@suse.de>
385 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
388 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
390 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
391 arrays instead of numbered variables.
393 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
395 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
396 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
398 2018-07-13 Richard Biener <rguenther@suse.de>
401 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
402 instead of get_context_die.
404 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
405 Richard Biener <rguenther@suse.de>
408 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
409 that the loop latch destination where phi is defined.
411 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
413 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
414 (riscv_expand_epilogue): Add assertion to check interrupt mode.
415 (riscv_set_current_function): Extract getting interrupt type to new
417 (riscv_get_interrupt_type): New function.
418 (riscv_merge_decl_attributes): New function, checking interrupt type is
420 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
422 2018-07-12 Paul Koning <ni1d@arrl.net>
424 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
427 2018-07-12 Paul Koning <ni1d@arrl.net>
429 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
430 zero reference, add doloop_end instead.
431 * doc/md.texi (decrement_and_branch_until_zero): Remove.
432 (Looping patterns): Remove decrement_and_branch_until_zero. Add
433 detail for doloop_end.
435 2018-07-12 Martin Sebor <msebor@redhat.com>
438 * attribs.c (decl_attributes): Reject conflicting attributes before
439 calling attribute handlers.
441 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
443 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
445 (gcc::dump_manager::get_dump_file_name): likewise.
446 (dump_begin): Likewise.
447 * dumpfile.h (dump_begin): Update prototype.
448 (gcc::dump_manager::get_dump_file_name,
449 gcc::dump_manager::get_dump_file_name): Update prototype.
451 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
453 * internal-fn.h (vectorizable_internal_fn_p): New function.
454 * tree-vect-slp.c (compatible_calls_p): Likewise.
455 (vect_build_slp_tree_1): Remove nops argument. Handle calls
456 to internal functions.
457 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
459 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
461 * fold-const.h (inverse_conditions_p): Declare.
462 * fold-const.c (inverse_conditions_p): New function.
463 * match.pd: Use inverse_conditions_p. Add folds of view_converts
464 that test the inverse condition of a conditional internal function.
465 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
466 * internal-fn.c (internal_fn_mask_index): Handle conditional
468 (vectorized_internal_fn_supported_p): New function.
469 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
470 (any_pred_load_store): Replace with...
471 (need_to_predicate): ...this new variable.
472 (redundant_ssa_names): New variable.
473 (ifcvt_can_use_mask_load_store): Move initial checks to...
474 (ifcvt_can_predicate): ...this new function. Handle tree codes
475 for which a conditional internal function exists.
476 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
477 instead of ifcvt_can_use_mask_load_store. Update after variable
479 (predicate_load_or_store): New function, split out from
480 predicate_mem_writes.
481 (check_redundant_cond_expr): New function.
482 (value_available_p): Likewise.
483 (predicate_rhs_code): Likewise.
484 (predicate_mem_writes): Rename to...
485 (predicate_statements): ...this. Use predicate_load_or_store
486 and predicate_rhs_code.
487 (combine_blocks, tree_if_conversion): Update after above name changes.
488 (ifcvt_local_dce): Handle redundant_ssa_names.
489 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
490 general conditional functions.
491 * tree-vect-stmts.c (vectorizable_call): Likewise.
493 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
494 Alan Hayward <alan.hayward@arm.com>
495 David Sherwood <david.sherwood@arm.com>
497 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
498 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
499 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
500 plus and minus and convert them into IFN_COND_FMA-based sequences.
501 (convert_mult_to_fma): Handle conditional plus and minus.
503 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
505 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
506 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
507 (cond_fnms_optab): New optabs.
508 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
510 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
511 * internal-fn.h (get_conditional_internal_fn): Declare.
512 (get_unconditional_internal_fn): Likewise.
513 * internal-fn.c (cond_ternary_direct): New macro.
514 (expand_cond_ternary_optab_fn): Likewise.
515 (direct_cond_ternary_optab_supported_p): Likewise.
516 (FOR_EACH_COND_FN_PAIR): Likewise.
517 (get_conditional_internal_fn): New function.
518 (get_unconditional_internal_fn): Likewise.
519 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
520 (gimple_match_op::gimple_match_op): Add a new overload for 5
522 (gimple_match_op::set_op): Likewise.
523 (gimple_resimplify5): Declare.
524 * genmatch.c (decision_tree::gen): Generate simplifications for
526 * gimple-match-head.c (gimple_simplify): Define an overload for
527 5 operands. Handle calls with 5 arguments in the top-level overload.
528 (convert_conditional_op): Handle conversions from unconditional
529 internal functions to conditional ones.
530 (gimple_resimplify5): New function.
531 (build_call_internal): Pass a fifth operand.
532 (maybe_push_res_to_seq): Likewise.
533 (try_conditional_simplification): Try converting conditional
534 internal functions to unconditional internal functions.
535 Handle 3-operand unconditional forms.
536 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
537 Define ternary equivalents of the current rules for binary conditional
539 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
541 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
542 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
543 (optab): Handle them.
544 (SVE_COND_FP_TERNARY): New int iterator.
545 (sve_fmla_op, sve_fmad_op): New int attributes.
546 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
547 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
548 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
550 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
552 * target.def (preferred_else_value): New target hook.
553 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
554 * doc/tm.texi: Regenerate.
555 * targhooks.h (default_preferred_else_value): Declare.
556 * targhooks.c (default_preferred_else_value): New function.
557 * internal-fn.h (conditional_internal_fn_code): Declare.
558 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
559 (get_conditional_internal_fn): Use it.
560 (conditional_internal_fn_code): New function.
561 * gimple-match.h (gimple_match_cond): New struct.
562 (gimple_match_op): Add a cond member function.
563 (gimple_match_op::gimple_match_op): Update all forms to take a
565 * genmatch.c (expr::gen_transform): Use the same condition as res_op
566 for the suboperation, but don't specify a particular else_value.
567 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
568 (visit_nary_op, visit_reference_op_load): Pass
569 gimple_match_cond::UNCOND to the gimple_match_op constructor.
570 * gimple-match-head.c: Include tree-eh.h
571 (convert_conditional_op): New function.
572 (maybe_resimplify_conditional_op): Likewise.
573 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
574 (gimple_resimplify2): Likewise.
575 (gimple_resimplify3): Likewise.
576 (gimple_resimplify4): Likewise.
577 (maybe_push_res_to_seq): Return null for conditional operations.
578 (try_conditional_simplification): New function.
579 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
581 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
583 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
585 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
587 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
589 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
591 (hash_tree): Do not hash DECL_FCONTEXT
592 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
593 Do not stream DECL_FCONTEXT.
594 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
595 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
597 2018-07-12 Richard Biener <rguenther@suse.de>
600 * dwarf2out.c (gen_block_die): Only output blocks when they have
601 at least one !DECL_IGNORED_P variable.
603 2018-07-12 Richard Biener <rguenther@suse.de>
606 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
607 Remove -mieee-fp handling.
609 2018-07-12 Richard Biener <rguenther@suse.de>
611 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
612 left-over from last patch.
614 2018-07-12 Jakub Jelinek <jakub@redhat.com>
616 PR tree-optimization/86492
617 * gimple-ssa-store-merging.c
618 (imm_store_chain_info::coalesce_immediate_stores): Call
619 check_no_overlap even for the merge_overlapping case. Formatting fix.
621 2018-07-12 Richard Biener <rguenther@suse.de>
624 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
625 move possibly trapping operations into the conditional.
627 2018-07-12 Richard Biener <rguenther@suse.de>
629 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
630 (vn_lookup_simplify_result): Remove recursion limit applied
632 (vn_nary_build_or_lookup_1): Adjust.
633 (try_to_simplify): Likewise.
634 * gimple-match-head.c (gimple_resimplify1): Instead apply one
636 (gimple_resimplify2): Likewise.
637 (gimple_resimplify3): Likewise.
638 (gimple_resimplify4): Likewise.
640 2018-07-11 Jakub Jelinek <jakub@redhat.com>
642 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
643 Use __mmask64 type instead of __mmask8 for __M argument.
644 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
645 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
647 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
648 __mmask16 for __M argument.
649 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
650 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
651 to __mmask16 instead of __mmask8.
652 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
653 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
654 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
655 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
656 instead of __mmask16 for __U argument.
657 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
658 __mmask16 instead of __mmask8 for __U argument.
659 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
661 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
663 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
665 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
667 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
669 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
671 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
673 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
675 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
677 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
679 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
680 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
681 return type as well as __M argument type and all casts from __mmask8
683 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
684 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
685 return type as well as __M argument type and all casts from __mmask8
687 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
688 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
689 return type as well as __M argument type and all casts from __mmask8
691 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
692 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
693 return type as well as __M argument type and all casts from __mmask8
695 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
696 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
699 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
701 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
702 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
705 2018-07-11 Paul Koning <ni1d@arrl.net>
707 * doc/md.texi (define_subst): Document how multiple occurrences of
708 the same argument in the replacement pattern are handled.
710 2018-07-11 Paul Koning <ni1d@arrl.net>
712 * doc/extend.texi (Common Variable Attributes): Move "mode" into
714 (Common Type Attributes): Add "mode" attribute.
716 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
718 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
719 stream DECL_ORIGINAL_TYPE.
720 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
721 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
722 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
723 Do not walk original type.
724 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
726 (write_ts_decl_non_common_tree_pointers): Do not stream
728 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
729 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
731 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
733 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
734 thread twice from the same starting edge.
736 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
738 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
739 * gimple.c (gimple_call_nonnull_result_p): ...here...
740 (gimple_call_nonnull_arg): ...and here.
741 * gimple.h (gimple_call_nonnull_result_p): New.
742 (gimple_call_nonnull_arg): New.
744 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
746 * config/arm/arm-cpus.in: Move information from fpu field of each
747 cpu definition to the isa field.
748 * config/arm/parsecpu.awk (fpu): Delete match rule.
749 (gen_comm_data): Don't add bits from the CPU's FPU entry.
751 2018-07-11 Richard Biener <rguenther@suse.de>
754 * dwarf2out.c (init_sections_and_labels): Use
755 output_asm_line_debug_info consistently.
756 (dwarf2out_early_finish): Likewise.
757 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
760 2018-07-11 Richard Biener <rguenther@suse.de>
762 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
763 Rework father_bb setting in a way to avoid propagating constants
764 multiple times on a loop body.
766 2018-07-10 Mark Wielaard <mark@klomp.org>
769 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
771 2018-07-10 Richard Biener <rguenther@suse.de>
773 * hash-map.h (hash_map::iterator::operator*): Return
774 references to key and value.
776 2018-07-10 Jakub Jelinek <jakub@redhat.com>
779 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
780 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
781 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
782 (gimplify_omp_for): For composite loops, move outer
783 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
784 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
785 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
786 TREE_LIST for both the original class iterator and the "last" helper
787 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
788 loop, remember has_decl_expr from outer composite loops for the
789 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
791 2018-07-09 Martin Sebor <msebor@redhat.com>
795 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
796 accessing implicitly initialized array elements.
797 * expr.c (string_constant): Handle string initializers of
798 character arrays within aggregates.
799 * gimple-fold.c (fold_array_ctor_reference): Add argument.
800 Store element offset. As a special case, handle zero size.
801 (fold_nonarray_ctor_reference): Same.
802 (fold_ctor_reference): Add argument. Store subobject offset.
803 * gimple-fold.h (fold_ctor_reference): Add argument.
805 2018-07-09 Paul Koning <ni1d@arrl.net>
807 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
808 (pdp11_insn_cost): New function.
809 (pdp11_md_asm_adjust): New function.
810 (TARGET_INVALID_WITHIN_DOLOOP): Define.
811 (pdp11_rtx_costs): Update to match machine better.
812 (output_addr_const_pdp11): Correct format mismatch warnings.
813 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
814 * config/pdp11/pdp11.md: General change to add base_cost and/or
815 length attributes for use by new pdp11_insn_cost function.
816 (MIN_BRANCH): Correct definition.
818 (doloop_end): Use standard pattern name for looping pattern.
819 (doloop_end_nocc): New.
820 (movsf): Add another constraint alternative.
821 (zero_extendqihi2): Add constraint alternatives for not in place
823 (zero_extendhisi2): Remove.
824 (shift patterns): Add CC handling variants.
828 (define_peephole2): New peephole to recognize mov that sets CC for
831 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
833 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
834 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
837 2018-07-09 Jakub Jelinek <jakub@redhat.com>
840 * real.c (real_nextafter): Return true if result is denormal.
842 2018-07-09 Martin Liska <mliska@suse.cz>
844 * common.opt: Add back wrongly removed attribute.
846 2018-07-09 Richard Biener <rguenther@suse.de>
849 * dwarf2out.c (gen_block_die): For an early generated DIE
850 always output high/low PC attributes.
852 2018-07-09 Tom de Vries <tdevries@suse.de>
854 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
855 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
856 onto VAR_DECL with abstract origin.
858 2018-07-07 Jim Wilson <jimw@sifive.com>
860 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
862 2018-07-07 Tom de Vries <tdevries@suse.de>
864 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
866 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
868 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
870 (combine_bound): Use wide-int overflow calculation instead of
872 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
874 * fold-const.c (int_const_binop_2): Same.
875 (extract_muldiv_1): Same.
876 (fold_div_compare): Same.
877 (fold_abs_const): Same.
879 * poly-int.h (add): Same.
883 * predict.c (predict_iv_comparison): Same.
884 * profile-count.c (slow_safe_scale_64bit): Same.
885 * simplify-rtx.c (simplify_const_binary_operation): Same.
886 * tree-chrec.c (tree_fold_binomial): Same.
887 * tree-data-ref.c (split_constant_offset_1): Same.
888 * tree-if-conv.c (idx_within_array_bound): Same.
889 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
890 * tree-ssa-phiopt.c (minmax_replacement): Same.
891 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
892 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
893 * vr-values.c (vr_values::adjust_range_with_scev): Same.
894 * wide-int.cc (wi::add_large): Same.
895 (wi::mul_internal): Same.
896 (wi::sub_large): Same.
897 (wi::divmod_internal): Same.
898 * wide-int.h: Change overflow type to overflow_type for neg, add,
899 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
900 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
901 mul_internal, divmod_internal.
902 (overflow_type): New enum.
903 (accumulate_overflow): New.
905 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
907 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
908 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
910 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
912 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
913 argument is checked for zero before entering loop, avoid checking again.
915 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
917 * gimplify.h (generic_expr_could_trap_p): Set as global function.
918 * gimplify.h (generic_expr_could_trap_p): Likwise.
919 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
921 2018-07-06 Jakub Jelinek <jakub@redhat.com>
923 PR tree-optimization/86401
924 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
925 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
926 (fold_bit_and_mask): ... here. New helper function for match.pd.
927 * fold-const.h (fold_bit_and_mask): Declare.
928 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
930 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
933 * target.def (translate_mode_attribute): New hook.
934 * targhooks.h (default_translate_mode_attribute): Declare.
935 * targhooks.c (default_translate_mode_attribute): New function.
936 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
937 * doc/tm.texi: Regenerate.
938 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
939 (rs6000_translate_mode_attribute): New function.
941 2018-07-06 Paul Koning <ni1d@arrl.net>
943 * doc/md.texi (define_split): Document DONE and FAIL.
944 (define_peephole2): Ditto.
946 2018-07-05 Jeff Law <law@redhat.com>
948 PR tree-optimization/86010
949 * tree-ssa-dse.c (compute_trims): More aggressively trim at
950 both the head and tail of mem* and str* calls.
952 2018-07-05 Jim Wilson <jimw@sifive.com>
954 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
956 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
958 * config/aarch64/aarch64-simd.md: correct flags text for
959 MIN_EXPR replacement.
961 2018-07-05 James Clarke <jrtc27@jrtc27.com>
963 * configure: Regenerated.
965 2018-07-05 Carl Love <cel@us.ibm.com>
967 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
968 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
969 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
970 VSX_BUILTIN_DOUBLEL_V4SF.
972 2018-07-05 Martin Sebor <msebor@redhat.com>
975 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
976 than its domain to compute its the upper bound of a char array.
978 2018-07-05 Nathan Sidwell <nathan@acm.org>
980 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
981 * doc/cpp.texi: Update comment.
982 * doc/tm.texi: Rebuilt.
983 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
984 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
985 * doc/extend.texi (Backwards Compatibility): Clarify it is system
986 headers affected by extern "C".
987 * system.h: Poison NO_IMPLICIT_EXTERN_C.
988 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
989 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
990 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
991 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
992 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
993 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
994 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
995 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
996 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
997 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
998 config/sparc/sp64-elf.h, config/spu/spu.h,
999 config/stormy16/stormy16.h, config/v850/v850.h,
1000 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
1001 define NO_IMPLICIT_EXTERN_C.
1002 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
1004 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1007 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1008 instead of GET_MODE_SIZE when comparing Units.
1010 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1013 * rtlanal.c (set_noop_p): Constrain on mode change,
1014 include hard-reg-set.h
1016 2018-07-05 Tamar Christina <tamar.christina@arm.com>
1018 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
1020 2018-07-05 Jakub Jelinek <jakub@redhat.com>
1023 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1026 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1028 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1030 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
1033 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1035 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1037 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1040 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1042 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1044 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1045 neg_*_op* variables.
1047 2018-07-04 Martin Liska <mliska@suse.cz>
1049 * tree-switch-conversion.c: Define
1050 max_ratio_for_speed and max_ratio_for_size constants.
1052 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1053 Martin Liska <mliska@suse.cz>
1058 * common.opt: Rename align options with 'str_' prefix.
1059 * common/config/i386/i386-common.c (set_malign_value): New
1061 (ix86_handle_option): Use it to set -falign-* options/
1062 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1063 type from int to string.
1064 * config/aarch64/aarch64.c: Update default values from int
1066 * config/alpha/alpha.c (alpha_override_options_after_change):
1068 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1069 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1070 max skip conditionally.
1071 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1072 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1073 max skip conditionally.
1074 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1075 max skip conditionally.
1076 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1077 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1078 max skip conditionally.
1079 * config/i386/i386.c (struct ptt): Change type from int to
1081 (ix86_default_align): Set default values.
1082 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1083 max skip conditionally.
1084 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1085 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1086 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1087 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1088 max skip conditionally.
1089 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1090 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1091 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1092 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1093 max skip conditionally.
1094 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1095 * config/ia64/ia64.c (ia64_option_override): Set default values
1096 for alignment options.
1097 * config/m68k/m68k.c: Handle new str_align_* options.
1098 * config/mips/mips.c (mips_set_compression_mode): Change
1100 (mips_option_override): Set default values for options.
1101 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1103 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1105 * config/rx/rx.c (rx_option_override): Likewise.
1106 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1107 (LABEL_ALIGN): Use align_labels_log.
1108 (LOOP_ALIGN): Use align_loops_align.
1109 * config/s390/s390.c (s390_asm_output_function_label): Use new
1111 * config/sh/sh.c (sh_override_options_after_change):
1112 Change type of constants.
1113 * config/spu/spu.c (spu_sched_init): Likewise.
1114 * config/sparc/sparc.c (sparc_option_override): Set default
1116 * config/visium/visium.c (visium_option_override): Likewise.
1117 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1118 emit p2align format with last argument if it's not needed.
1119 * doc/invoke.texi: Document extended format of -falign-*.
1120 * final.c: Use align_labels alignment.
1121 * flags.h (struct target_flag_state): Change type to use
1123 (struct align_flags_tuple): New.
1124 (struct align_flags): Likewise.
1125 (align_loops_log): Redefine macro to use new types.
1126 (align_loops_max_skip): Redefine macro to use new types.
1127 (align_jumps_log): Redefine macro to use new types.
1128 (align_jumps_max_skip): Redefine macro to use new types.
1129 (align_labels_log): Redefine macro to use new types.
1130 (align_labels_max_skip): Redefine macro to use new types.
1131 (align_functions_log): Redefine macro to use new types.
1132 (align_loops): Redefine macro to use new types.
1133 (align_jumps): Redefine macro to use new types.
1134 (align_labels): Redefine macro to use new types.
1135 (align_functions): Redefine macro to use new types.
1136 (align_functions_max_skip): Redefine macro to use new types.
1137 (align_loops_value): New macro.
1138 (align_jumps_value): New macro.
1139 (align_labels_value): New macro.
1140 (align_functions_value): New macro.
1141 * function.c (invoke_set_current_function_hook): Propagate
1142 alignment values from flags to global variables default in
1144 * ipa-icf.c (sem_function::equals_wpa): Use
1145 cl_optimization_option_eq instead of memcmp.
1146 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1148 (cl_optimization_stream_in): Likewise.
1149 * optc-save-gen.awk: Support strings in cl_optimization.
1150 * opth-gen.awk: Likewise.
1151 * opts.c (finish_options): Remove error checking of invalid
1153 (MAX_CODE_ALIGN): Remove.
1154 (MAX_CODE_ALIGN_VALUE): Likewise.
1155 (parse_and_check_align_values): New function.
1156 (check_alignment_argument): Likewise.
1157 (common_handle_option): Use check_alignment_argument.
1158 * opts.h (parse_and_check_align_values): Declare.
1159 * toplev.c (init_alignments): Remove.
1160 (read_log_maxskip): New.
1161 (parse_N_M): Likewise.
1162 (parse_alignment_opts): Likewise.
1163 (backend_init_target): Remove usage of init_alignments.
1164 * toplev.h (parse_alignment_opts): Declare.
1165 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1167 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1168 * tree.c (cl_option_hasher::equal): New.
1169 * varasm.c: Use new global macros.
1171 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1173 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1174 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1175 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1176 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1177 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1178 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1179 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1180 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1181 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1182 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1183 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1184 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1185 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1187 2018-07-04 Martin Liska <mliska@suse.cz>
1188 Jonathan Wakely <jwakely@redhat.com>
1190 * coverage.c: Use correct type.
1191 * doc/invoke.texi: Language correction.
1193 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1196 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1197 ENDBRANCH for non-tail call which may return via indirect branch.
1198 * doc/extend.texi: Document indirect_return attribute.
1200 2018-07-03 Martin Sebor <msebor@redhat.com>
1202 PR tree-optimization/86274
1203 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1205 (format_floating): Correct handling of infinities and NaNs.
1207 2018-07-03 Martin Sebor <msebor@redhat.com>
1209 * print-tree.c (print_real_cst): New function.
1210 (print_node_brief): Call it.
1211 (print_node): Ditto.
1213 2018-07-03 Jeff Law <law@redhat.com>
1215 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1216 into a single pattern.
1218 * config/h8300/h8300.md (ors code_iterator): New.
1219 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1220 a single pattern and single splitter.
1221 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1222 (iorqi3_1, xorqi3_1): Likewise.
1223 (iorqi3, xorqi3 expanders): Similarly.
1225 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1226 (movmd_internal) into a single pattern using the P mode iterator.
1227 (movmd splitters): Similarly.
1228 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1229 (movsd splitters): Similarly.
1231 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1232 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1233 right table based on the size of the operand.
1234 * config/h8300/h8300.md (length_table): Corresponding changes. All
1235 references to "addb", "addw" and "addl" changed to "add".
1236 (btst patterns): Merge two variants into a single pattern.
1237 (tstqi, tsthi): Likewise.
1238 (addhi3_incdec, addsi3_incdec): Likewise.
1239 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1240 (mulhi3, mulsi3): Likewise.
1241 (udivhi3, udivsi3): Likewise.
1242 (divhi3, divsi3): Likewise.
1243 (andorqi3, andorhi3, andorsi3): Likewise.
1245 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1248 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1249 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1251 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1253 PR tree-optimization/85694
1254 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1255 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1257 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1258 (<u>avg<mode>3_ceil): New patterns.
1260 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1262 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1263 scan-tree-dump directive.
1264 * gcc.dg/vect/slp-perm-2.c: Likewise.
1265 * gcc.dg/vect/slp-perm-3.c: Likewise.
1266 * gcc.dg/vect/slp-perm-5.c: Likewise.
1267 * gcc.dg/vect/slp-perm-6.c: Likewise.
1268 * gcc.dg/vect/slp-perm-7.c: Likewise.
1269 * gcc.dg/vect/slp-perm-8.c: Likewise.
1271 2018-07-03 Marek Polacek <polacek@redhat.com>
1274 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1276 2018-07-03 Richard Biener <rguenther@suse.de>
1279 * tree-ssa-structalias.c (find_func_clobbers): Properly
1280 handle indirect calls.
1282 2018-07-03 Jeff Law <law@redhat.com>
1284 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1285 (shifts): New code iterator.
1286 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1287 expander. Fix HImode handling on H8/SX.
1288 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1289 (subqi3, subhi3, subsi3 expanders): Likewise.
1290 (andqi3, andhi3, andsi3 expanders): Likewise.
1291 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1292 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1293 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1294 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1295 (zero_extendqihi2, zero_extendqisi2): Likewise.
1296 (extendqihi2, extendqisi2): Likewise.
1297 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1298 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1299 (rotlqi3_1, rotlhi3_1): Likewise.
1300 (logicalhi3_sn, logicalsi3_sn): Likewise.
1301 (logicalhi3, logicalsi3): Likewise.
1303 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1305 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1306 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1307 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1308 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1309 type to append_pattern_def_seq instead of creating a stmt_vec_info
1311 (build_mask_conversion): Likewise. Remove vinfo argument.
1312 (vect_add_conversion_to_patterm): Likewise, renaming to...
1313 (vect_add_conversion_to_pattern): ...this.
1314 (vect_recog_mask_conversion_pattern): Update call to
1315 build_mask_conversion. Pass the vector type to
1316 append_pattern_def_seq here too.
1317 (vect_recog_gather_scatter_pattern): Update call to
1318 vect_add_conversion_to_pattern.
1320 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1322 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1323 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1324 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1325 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1326 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1327 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1328 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1329 append_pattern_def_seq instead of new_pattern_def_seq.
1330 (vect_recog_divmod_pattern): Do both of the above.
1331 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1334 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1336 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1337 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1338 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1339 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1340 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1341 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1342 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1343 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1344 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1345 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1346 parameter with a single stmt_vec_info.
1347 (vect_recog_func_ptr): Likewise.
1348 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1349 (vect_try_gather_scatter_pattern): ...this.
1350 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1351 the stmt_vec_info of the statement to be matched. Don't clear
1352 STMT_VINFO_RELATED_STMT.
1353 (vect_pattern_recog): Update call accordingly.
1355 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1357 PR tree-optimization/85694
1358 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1359 (uavgM3_ceil): Document new optabs.
1360 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1361 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1363 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1364 (savg_ceil_optab): New optabs.
1365 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1366 (vect_vect_recog_func_ptrs): Add it.
1367 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1368 constant directly from the associated lhs.
1370 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1372 * tree-vect-patterns.c (vect_split_statement): New function.
1373 (vect_convert_input): Use it to try to split an existing cast.
1375 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1377 * poly-int.h (print_hex): New function.
1378 * dumpfile.h (dump_dec, dump_hex): Declare.
1379 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1380 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1381 min_input_precision, operation_precision and operation_sign.
1382 * tree-vect-patterns.c (vect_get_range_info): New function.
1383 (vect_same_loop_or_bb_p, vect_single_imm_use)
1384 (vect_operation_fits_smaller_type): Delete.
1385 (vect_look_through_possible_promotion): Add an optional
1386 single_use_p parameter.
1387 (vect_recog_over_widening_pattern): Rewrite to use new
1388 stmt_vec_info infomration. Handle one operation at a time.
1389 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1390 (vect_truncatable_operation_p, vect_set_operation_type)
1391 (vect_set_min_input_precision): New functions.
1392 (vect_determine_min_output_precision_1): Likewise.
1393 (vect_determine_min_output_precision): Likewise.
1394 (vect_determine_precisions_from_range): Likewise.
1395 (vect_determine_precisions_from_users): Likewise.
1396 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1397 (vect_vect_recog_func_ptrs): Put over_widening first.
1399 (vect_pattern_recog): Call vect_determine_precisions.
1401 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1403 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1404 statements that have been replaced by further pattern statements.
1405 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1407 2018-07-03 Richard Biener <rguenther@suse.de>
1409 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1410 always set *dt. Dump vectype in vectype overload.
1411 * dumpfile.h (dump_gimple_expr): New function.
1412 (dump_gimple_expr_loc): Likewise.
1413 * dumpfile.c (dump_gimple_expr): New function.
1414 (dump_gimple_expr_loc): Likewise.
1416 2018-07-02 Jeff Law <law@redhat.com>
1418 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1419 the H8/300, H8/300H and H8/S variants into a single pattern.
1420 (movhi_h8300, movqi_h8300hs): Similarly.
1421 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1422 (QHI mode iterator): New.
1424 * config/h8300/h8300.md: Remove trailing whitespace.
1426 2018-07-02 Jim Wilson <jimw@sifive.com>
1428 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1429 instead of emit_insn for interrupt returns.
1430 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1431 (riscv_sret, riscv_uret): Likewise.
1433 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1435 * pretty-print.c (selftest::test_pp_format): Move save and restore
1436 of quotes to class auto_fix_quotes, and add an instance.
1437 * selftest.c: Include "intl.h".
1438 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1439 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1440 * selftest.h (selftest::auto_fix_quotes): New class.
1442 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1444 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1445 (aarch64_sve_prepare_conditional_op): Remove.
1446 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1447 Allow aarch64_simd_reg_or_zero as select operand; remove
1448 the aarch64_sve_prepare_conditional_op call.
1449 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1450 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1451 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1452 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1453 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1454 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1455 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1456 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1457 and a splitters to match all of the *_any patterns.
1458 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1460 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1461 (SVE_COND_FP_BINARY_REV): Remove.
1462 (sve_int_op_rev, sve_fp_op_rev): New.
1463 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1464 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1465 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1466 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1467 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1468 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1469 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1470 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1471 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1473 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1474 Remove match_dup 1 from the inner unspec.
1475 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1477 * config/aarch64/aarch64.md (movprfx): New attr.
1478 (length): Default movprfx to 8.
1479 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1480 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1481 (*<su>mul<SVE_I>3_highpart): Likewise.
1482 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1483 (*v<ASHIFT><SVE_I>3): Likewise.
1484 (*<su><MAXMIN><SVE_I>3): Likewise.
1485 (*<su><MAXMIN><SVE_F>3): Likewise.
1486 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1487 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1488 (*div<SVE_F>4): Likewise.
1490 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1492 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1495 2018-07-02 Richard Biener <rguenther@suse.de>
1497 PR tree-optimization/86363
1498 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1499 memset argument refers to a non-variable address.
1501 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1503 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1504 {PLUS,MINUS}_EXPR code to...
1505 (adjust_symbolic_bound): ...here,
1506 (combine_bound): ...here,
1507 (set_value_range_with_overflow): ...and here.
1509 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1511 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1513 (extract_range_from_abs_expr): ...here.
1515 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1517 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1518 -fno-omit-frame-pointer when not optimizing.
1520 2018-07-02 Martin Liska <mliska@suse.cz>
1523 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1524 (propagate_nothrow): Likewise.
1526 2018-07-02 Martin Liska <mliska@suse.cz>
1529 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1531 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1533 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1534 function in r262149, changing "loc" param from source_location to
1535 const dump_location_t &.
1536 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1537 declaration, as above.
1539 2018-07-01 Paul Koning <ni1d@arrl.net>
1541 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1542 -munit-asm, -mgnu-asm, -mdec-asm.
1543 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1544 (pdp11_output_labelref): New.
1545 (pdp11_output_def): New.
1546 (pdp11_output_addr_vec_elt): New.
1547 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1548 %# and %@ format codes.
1549 (pdp11_option_override): New.
1550 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1551 (pdp11_output_ident): New.
1552 (pdp11_asm_named_section): New.
1553 (pdp11_asm_init_sections): New.
1554 (pdp11_file_start): New.
1555 (pdp11_file_end): New.
1556 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1557 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1558 %o, like %c but octal.
1559 (pdp11_option_override): New.
1560 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1562 (DATA_SECTION_ASM_OP): Ditto.
1563 (READONLY_DATA_SECTION_ASM_OP): New.
1564 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1565 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1566 (ASM_OUTPUT_LABELREF): Ditto.
1567 (ASM_OUTPUT_DEF): Ditto.
1568 (ASM_OUTPUT_EXTERNAL): New.
1569 (ASM_OUTPUT_SOURCE_FILENAME): New.
1570 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1571 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1572 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1573 %# and %@ format codes.
1574 * config/pdp11/pdp11.opt (mgnu-asm): New.
1575 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1576 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1577 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1579 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1581 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1582 dereferencing path[] beyond its length.
1584 (debug_all_paths): New.
1585 (rewire_first_differing_edge): New.
1586 (adjust_paths_after_duplication): New.
1587 (duplicate_thread_path): Call adjust_paths_after_duplication.
1589 (thread_through_all_blocks): Add new argument to
1590 duplicate_thread_path.
1592 2018-06-30 Jim Wilson <jimw@sifive.com>
1594 * config/riscv/predicates.md (p2m1_shift_operand): New.
1595 (high_mask_shift_operand): New.
1596 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1597 pattern using p2m1_shift_operand.
1598 (lshsi3_zero_extend_3+2): New combiner pattern using
1599 high_mask_shift_operand.
1601 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1603 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1605 (vect_recog_rotate_pattern): ...here.
1606 (vect_convert_input): Try to insert casts of invariants in the
1608 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1609 preheader to be empty.
1611 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1613 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1614 vector type. If given, install it in the new statement's
1616 (vect_element_precision): New function.
1617 (vect_unpromoted_value): New struct.
1618 (vect_unpromoted_value::vect_unpromoted_value): New function.
1619 (vect_unpromoted_value::set_op): Likewise.
1620 (vect_look_through_possible_promotion): Likewise.
1621 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1622 (vect_widened_op_tree, vect_convert_input): Likewise.
1623 (vect_convert_inputs, vect_convert_output): Likewise.
1624 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1625 to handle the optional cast of the multiplication result and
1626 vect_widened_op_tree to detect the widened multiplication itself.
1627 Do not require the input and output of promotion casts to have
1628 the same sign, but base the signedness of the operation on the
1629 input rather than the result. If the pattern includes two
1630 promotions, check that those promotions have the same sign.
1631 Do not restrict the MULT_EXPR handling to a double-width result;
1632 handle quadruple-width results and wider. Use vect_convert_inputs
1633 to convert the inputs to the common type.
1634 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1635 to handle the optional cast of the ABS result. Also allow a sign
1636 change or a sign extension between the ABS and MINUS.
1637 Use vect_widened_op_tree to detect the widened subtraction and use
1638 vect_convert_inputs to convert the inputs to the common type.
1639 (vect_handle_widen_op_by_const): Delete.
1640 (vect_recog_widen_op_pattern): New function.
1641 (vect_recog_widen_mult_pattern): Use it.
1642 (vect_recog_widen_shift_pattern): Likewise.
1643 (vect_recog_widen_sum_pattern): Use
1644 vect_look_through_possible_promotion to handle the promoted
1647 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1649 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1650 the containing gimple_seq *.
1651 * gimple-iterator.h (gsi_for_stmt): Declare it.
1652 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1653 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1654 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1655 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1656 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1658 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1660 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1661 statement being replaced is part of an existing pattern
1662 definition sequence, inserting the new pattern statements before
1664 (vect_pattern_recog_1): Don't return a bool. If the statement
1665 is already part of a pattern, instead apply pattern matching
1666 to the pattern definition statements. Don't clear the
1667 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1668 (vect_pattern_recog): Don't break after the first match;
1669 continue processing the pattern definition statements instead.
1670 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1672 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1674 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1675 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1676 (vect_recog_widen_sum_pattern): Use it.
1678 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1680 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1681 phi is not a pattern statement and has not been replaced by
1682 a pattern statement.
1683 * tree-vect-patterns.c (type_conversion_p): Don't check
1684 STMT_VINFO_IN_PATTERN_P.
1685 (vect_recog_vector_vector_shift_pattern): Likewise.
1686 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1687 the pattern statement rather than the original statement; check
1688 directly for a WIDEN_MULT_EXPR here.
1689 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1690 vect_is_simple_use to return the pattern statement rather
1691 than the original statement; use is_pattern_stmt_p to check
1692 for such a pattern statement.
1693 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1694 to return the pattern statement rather than the original statement;
1695 don't do the same transformation here.
1696 (vect_is_simple_use): If the defining statement has been replaced
1697 by a pattern statement, return the pattern statement instead.
1698 Remove the corresponding (local) transformation from the vectype
1701 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1703 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1704 end and default to null.
1705 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1706 (vectorizable_reduction): Update calls accordingly, dropping the
1707 gimple ** argument if the passed-back statement isn't needed.
1708 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1709 (vect_recog_rotate_pattern): Likewise.
1710 (vect_recog_mask_conversion_pattern): Likewise.
1711 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1712 (vect_mask_constant_operand_p): Likewise.
1713 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1714 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1715 (get_group_load_store_type, get_load_store_type): Likewise.
1716 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1717 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1718 (vectorizable_conversion, vectorizable_assignment): Likewise.
1719 (vectorizable_shift, vectorizable_operation): Likewise.
1720 (vectorizable_store, vect_is_simple_cond): Likewise.
1721 (vectorizable_condition, vectorizable_comparison): Likewise.
1722 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1723 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1724 and move it to the end. Cope with null def_stmt_outs.
1726 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1728 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1730 2018-06-29 Jeff Law <law@redhat.com>
1732 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1733 displacements for TARGET_V850E2V3 and newer.
1734 (TARGET_LRA_P): Remove. Defaults to LRA now.
1735 * config/v850/v850.md (sign23byte_load): Remove.
1736 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1737 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1739 2018-06-29 Martin Liska <mliska@suse.cz>
1742 * coverage.c (coverage_init): Mangle full path name.
1743 * doc/invoke.texi: Document the change.
1744 * gcov-io.c (mangle_path): New.
1745 * gcov-io.h (mangle_path): Likewise.
1746 * gcov.c (mangle_name): Use mangle_path for path mangling.
1748 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1750 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1751 if starting source register is not even.
1753 2018-06-29 Martin Liska <mliska@suse.cz>
1755 PR tree-optimization/86263
1756 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1757 Make edge redirection.
1759 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1761 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1762 (dump_scope_depth): New variable.
1763 (get_dump_scope_depth): New function.
1764 (dump_begin_scope): New function.
1765 (dump_end_scope): New function.
1766 * dumpfile.h (get_dump_scope_depth): New declaration.
1767 (dump_begin_scope): New declaration.
1768 (dump_end_scope): New declaration.
1769 (class auto_dump_scope): New class.
1770 (AUTO_DUMP_SCOPE): New macro.
1771 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1774 2018-06-29 Richard Biener <rguenther@suse.de>
1776 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1777 compute_all_dependences succeeds.
1778 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1779 exceed --param loop-max-datarefs-for-datadeps.
1781 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1783 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1785 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1788 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1789 alternative 0 in preferred_for_speed attribute.
1791 2018-06-28 Paul Koning <ni1d@arrl.net>
1793 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1794 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1795 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1796 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1797 for shift insn patterns.
1799 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1801 * cgraph.c (cgraph_node::get_body): Replace assignments to
1802 "dump_file" with calls to set_dump_file.
1803 * dumpfile.c (alt_dump_file): Make static, and group with...
1804 (alt_flags): ...this definition.
1805 (dumps_are_enabled): New variable.
1806 (refresh_dumps_are_enabled): New function.
1807 (set_dump_file): New function.
1808 (set_alt_dump_file): New function.
1809 (gcc::dump_manager::dump_start): Replace assignments to
1810 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1812 (gcc::dump_manager::dump_finish): Likewise.
1813 * dumpfile.h (alt_dump_file): Delete decl.
1814 (dumps_are_enabled): New variable decl.
1815 (set_dump_file): New function decl.
1816 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1818 * tree-nested.c (lower_nested_functions): Replace assignments to
1819 "dump_file" with calls to set_dump_file.
1821 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1823 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1824 goto_locus of each outgoing edge of each basic block.
1826 2018-06-28 Richard Biener <rguenther@suse.de>
1828 * dwarf2out.c (decl_scope_table): Remove.
1829 (push_decl_scope): Likewise.
1830 (pop_decl_scope): Likewise.
1831 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1832 (gen_struct_or_union_type_die): Likewise.
1833 (gen_tagged_type_die): Likewise.
1834 (dwarf2out_init): Do not initialize decl_scope_table.
1835 (dwarf2out_c_finalize): Do not free it.
1837 2018-06-28 Richard Biener <rguenther@suse.de>
1839 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1840 deciding whether to not re-use a DIE.
1842 2018-06-28 Richard Biener <rguenther@suse.de>
1844 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1845 DW_AT_abstract_origin attribute.
1847 2018-06-28 Martin Liska <mliska@suse.cz>
1849 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1850 Use newly introduced constants.
1851 * tree-switch-conversion.h (struct jump_table_cluster):
1852 Define max_ratio_for_size and max_ratio_for_speed.
1854 2018-06-28 Martin Liska <mliska@suse.cz>
1856 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1857 Add new checking assert to catch invalid state.
1858 (jump_table_cluster::can_be_handled): Handle single case
1860 (jump_table_cluster::is_beneficial): Bail out for such case.
1861 (bit_test_cluster::find_bit_tests):
1862 Add new checking assert to catch invalid state.
1863 (bit_test_cluster::can_be_handled): Handle single case
1865 (bit_test_cluster::is_beneficial): Bail out for such case.
1866 (switch_decision_tree::analyze_switch_statement):
1869 2018-06-28 Martin Liska <mliska@suse.cz>
1871 * common.opt: Introduce -completion option.
1872 * gcc.c (driver_handle_option): Handle it.
1873 (driver::main): Print completions if completion
1875 * opt-suggestions.c (option_proposer::get_completions):
1877 (option_proposer::suggest_completion): Likewise.
1878 (option_proposer::find_param_completions): Likewise.
1879 (verify_autocompletions): Likewise.
1880 (test_completion_valid_options): Likewise.
1881 (test_completion_valid_params): Likewise.
1882 (in_completion_p): Likewise.
1883 (empty_completion_p): Likewise.
1884 (test_completion_partial_match): Likewise.
1885 (test_completion_garbage): Likewise.
1886 (opt_proposer_c_tests): Likewise.
1887 * opt-suggestions.h: Declare new functions.
1888 * opts.c (common_handle_option): Handle OPT__completion_.
1889 * selftest-run-tests.c (selftest::run_tests): Add
1890 opt_proposer_c_tests.
1891 * selftest.c (assert_str_startswith): New.
1892 * selftest.h (assert_str_startswith): Likewise.
1893 (opt_proposer_c_tests): New.
1894 (ASSERT_STR_STARTSWITH): Likewise.
1896 2018-06-28 Martin Liska <mliska@suse.cz>
1898 * Makefile.in: Add opt-suggestions.o.
1899 * gcc-main.c: Include opt-suggestions.h.
1900 * gcc.c (driver::driver): Likewise.
1901 (driver::~driver): Remove m_option_suggestions.
1902 (driver::build_option_suggestions): Moved to option_proposer.
1903 (driver::suggest_option): Likewise.
1904 (driver::handle_unrecognized_options): Use option_proposer.
1905 * gcc.h (class driver): Add new memver m_option_proposer.
1906 * opt-suggestions.c: New file.
1907 * opt-suggestions.h: New file.
1909 2018-06-28 Martin Liska <mliska@suse.cz>
1911 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1912 (auto_string_vec::~auto_string_vec): Likewise.
1914 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1916 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1917 prevent_decl_creation_for_types fields up and add reset_location field.
1918 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1919 statement if id->reset_location is true.
1920 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1921 id->reset_location is true.
1922 (copy_phis_for_bb): Force input_location on the arguments if
1923 id->reset_location is true.
1924 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1925 is set on the function to be inlined.
1927 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1929 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1931 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1933 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1934 registers for Pmode.
1935 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1936 hard registers for the clobbered pseudo.
1938 2018-06-27 Paul Koning <ni1d@arrl.net>
1940 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1941 mutually exclusive options.
1942 * config/pdp11/constraints.md (h): New constraint.
1943 (O): Update definition to match shift code generation.
1944 (D): New constraint.
1945 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1947 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1949 (output_jump): Change arguments.
1950 (pdp11_fixed_cc_regs): New function.
1951 (pdp11_cc_mode): Ditto.
1952 (pdp11_expand_shift): Ditto.
1953 (pdp11_assemble_shift): Ditto.
1954 (pdp11_small_shift): Ditto.
1955 (pdp11_branch_cost): Remove.
1956 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1958 (pdp11_register_move_cost): Update for CC registers.
1959 (pdp11_rtx_costs): Add case for LSHIFTRT.
1960 (pdp11_output_jump): Add CCNZ mode conditional branches.
1961 (notice_update_cc_on_set): Remove.
1962 (pdp11_cc_mode): New function.
1963 (simple_memory_operand): Correct pre/post decrement case.
1964 (no_side_effect_operand): New function.
1965 (pdp11_regno_reg_class): Add CC_REGS class.
1966 (pdp11_fixed_cc_regs): New function.
1967 (pdp11_small_shift): New function.
1968 (pdp11_expand_shift): New function to expand shift insns.
1969 (pdp11_assemble_shift): New function to output shifts.
1970 (pdp11_branch_cost): Remove.
1971 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1972 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1973 (WCHAR_TYPE): Ditto.
1974 (PTRDIFF_TYPE): Ditto.
1975 (ADJUST_INSN_LENGTH): New macro.
1976 (FIXED_REGISTERS): Add CC registers.
1977 (CALL_USED_REGISTERS): Ditto.
1979 (REG_CLASS_NAMES): Ditto.
1980 (REG_CLASS_CONTENTS): Ditto.
1981 (SELECT_CC_MODE): Use new function.
1982 (TARGET_FLAGS_REGNUM): New macro.
1983 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1984 (cc0_reg_rtx): Remove.
1985 (CC_STATUS_MDEP): Remove.
1986 (CC_STATUS_MDEFP_INIT): Remove.
1987 (CC_IN_FPU): Remove.
1988 (NOTICE_UPDATE_CC): Remove.
1989 (REGISTER_NAMES): Add CC registers.
1990 (BRANCH_COST): Change to constant 1.
1991 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1993 * config/pdp11/pdp11.opt (mbcopy): Remove.
1994 (mbcopy-builtin): Remove.
1995 (mbranch-cheap): Remove.
1996 (mbranch-expensive): Remove.
1997 * config/pdp11/predicates.md (expand_shift_operand): Update to
1998 match shift code generation.
1999 (ccnz_operator): New predicate.
2000 * doc/invoke.texi (PDP-11 Options): Remove deleted options
2001 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
2002 Remove non-existent option -mabshi, -mno-abshi. Document mutually
2004 * doc/md.texi (PDP-11): Document new D and h constraints. Update
2005 description of O constraint.
2007 2018-06-27 Jeff Law <law@redhat.com>
2008 Austin Law <austinklaw@gmail.com>
2010 * config/v850/v850.md (addsi3_set_flags): New pattern.
2011 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
2012 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
2013 (zero_extendhisi2_v850_set_flags): Likewise.
2014 (zero_extendqisi2_v850_set_flags): Likewise.
2015 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
2016 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
2017 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
2019 * config/v850/v850-protos.h (notice_update_cc): Remove.
2020 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
2021 (v850_print_operand): Handle 'D' and "d".
2022 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
2023 Add handling of arithmetic/logical operations compared against zero.
2024 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
2025 Do not look at v850_compare_op, instead get mode from last argument.
2026 (v850_gen_compare): Remove
2027 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
2028 after reload for prologue insns.
2029 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
2031 (construct_save_jarl): Likewise.
2032 (TARGET_FLAGS_REGNUM): Define.
2033 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2034 (NOTICE_UPDATE_CC): Remove.
2035 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2036 than cc0. Conditionalize on reload_completed.
2037 (cmpsi_insn, setfcc_insn): Likewise.
2038 (tst1 splitter): Turn into define_and_split which sets the flags
2040 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2041 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2042 (cstoresf4, cstoredf4): Clobber the flags.
2043 (cmpsi, cmpsf, cmpdf): Remove expanders.
2044 (setf_insn): Remove pattern.
2045 (addsi3): Turn into define_and_split which clobbers the flags after
2046 reload and a suitable pattern (addsi3_clobber_flags) for use after
2048 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2049 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2050 (ashrsi3, ashrsi3_v850e2): Likewise.
2051 (bins): Clobber the flags.
2052 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2053 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2054 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2055 (call_value_internal_short, call_value_internal_long): Likewise.
2056 (callt_save_interrupt, callt_return_interrupt): Likewise.
2057 (save_interrupt, return_interrupt): Likewise.
2058 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2059 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2060 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2061 (All FP comparisons): Only allow after reload has completed.
2063 (divh, divhu): Tweak output template.
2064 (branch_z_normal, branch_z_invert): Remove
2065 (branch_nz_normal, branch_nz_invert): Likewise.
2066 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2068 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2069 * config/v850/v850.c (notice_update_cc): Remove.
2070 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2071 (CC_NO_CARRY): Likewise.
2072 (NOTICE_UPDATE_CC): Define to nothing.
2073 * config/v850/v850.md: Remove block comment on cc0 handling
2074 Remove "cc" attribute from all patterns. Remove cc_status handling
2075 from all patterns. Minor formatting fixes.
2077 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2079 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2080 (cortex-a76.cortex-a55): Likewise.
2081 * config/aarch64/aarch64-tune.md: Regenerate.
2082 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2083 cortex-a76.cortex-a55.
2085 2018-06-27 Jeff Law <law@redhat.com>
2087 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2088 (MULTILIB_DIRNAMES): Similarly.
2090 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2092 * gimple.h (gimple_return_retbnd): Delete.
2093 (gimple_return_set_retbnd): Likewise.
2094 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2095 gimple_return_set_retbnd.
2096 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2097 gimple_return_retbnd and adjust.
2098 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2099 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2100 Explicitly return NULL in a couple more cases. Move assertion
2101 on debug statements and remove unreachable code.
2102 (reset_debug_binding): Do not test id->retbnd.
2103 (expand_call_inline): Do not set it.
2105 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2107 * configure.ac: Add --disable-gcov option.
2108 * configure: Regenerate.
2109 * Makefile.in: Honour @enable_gcov@.
2110 * doc/install.texi: Document --disable-gcov.
2112 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2114 * config/arm/arm-cpus.in (cortex-a76): New entry.
2115 (cortex-a76.cortex-a55): Likewise.
2116 * config/arm/arm-tables.opt: Regenerate.
2117 * config/arm/arm-tune.md: Likewise.
2118 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2119 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2120 cortex-a76.cortex-a55.
2122 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2125 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2127 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2129 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2131 (EPILOGUE_USES): Likewise.
2133 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2135 * tree-inline.c (remap_location): New function extracted from...
2136 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2137 (copy_phis_for_bb): ...here. Call remap_location.
2138 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2140 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2142 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2143 unaligned vsx for 16B memset.
2145 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2148 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2149 ieee128_float_type_node to long_double_type_node unless
2150 TARGET_LONG_DOUBLE_128 is set.
2152 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2154 * cfgloop.c (get_loop_location): Convert return type from
2155 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2156 by implicit construction from rtx_insn *, and using
2157 dump_user_location_t::from_function_decl for the fallback case.
2158 * cfgloop.h (get_loop_location): Convert return type from
2159 location_t to dump_user_location_t.
2160 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2161 dump_printf_loc to pass in a dump_location_t rather than a
2162 location_t, via the gimple stmt.
2163 * coverage.c (get_coverage_counts): Update calls to
2164 dump_printf_loc to pass in dump_location_t rather than a
2166 * doc/optinfo.texi (Dump types): Convert example of
2167 dump_printf_loc from taking "locus" to taking "insn". Update
2168 description of the "_loc" calls to cover dump_location_t.
2169 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
2171 (dump_user_location_t::dump_user_location_t): New constructors,
2172 from gimple *stmt and rtx_insn *.
2173 (dump_user_location_t::from_function_decl): New function.
2174 (dump_loc): Make static.
2175 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2176 const dump_location_t &.
2177 (dump_generic_expr_loc): Delete.
2178 (dump_printf_loc): Convert param "loc" from location_t to
2179 const dump_location_t &.
2180 (selftest::test_impl_location): New function.
2181 (selftest::dumpfile_c_tests): New function.
2182 * dumpfile.h: Include "profile-count.h".
2183 (class dump_user_location_t): New class.
2184 (struct dump_impl_location_t): New struct.
2185 (class dump_location_t): New class.
2186 (dump_printf_loc): Convert 2nd param from source_location to
2187 const dump_location_t &.
2188 (dump_generic_expr_loc): Delete.
2189 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2190 const dump_location_t &.
2191 * gimple-fold.c (fold_gimple_assign): Update call to
2192 dump_printf_loc to pass in a dump_location_t rather than a
2193 location_t, via the gimple stmt.
2194 (gimple_fold_call): Likewise.
2195 * gimple-loop-interchange.cc
2196 (loop_cand::analyze_iloop_reduction_var): Update for change to
2197 check_reduction_path.
2198 (tree_loop_interchange::interchange): Update for change to
2200 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2201 change in return-type of find_loop_location.
2202 (graphite_regenerate_ast_isl): Likewise.
2203 * graphite-optimize-isl.c (optimize_isl): Likewise.
2204 * graphite.c (graphite_transform_loops): Likewise.
2205 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2206 pass in a dump_location_t rather than a location_t, via the
2208 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2209 * ipa.c (walk_polymorphic_call_targets): Likewise.
2210 * loop-unroll.c (report_unroll): Convert "locus" param from
2211 location_t to dump_location_t.
2212 (decide_unrolling): Update for change to get_loop_location's
2214 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2215 location_t to dump_user_location_t.
2216 (grid_find_single_omp_among_assignments_1): Updates calls to
2217 dump_printf_loc to pass in a dump_location_t rather than a
2218 location_t, via the gimple stmt.
2219 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2220 location_t to dump_location_t. Updates calls to dump_printf_loc
2221 to pass in a dump_location_t rather than a location_t, via the
2223 (grid_inner_loop_gridifiable_p): Likewise.
2224 (grid_dist_follows_simple_pattern): Likewise.
2225 (grid_gfor_follows_tiling_pattern): Likewise.
2226 (grid_target_follows_gridifiable_pattern): Likewise.
2227 (grid_attempt_target_gridification): Convert initialization
2228 of local "grid" from memset to zero-initialization; FIXME: does
2229 this require C++11? Update call to dump_printf_loc to pass in a
2230 optinfo_location rather than a location_t, via the gimple stmt.
2231 * profile.c (read_profile_edge_counts): Updates call to
2232 dump_printf_loc to pass in a dump_location_t rather than a
2234 (compute_branch_probabilities): Likewise.
2235 * selftest-run-tests.c (selftest::run_tests): Call
2237 * selftest.h (dumpfile_c_tests): New decl.
2238 * tree-loop-distribution.c (pass_loop_distribution::execute):
2239 Update for change in return type of find_loop_location.
2240 * tree-parloops.c (parallelize_loops): Likewise.
2241 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2242 "locus" from location_t to dump_user_location_t.
2243 (canonicalize_loop_induction_variables): Likewise.
2244 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2245 for change in return type of find_loop_location.
2246 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2247 to dump_printf_loc to pass in a dump_location_t rather than a
2248 location_t, via the stmt.
2249 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2251 * tree-vect-loop-manip.c (find_loop_location): Convert return
2252 type from source_location to dump_user_location_t.
2253 (vect_do_peeling): Update for above change.
2254 (vect_loop_versioning): Update for change in type of
2256 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2257 from location_t to dump_user_location_t.
2258 (vect_estimate_min_profitable_iters): Update for change in type
2260 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2261 location_t to dump_location_t.
2262 (vect_slp_bb): Update for change in type of vect_location.
2263 * tree-vectorizer.c (vect_location): Convert from source_location
2264 to dump_user_location_t.
2265 (try_vectorize_loop_1): Update for change in vect_location's type.
2266 (vectorize_loops): Likewise.
2267 (increase_alignment): Likewise.
2268 * tree-vectorizer.h (vect_location): Convert from source_location
2269 to dump_user_location_t.
2270 (find_loop_location): Convert return type from source_location to
2271 dump_user_location_t.
2272 (check_reduction_path): Convert 1st param from location_t to
2273 dump_user_location_t.
2274 * value-prof.c (check_counter): Update call to dump_printf_loc to
2275 pass in a dump_user_location_t rather than a location_t; update
2276 call to error_at for change in type of "locus".
2277 (check_ic_target): Update call to dump_printf_loc to
2278 pass in a dump_user_location_t rather than a location_t, via the
2281 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2283 * config/s390/s390.h (enum processor_flags): Do not use
2284 default tune parameter when -march was specified.
2286 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2289 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2290 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2293 2018-06-26 Richard Biener <rguenther@suse.de>
2295 PR tree-optimization/86287
2297 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2298 (vect_analyze_loop): Initialize n_stmts.
2300 2018-06-26 Richard Biener <rguenther@suse.de>
2303 * fold-const.c (fold_convertible_p): Pointer extension
2306 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2309 * dwarf2out.c (loc_list_has_views): Adjust comments.
2310 (dw_loc_list): Split single cross-partition range with
2313 2018-06-25 Jeff Law <law@redhat.com>
2315 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2316 on -mbig-switch by default.
2318 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2320 (const_float_0_operand): Remove unused predicate.
2321 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2322 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2323 (recipsf2): New expander. Original pattern now called
2325 (recipdf2, recipdf2_insn): Similarly.
2326 (rsqrtsf2, rsqrtsf2_insn): Similarly
2327 (rsqrtdf2, rsqrtdf2_insn): Similarly
2329 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2331 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2332 Simplify logic for FreeBSD (twice).
2334 2018-06-25 Martin Sebor <msebor@redhat.com>
2336 PR tree-optimization/86204
2337 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2338 a strnlen result if it's less than the length of the string.
2340 2018-06-25 Martin Sebor <msebor@redhat.com>
2342 PR tree-optimization/85700
2343 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2344 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2345 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2347 2018-06-25 Martin Sebor <msebor@redhat.com>
2349 * doc/extend.texi (Zero-length arrays): Update and clarify.
2351 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2353 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2354 added IEEE/IBM long double multilib support on PowerPC little
2355 endian Linux systems.
2356 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2357 (MULTILIB_DEFAULTS): Likewise.
2358 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2360 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2361 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2362 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2364 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2367 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2368 (REORDER_45): Likewise.
2370 2018-06-25 Jeff Law <law@redhat.com>
2372 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2373 dividend to 32 bits. Adjust length.
2374 (udivmodhi4): Cleanup output template. Fix length.
2376 2018-06-25 Carl Love <cel@us.ibm.com>
2378 * config/rs6000/vsx.md: Change word selector to prefered location.
2380 2018-06-25 Richard Biener <rguenther@suse.de>
2382 PR tree-optimization/86304
2383 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2384 epilogue-if-converted loops as well.
2386 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2388 * lto-section-out.c (lto_begin_section): Do not print section
2389 name for noaddr and unnumbered dumps.
2391 2018-06-25 Richard Biener <rguenther@suse.de>
2393 * tree-vectorizer.h (struct vec_info_shared): New structure
2394 with parts split out from struct vec_info and loop_nest from
2395 struct _loop_vec_info.
2396 (struct vec_info): Adjust accordingly.
2397 (struct _loop_vec_info): Likewise.
2398 (LOOP_VINFO_LOOP_NEST): Adjust.
2399 (LOOP_VINFO_DATAREFS): Likewise.
2400 (LOOP_VINFO_DDRS): Likewise.
2401 (struct _bb_vec_info): Likewise.
2402 (BB_VINFO_DATAREFS): Likewise.
2403 (BB_VINFO_DDRS): Likewise.
2404 (struct _stmt_vec_info): Add dr_aux member.
2405 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2406 (DR_MISALIGNMENT_UNINITIALIZED): New.
2407 (set_dr_misalignment): Adjust.
2408 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2409 (vect_analyze_loop): Adjust prototype.
2410 (vect_analyze_loop_form): Likewise.
2411 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2412 Compute dependences lazily.
2413 (vect_record_base_alignments): Use shared datarefs/ddrs.
2414 (vect_verify_datarefs_alignment): Likewise.
2415 (vect_analyze_data_refs_alignment): Likewise.
2416 (vect_analyze_data_ref_accesses): Likewise.
2417 (vect_analyze_data_refs): Likewise.
2418 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2419 constructor parameter for shared part.
2420 (vect_analyze_loop_form): Pass in shared part and adjust.
2421 (vect_analyze_loop_2): Pass in storage for the number of
2422 stmts. Move loop nest finding to the caller. Compute
2424 (vect_analyze_loop): Pass in shared part.
2425 (vect_transform_loop): Verify shared datarefs are unchanged.
2426 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2427 constructor parameter for shared part.
2428 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2429 (vect_slp_bb): Verify shared datarefs are unchanged before
2431 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2433 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2434 DR_MISALIGNMENT_UNINITIALIZED.
2435 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2436 parameter for shared part.
2437 (vec_info::~vec_info): Adjust.
2438 (vec_info_shared::vec_info_shared): New.
2439 (vec_info_shared::~vec_info_shared): Likewise.
2440 (vec_info_shared::save_datarefs): Likewise.
2441 (vec_info_shared::check_datarefs): Likewise.
2442 (try_vectorize_loop_1): Construct shared part live for analyses
2443 of a single loop for multiple vector sizes.
2444 * tree-parloops.c (gather_scalar_reductions): Adjust.
2446 2018-06-25 Richard Biener <rguenther@suse.de>
2448 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2449 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2450 (vect_analyze_data_refs): Remove similar code from here and
2451 simplify accordingly.
2453 2018-06-25 Richard Biener <rguenther@suse.de>
2455 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2456 for reverse storage order accesses rather than asserting
2457 they cannot happen here.
2459 2018-06-25 Tom de Vries <tdevries@suse.de>
2462 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2463 Use data16 instead of .byte for insn prefix.
2465 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2468 * parser.c (make_char_string_pack): Pass this literal chars
2469 through cpp_interpret_string.
2470 (cp_parser_userdef_numeric_literal): Check the result of
2471 make_char_string_pack.
2473 2018-06-24 Maya Rashish <coypu@sdf.org>
2475 * ginclude/stddef.h: Simplify conditions around avoiding
2476 re-definition of __size_t.
2478 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2480 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2481 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2483 2018-06-22 Maya Rashish <coypu@sdf.org>
2485 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2486 made to OpenBSD and NetBSD through the years.
2488 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2490 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2491 behavior of vec_pack (vector double, vector double) to match
2492 behavior of vec_float2 (vector double, vector double).
2494 2018-06-22 Olivier Hainque <hainque@adacore.com>
2496 * gimplify.c (gimplify_function_tree): Prevent creation
2497 of a trampoline for the address of the current function
2498 passed to entry/exit instrumentation hooks.
2500 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2503 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2506 2018-06-22 Martin Liska <mliska@suse.cz>
2508 PR tree-optimization/86263
2509 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2510 Bail out if is_enabled is false.
2511 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2513 (jump_table_cluster::is_enabled): New function.
2515 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2517 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2518 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2519 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2520 (lto_input_ts_binfo_tree_pointers): Likewise.
2521 * tree-streamer-out.c (streamer_write_tree_bitfields,
2522 write_ts_binfo_tree_pointers): Likewise.
2523 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2525 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2527 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2529 2018-06-22 Martin Liska <mliska@suse.cz>
2531 * symbol-summary.h (get): Make it pure and inline move
2532 functionality from ::get function.
2533 (get): Remove and inline into ::get and ::get_create.
2534 (get_create): Move code from ::get function.
2536 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2539 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2540 -x assembler-with-cpp.
2542 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2544 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2545 _FILE_OFFSET_BITS=64 for C++.
2547 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2549 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2550 conversion insn that shows up when pr85657-3.c is compiled using
2551 IEEE 128-bit long double.
2552 (neg<mode>2_internal): Use the correct mode to check whether the
2553 mode is IBM extended.
2554 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2555 multiply and divide external functions from being created more
2558 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2560 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2562 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2563 the edge can be forwarded.
2564 (cfg_layout_merge_blocks): Likewise.
2566 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2568 * except.c (finish_eh_generation): Commit edge insertions only after
2569 the EH edges have been redirected from post-landing to landing pads.
2571 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2573 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2574 create_tmp_var_for to create the FRAME decl.
2575 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2577 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2579 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2580 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2581 debug statement when resetting its value.
2582 (expand_call_inline): Copy the locus of the call onto the assignment
2583 of the return value, if any. Use local variable in more cases.
2585 2018-06-21 Martin Liska <mliska@suse.cz>
2587 * ipa-pure-const.c (propagate_nothrow): Use
2588 funct_state_summaries->get.
2589 (dump_malloc_lattice): Likewise.
2590 (propagate_malloc): Likewise.
2592 2018-06-21 Richard Biener <rguenther@suse.de>
2594 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2595 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2596 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2598 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2599 BLOCK_ABSTRACT_ORIGIN unconditionally.
2601 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2603 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2605 * ipa-reference.c (ipa_reference_c_finalize): Delete
2606 ipa_ref_opt_sum_summaries and set it to NULL.
2608 2018-06-21 Tom de Vries <tdevries@suse.de>
2610 PR tree-optimization/85859
2611 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2612 test with comment from bb_no_side_effects_p.
2614 2018-06-21 Richard Biener <rguenther@suse.de>
2616 PR tree-optimization/86232
2617 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2618 max for constant niter.
2620 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2622 * config/aarch64/aarch64-simd.md
2623 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2625 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2627 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2628 Make opernads of the unspec commutative.
2630 2018-06-21 Richard Biener <rguenther@suse.de>
2632 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2633 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2634 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2635 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2637 (vect_record_base_alignments): Do not record base alignment
2638 for gathers or scatters.
2639 (vect_compute_data_ref_alignment): Drop return value that is always
2640 true. Bail out early for gathers or scatters.
2641 (vect_enhance_data_refs_alignment): Bail out early for gathers
2643 (vect_find_same_alignment_drs): Likewise.
2644 (vect_analyze_data_refs_alignment): Remove dead code.
2645 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2646 (vect_analyze_data_refs): For possible gathers or scatters do
2647 not create an alternate DR, just check their possible validity
2648 and mark them. Adjust DECL_NONALIASED handling to not rely
2650 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2651 update inits of gathers or scatters.
2652 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2653 Also copy gather/scatter flag to pattern vinfo.
2655 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2657 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2658 behavior of vec_packsu (vector unsigned long long, vector unsigned
2659 long long) to match behavior of vec_packs with same signature.
2661 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2662 Thomas Schwinge <thomas@codesourcery.com>
2663 Cesar Philippidis <cesar@codesourcery.com>
2665 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2666 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2667 (gimplify_adjust_omp_clauses): Likewise.
2668 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2669 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2670 (gimplify_omp_target_update): Update handling of acc update and
2672 * omp-low.c (install_var_field): Remove unused parameter
2673 base_pointers_restrict.
2674 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2675 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2677 (omp_target_base_pointers_restrict_p): Delete.
2678 (scan_omp_target): Update call to scan_sharing_clauses.
2679 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2681 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2682 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2683 (convert_local_omp_clauses): Likewise.
2684 * tree-pretty-print.c (dump_omp_clause): Likewise.
2685 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2687 (omp_clause_code_name): Likewise.
2689 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2692 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2695 PR tree-optimization/86231
2696 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2697 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2699 2018-06-20 Tom de Vries <tdevries@suse.de>
2701 PR tree-optimization/86097
2702 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2703 iv type if signedness of iv type is not the same as that of *nit.
2705 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2707 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2708 EDGE_EH edges, verify they are all EDGE_EH.
2710 2018-06-20 Maya Rashish <coypu@sdf.org>
2712 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2714 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2716 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2717 * config/aarch64/aarch64.c (xgene1_tunings): Add
2718 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2719 (aarch64_mode_valid_for_sched_fusion_p):
2720 Allow 16-byte modes.
2721 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2722 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2724 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2726 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2727 * config/aarch64/iterators.md (VQ2): New mode iterator.
2729 2018-06-20 Martin Liska <mliska@suse.cz>
2731 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2732 Change default ratio from 10 to 8.
2734 2018-06-20 Martin Liska <mliska@suse.cz>
2736 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2738 (bit_test_cluster::find_bit_tests): Likewise.
2739 (switch_decision_tree::analyze_switch_statement): Find clusters.
2740 * tree-switch-conversion.h (struct jump_table_cluster): Document
2743 2018-06-20 Martin Liska <mliska@suse.cz>
2745 * tree-switch-conversion.c (switch_conversion::collect):
2746 Record m_uniq property.
2747 (switch_conversion::expand): Bail out for special conditions.
2748 (group_cluster::~group_cluster): New.
2749 (group_cluster::group_cluster): Likewise.
2750 (group_cluster::dump): Likewise.
2751 (jump_table_cluster::emit): New.
2752 (switch_decision_tree::fix_phi_operands_for_edges): New.
2753 (struct case_node): Remove struct.
2754 (jump_table_cluster::can_be_handled): New.
2755 (case_values_threshold): Moved to header.
2756 (reset_out_edges_aux): Likewise.
2757 (jump_table_cluster::is_beneficial): New.
2758 (bit_test_cluster::can_be_handled): Likewise.
2759 (add_case_node): Remove.
2760 (bit_test_cluster::is_beneficial): New.
2761 (case_bit_test::cmp): New.
2762 (bit_test_cluster::emit): New.
2763 (expand_switch_as_decision_tree_p): Remove.
2764 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2765 (fix_phi_operands_for_edge): Likewise.
2766 (switch_decision_tree::analyze_switch_statement): New.
2767 (compute_cases_per_edge): Move ...
2768 (switch_decision_tree::compute_cases_per_edge): ... here.
2769 (try_switch_expansion): Likewise.
2770 (switch_decision_tree::try_switch_expansion): Likewise.
2771 (record_phi_operand_mapping): Likewise.
2772 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2773 (emit_case_decision_tree): Likewise.
2774 (switch_decision_tree::emit): Likewise.
2775 (balance_case_nodes): Likewise.
2776 (switch_decision_tree::balance_case_nodes): Likewise.
2777 (dump_case_nodes): Likewise.
2778 (switch_decision_tree::dump_case_nodes): Likewise.
2779 (emit_jump): Likewise.
2780 (switch_decision_tree::emit_jump): Likewise.
2781 (emit_cmp_and_jump_insns): Likewise.
2782 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2783 (emit_case_nodes): Likewise.
2784 (switch_decision_tree::emit_case_nodes): Likewise.
2785 (conditional_probability): Remove.
2786 * tree-switch-conversion.h (enum cluster_type): New.
2788 (struct cluster): Likewise.
2789 (cluster::cluster): Likewise.
2790 (struct simple_cluster): Likewise.
2791 (simple_cluster::simple_cluster): Likewise.
2792 (struct group_cluster): Likewise.
2793 (struct jump_table_cluster): Likewise.
2794 (struct bit_test_cluster): Likewise.
2795 (struct min_cluster_item): Likewise.
2796 (struct case_tree_node): Likewise.
2797 (case_tree_node::case_tree_node): Likewise.
2798 (jump_table_cluster::case_values_threshold): Likewise.
2799 (struct case_bit_test): Likewise.
2800 (struct switch_decision_tree): Likewise.
2801 (struct switch_conversion): Likewise.
2802 (switch_decision_tree::reset_out_edges_aux): Likewise.
2804 2018-06-20 Martin Liska <mliska@suse.cz>
2806 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2807 (hoist_edge_and_branch_if_true): Likewise.
2808 (expand_switch_using_bit_tests_p): Likewise.
2809 (struct case_bit_test): Likewise.
2810 (case_bit_test_cmp): Likewise.
2811 (emit_case_bit_tests): Likewise.
2812 (switch_conversion::switch_conversion): New class.
2813 (struct switch_conv_info): Remove old struct.
2814 (collect_switch_conv_info): More to ...
2815 (switch_conversion::collect): ... this.
2816 (check_range): Likewise.
2817 (switch_conversion::check_range): Likewise.
2818 (check_all_empty_except_final): Likewise.
2819 (switch_conversion::check_all_empty_except_final): Likewise.
2820 (check_final_bb): Likewise.
2821 (switch_conversion::check_final_bb): Likewise.
2822 (create_temp_arrays): Likewise.
2823 (switch_conversion::create_temp_arrays): Likewise.
2824 (free_temp_arrays): Likewise.
2825 (gather_default_values): Likewise.
2826 (switch_conversion::gather_default_values): Likewise.
2827 (build_constructors): Likewise.
2828 (switch_conversion::build_constructors): Likewise.
2829 (constructor_contains_same_values_p): Likewise.
2830 (switch_conversion::contains_same_values_p): Likewise.
2831 (array_value_type): Likewise.
2832 (switch_conversion::array_value_type): Likewise.
2833 (build_one_array): Likewise.
2834 (switch_conversion::build_one_array): Likewise.
2835 (build_arrays): Likewise.
2836 (switch_conversion::build_arrays): Likewise.
2837 (gen_def_assigns): Likewise.
2838 (switch_conversion::gen_def_assigns): Likewise.
2839 (prune_bbs): Likewise.
2840 (switch_conversion::prune_bbs): Likewise.
2841 (fix_phi_nodes): Likewise.
2842 (switch_conversion::fix_phi_nodes): Likewise.
2843 (gen_inbound_check): Likewise.
2844 (switch_conversion::gen_inbound_check): Likewise.
2845 (process_switch): Use the newly created class.
2846 (switch_conversion::expand): New.
2847 (switch_conversion::~switch_conversion): New.
2848 * tree-switch-conversion.h: New file.
2850 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2852 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2853 tree-vect-patterns.c.
2854 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2855 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2856 (vect_recog_sad_pattern): Likewise.
2857 (vect_recog_widen_sum_pattern): Likewise.
2858 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2859 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2860 (vect_recog_rotate_pattern): Likewise.
2861 (vect_recog_mult_pattern): Likewise.
2862 (vect_recog_vector_vector_shift_pattern): Likewise.
2863 (vect_recog_divmod_pattern): Likewise.
2864 (vect_recog_mixed_size_cond_pattern): Likewise.
2865 (vect_recog_bool_pattern): Likewise.
2866 (vect_recog_mask_conversion_pattern): Likewise.
2867 (vect_try_gather_scatter_pattern): Likewise.
2868 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2869 (vect_recog_over_widening_pattern): Likewise.
2870 (vect_recog_gather_scatter_pattern): Likewise.
2871 (vect_recog_func_ptr): Move from tree-vectorizer.h
2872 (vect_vect_recog_func_ptrs): Move further down the file.
2873 (vect_recog_func): Likewise. Remove the third argument.
2874 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2875 (vect_pattern_recog_1): Expect the pattern function to do any
2876 necessary target tests. Also expect it to provide a vector type.
2877 Remove the type_in handling.
2879 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2881 * tree-vect-patterns.c (vect_pattern_detected): New function.
2882 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2883 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2884 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2885 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2886 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2887 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2888 (vect_recog_mask_conversion_pattern)
2889 (vect_try_gather_scatter_pattern): Likewise.
2891 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2893 * tree-vect-patterns.c (vect_get_internal_def): New function.
2894 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2895 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2896 (search_type_for_mask_1): Use it.
2898 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2900 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2901 redundant WIDEN_SUM_EXPR handling.
2902 (vect_recog_sad_pattern): Likewise.
2904 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2906 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2907 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2908 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2909 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2910 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2912 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2914 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2915 use the stmt_vec_info of the original bb statement for the
2916 new zero assignment, even if the call is part of a pattern.
2918 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2920 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2921 that the sequence is attached to the original statement rather
2922 than the pattern statement.
2923 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2924 PATTERN_DEF_SEQ from the original statement rather than
2925 the main pattern statement.
2926 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2927 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2928 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2930 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2932 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2933 definition statements before the early exit for statements that aren't
2935 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2937 (vect_transform_loop): ...here. Process pattern definition
2938 statements without first checking whether the main pattern
2939 statement is live or relevant.
2941 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2943 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2944 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2946 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2948 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2949 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2950 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2951 (expand_block_compare): Change select_block_compare_mode call.
2952 (expand_strncmp_align_check): Use new functions, fix comment.
2953 (emit_final_str_compare_gpr): New function.
2954 (expand_strn_compare): Refactor and clean up code.
2955 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2957 2018-06-19 Tony Reix <tony.reix@atos.com>
2958 Damien Bergamini <damien.bergamini@atos.com>
2959 David Edelsohn <dje.gcc@gmail.com>
2961 * collect2.c (static_obj): New variable.
2962 (static_libs): New variable.
2963 (is_in_list): Uncomment declaration.
2964 (main): Track AIX libraries linked statically.
2965 (is_in_list): Uncomment definition.
2966 (scan_prog_file): Don't add AIX shared libraries initializer
2967 to constructor list if linking statically.
2969 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2971 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2973 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2975 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2977 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2980 2018-06-19 Martin Liska <mliska@suse.cz>
2982 * config/i386/i386.c (ix86_can_inline_p): Do not use
2983 ipa_fn_summaries::get_create.
2984 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2986 (devirtualization_time_bonus): Likewise.
2987 (ipcp_propagate_stage): Likewise.
2988 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2989 (edge_set_predicate): Likewise.
2990 (evaluate_conditions_for_known_args): Likewise.
2991 (evaluate_properties_for_edge): Likewise.
2992 (ipa_call_summary::reset): Tranform to ...
2993 (ipa_call_summary::~ipa_call_summary): ... this.
2994 (ipa_fn_summary::reset): Transform to ...
2995 (ipa_fn_summary::~ipa_fn_summary): ... this.
2996 (ipa_fn_summary_t::remove): Rename to ...
2997 (ipa_fn_summary_t::remove_callees): ... this.
2998 (ipa_fn_summary_t::duplicate): Use placement new
2999 instead of memory copy.
3000 (ipa_call_summary_t::duplicate): Likewise.
3001 (ipa_call_summary_t::remove): Remove.
3002 (dump_ipa_call_summary): Change get_create to get.
3003 (ipa_dump_fn_summary): Dump only when summary exists.
3004 (analyze_function_body): Use symbol_summary::get instead
3006 (compute_fn_summary): Likewise.
3007 (estimate_edge_devirt_benefit): Likewise.
3008 (estimate_edge_size_and_time): Likewise.
3009 (inline_update_callee_summaries): Likewise.
3010 (remap_edge_change_prob): Likewise.
3011 (remap_edge_summaries): Likewise.
3012 (ipa_merge_fn_summary_after_inlining): Likewise.
3013 (write_ipa_call_summary): Likewise.
3014 (ipa_fn_summary_write): Likewise.
3015 (ipa_free_fn_summary): Likewise.
3016 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
3017 (struct ipa_call_summary): Likewise.
3018 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
3020 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
3021 (estimate_size_after_inlining): Likewise.
3022 (estimate_growth): Likewise.
3023 (growth_likely_positive): Likewise.
3024 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
3025 (inline_call): Likewise.
3026 * ipa-inline.c (caller_growth_limits): Likewise.
3027 (can_inline_edge_p): Likewise.
3028 (can_inline_edge_by_limits_p): Likewise.
3029 (compute_uninlined_call_time): Likewise.
3030 (compute_inlined_call_time): Likewise.
3031 (want_inline_small_function_p): Likewise.
3032 (edge_badness): Likewise.
3033 (update_caller_keys): Likewise.
3034 (update_callee_keys): Likewise.
3035 (inline_small_functions): Likewise.
3036 (inline_to_all_callers_1): Likewise.
3037 (dump_overall_stats): Likewise.
3038 (early_inline_small_functions): Likewise.
3039 (early_inliner): Likewise.
3040 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3041 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3042 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3043 * ipa-split.c (execute_split_functions): Likewise.
3044 * symbol-summary.h: Likewise.
3045 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3047 2018-06-19 Richard Biener <rguenther@suse.de>
3049 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3050 (vectorize_loops): ... here. Fix dbgcnt handling.
3051 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3053 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3056 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3057 ieee128 argument takes up only one (vector) register, not two (floating
3060 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3062 * gimplify.c (gimplify_init_constructor): Really never clear for an
3063 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3065 2018-06-19 Richard Biener <rguenther@suse.de>
3067 PR tree-optimization/86179
3068 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3069 after failed recognition.
3071 2018-06-18 Martin Sebor <msebor@redhat.com>
3074 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3075 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3076 Handle integer subtraction.
3077 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3078 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3080 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3082 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3083 param from rtx to rtx_insn *.
3084 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3086 (frv_ifcvt_modify_insn): Likwise.
3087 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3088 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3089 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3090 as_a <rtx_insn *> cast to local "unprotected_region" once
3091 it's been established that it's not NULL or pc_rtx.
3092 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3093 param "sethi" from rtx to rtx_insn *.
3094 (nds32_group_float_insns): Likewise for param "insn".
3095 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3097 (vax_output_int_subtract): Likewise.
3098 * config/vax/vax.c (vax_output_int_add): Likewise for param
3100 (vax_output_int_subtract): Likewise.
3101 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3102 (emit_pattern_after): Likewise for param "after".
3103 (emit_insn_after): Likewise.
3104 (emit_jump_insn_after): Likewise.
3105 (emit_call_insn_after): Likewise.
3106 (emit_debug_insn_after): Likewise.
3107 (emit_pattern_before): Likewise for param "before".
3108 (emit_insn_before): Likewise.
3109 (emit_jump_insn_before): Likewise.
3110 * final.c (get_insn_template): Likewise for param "insn", removing
3112 * output.h (get_insn_template): Likewise for 2nd param.
3113 * rtl.h (emit_insn_before): Likewise.
3114 (emit_jump_insn_before): Likewise.
3115 (emit_debug_insn_before_noloc): Likewise.
3116 (emit_insn_after): Likewise.
3117 (emit_jump_insn_after): Likewise.
3118 (emit_call_insn_after): Likewise.
3119 (emit_debug_insn_after): Likewise.
3120 (set_insn_deleted): Likewise for param.
3122 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3125 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3126 floating point modes, so that IFmode is numerically greater than
3127 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3128 to declare the ordering. This prevents IFmode from being
3129 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3130 machine. Include rs6000-modes.h to share the fractional values
3131 between genmodes* and the rest of the compiler.
3135 * config/rs6000/rs6000-modes.h: New file.
3136 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3137 meaning of rs6000_long_double_size so that 126..128 selects an
3138 appropriate 128-bit floating point type.
3139 (rs6000_option_override_internal): Likewise.
3140 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3141 (TARGET_LONG_DOUBLE_128): Change the meaning of
3142 rs6000_long_double_size so that 126..128 selects an appropriate
3143 128-bit floating point type.
3144 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3145 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3146 source and destination to match the standard usage.
3147 (truncifkf2): Likewise.
3148 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3149 ISA 2.07 to use an explicit clobber, instead of passing in a
3151 (copysign<mode>3_soft): Likewise.
3153 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3155 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3156 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3157 (vect_slp_analyze_instance_dependence): Likewise.
3158 (vect_enhance_data_refs_alignment): Likewise.
3159 (vect_analyze_data_refs_alignment): Likewise.
3160 (vect_slp_analyze_and_verify_instance_alignment
3161 (vect_analyze_data_ref_accesses): Likewise.
3162 (vect_prune_runtime_alias_test_list): Likewise.
3163 (vect_analyze_data_refs): Likewise.
3164 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3165 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3166 (vect_analyze_scalar_cycles_1): Likewise.
3167 (vect_get_loop_niters): Likewise.
3168 (vect_analyze_loop_form_1): Likewise.
3169 (vect_update_vf_for_slp): Likewise.
3170 (vect_analyze_loop_operations): Likewise.
3171 (vect_analyze_loop): Likewise.
3172 (vectorizable_induction): Likewise.
3173 (vect_transform_loop): Likewise.
3174 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3175 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3176 (vect_make_slp_decision): Likewise.
3177 (vect_detect_hybrid_slp): Likewise.
3178 (vect_slp_analyze_operations): Likewise.
3179 (vect_slp_bb): Likewise.
3180 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3181 (vectorizable_bswap): Likewise.
3182 (vectorizable_call): Likewise.
3183 (vectorizable_simd_clone_call): Likewise.
3184 (vectorizable_conversion): Likewise.
3185 (vectorizable_assignment): Likewise.
3186 (vectorizable_shift): Likewise.
3187 (vectorizable_operation): Likewise.
3188 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3190 2018-06-18 Martin Sebor <msebor@redhat.com>
3192 PR tree-optimization/81384
3193 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3194 * builtins.c (expand_builtin_strnlen): New function.
3195 (expand_builtin): Call it.
3196 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3197 * builtins.def (BUILT_IN_STRNLEN): New.
3198 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3199 Warn for bounds in excess of maximum object size.
3200 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3201 single-value ranges. Handle strnlen.
3202 (handle_builtin_strlen): Handle strnlen.
3203 (strlen_check_and_optimize_stmt): Same.
3204 * doc/extend.texi (Other Builtins): Document strnlen.
3206 2018-06-18 Maya Rashish <coypu@sdf.org>
3208 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3209 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3210 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3212 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3214 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3216 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3218 * tree.c (escaped_string::escape): Replace cast to char * by
3219 const_cast<char *> (unescaped).
3221 2018-06-18 Nick Clifton <nickc@redhat.com>
3224 * tree.c (escaped_string): New class. Converts an unescaped
3225 string into its escaped equivalent.
3226 (warn_deprecated_use): Use the new class to convert the
3227 deprecation message, if present.
3228 (test_escaped_strings): New self test.
3229 (test_c_tests): Add test_escaped_strings.
3230 * doc/extend.texi (deprecated): Add a note that the
3231 deprecation message is affected by the -fmessage-length
3232 option, and that control characters will be escaped.
3233 (#pragma GCC error): Document this pragma.
3234 (#pragma GCC warning): Likewise.
3235 * doc/invoke.texi (-fmessage-length): Document this option's
3236 effect on the #warning and #error preprocessor directives and
3237 the deprecated attribute.
3239 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3241 * tree.c (decl_value_expr_lookup): Revert latest change.
3242 (decl_value_expr_insert): Likewise.
3244 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3246 * gimplify.c (nonlocal_vlas): Delete.
3247 (nonlocal_vla_vars): Likewise.
3248 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3250 (gimplify_body): Do not create and destroy nonlocal_vlas.
3251 * tree-nested.c: Include diagnostic.h.
3252 (use_pointer_in_frame): Tweak.
3253 (lookup_field_for_decl): Add assertion and declare the transformation.
3254 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3255 internal error when the reference is in a wrong context. Do not
3256 create a debug decl by default.
3257 (note_nonlocal_block_vlas): Delete.
3258 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3259 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3260 create a debug decl by default.
3261 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3262 call is in a wrong context.
3263 (fixup_vla_decls): New function.
3264 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3265 debug variables were created.
3266 * tree.c (decl_value_expr_lookup): Add checking assertion.
3267 (decl_value_expr_insert): Likewise.
3269 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3272 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3273 * tree-scalar-evolution.c (interpret_expr): Likewise.
3274 (expression_expensive_p): Likewise.
3275 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3276 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3277 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3278 (ssa_defined_by_minus_one_stmt_p): New.
3280 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3283 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3284 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3285 * dojump.c (do_jump): Likewise.
3286 * expr.c (expand_expr_real_2): Check operand type's sign.
3287 * fold-const.c (const_unop): Handle ABSU_EXPR.
3288 (fold_abs_const): Likewise.
3289 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3290 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3291 (strip_sign_op_1): Likesise.
3292 * match.pd: Add new pattern to generate ABSU_EXPR.
3293 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3294 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3295 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3296 * tree-inline.c (estimate_operator_cost): Likewise.
3297 * tree-pretty-print.c (dump_generic_node): Likewise.
3298 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3299 * tree.def (ABSU_EXPR): New.
3301 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3304 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3305 documented as preserved for backward compatibility only.
3306 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3308 PR rtl-optimization/86108
3309 * bb-reorder.c (create_forwarder_block): Renamed to ...
3310 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3311 jump from new landing pad to the second part.
3312 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3315 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3318 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3319 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3320 Only call store_expr for halves if the mode is the same.
3323 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3324 Fix up comment formatting.
3326 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3328 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3329 (typed_splay_tree::closure,
3330 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3331 (typed_splay_tree::typed_splay_tree,
3332 typed_splay_tree::operator =): Declared private.
3333 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3334 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3335 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3336 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3337 typed_splay_tree::splay_tree_splay,
3338 typed_splay_tree::splay_tree_foreach_helper,
3339 typed_splay_tree::splay_tree_insert,
3340 typed_splay_tree::splay_tree_remove,
3341 typed_splay_tree::splay_tree_lookup,
3342 typed_splay_tree::splay_tree_predecessor,
3343 typed_splay_tree::splay_tree_successor,
3344 typed_splay_tree::splay_tree_min,
3345 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3346 (typed_splay_tree::root, typed_splay_tree::comp,
3347 typed_splay_tree::delete_key,
3348 typed_splay_tree::delete_value): New data members.
3349 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3350 typed_splay_tree::remove.
3352 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3354 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3355 -mginv and -mno-ginv to the assembler.
3356 * config/mips/mips.opt (-mcrc): New option.
3358 * doc/invoke.text (-mcrc): Document.
3361 2018-06-15 Nick Clifton <nickc@redhat.com>
3364 * tree.c (escaped_string): New class. Converts an unescaped
3365 string into its escaped equivalent.
3366 (warn_deprecated_use): Use the new class to convert the
3367 deprecation message, if present.
3368 (test_escaped_strings): New self test.
3369 (test_c_tests): Add test_escaped_strings.
3370 * doc/extend.texi (deprecated): Add a note that the
3371 deprecation message is affected by the -fmessage-length
3372 option, and that control characters will be escaped.
3373 (#pragma GCC error): Document this pragma.
3374 (#pragma GCC warning): Likewise.
3375 * doc/invoke.texi (-fmessage-length): Document this option's
3376 effect on the #warning and #error preprocessor directives and
3377 the deprecated attribute.
3379 2018-06-15 Richard Biener <rguenther@suse.de>
3381 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3382 here, also noting vector size used.
3383 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3384 size used in MSG_OPTIMIZED_LOCATIONS dump.
3385 (pass_slp_vectorize::execute): Adjust.
3387 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3390 * config/arc/arc.c (arc_return_address_register): Fix
3393 2018-06-15 Richard Biener <rguenther@suse.de>
3396 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3397 leave useless conversion stripping to force_gimple_operand_gsi.
3398 (gimplify_build2): Likewise.
3399 (gimplify_build1): Likewise.
3401 2018-06-15 Richard Biener <rguenther@suse.de>
3404 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3405 before adjusting their block.
3407 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3409 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3410 multilibs for *-*-rtems*.
3411 * config/riscv/t-rtems: New file.
3413 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3416 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3417 unsigned_type_for returns NULL.
3420 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3421 subregs of multi-word pseudos unless the float mode has word size.
3423 2018-06-14 Richard Biener <rguenther@suse.de>
3426 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3427 duplicate and harmful type_hash_canon.
3428 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3430 2018-06-14 Richard Biener <rguenther@suse.de>
3433 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3436 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3438 * config/rtems.h (STDINT_LONG32): Define.
3440 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3441 Prachi Godbole <prachi.godbole@imgtec.com>
3443 * config/mips/mips-cpus.def: Define P6600.
3444 * config/mips/mips-tables.opt: Regenerate.
3445 * config/mips/mips.c (mips_ucbranch_type): New enum.
3446 (mips_rtx_cost_data): Add support for P6600.
3447 (mips_issue_rate): Likewise.
3448 (mips_multipass_dfa_lookahead): Likewise.
3449 (mips_avoid_hazard): Likewise.
3450 (mips_reorg_process_insns): Likewise.
3451 (mips_classify_branch_p6600): New function.
3452 * config/mips/mips.h (TUNE_P6600): New define.
3453 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3454 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3455 * config/mips/mips.md: Include p6600.md.
3456 (processor): Add p6600.
3457 * config/mips/p6600.md: New file.
3458 * doc/invoke.texi: Add p6600 to supported architectures.
3460 2018-06-13 Martin Sebor <msebor@redhat.com>
3462 PR tree-optimization/86114
3463 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3465 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3467 2018-06-13 Richard Biener <rguenther@suse.de>
3469 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3470 Properly set vector type of the intermediate stmt.
3471 * tree-vect-stmts.c (vectorizable_operation): The destination
3472 var always has vectype_out type.
3474 2018-06-13 Jeff Law <law@redhat.com>
3476 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3477 integer 0 for argument to print_rtl_with_bb.
3478 (rl78_reorg): Likewise.
3480 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3482 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3483 from rtx to rtx_insn *.
3484 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3486 (add_sched_insns_for_speculation): Likewise for local "target",
3487 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3488 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3489 from rtx_insn ** to rtx_code_label **.
3490 (reorg_emit_nops): Likewise.
3491 (c6x_reorg): Likewise for local "call_labels".
3492 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3494 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3495 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3496 the loops over LABEL_REFS.
3497 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3499 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3500 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3501 (split_branches): Strengthen local "olabel" from rtx to
3502 rtx_insn *, adding a safe_as_a cast.
3503 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3505 (add_insn_after): Likewise for first two params.
3506 (add_insn_before): Likewise.
3507 (remove_insn): Likewise for param.
3508 (emit_pattern_before_noloc): Likewise for second and third params.
3509 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3510 (emit_call_insn_before_noloc): Likewise.
3511 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3513 (emit_barrier_before): Likewise.
3514 (emit_label_before): Strengthen "label" param from "rtx" to
3515 "rtx_code_label *". Strengthen "before" param from "rtx" to
3517 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3519 (emit_pattern_after_noloc): Likewise.
3520 (emit_insn_after_noloc): Likewise.
3521 (emit_jump_insn_after_noloc): Likewise.
3522 (emit_call_insn_after_noloc): Likewise.
3523 (emit_debug_insn_after_noloc): Likewise.
3524 (emit_barrier_after): Likewise.
3525 (emit_label_after): Likewise for both params.
3526 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3527 "loc" param from "int" to "location_t".
3528 (emit_insn_after_setloc): Likewise.
3529 (emit_jump_insn_after_setloc): Likewise.
3530 (emit_call_insn_after_setloc): Likewise.
3531 (emit_debug_insn_after_setloc): Likewise.
3532 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3533 "loc" param from "int" to "location_t".
3534 (emit_pattern_before): Convert NULL_RTX to NULL.
3535 (emit_insn_before_setloc): Convert "loc" param from "int" to
3537 (emit_jump_insn_before_setloc): Likewise.
3538 (emit_call_insn_before_setloc): Likewise.
3539 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3540 rtx_insn *. Convert "loc" param from "int" to "location_t".
3541 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3542 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3543 Convert 3rd param from "int" to "location_t".
3544 (emit_barrier_before, emit_barrier_after, next_real_insn):
3545 Strengthen param from rtx to rtx_insn *.
3546 (emit_label_before): Strengthen 1st param from "rtx" to
3547 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3549 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3550 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3551 Strengthen 2nd param from "rtx" to "rtx_insn *".
3552 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3553 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3554 Likewise. Convert 3rd param from "int" to "location_t".
3555 (emit_label_after): Strengthen 1st param from "rtx" to
3557 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3559 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3560 from "rtx" to "rtx_insn *".
3562 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3564 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3565 bodies streamed in with -Q.
3566 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3567 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3568 * gimple-streamer-out.c: Include gimple-pretty-print.h
3569 (output_bb): Dump stmts streamed.
3570 * lto-section-out.c: Include print-tree.h
3571 (lto_begin_section): Dump sections created.
3572 (lto_output_decl_index): Dump decl encoded.
3573 * lto-streamer-out.c: Include print-tree.h
3574 (create_output_block): Dump output block created.
3575 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3576 (output_function): Dump function output.
3577 (output_constructor): Dump constructor streamed.
3578 (write_global_stream): Output indexes encoded.
3579 (produce_asm_for_decls): Dump streams encoded.
3580 * lto-streamer.c (streamer_dump_file): New global var.
3581 * lto-streamer.h (streamer_dump_file): Declare.
3582 * passes.c (ipa_write_summaries): Initialize streamer dump.
3583 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3586 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3589 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3590 offsets for register save directives. Emit a second batch of save
3591 directives, if need be, when the function accesses prior frames.
3593 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3595 * config/arc/fpu.md (fmasf4): Force operand to register.
3596 (fnmasf4): Likewise.
3598 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3600 * config/arc/arc-protos.h (arc_pad_return): Remove.
3601 * config/arc/arc.c (machine_function): Remove force_short_suffix
3603 (arc_print_operand): Adjust printing of '&'.
3604 (arc_verify_short): Remove conditional printing of short suffix.
3605 (arc_final_prescan_insn): Remove reference to size_reason.
3606 (pad_return): New function.
3607 (arc_reorg): Call pad_return.
3608 (arc_pad_return): Remove.
3609 (arc_init_machine_status): Remove reference to force_short_suffix.
3610 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3611 (attr length): When attribute iscompact is true force to 2
3612 regardless; in the case of maybe check if we want to force the
3613 instruction to have 4 bytes length.
3614 (nopv): Change it to generate 4 byte long nop as well.
3615 (blockage): New pattern.
3616 (simple_return): Remove call to arc_pad_return.
3617 (p_return_i): Likewise.
3619 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3621 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3623 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3625 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3628 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3630 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3631 for ARC700 and ARCv2.
3633 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3636 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3637 operands[2] instead of operands[1].
3640 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3642 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3643 case, check whether the outer register overlaps an unallocatable
3644 register, not just whether it fits the required class.
3646 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3648 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3650 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3651 (memrefs_conflict_p): Likewise.
3652 (init_alias_analysis): Likewise.
3653 * cfgexpand.c (expand_debug_expr): Likewise.
3654 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3655 * cse.c (fold_rtx): Likewise.
3656 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3657 * expr.c (emit_block_move_hints): Likewise.
3658 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3659 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3660 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3661 (emit_group_store): Likewise.
3662 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3663 to read the PRE/POST_MODIFY increment.
3664 * calls.c (store_one_arg): Use strip_offset.
3665 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3667 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3669 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3670 (simplify_binary_operation_1): Extend CONST_INT handling to
3672 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3673 than a HOST_WIDE_INT.
3674 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3676 (adjust_mems, add_stores): Update accodingly.
3677 (vt_canonicalize_addr): Track polynomial offsets.
3678 (emit_note_insn_var_location): Likewise.
3679 (vt_add_function_parameter): Likewise.
3680 (vt_initialize): Likewise.
3682 2018-06-12 Jeff Law <law@redhat.com>
3684 * config.gcc (alpha*-*-freebsd*): Remove.
3685 * config/alpha/freebsd.h: Remove.
3687 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3690 * spellcheck-tree.c (levenshtein_distance): Rename to...
3691 (get_edit_distance): ...this, and update for underlying renaming.
3692 * spellcheck-tree.h (levenshtein_distance): Rename to...
3693 (get_edit_distance): ...this.
3694 * spellcheck.c (levenshtein_distance): Rename to...
3695 (get_edit_distance): ...this. Convert from Levenshtein distance
3696 to Damerau-Levenshtein distance by supporting transpositions of
3697 adjacent characters. Rename "v1" to "v_next" and "v0" to
3699 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3700 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3701 update for underlying renaming.
3702 (selftest::levenshtein_distance_unit_test): Rename to...
3703 (selftest::test_get_edit_distance_unit): ...this, and update for
3704 underlying renaming.
3705 (selftest::test_find_closest_string): Add example from PR 69968
3706 where transposition helps
3707 (selftest::test_metric_conditions): Update for renaming.
3708 (selftest::test_metric_conditions): Likewise.
3709 (selftest::spellcheck_c_tests): Likewise.
3710 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3711 (get_edit_distance): ...this.
3712 (best_match::consider): Update for renaming.
3714 2018-06-12 Martin Sebor <msebor@redhat.com>
3716 PR tree-optimization/85259
3717 * builtins.c (compute_objsize): Handle constant offsets.
3718 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3719 true iff a warning has been issued.
3720 * gimple.h (gimple_nonartificial_location): New function.
3721 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3722 gimple_nonartificial_location and handle -Wno-system-headers.
3723 (handle_builtin_stxncpy): Same.
3725 2018-06-12 Martin Sebor <msebor@redhat.com>
3728 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3730 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3732 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3733 BUILTIN_VEC_XST entries for pointer to double and long long.
3735 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3738 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3740 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3743 2018-06-12 Martin Liska <mliska@suse.cz>
3745 * doc/options.texi: Document IntegerRange.
3747 2018-06-12 Martin Liska <mliska@suse.cz>
3749 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3750 * opt-functions.awk: Handle Deprecated flag.
3751 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3753 (read_cmdline_option): Report warning for a deprecated option.
3754 * opts.h (struct cl_option): Add new field cl_deprecated.
3755 (CL_ERR_DEPRECATED): New.
3757 2018-06-12 Martin Liska <mliska@suse.cz>
3759 * doc/options.texi: Document Deprecated option flag.
3761 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3763 * config/arc/arc-arch.h (arc_extras): New enum.
3764 (arc_cpu_t):Add field extra.
3765 (arc_cpu_types): Consider the extras.
3766 * config/arc/arc-cpus.def: Add extras info.
3767 * config/arc/arc-opts.h (processor_type): Consider extra field.
3768 * config/arc/arc.c (arc_override_options): Handle extra field.
3770 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3772 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3773 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3775 * config/arc/arc.c (arc_init): Update pic warning.
3776 (irq_range): Update irq range parsing warnings.
3777 (arc_override_options): Update various warning messages.
3778 (arc_handle_aux_attribute): Likewise.
3780 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3782 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3784 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3786 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3787 compiler messages directives.
3789 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3791 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3792 * config/mips/mips-tables.opt: Regenerate.
3793 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3795 * doc/invoke.texi: Document -march=i6500.
3797 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3799 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3800 (i6400_gpmul): Add cpu_unit.
3801 (i6400_gpdiv): Likewise.
3802 (i6400_msa_add_d): Update reservations.
3803 (i6400_msa_int_add) Likewise.
3804 (i6400_msa_short_logic3) Likewise.
3805 (i6400_msa_short_logic2) Likewise.
3806 (i6400_msa_short_logic) Likewise.
3807 (i6400_msa_move) Likewise.
3808 (i6400_msa_cmp) Likewise.
3809 (i6400_msa_short_float2) Likewise.
3810 (i6400_msa_div_d) Likewise.
3811 (i6400_msa_long_logic1) Likewise.
3812 (i6400_msa_long_logic2) Likewise.
3813 (i6400_msa_mult) Likewise.
3814 (i6400_msa_long_float2) Likewise.
3815 (i6400_msa_long_float4) Likewise.
3816 (i6400_msa_long_float5) Likewise.
3817 (i6400_msa_long_float8) Likewise.
3818 (i6400_fpu_fadd): Include frint type.
3819 (i6400_fpu_store): New define_insn_reservation.
3820 (i6400_fpu_load): Likewise.
3821 (i6400_fpu_move): Likewise.
3822 (i6400_fpu_fcmp): Likewise.
3823 (i6400_fpu_fmadd): Likewise.
3824 (i6400_int_mult): Include imul3nc type and update reservation.
3825 (i6400_int_div): Include idiv3 type and update reservation.
3826 (i6400_int_load): Update to check type not move_type.
3827 (i6400_int_store): Likewise.
3828 (i6400_int_prefetch): Set zero latency.
3830 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3832 * gcc.c: Document new %@{...} sequence.
3833 (LINK_COMMAND_SPEC): Use it for the -L switches.
3834 (cpp_unique_options): Use it for the -I switches.
3835 (at_file_argbuf): New global variable.
3836 (in_at_file): Likewise.
3837 (alloc_args): Create at_file_argbuf.
3838 (clear_args): Truncate at_file_argbuf.
3839 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3840 (open_at_file): New function.
3841 (close_at_file): Likewise.
3842 (create_at_file): Delete.
3843 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3846 (validate_switches_from_spec): Deal with %@{...} sequence.
3847 (validate_switches): Likewise.
3848 (driver::finalize): Call clear_args.
3850 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3852 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3854 2018-06-11 Martin Sebor <msebor@redhat.com>
3856 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3857 (Wno-class-memaccess): Add @opindex.
3858 (Wno-templates, Wno-multiple-inheritance): Same.
3859 (Wno-virtual-inheritance, Wno-namespaces): Same.
3860 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3861 (Wno-format-overflow, Wno-format-truncation): Same.
3862 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3863 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3864 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3865 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3866 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3867 (Wno-misspelled-isr): Same.
3869 2018-06-11 Martin Sebor <msebor@redhat.com>
3871 * PR tree-optimization/86083
3872 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3874 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3876 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3878 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3881 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3882 on the correct operand.
3883 (*movdi_internal64): Ditto.
3885 2018-06-11 Martin Liska <mliska@suse.cz>
3887 PR tree-optimization/86089
3888 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3890 2018-06-11 Julia Koval <julia.koval@intel.com>
3892 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3893 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3894 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3896 2018-06-11 Olivier Hainque <hainque@adacore.com>
3898 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3899 for Ada with strict dwarf2.
3901 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3904 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3907 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3909 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3911 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3913 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3916 2018-06-08 Martin Liska <mliska@suse.cz>
3918 * tree-cfg.h (debug_function): Fix argument type to match
3921 2018-06-08 Martin Liska <mliska@suse.cz>
3923 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3924 Remove usage of MPX-related (and removed) fields.
3925 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3927 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3929 * cfg.c (debug): Use TDF_NONE rather than 0.
3930 * cfghooks.c (debug): Likewise.
3931 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3932 (struct dump_option_value_info): Convert to...
3933 (struct kv_pair): ...this template type.
3934 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3936 (optinfo_verbosity_options): Likewise.
3937 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3939 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3940 than int for "optgroup_flags" param.
3941 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3943 (dump_dec): Likewise.
3944 (dump_finish): Use TDF_NONE rather than 0.
3945 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3946 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3947 than 0. Update for change to option_ptr.
3948 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3949 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3950 0. Update for changes to optinfo_verbosity_options and
3952 (opt_info_switch_p): Convert optgroup_flags from int to
3954 (dump_basic_block): Use dump_flags_t rather than int
3955 for "dump_kind" param.
3956 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3957 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3958 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3959 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3960 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3961 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3962 TDF_NONE): Convert from macros to...
3963 (enum dump_flag): ...this new enum.
3964 (dump_flags_t): Update to use enum.
3965 (operator|, operator&, operator~, operator|=, operator&=):
3966 Implement for dump_flags_t.
3967 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3968 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3969 Convert from macros to...
3970 (enum optgroup_flag): ...this new enum.
3971 (optgroup_flags_t): New typedef.
3972 (operator|, operator|=): Implement for optgroup_flags_t.
3973 (struct dump_file_info): Convert field "alt_flags" to
3974 dump_flags_t. Convert field "optgroup_flags" to
3976 (dump_basic_block): Use dump_flags_t rather than int for param.
3977 (dump_generic_expr_loc): Likewise.
3978 (dump_dec): Likewise.
3979 (dump_register): Convert param "optgroup_flags" to
3981 (opt_info_enable_passes): Likewise.
3982 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3984 * gimple-pretty-print.c (debug): Likewise.
3985 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3986 (merged_store_group::apply_stores): Likewise.
3987 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3988 * gimple.c (verify_gimple_pp): Likewise.
3989 * graphite-poly.c (print_pbb_body): Likewise.
3990 * passes.c (pass_manager::register_one_dump_file): Convert
3991 local "optgroup_flags" to optgroup_flags_t.
3992 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3994 (debug_body): Likewise.
3995 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3996 to optgroup_flags_t.
3997 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3999 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
4000 (convert_mult_to_fma): Likewise.
4001 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
4002 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
4003 * tree-vect-data-refs.c (dump_lower_bound): Convert param
4004 "dump_kind" to dump_flags_t.
4006 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4008 * config/rs6000/rs6000.c (min, max): Delete.
4010 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
4012 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
4013 -mabi=spe and -mabi=no-spe.
4015 2018-06-08 Martin Liska <mliska@suse.cz>
4017 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
4018 where we expect an existing summary.
4020 2018-06-08 Martin Liska <mliska@suse.cz>
4022 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
4023 * ipa-inline.h (estimate_edge_growth): Likewise.
4025 2018-06-08 Martin Liska <mliska@suse.cz>
4027 * cgraph.c (function_version_hasher::hash): Use
4028 cgraph_node::get_uid ().
4029 (function_version_hasher::equal):
4030 * cgraph.h (cgraph_node::get_uid): New method.
4031 * ipa-inline.c (update_caller_keys): Use
4032 cgraph_node::get_uid ().
4033 (update_callee_keys): Likewise.
4034 * ipa-utils.c (searchc): Likewise.
4035 (ipa_reduced_postorder): Likewise.
4036 * lto-cgraph.c (input_node): Likewise.
4037 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4038 * symbol-summary.h (symtab_insertion): Likewise.
4039 (symtab_removal): Likewise.
4040 (symtab_duplication): Likewise.
4041 * tree-pretty-print.c (dump_function_header): Likewise.
4042 * tree-sra.c (convert_callers_for_node): Likewise.
4044 2018-06-08 Martin Liska <mliska@suse.cz>
4046 * cgraph.c (symbol_table::create_edge): Always assign a new
4048 (symbol_table::free_edge): Do not recycle numbers.
4049 * cgraph.h (cgraph_edge::get): New method.
4050 * symbol-summary.h (symtab_removal): Use it.
4051 (symtab_duplication): Likewise.
4052 (call_summary::hashable_uid): Remove.
4054 2018-06-08 Martin Liska <mliska@suse.cz>
4056 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4057 (initialize_growth_caches): Remove.
4058 (free_growth_caches): Likewise.
4059 (do_estimate_edge_time): Use edge_growth_cache.
4060 (do_estimate_edge_size): Likewise.
4061 (do_estimate_edge_hints): Likewise.
4062 * ipa-inline.c (reset_edge_caches): Likewise.
4063 (recursive_inlining): Likewise.
4064 (inline_small_functions): Likewise.
4065 * ipa-inline.h (initialize_growth_caches): Remove.
4066 (estimate_edge_size): Likewise.
4067 (estimate_edge_time): Likewise.
4068 (estimate_edge_hints): Likewise.
4069 (reset_edge_growth_cache): Likewise.
4070 * symbol-summary.h (call_summary::remove): New method.
4072 2018-06-08 Martin Liska <mliska@suse.cz>
4074 * ipa-cp.c (class edge_clone_summary): New summary.
4075 (grow_edge_clone_vectors): Remove.
4076 (ipcp_edge_duplication_hook): Remove.
4077 (class edge_clone_summary_t): New call_summary class.
4078 (ipcp_edge_removal_hook): Remove.
4079 (edge_clone_summary_t::duplicate): New function.
4080 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4081 (create_specialized_node): Likewise.
4082 (ipcp_driver): Initialize edge_clone_summaries and do not
4085 2018-06-08 Martin Liska <mliska@suse.cz>
4087 * symbol-summary.h (get): New function.
4088 (call_summary::m_initialize_when_cloning): New class member.
4090 2018-06-08 Martin Liska <mliska@suse.cz>
4092 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4093 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4094 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4095 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4096 (remove_cgraph_node_from_order): Use the removed_nodes set.
4097 (do_per_function_toporder): Likwise.
4098 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4099 instead of summary_uid.
4100 (symtab_removal): Likewise.
4101 (symtab_duplication): Likewise.
4103 2018-06-08 Martin Liska <mliska@suse.cz>
4105 * ipa-cp.c (ipcp_store_bits_results): Use
4106 ipcp_transformation_sum.
4107 (ipcp_store_vr_results): Likewise.
4108 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4110 (ipcp_transformation_initialize): ... this.
4111 (ipa_set_node_agg_value_chain):
4112 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4113 (write_ipcp_transformation_info): Likewise.
4114 (read_ipcp_transformation_info): Likewise.
4115 (ipcp_update_bits): Likewise.
4116 (ipcp_update_vr): Likewise.
4117 (ipcp_transform_function): Likewise.
4118 * ipa-prop.h: Rename ipcp_transformation_summary to
4119 ipcp_transformation.
4120 (class ipcp_transformation_t): New function summary.
4121 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4122 (ipa_get_agg_replacements_for_node): Likewise.
4124 2018-06-08 Martin Liska <mliska@suse.cz>
4126 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4128 (class funct_state_summary_t): New function_summary class.
4129 (has_function_state): Remove.
4130 (get_function_state): Likewise.
4131 (set_function_state): Likewise.
4132 (add_new_function): Likewise.
4133 (funct_state_summary_t::insert): New function.
4134 (duplicate_node_data): Remove.
4135 (remove_node_data): Remove.
4136 (funct_state_summary_t::duplicate): New function.
4137 (register_hooks): Create new funct_state_summaries.
4138 (pure_const_generate_summary): Use it.
4139 (pure_const_write_summary): Likewise.
4140 (pure_const_read_summary): Likewise.
4141 (propagate_pure_const): Likewise.
4142 (propagate_nothrow): Likewise.
4143 (dump_malloc_lattice): Likewise.
4144 (propagate_malloc): Likewise.
4145 (execute): Do not register hooks, just remove summary
4147 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4150 2018-06-08 Martin Liska <mliska@suse.cz>
4152 * ipa-reference.c (remove_node_data): Remove.
4153 (duplicate_node_data): Likewise.
4154 (class ipa_ref_var_info_summary_t): New class.
4155 (class ipa_ref_opt_summary_t): Likewise.
4156 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4157 (get_reference_optimization_summary): Use
4158 ipa_ref_opt_sum_summaries.
4159 (set_reference_vars_info): Remove.
4160 (set_reference_optimization_summary): Likewise.
4161 (ipa_init): Create summaries.
4162 (init_function_info): Use function summary.
4163 (ipa_ref_opt_summary_t::duplicate): New function.
4164 (ipa_ref_opt_summary_t::remove): New function.
4165 (get_read_write_all_from_node): Fix GNU coding style.
4166 (propagate): Use function summary.
4167 (write_node_summary_p): Fix GNU coding style.
4168 (stream_out_bitmap): Likewise.
4169 (ipa_reference_read_optimization_summary): Use function summary.
4170 (ipa_reference_c_finalize): Do not release hooks.
4172 2018-06-08 Martin Liska <mliska@suse.cz>
4174 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4175 (analyze_function_body): Extract multiple calls of get_create.
4176 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4177 * ipa-inline.c (recursive_inlining): Use ::get method.
4178 * ipa-inline.h (estimate_edge_growth): Likewise.
4180 2018-06-08 Martin Liska <mliska@suse.cz>
4182 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4184 (hsa_function_summary::hsa_function_summary): Use the new enum
4186 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4187 * hsa-gen.c (hsa_get_host_function): Likewise.
4188 (get_brig_function_name): Likewise.
4189 * ipa-hsa.c (process_hsa_functions): Likewise.
4190 (ipa_hsa_write_summary): Likewise.
4191 * symbol-summary.h (symtab_duplication): Use ::get function/
4192 (get): New function.
4194 2018-06-08 Martin Liska <mliska@suse.cz>
4196 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4198 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4199 (hsa_register_kernel): Likewise.
4200 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4201 * hsa-gen.c (hsa_get_host_function): Likewise.
4202 (get_brig_function_name): Likewise.
4203 (generate_hsa): Likewise.
4204 (pass_gen_hsail::execute): Likewise.
4205 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4206 (devirtualization_time_bonus): Likewise.
4207 (ipcp_propagate_stage): Likewise.
4208 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4209 (edge_set_predicate): Likewise.
4210 (evaluate_conditions_for_known_args): Likewise.
4211 (evaluate_properties_for_edge): Likewise.
4212 (ipa_fn_summary::reset): Likewise.
4213 (ipa_fn_summary_t::duplicate): Likewise.
4214 (dump_ipa_call_summary): Likewise.
4215 (ipa_dump_fn_summary): Likewise.
4216 (analyze_function_body): Likewise.
4217 (compute_fn_summary): Likewise.
4218 (estimate_edge_devirt_benefit): Likewise.
4219 (estimate_edge_size_and_time): Likewise.
4220 (estimate_calls_size_and_time): Likewise.
4221 (estimate_node_size_and_time): Likewise.
4222 (inline_update_callee_summaries): Likewise.
4223 (remap_edge_change_prob): Likewise.
4224 (remap_edge_summaries): Likewise.
4225 (ipa_merge_fn_summary_after_inlining): Likewise.
4226 (ipa_update_overall_fn_summary): Likewise.
4227 (read_ipa_call_summary): Likewise.
4228 (inline_read_section): Likewise.
4229 (write_ipa_call_summary): Likewise.
4230 (ipa_fn_summary_write): Likewise.
4231 (ipa_free_fn_summary): Likewise.
4232 * ipa-hsa.c (process_hsa_functions): Likewise.
4233 (ipa_hsa_write_summary): Likewise.
4234 (ipa_hsa_read_section): Likewise.
4235 * ipa-icf.c (sem_function::merge): Likewise.
4236 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4237 (do_estimate_edge_time): Likewise.
4238 (estimate_size_after_inlining): Likewise.
4239 (estimate_growth): Likewise.
4240 (growth_likely_positive): Likewise.
4241 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4242 (inline_call): Likewise.
4243 * ipa-inline.c (caller_growth_limits): Likewise.
4244 (can_inline_edge_p): Likewise.
4245 (can_inline_edge_by_limits_p): Likewise.
4246 (compute_uninlined_call_time): Likewise.
4247 (compute_inlined_call_time): Likewise.
4248 (want_inline_small_function_p): Likewise.
4249 (edge_badness): Likewise.
4250 (update_caller_keys): Likewise.
4251 (update_callee_keys): Likewise.
4252 (recursive_inlining): Likewise.
4253 (inline_small_functions): Likewise.
4254 (inline_to_all_callers_1): Likewise.
4255 (dump_overall_stats): Likewise.
4256 (early_inline_small_functions): Likewise.
4257 (early_inliner): Likewise.
4258 * ipa-inline.h (estimate_edge_growth): Likewise.
4259 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4260 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4261 * ipa-prop.h (IPA_NODE_REF): Likewise.
4262 (IPA_EDGE_REF): Likewise.
4263 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4264 (propagate_malloc): Likewise.
4265 * ipa-split.c (execute_split_functions): Likewise.
4266 * symbol-summary.h: Rename get to get_create.
4268 (get_create): Likewise.
4269 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4271 2018-06-08 Martin Liska <mliska@suse.cz>
4273 * symbol-summary.h (release): Move definition out of class
4275 (symtab_removal): Likewise.
4276 (symtab_duplication): Likewise.
4278 2018-06-08 Martin Liska <mliska@suse.cz>
4280 * symbol-summary.h (function_summary): Move constructor
4281 implementation out of class declaration.
4282 (release): Likewise.
4283 (symtab_insertion): Likewise.
4284 (symtab_removal): Likewise.
4285 (symtab_duplication): Likewise.
4288 2018-06-08 Martin Liska <mliska@suse.cz>
4290 * Makefile.in: Remove support for MPX (macros, related functions,
4291 fields in cgraph_node, ...).
4292 * builtin-types.def (BT_BND): Likewise.
4293 (BT_FN_BND_CONST_PTR): Likewise.
4294 (BT_FN_CONST_PTR_BND): Likewise.
4295 (BT_FN_VOID_PTR_BND): Likewise.
4296 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4297 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4298 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4299 (expand_builtin_mempcpy_with_bounds): Likewise.
4300 (expand_builtin_memset_with_bounds): Likewise.
4301 (expand_builtin_memset_args): Likewise.
4302 (std_expand_builtin_va_start): Likewise.
4303 (expand_builtin): Likewise.
4304 (expand_builtin_with_bounds): Likewise.
4305 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4306 (DEF_LIB_BUILTIN_CHKP): Likewise.
4307 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4308 (DEF_CHKP_BUILTIN): Likewise.
4309 (BUILT_IN_MEMCPY): Likewise.
4310 (BUILT_IN_MEMMOVE): Likewise.
4311 (BUILT_IN_MEMPCPY): Likewise.
4312 (BUILT_IN_MEMSET): Likewise.
4313 (BUILT_IN_STPCPY): Likewise.
4314 (BUILT_IN_STRCAT): Likewise.
4315 (BUILT_IN_STRCHR): Likewise.
4316 (BUILT_IN_STRCPY): Likewise.
4317 (BUILT_IN_STRLEN): Likewise.
4318 (BUILT_IN_MEMCPY_CHK): Likewise.
4319 (BUILT_IN_MEMMOVE_CHK): Likewise.
4320 (BUILT_IN_MEMPCPY_CHK): Likewise.
4321 (BUILT_IN_MEMSET_CHK): Likewise.
4322 (BUILT_IN_STPCPY_CHK): Likewise.
4323 (BUILT_IN_STRCAT_CHK): Likewise.
4324 (BUILT_IN_STRCPY_CHK): Likewise.
4325 * calls.c (store_bounds): Likewise.
4326 (emit_call_1): Likewise.
4327 (special_function_p): Likewise.
4328 (maybe_warn_nonstring_arg): Likewise.
4329 (initialize_argument_information): Likewise.
4330 (finalize_must_preallocate): Likewise.
4331 (compute_argument_addresses): Likewise.
4332 (expand_call): Likewise.
4333 * cfgexpand.c (expand_call_stmt): Likewise.
4334 (expand_return): Likewise.
4335 (expand_gimple_stmt_1): Likewise.
4336 (pass_expand::execute): Likewise.
4337 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4338 (cgraph_node::remove): Likewise.
4339 (cgraph_node::dump): Likewise.
4340 (cgraph_node::verify_node): Likewise.
4341 * cgraph.h (chkp_function_instrumented_p): Likewise.
4342 (symtab_node::get_alias_target): Likewise.
4343 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4344 (cgraph_local_p): Likewise.
4345 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4346 (cgraph_edge::rebuild_references): Likewise.
4347 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4348 (walk_polymorphic_call_targets): Likewise.
4349 (cgraph_node::expand_thunk): Likewise.
4350 (symbol_table::output_weakrefs): Likewise.
4351 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4352 (ix86_handle_option): Likewise.
4353 * config/i386/constraints.md: Likewise.
4354 * config/i386/i386-builtin-types.def (BND): Likewise.
4358 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4359 (BDESC_FIRST): Likewise.
4361 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4362 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4363 * config/i386/i386.c (enum reg_class): Likewise.
4364 (ix86_target_string): Likewise.
4365 (ix86_option_override_internal): Likewise.
4366 (ix86_conditional_register_usage): Likewise.
4367 (ix86_valid_target_attribute_inner_p): Likewise.
4368 (ix86_set_indirect_branch_type): Likewise.
4369 (ix86_set_current_function): Likewise.
4370 (ix86_function_arg_regno_p): Likewise.
4371 (init_cumulative_args): Likewise.
4372 (ix86_function_arg_advance): Likewise.
4373 (ix86_function_arg): Likewise.
4374 (ix86_pass_by_reference): Likewise.
4375 (ix86_function_value_regno_p): Likewise.
4376 (ix86_function_value_1): Likewise.
4377 (ix86_function_value_bounds): Likewise.
4378 (ix86_return_in_memory): Likewise.
4379 (ix86_setup_incoming_vararg_bounds): Likewise.
4380 (ix86_va_start): Likewise.
4381 (indirect_thunk_need_prefix): Likewise.
4382 (print_reg): Likewise.
4383 (ix86_print_operand): Likewise.
4384 (ix86_expand_call): Likewise.
4385 (ix86_output_function_return): Likewise.
4386 (reg_encoded_number): Likewise.
4387 (BDESC_VERIFYS): Likewise.
4388 (ix86_init_mpx_builtins): Likewise.
4389 (ix86_init_builtins): Likewise.
4390 (ix86_emit_cmove): Likewise.
4391 (ix86_emit_move_max): Likewise.
4392 (ix86_expand_builtin): Likewise.
4393 (ix86_builtin_mpx_function): Likewise.
4394 (ix86_get_arg_address_for_bt): Likewise.
4395 (ix86_load_bounds): Likewise.
4396 (ix86_store_bounds): Likewise.
4397 (ix86_load_returned_bounds): Likewise.
4398 (ix86_store_returned_bounds): Likewise.
4399 (ix86_class_likely_spilled_p): Likewise.
4400 (ix86_hard_regno_mode_ok): Likewise.
4401 (x86_order_regs_for_local_alloc): Likewise.
4402 (ix86_mitigate_rop): Likewise.
4403 (ix86_bnd_prefixed_insn_p): Likewise.
4404 (ix86_mpx_bound_mode): Likewise.
4405 (ix86_make_bounds_constant): Likewise.
4406 (ix86_initialize_bounds): Likewise.
4407 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4408 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4409 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4410 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4411 (TARGET_CHKP_BOUND_MODE): Likewise.
4412 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4413 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4414 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4415 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4416 * config/i386/i386.h (TARGET_MPX): Likewise.
4417 (TARGET_MPX_P): Likewise.
4418 (VALID_BND_REG_MODE): Likewise.
4419 (FIRST_BND_REG): Likewise.
4420 (LAST_BND_REG): Likewise.
4421 (enum reg_class): Likewise.
4422 (BND_REG_P): Likewise.
4423 (BND_REGNO_P): Likewise.
4424 (BNDmode): Likewise.
4425 (ADJUST_INSN_LENGTH): Likewise.
4426 * config/i386/i386.md: Likewise.
4427 * config/i386/i386.opt: Likewise.
4428 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4429 (defined): Likewise.
4430 (LINK_MPX): Likewise.
4431 (MPX_SPEC): Likewise.
4432 (LIBMPX_SPEC): Likewise.
4433 (LIBMPXWRAPPERS_SPEC): Likewise.
4434 (CHKP_SPEC): Likewise.
4435 * config/i386/predicates.md: Likewise.
4436 * dbxout.c (dbxout_type): Likewise.
4437 * doc/extend.texi: Likewise.
4438 * doc/invoke.texi: Likewise.
4439 * doc/md.texi: Likewise.
4440 * doc/tm.texi: Likewise.
4441 * doc/tm.texi.in: Likewise.
4442 * dwarf2out.c (is_base_type): Likewise.
4443 (gen_formal_types_die): Likewise.
4444 (gen_subprogram_die): Likewise.
4445 (gen_type_die_with_usage): Likewise.
4446 (gen_decl_die): Likewise.
4447 (dwarf2out_late_global_decl): Likewise.
4448 * expr.c (expand_assignment): Likewise.
4449 (emit_storent_insn): Likewise.
4450 (store_expr_with_bounds): Likewise.
4451 (store_expr): Likewise.
4452 (expand_expr_real_1): Likewise.
4453 * expr.h (store_expr_with_bounds): Likewise.
4454 * function.c (use_register_for_decl): Likewise.
4455 (struct bounds_parm_data): Likewise.
4456 (assign_parms_augmented_arg_list): Likewise.
4457 (assign_parm_find_entry_rtl): Likewise.
4458 (assign_parm_is_stack_parm): Likewise.
4459 (assign_parm_load_bounds): Likewise.
4460 (assign_bounds): Likewise.
4461 (assign_parms): Likewise.
4462 (expand_function_start): Likewise.
4463 * gcc.c (CHKP_SPEC): Likewise.
4464 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4465 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4466 (wrestrict_dom_walker::check_call): Likewise.
4467 * gimple.c (gimple_build_call_from_tree): Likewise.
4468 * gimple.h (enum gf_mask): Likewise.
4469 (gimple_call_with_bounds_p): Likewise.
4470 (gimple_call_set_with_bounds): Likewise.
4471 * gimplify.c (gimplify_init_constructor): Likewise.
4472 * ipa-cp.c (initialize_node_lattices): Likewise.
4473 (propagate_constants_across_call): Likewise.
4474 (find_more_scalar_values_for_callers_subset): Likewise.
4475 * ipa-hsa.c (process_hsa_functions): Likewise.
4476 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4477 * ipa-icf.c (sem_function::merge): Likewise.
4478 * ipa-inline.c (early_inliner): Likewise.
4479 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4480 (warn_function_cold): Likewise.
4481 (propagate_pure_const): Likewise.
4482 * ipa-ref.h (enum GTY): Likewise.
4483 * ipa-split.c (find_retbnd): Likewise.
4484 (consider_split): Likewise.
4485 (split_function): Likewise.
4486 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4487 * ipa.c (walk_polymorphic_call_targets): Likewise.
4488 (symbol_table::remove_unreachable_nodes): Likewise.
4489 (process_references): Likewise.
4490 (cgraph_build_static_cdtor_1): Likewise.
4491 * lto-cgraph.c (lto_output_node): Likewise.
4492 (output_refs): Likewise.
4493 (compute_ltrans_boundary): Likewise.
4494 (input_overwrite_node): Likewise.
4495 (input_node): Likewise.
4496 (input_cgraph_1): Likewise.
4497 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4498 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4499 (class pass_chkp_instrumentation_passes): Likewise.
4500 (make_pass_chkp_instrumentation_passes): Likewise.
4501 * passes.def: Likewise.
4502 * rtl.h (struct GTY): Likewise.
4503 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4504 * stor-layout.c (layout_type): Likewise.
4505 * symtab.c: Likewise.
4506 * target.def: Likewise.
4507 * targhooks.c (default_chkp_bound_type): Likewise.
4508 (default_chkp_bound_mode): Likewise.
4509 (default_builtin_chkp_function): Likewise.
4510 (default_chkp_function_value_bounds): Likewise.
4511 (default_chkp_make_bounds_constant): Likewise.
4512 (default_chkp_initialize_bounds): Likewise.
4513 * targhooks.h (default_chkp_bound_type): Likewise.
4514 (default_chkp_bound_mode): Likewise.
4515 (default_builtin_chkp_function): Likewise.
4516 (default_chkp_function_value_bounds): Likewise.
4517 (default_chkp_make_bounds_constant): Likewise.
4518 (default_chkp_initialize_bounds): Likewise.
4519 * toplev.c (compile_file): Likewise.
4520 (process_options): Likewise.
4521 * tree-core.h (DEF_BUILTIN): Likewise.
4522 (DEF_BUILTIN_CHKP): Likewise.
4523 * tree-inline.c (declare_return_variable): Likewise.
4524 (remap_gimple_stmt): Likewise.
4525 (copy_bb): Likewise.
4526 (initialize_inlined_parameters): Likewise.
4527 (expand_call_inline): Likewise.
4528 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4529 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4530 (make_pass_ipa_chkp_produce_thunks): Likewise.
4531 (make_pass_chkp): Likewise.
4532 (make_pass_chkp_opt): Likewise.
4533 (make_pass_chkp_instrumentation_passes): Likewise.
4534 * tree-pretty-print.c (dump_generic_node): Likewise.
4535 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4536 * tree-ssa-dce.c (propagate_necessity): Likewise.
4537 (eliminate_unnecessary_stmts): Likewise.
4538 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4539 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4540 * tree-ssa-sccvn.h: Likewise.
4541 * tree-ssa-strlen.c (get_string_length): Likewise.
4542 (valid_builtin_call): Likewise.
4543 (adjust_last_stmt): Likewise.
4544 (handle_builtin_strchr): Likewise.
4545 (handle_builtin_strcpy): Likewise.
4546 (handle_builtin_stxncpy): Likewise.
4547 (handle_builtin_memcpy): Likewise.
4548 (handle_builtin_strcat): Likewise.
4549 (strlen_check_and_optimize_stmt): Likewise.
4550 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4551 * tree-streamer-in.c: Likewise.
4552 * tree-streamer.c (record_common_node): Likewise.
4553 * tree.c (tree_code_size): Likewise.
4554 (wide_int_to_tree_1): Likewise.
4555 (type_contains_placeholder_1): Likewise.
4556 (build_common_tree_nodes): Likewise.
4557 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4558 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4559 (POINTER_BOUNDS_P): Likewise.
4560 (BOUNDED_TYPE_P): Likewise.
4561 (BOUNDED_P): Likewise.
4562 (CALL_WITH_BOUNDS_P): Likewise.
4563 (pointer_bounds_type_node): Likewise.
4564 * value-prof.c (gimple_ic): Likewise.
4565 * var-tracking.c (vt_add_function_parameters): Likewise.
4566 * varasm.c (make_decl_rtl): Likewise.
4567 (assemble_start_function): Likewise.
4568 (output_constant): Likewise.
4569 (maybe_assemble_visibility): Likewise.
4570 * varpool.c (ctor_for_folding): Likewise.
4571 * chkp-builtins.def: Remove.
4572 * ipa-chkp.c: Remove.
4573 * ipa-chkp.h: Remove.
4574 * rtl-chkp.c: Remove.
4575 * rtl-chkp.h: Remove.
4576 * tree-chkp-opt.c: Remove.
4577 * tree-chkp.c: Remove.
4578 * tree-chkp.h: Remove.
4580 2018-06-07 Carl Love <cel@us.ibm.com>
4582 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4583 vextract_fp_from_shortl): Add BE support.
4585 2018-06-07 Paul Koning <ni1d@arrl.net>
4587 * compare-elim.c (try_merge_compare): Don't merge compare if
4588 address contains a side effect.
4589 (try_eliminate_compare): Likewise.
4591 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4593 * config.gcc: Support "tremont".
4594 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4595 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4597 * config/i386/i386.c (m_TREMONT): Define.
4598 (processor_target_table): Add "tremont".
4599 (PTA_TREMONT): Define.
4600 (ix86_lea_outperforms): Add TARGET_TREMONT.
4601 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4602 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4603 and M_INTEL_GOLDMONT_PLUS.
4604 (fold_builtin_cpu): Add "tremont".
4605 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4606 (ix86_option_override_internal): Add "tremont".
4607 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4608 (processor_type): Add PROCESSOR_TREMONT.
4609 * config/i386/x86-tune.def: Add m_TREMONT.
4610 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4612 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4614 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4615 symbol defined for msp430i* devices to be lower case.
4617 gcc/testsuite/gcc.target/msp430/
4618 * msp430i-device-symbol.c: New test.
4619 * msp430f-device-symbol.c: New test.
4620 * msp430.h: New test header file.
4622 2018-06-07 Richard Biener <rguenther@suse.de>
4624 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4625 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4626 Properly wrap signed arithmetic if overflow wraps.
4628 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4630 PR tree-optimization/69615
4631 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4632 of a cast from a same precision integral SSA_NAME in a bb dominated
4633 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4634 cast to utype if rhs2 has already a compatible type.
4636 2018-06-07 Richard Biener <rguenther@suse.de>
4638 PR tree-optimization/85935
4639 * graphite-scop-detection.c (find_params_in_bb): Analyze
4640 condition operands with respect to the correct loop. Assert
4641 the analysis doesn't fail.
4643 2018-06-04 Carl Love <cel@us.ibm.com>
4645 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4646 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4647 as it is slightly cheaper.
4648 (first_match_or_eos_index_<mode>):
4649 Calculate index using natural element order.
4650 (first_match_index_<mode>):
4651 Calculate index using natural element order.
4652 (first_match_or_eos_index_<mode>):
4653 Calculate index using natural order.
4654 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4655 for BE and LE modes.
4656 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4657 P9V_BUILTIN_VCLZLSBB_V16QI.
4658 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4661 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4663 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4664 indentation and line wrap for many prototypes. Add missing
4665 @smallexample directives around block of prototypes for vec_xl and
4668 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4670 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4671 track if we pass or return IEEE 128-bit floating point.
4672 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4673 C++ mangling that is compatible with GCC 8.1.
4674 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4675 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4676 floating point types.
4677 (rs6000_function_arg_advance_1): Likewise.
4678 (rs6000_mangle_type): Optionally generate mangled names that match
4679 what GCC 8.1 generated for IEEE 128-bit floating point types.
4680 (rs6000_globalize_decl_name): If we have an external function that
4681 passes or returns IEEE 128-bit types, generate a weak reference
4682 from the mangled name used in GCC 8.1 to the current mangled
4684 (rs6000_init_builtins): Make __ibm128 use the long double type if
4685 long double is IBM extended double. Make __float128 use the long
4686 double type if long double is IEEE 128-bit.
4689 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4690 macro for __ibm128 built-in functions.
4691 (PACK_IF): Add __ibm128 pack/unpack functions.
4692 (UNPACK_IF): Likewise.
4693 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4694 enable long double built-in functions if long double is IEEE
4695 128-bit floating point.
4696 (rs6000_invalid_builtin): Update long double built-in function
4698 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4699 functions, adjust the built-in function to use the long double
4700 built-in function if __ibm128 and long double are the same type.
4701 * doc/extend.texi (PowerPC builtins): Update documention for
4702 __builtin_{,un}pack_longdouble. Add documentation for
4703 __builtin_{,un}pack_ibm128.
4705 2018-06-06 Jim Wilson <jimw@sifive.com>
4707 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4708 (struct machine_function): New field interrupt_mode.
4709 (riscv_handle_type_attribute): New function. Add forward declaration.
4710 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4711 (riscv_expand_epilogue): Check interrupt_mode field.
4712 (riscv_set_current_function): Check interrupt attribute args and
4713 set interrupt_mode field.
4714 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4715 (riscv_sret, riscv_uret): New.
4716 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4717 new arguments to interrupt attribute.
4719 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4722 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4723 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4725 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4727 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4728 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4729 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4730 several redundant entries.
4732 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4734 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4735 type from "rtx" to "rtx_insn *".
4736 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4737 for local "call_insn", removing cast.
4738 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4740 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4742 PR tree-optimization/86066
4743 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4744 for BIT_INSERT_EXPR stores.
4746 2018-06-06 Richard Biener <rguenther@suse.de>
4748 PR tree-optimization/86062
4749 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4750 component refs ontop
4751 of to be offsetted base.
4753 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4755 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4756 to be static and remove check on interrupt attribute name.
4758 gcc/testsuite/gcc.target/msp430/
4759 * function-attributes-4.c: New test.
4760 * static-interrupts.c: New test.
4762 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4764 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4765 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4767 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4770 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4772 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4773 Remove second argument, change how error is called.
4774 (aarch64_layout_arg): Remove second argument from
4775 aarch64_err_no_fpadvsimd call.
4776 (aarch64_init_cumulative_args): Ditto.
4777 (aarch64_gimplify_va_arg_expr): Ditto.
4778 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4780 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4782 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4783 (*simple_return_indirect_internal<mode>): Rename from
4784 simple_return_indirect_internal. Use W mode iterator.
4785 (rstorssp): New expander.
4786 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4787 (clrssbsy): New expander.
4788 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4790 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4792 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4794 (cmse_check_pointed_object): Likewise.
4796 2018-06-05 Martin Liska <mliska@suse.cz>
4798 PR gcov-profile/47618
4799 * doc/invoke.texi: Document how -fprofile-dir format
4802 2018-06-05 Richard Biener <rguenther@suse.de>
4804 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4805 removal pretend DOM info isn't available so we do not update
4806 it and only remove edges, not dominated blocks. Actually free
4807 DOM info in case we removed something. Remove unreachable blocks.
4808 (mfb_keep_latches): Work with either DOM info or marked backedges.
4809 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4810 first. Mark backedges if DOM info isn't available.
4811 (Re-)compute DOM info after cleanup_control_flow_pre.
4813 2018-06-05 Richard Biener <rguenther@suse.de>
4815 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4816 (locus_discrim_hasher::hash): Adjust.
4817 (locus_discrim_hasher::equal): Likewise.
4818 (next_discriminator_for_locus): Work on line directly.
4819 (same_line_p): Pass in expanded locus1 as well.
4820 (assign_discriminators): Avoid redundant location expansions.
4822 2018-06-05 Richard Biener <rguenther@suse.de>
4824 PR tree-optimization/86046
4825 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4826 if required after clearing TREE_ADDRESSABLE.
4828 2018-06-05 Richard Biener <rguenther@suse.de>
4830 PR tree-optimization/86047
4831 * tree-ssa-loop.c (for_each_index): Glob handling of all
4832 decls and constants and really handle all of them.
4834 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4837 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4838 qualifier_void_pointer and qualifier_const_void_pointer.
4839 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4840 (arm_init_builtins): Handle the above.
4841 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4842 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4843 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4846 2018-06-05 Martin Liska <mliska@suse.cz>
4848 * auto-profile.c (read_autofdo_file): Do not use
4849 gcov_ctr_summary struct.
4850 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4851 * coverage.c (struct counts_entry): Likewise.
4852 (read_counts_file): Read just single summary entry.
4853 (get_coverage_counts): Use gcov_summary struct.
4854 * coverage.h (get_coverage_counts): Likewise.
4855 * gcov-dump.c (dump_working_sets): Likewise.
4856 (tag_summary): Dump just single summary.
4857 * gcov-io.c (gcov_write_summary): Write just histogram
4859 (gcov_read_summary): Read just single summary.
4860 (compute_working_sets): Use gcov_summary struct.
4861 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4862 of GCOV_COUNTERS_SUMMABLE.
4863 (GCOV_COUNTERS_SUMMABLE): Remove.
4864 (GCOV_FIRST_VALUE_COUNTER): Replace with
4865 GCOV_COUNTER_V_INTERVAL.
4866 (struct gcov_ctr_summary): Remove.
4867 (struct gcov_summary): Directly use fields of former
4869 (compute_working_sets): Use gcov_summary struct.
4870 * gcov.c (read_count_file): Do not use ctrs fields.
4871 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4873 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4875 * profile.c: Likewise.
4876 * profile.h: Likewise.
4878 2018-06-05 Martin Liska <mliska@suse.cz>
4880 PR gcov-profile/84846
4881 * gcov.c (output_lines): Print working directory only
4882 in intermediate format.
4884 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4886 * config/s390/s390-builtin-types.def: Add void function type.
4887 * config/s390/s390-builtins.def: Use the function type for the
4890 2018-06-04 Jim Wilson <jimw@sifive.com>
4892 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4894 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4895 and maybe_eh_return. Change regno to unsigned int. Use new args to
4896 handle EH_RETURN_DATA_REGNO registers properly.
4897 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4898 (riscv_expand_epilogue): Update comment. Change argument name and
4899 type. Update code to use new name and type. Pass new args to
4900 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4902 * config/riscv/riscv.md (NORMAL_RETURN): New.
4903 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4904 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4905 (eh_return): Call gen_eh_return_internal and emit barrier.
4906 (eh_return_internal): Call riscv_expand_epilogue.
4908 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4910 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4911 bit_insertion field and declare can_be_merged_into method.
4912 (merged_store_group::can_be_merged_into): New method.
4913 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4914 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4915 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4917 2018-06-04 Richard Biener <rguenther@suse.de>
4919 PR tree-optimization/85955
4920 * builtins.c (fold_builtin_sincos): Convert pointers to
4921 destination to appropriate type before dereferencing.
4923 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4925 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4927 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4929 * expr.c (expand_expr_real_1): Force the operand into memory if
4930 its TYPE_MODE is BLKmode and if there is no integer mode for
4931 the number of bits being extracted.
4933 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4937 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4938 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4940 2018-06-04 Richard Biener <rguenther@suse.de>
4942 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4943 (cleanup_tree_cfg_noloop): ... single caller. Do
4944 start_recording_case_labels later.
4946 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4948 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4949 to _IMMINTRIN_H_INCLUDED.
4950 * config/i386/pconfigintrin.h: Ditto.
4951 * config/i386/waitpkgintrin.h: Ditto.
4952 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4953 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4954 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4955 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4956 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4957 waitpkgintrin.h and cldemoteintrin.h.
4959 2018-06-04 Richard Biener <rguenther@suse.de>
4961 PR tree-optimization/86038
4962 * tracer.c (find_best_successor): Check probability for
4963 being initialized, bail out if not.
4965 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4968 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4969 of bits to ignore when comparing architectures.
4971 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4973 PR tree-optimization/69615
4974 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4975 maximum or minimum of the type, try to merge it also as if
4976 range1 is + [-, x - 1] or + [x + 1, -].
4979 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4981 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4983 PR tree-optimization/86034
4984 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4985 the unsigned bitfield type in a bit insertion sequence if it does not
4986 have a larger precision than the bitfield size.
4987 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4989 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4991 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4993 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4995 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4996 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4997 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4998 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
5000 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
5002 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
5003 Disable -fdelete-null-pointer-checks for ELF toolchain.
5005 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5006 Kito Cheng <kito.cheng@gmail.com>
5008 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
5009 (nds32le-*-*, nds32be-*-*): Integrate checking process.
5010 (nds32*-*-*): Add glibc and uclibc conditions.
5011 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
5012 (TARGET_EXCEPT_UNWIND_INFO): Define.
5013 * config/nds32/elf.h: New file.
5014 * config/nds32/linux.h: New file.
5015 * config/nds32/nds32-elf.opt: New file.
5016 * config/nds32/nds32-linux.opt: New file.
5017 * config/nds32/nds32-fp-as-gp.c
5018 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
5019 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
5021 (nds32_asm_file_end): Ditto.
5022 (nds32_print_operand): Ditto.
5023 (nds32_insert_attributes): Ditto.
5024 (nds32_init_libfuncs): New function.
5025 (TARGET_HAVE_TLS): Define.
5026 (TARGET_INIT_LIBFUNCS): Define.
5027 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
5029 (TARGET_ELF): Apply different mcmodel setting.
5030 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
5031 been migrated into elf.h and linux.h files.
5032 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
5033 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5034 (mcmodel): The content has been migrated into nds32-elf.opt and
5035 nds32-linux.opt files.
5036 * config/nds32/t-elf: New file.
5037 * config/nds32/t-linux: New file.
5039 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5040 Shiva Chen <shiva0217@gmail.com>
5042 * config/nds32/constants.md (unspec_volatile_element): Add
5043 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5044 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5046 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5047 (make_pass_nds32_fp_as_gp): Declare.
5048 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5050 (nds32_asm_function_end_prologue): Remove unused asm output.
5051 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5052 (nds32_asm_file_start): Output necessary fp_as_gp information.
5053 (nds32_option_override): Adjust register usage.
5054 (nds32_expand_prologue): Consider fp_as_gp situation.
5055 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5056 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5059 (simple_return): Ditto.
5060 (omit_fp_begin): Output special directive for fp_as_gp.
5061 (omit_fp_end): Output special directive for fp_as_gp.
5062 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5063 mforbid-fp-as-gp): New options.
5065 2018-06-01 Mark Wielaard <mark@klomp.org>
5067 * dwarf2out.c (dwarf2out_finish): Remove generation of
5068 DW_AT_loclists_base.
5070 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5072 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5073 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5074 (struct merged_store_group): Add bit_insertion field.
5075 (dump_char_array): Use standard hexadecimal format.
5076 (merged_store_group::merged_store_group): Set bit_insertion to false.
5077 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5078 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5079 also print the mask in the dump file.
5080 (pass_store_merging::gate): Minor tweak.
5081 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5082 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5083 stores with INTEGER_CST stores.
5084 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5085 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5086 and use it throughout. Generate bit insertion sequences if need be.
5087 (pass_store_merging::process_store): Remove redundant condition.
5088 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5090 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5092 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5093 the 128-bit floating point types. Fix function comment.
5095 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5097 * config/aarch64/aarch64-simd.md
5098 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5100 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5103 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5105 PR tree-optimization/85989
5106 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5108 (backprop::intersect_uses): Check it when deciding whether this
5109 is a backedge reference.
5110 (backprop::process_block): Add each phi to m_visited_phis
5111 after visiting it, then clear it at the end.
5113 2018-06-01 Richard Biener <rguenther@suse.de>
5115 * tree-vectorizer.h (vect_dr_stmt): New function.
5116 (vect_get_load_cost): Adjust.
5117 (vect_get_store_cost): Likewise.
5118 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5119 Use vect_dr_stmt instead of DR_SMTT.
5120 (vect_record_base_alignments): Likewise.
5121 (vect_calculate_target_alignment): Likewise.
5122 (vect_compute_data_ref_alignment): Likewise and make static.
5123 (vect_update_misalignment_for_peel): Likewise.
5124 (vect_verify_datarefs_alignment): Likewise.
5125 (vector_alignment_reachable_p): Likewise.
5126 (vect_get_data_access_cost): Likewise. Pass down
5127 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5128 (vect_get_peeling_costs_all_drs): Likewise.
5129 (vect_peeling_hash_get_lowest_cost): Likewise.
5130 (vect_enhance_data_refs_alignment): Likewise.
5131 (vect_find_same_alignment_drs): Likewise.
5132 (vect_analyze_data_refs_alignment): Likewise.
5133 (vect_analyze_group_access_1): Likewise.
5134 (vect_analyze_group_access): Likewise.
5135 (vect_analyze_data_ref_access): Likewise.
5136 (vect_analyze_data_ref_accesses): Likewise.
5137 (vect_vfa_segment_size): Likewise.
5138 (vect_small_gap_p): Likewise.
5139 (vectorizable_with_step_bound_p): Likewise.
5140 (vect_prune_runtime_alias_test_list): Likewise.
5141 (vect_analyze_data_refs): Likewise.
5142 (vect_supportable_dr_alignment): Likewise.
5143 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5144 (vect_gen_prolog_loop_niters): Likewise.
5145 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5146 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5148 (vect_recog_mask_conversion_pattern): Likewise.
5149 (vect_try_gather_scatter_pattern): Likewise.
5150 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5151 to vect_get_store_cost.
5152 (vect_get_store_cost): Get stmt_info instead of DR.
5153 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5154 (vect_get_load_cost): Get stmt_info instead of DR.
5156 2018-06-01 Richard Biener <rguenther@suse.de>
5159 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5160 vars wrapped in MEM_REFs.
5162 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5164 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5165 Fix subreg tests so that we only return a choice between
5166 GENERAL_REGS and FP_REGS if the original classes included both.
5168 2018-06-01 Richard Biener <rguenther@suse.de>
5171 * tree-ssa-structalias.c (get_function_part_constraint):
5172 Handle NULL fi->decl.
5173 (find_func_aliases_for_call): Properly handle indirect
5174 fi from direct call.
5175 (find_func_clobbers): Likewise.
5176 (ipa_pta_execute): Likewise.
5177 (create_variable_info_for): For functions that are ifunc_resolver
5178 resolve to a varinfo that contains the result of the resolver call.
5179 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5182 2018-05-31 Michael Collison <michael.collison@arm.com>
5184 * config/aarch64/aarch64.md:
5185 (*fix_to_zero_extenddfdi2): New pattern.
5186 * gcc.target/aarch64/fix_extend1.c: New testcase.
5188 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5192 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5193 and BUILT_IN_STRNCMP_EQ.
5194 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5195 BUILT_IN_STRNCMP_EQ.
5196 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5197 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5198 (gimple_fold_builtin): Likewise.
5199 * tree-ssa-strlen.c (compute_string_length): New function.
5200 (determine_min_obsize): New function.
5201 (handle_builtin_string_cmp): New function to handle calls to
5202 string compare functions.
5203 (strlen_optimize_stmt): Add handling to builtin string compare
5205 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5206 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5207 * tree.c (build_common_builtin_nodes): Add new defines of
5208 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5210 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5213 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5214 functions with naked attribute.
5216 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5218 * config/i386/sse.md (avx_vec_concat<mode>):
5219 Substitute concat_tg_mode mode attribute with xtg_mode.
5220 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5221 (concat_tg_mode): Remove mode attribute.
5223 2018-05-31 Martin Sebor <msebor@redhat.com>
5226 * calls.c (alloc_max_size): Correct a logic error/typo.
5227 Treat excessive arguments as infinite. Warn for invalid arguments.
5228 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5230 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5233 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5234 and movx for Haswell.
5236 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5237 Cesar Philippidis <cesar@codesourcery.com>
5240 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5241 when emitting error on private/firstprivate reductions.
5242 * omp-low.c (lower_omp_target): Avoid reference-type processing
5243 on pointers for firstprivate clause.
5245 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5247 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5250 * config/aarch64/aarch64-simd.md
5251 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5252 (aarch64_ld1_x3_<mode>): Likewise
5253 (aarch64_st1x2<VALLDIF:mode>): Likewise
5254 (aarch64_st1_x2_<mode>): Likewise
5255 (aarch64_st1x3<VALLDIF:mode>): Likewise
5256 (aarch64_st1_x3_<mode>): Likewise
5257 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5258 (vld1_s8_x3): Likewise.
5259 (vld1_u16_x3): Likewise.
5260 (vld1_s16_x3): Likewise.
5261 (vld1_u32_x3): Likewise.
5262 (vld1_s32_x3): Likewise.
5263 (vld1_u64_x3): Likewise.
5264 (vld1_s64_x3): Likewise.
5265 (vld1_f16_x3): Likewise.
5266 (vld1_f32_x3): Likewise.
5267 (vld1_f64_x3): Likewise.
5268 (vld1_p8_x3): Likewise.
5269 (vld1_p16_x3): Likewise.
5270 (vld1_p64_x3): Likewise.
5271 (vld1q_u8_x3): Likewise.
5272 (vld1q_s8_x3): Likewise.
5273 (vld1q_u16_x3): Likewise.
5274 (vld1q_s16_x3): Likewise.
5275 (vld1q_u32_x3): Likewise.
5276 (vld1q_s32_x3): Likewise.
5277 (vld1q_u64_x3): Likewise.
5278 (vld1q_s64_x3): Likewise.
5279 (vld1q_f16_x3): Likewise.
5280 (vld1q_f32_x3): Likewise.
5281 (vld1q_f64_x3): Likewise.
5282 (vld1q_p8_x3): Likewise.
5283 (vld1q_p16_x3): Likewise.
5284 (vld1q_p64_x3): Likewise.
5285 (vst1_s64_x2): Likewise.
5286 (vst1_u64_x2): Likewise.
5287 (vst1_f64_x2): Likewise.
5288 (vst1_s8_x2): Likewise.
5289 (vst1_p8_x2): Likewise.
5290 (vst1_s16_x2): Likewise.
5291 (vst1_p16_x2): Likewise.
5292 (vst1_s32_x2): Likewise.
5293 (vst1_u8_x2): Likewise.
5294 (vst1_u16_x2): Likewise.
5295 (vst1_u32_x2): Likewise.
5296 (vst1_f16_x2): Likewise.
5297 (vst1_f32_x2): Likewise.
5298 (vst1_p64_x2): Likewise.
5299 (vst1q_s8_x2): Likewise.
5300 (vst1q_p8_x2): Likewise.
5301 (vst1q_s16_x2): Likewise.
5302 (vst1q_p16_x2): Likewise.
5303 (vst1q_s32_x2): Likewise.
5304 (vst1q_s64_x2): Likewise.
5305 (vst1q_u8_x2): Likewise.
5306 (vst1q_u16_x2): Likewise.
5307 (vst1q_u32_x2): Likewise.
5308 (vst1q_u64_x2): Likewise.
5309 (vst1q_f16_x2): Likewise.
5310 (vst1q_f32_x2): Likewise.
5311 (vst1q_f64_x2): Likewise.
5312 (vst1q_p64_x2): Likewise.
5313 (vst1_s64_x3): Likewise.
5314 (vst1_u64_x3): Likewise.
5315 (vst1_f64_x3): Likewise.
5316 (vst1_s8_x3): Likewise.
5317 (vst1_p8_x3): Likewise.
5318 (vst1_s16_x3): Likewise.
5319 (vst1_p16_x3): Likewise.
5320 (vst1_s32_x3): Likewise.
5321 (vst1_u8_x3): Likewise.
5322 (vst1_u16_x3): Likewise.
5323 (vst1_u32_x3): Likewise.
5324 (vst1_f16_x3): Likewise.
5325 (vst1_f32_x3): Likewise.
5326 (vst1_p64_x3): Likewise.
5327 (vst1q_s8_x3): Likewise.
5328 (vst1q_p8_x3): Likewise.
5329 (vst1q_s16_x3): Likewise.
5330 (vst1q_p16_x3): Likewise.
5331 (vst1q_s32_x3): Likewise.
5332 (vst1q_s64_x3): Likewise.
5333 (vst1q_u8_x3): Likewise.
5334 (vst1q_u16_x3): Likewise.
5335 (vst1q_u32_x3): Likewise.
5336 (vst1q_u64_x3): Likewise.
5337 (vst1q_f16_x3): Likewise.
5338 (vst1q_f32_x3): Likewise.
5339 (vst1q_f64_x3): Likewise.
5340 (vst1q_p64_x3): Likewise.
5342 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5344 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5345 user_label_prefix to name.
5347 * tree-core.h: Update comment about the format of NAME string
5348 passed to handler in attribute_spec.
5350 * config/msp430/msp430.md: Remove erroneous subreg expression from
5351 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5352 zero_extend{q,h}isi2.
5354 2018-05-30 Borislav Petkov <bp@suse.de>
5356 * doc/extend.texi: Document some architecture specific
5357 constraints and sort entries.
5359 2018-05-30 Martin Sebor <msebor@redhat.com>
5362 * builtins.c (expand_builtin_stpcpy_1): New function.
5363 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5364 only if the former succeeds.
5366 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5368 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5371 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5373 * doc/invoke.texi (-flinker-output): Document
5375 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5377 * passes.c (ipa_write_summaries): Only modify statements if body
5379 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5380 incrementally linking.
5381 (ipa_passes): Likewise.
5382 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5383 pass down resolution info.
5384 * common.opt (flag_incremental_link): Update info.
5385 * gcc.c (plugin specs): Turn flinker-output=* to
5386 -plugin-opt=-linker-output-known
5387 * toplev.c (compile_file): Also cut compilation when doing incremental
5389 * flag-types. (enum lto_partition_model): Add
5390 LTO_LINKER_OUTPUT_NOLTOREL.
5391 (invoke.texi): Add -flinker-output docs.
5392 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5393 link same way as WPA; do not stream in dead initializers.
5395 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5396 darf2out_register_external_decl): Support incremental link.
5398 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5400 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5402 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5404 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5405 it down to simple_object_copy_lto_debug_sections.
5406 (run_gcc): Determine incremental LTO link time and configure
5407 lto1 into non-wpa mode, disable renaming of debug sections.
5409 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5411 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5412 descriptions of various incorrectly documented functions.
5414 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5417 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5418 address check not strict.
5420 2018-05-30 Richard Biener <rguenther@suse.de>
5422 PR tree-optimization/85964
5423 * tracer.c (better_p): Drop initialized count check, we only
5424 call the function with initialized counts now.
5425 (find_best_successor): Do find a best edge if one
5426 has uninitialized count.
5427 (find_best_predecessor): Likewise. Do BB frequency check only
5428 if count is initialized.
5430 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5432 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5433 (aarch64_ldrstr_offset_compare): New.
5434 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5435 load/store orderings.
5436 (aarch64_gen_adjusted_ldpstp): Likewise.
5438 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5440 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5441 Check for subset of GENERAL_REGS and FP_REGS.
5442 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5445 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5447 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5448 and wi::to_poly_offset. Add the current offset and then check
5449 whether the sum fits, rather than using an unchecked addition of
5450 a checked term. Check for a shwi rather than a uhwi.
5451 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5452 (store_constructor): Use poly_int_tree_p.
5453 (expand_expr_real_1): Likewise.
5454 * function.c (assign_temp): Likewise.
5455 * fold-const.c (const_binop): Use poly_int_tree_p and
5457 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5459 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5460 to_poly_offset for MEM offsets.
5461 * ipa-icf.c (sem_variable::equals): Likewise.
5462 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5463 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5464 wi::to_poly_offset for BIT_FIELD_REF offsets.
5465 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5467 * var-tracking.c (emit_note_insn_var_location): Use
5468 tree_to_poly_uint64.
5470 2018-05-29 Jim Wilson <jimw@sifive.com>
5472 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5474 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5477 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5478 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5480 (sse4_1_round<mode>2): Use nonimmediate_operand
5481 for operand 1 predicate.
5483 2018-05-29 Martin Sebor <msebor@redhat.com>
5484 Richard Biener <rguenther@suse.de>
5487 * calls.c (get_size_range): Call determine_value_range instead
5488 of get_value_range..
5489 * tree-vrp.h (determine_value_range): Declared new function.
5490 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5492 2018-05-29 Richard Biener <rguenther@suse.de>
5494 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5495 sure to use non-pattern stmts for get_earlier_stmt arguments.
5496 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5497 called on pattern stmts.
5498 (get_later_stmt): Likewise.
5500 2018-05-29 Martin Liska <mliska@suse.cz>
5502 PR gcov-profile/85759
5503 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5506 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5508 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5510 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5514 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5515 VEC_PACK_FLOAT_EXPR): New tree codes.
5516 * tree-pretty-print.c (op_code_prio): Handle
5517 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5518 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5519 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5520 * tree-inline.c (estimate_operator_cost): Likewise.
5521 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5522 * fold-const.c (const_binop): Likewise.
5523 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5524 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5525 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5526 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5527 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5528 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5529 * expr.c (expand_expr_real_2): Likewise.
5530 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5531 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5532 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5534 * optabs.c (expand_widen_pattern_expr): For
5535 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5536 sign from result type rather than operand's type.
5537 (expand_binop_directly): For vec_packu_float_optab and
5538 vec_packs_float_optab allow result type to be different from operand's
5540 * optabs-tree.c (optab_for_tree_code): Handle
5541 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5542 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5543 * tree-vect-generic.c (expand_vector_operations_1): Handle
5544 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5545 VEC_PACK_FLOAT_EXPR.
5546 * tree-vect-stmts.c (supportable_widening_operation): Handle
5548 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5549 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5550 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5551 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5552 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5554 (vec_pack<floatprefix>_float_<mode>): New expander.
5555 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5557 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5558 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5559 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5560 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5561 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5563 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5564 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5565 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5566 VEC_PACK_FLOAT_EXPR): Document.
5568 2018-05-29 Richard Biener <rguenther@suse.de>
5570 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5572 (stmt_vec_info_vec): Make pointer.
5573 (init_stmt_vec_info_vec): Remove.
5574 (free_stmt_vec_info_vec): Likewise.
5575 (set_stmt_vec_info_vec): New function.
5576 (free_stmt_vec_infos): Likewise.
5577 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5578 (set_vinfo_for_stmt): Likewise.
5579 (get_earlier_stmt): Likewise.
5580 (get_later_stmt): Likewise.
5581 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5582 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5583 (vec_info::~vec_info): Free stmt_vec_infos.
5584 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5585 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5586 (pass_slp_vectorize::execute): Likewise.
5587 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5588 (free_stmt_vec_info_vec): Likewise.
5589 (set_stmt_vec_info_vec): New function.
5590 (free_stmt_vec_infos): Likewise.
5591 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5592 the global stmt_vec_info_vec.
5593 * tree-parloops.c (gather_scalar_reductions): Use
5594 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5597 2018-05-29 Richard Biener <rguenther@suse.de>
5599 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5601 2018-05-29 Martin Liska <mliska@suse.cz>
5602 David Malcolm <dmalcolm@redhat.com>
5604 * vec.c (test_reverse): New.
5605 (vec_c_tests): Add new test.
5606 * vec.h (vl_ptr>::reverse): New function.
5608 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5610 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5612 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5615 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5617 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5619 2018-05-28 Richard Biener <rguenther@suse.de>
5621 PR tree-optimization/85933
5622 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5623 look at stmts marked as vectorizable.
5625 2018-05-28 Richard Biener <rguenther@suse.de>
5627 PR tree-optimization/85934
5628 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5629 vector boolean check before scalar optimization.
5631 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5633 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5636 2018-05-28 Mark Wielaard <mark@klomp.org>
5638 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5639 if it is an expression containing a minus sign.
5641 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5643 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5645 2018-05-27 Paul Koning <ni1d@arrl.net>
5647 * config/pdp11/pdp11.md (truncsihi2): Remove.
5649 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5650 Chung-Ju Wu <jasonwucj@gmail.com>
5652 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5654 (unaligned_store_dw): Ditto.
5655 * config/nds32/nds32-memory-manipulation.c
5656 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5657 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5658 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5659 (emit_setmem_word_loop): Rename to ...
5660 (emit_setmem_doubleword_loop): ... this.
5661 (nds32_gen_dup_4_byte_to_word_value): New function.
5662 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5663 (nds32_expand_setmem_loop): Refine implementation.
5664 (nds32_expand_setmem_loop_v3m): Ditto.
5665 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5668 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5670 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5672 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5674 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5675 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5676 machine->attr_no_prologue_p.
5677 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5678 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5679 (nds32_expand_epilogue): Consider attr_naked_p.
5680 (nds32_expand_epilogue_v3pop): Likewise.
5681 (nds32_can_use_return_insn): Likewise.
5682 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5683 attr_no_prologue_p fields.
5684 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5686 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5689 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5691 * config/i386/sse.md
5692 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5694 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5696 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5698 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5700 (*<floatsuffix>floatv2div2sf2): Rename to ...
5701 (*float<floatunssuffix>v2div2sf2): ... this.
5702 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5703 (float<floatunssuffix>v2div2sf2_mask): ... this.
5704 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5705 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5706 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5708 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5710 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5712 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5714 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5716 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5718 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5719 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5720 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5721 gen_ufix_truncv8dfv8si2.
5722 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5723 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5724 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5725 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5726 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5727 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5728 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5729 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5731 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5735 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5737 2018-05-25 Jim Wilson <jimw@sifive.com>
5739 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5740 * config/riscv/riscv.c (struct machine_function): Add
5741 interrupt_handler_p and attribute_checked_p fields.
5742 (riscv_attribute_table): Add interrupt.
5743 (riscv_interrupt_type_p): New.
5744 (riscv_save_reg_p): Save extra regs for interrupt handler.
5745 (riscv_use_save_libcall): Return false for interrupt handler.
5746 (riscv_first_stack_step): Add forward declaration.
5747 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5748 for interrupt handler with large frame. Use it for saved reg list.
5749 (riscv_expand_prologue): Move flag_stack_usage_info support to
5750 eliminate duplication.
5751 (riscv_expand_epilogue): Generate mret for interrupt handler.
5752 (riscv_epilogue_uses): New.
5753 (riscv_can_use_return_insn): Return false for interrupt handler.
5754 (riscv_function_ok_for_sibcall): Likewise.
5755 (riscv_set_current_function): Add interrupt handler support.
5756 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5757 * config/riscv/riscv.md (UNSPECV_MRET): New.
5759 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5761 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5763 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5765 PR tree-optimization/85712
5766 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5767 this candidate has already been replaced in-situ by a copy.
5769 2018-05-25 Jason Merrill <jason@redhat.com>
5771 PR c++/80485 - inline function non-zero address.
5772 * symtab.c (nonzero_address): Check DECL_COMDAT.
5774 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5777 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5778 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5779 extension of SImode operation. Use const123_operand predicate.
5780 (*saddsi_1): Remove.
5781 (*saddl_se_1): Ditto.
5783 (*ssubl_se_1): Ditto.
5784 * config/alpha/predicates.md (const123_operand): New predicate.
5785 * config/alpha/constraints.md (P): Use IN_RANGE.
5787 2018-05-25 Richard Biener <rguenther@suse.de>
5789 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5791 (ref_maybe_used_by_stmt_p): Likewise.
5792 (stmt_may_clobber_ref_p): Likewise.
5793 (stmt_may_clobber_ref_p_1): Likewise.
5794 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5796 (ref_maybe_used_by_stmt_p): Likewise.
5797 (stmt_may_clobber_ref_p): Likewise.
5798 (stmt_may_clobber_ref_p_1): Likewise.
5799 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5800 the alias oracle to disambiguate DRs with stmts DR analysis
5802 (vect_analyze_data_refs): Do not give up on not analyzable
5803 DRs for BB vectorization. Remove code truncating the dataref
5806 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5809 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5810 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5811 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5813 2018-05-25 Richard Biener <rguenther@suse.de>
5815 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5816 function, combining stmt data ref gathering and fatal analysis
5818 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5819 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5820 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5821 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5822 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5824 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5826 PR tree-optimization/85720
5827 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5828 SCC if all partitions are builtins.
5829 (version_loop_by_alias_check): New parameter. Generate cancelable
5830 runtime alias check if all partitions are builtins.
5831 (distribute_loop): Update call to above function.
5833 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5835 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5836 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5837 (parm_default_def_partition_arg): Ditto.
5838 (set_parm_default_def_partition): Ditto.
5839 (get_parm_default_def_partitions): Ditto and make it static.
5840 (get_undefined_value_partitions): Ditto and make it static.
5841 (remove_ssa_form): Refactor call to init_var_map here.
5842 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5843 computation for loop region.
5844 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5845 (register_default_def): Delete.
5846 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5847 (parm_default_def_partition_arg): Ditto.
5848 (set_parm_default_def_partition): Ditto.
5849 (get_parm_default_def_partitions): Ditto and make it static.
5850 (get_undefined_value_partitions): Ditto and make it static.
5851 (coalesce_with_default, coalesce_with_default): Update comment.
5852 (create_coalesce_list_for_region): New func factored out from
5853 create_outofssa_var_map.
5854 (populate_coalesce_list_for_outofssa): New func factored out from
5855 create_outofssa_var_map and coalesce_ssa_name.
5856 (create_outofssa_var_map): Delete.
5857 (coalesce_ssa_name): Refactor to support live range computation.
5858 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5859 (get_parm_default_def_partitions): Delete.
5860 (get_undefined_value_partitions): Ditto.
5861 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5862 computation for loop region.
5863 (new_tree_live_info, loe_visit_block): Ditto.
5864 (live_worklist, set_var_live_on_entry): Ditto.
5865 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5866 * tree-ssa-live.h (struct _var_map): New fields.
5867 (init_var_map): Change decl.
5868 (region_contains_p): New.
5870 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5872 * tree-ssa-live.h (live_merge_and_clear): Delete.
5874 2018-05-25 Richard Biener <rguenther@suse.de>
5877 * tree-dump.c (dequeue_and_dump): Remove access to removed
5878 operand 2 of a SWITCH_EXPR.
5880 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5882 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5883 multiplication and division.
5884 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5885 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5886 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5887 (cond_udiv_optab, cond_umod_optab): New optabs.
5888 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5889 (IFN_COND_RDIV): New internal functions.
5890 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5891 TRUNC_MOD_EXPR and RDIV_EXPR.
5892 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5893 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5895 (SVE_INT_BINARY): Include mult.
5896 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5897 (optab, sve_int_op): Handle mult.
5898 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5900 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5901 for SVE_INT_BINARY_SD.
5903 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5905 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5906 (optab, sve_int_op): Handle div and udiv.
5907 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5908 for SVE_INT_BINARY_SD.
5909 (*<optab><mode>3): New insn for the same.
5911 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5913 * tree-vect-patterns.c: Include predict.h.
5914 (vect_recog_divmod_pattern): Restrict check for division support
5915 to when optimizing for size.
5917 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5919 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5920 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5921 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5922 (gimple_match_op::set_op): Likewise.
5923 (gimple_resimplify4): Declare.
5924 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5925 (expr::gen_transform): Likewise.
5926 (decision_tree::gen): Generate a simplification routine for 4 operands.
5927 * gimple-match-head.c (gimple_simplify): Add an overload for
5928 4 operands. In the top-level function, handle up to 4 call
5929 arguments and call gimple_resimplify4.
5930 (gimple_resimplify4): New function.
5931 (build_call_internal): Pass a fourth operand.
5932 (maybe_push_to_seq): Likewise.
5933 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5934 Fold VEC_COND_EXPRs of an operation and a default value into
5935 an IFN_COND_* function if possible.
5936 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5938 (SVE_COND_FP_BINARY): Include them.
5939 (optab, sve_fp_op): Handle them.
5940 (SVE_INT_BINARY_REV): New code iterator.
5941 (SVE_COND_FP_BINARY_REV): New int iterator.
5942 (commutative): New int attribute.
5943 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5945 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5947 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5948 (*cond_<optab><mode>): New patterns for reversed operands.
5950 2018-05-25 Richard Biener <rguenther@suse.de>
5952 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5953 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5954 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5955 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5956 * tree-vect-data-refs.c (everywhere): Adjust users.
5957 * tree-vect-loop.c (everywhere): Likewise.
5958 * tree-vect-slp.c (everywhere): Likewise.
5959 * tree-vect-stmts.c (everywhere): Likewise.
5960 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5962 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5964 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5966 (gcc_cv_as_section_exclude): ... this.
5967 Try Solaris as #exclude syntax.
5968 * configure: Regenerate.
5969 * config.in: Regenerate.
5970 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5972 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5973 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5975 * varasm.c (default_elf_asm_named_section): Don't check if
5976 HAVE_GAS_SECTION_EXCLUDE is defined.
5978 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5980 * doc/md.texi: Update the documentation of the cond_* optabs
5981 to mention the new final operand. Fix GET_MODE_NUNITS call.
5982 Describe the scalar case too.
5983 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5984 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5986 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5987 (get_conditional_internal_fn): Update comment.
5988 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5989 accumulator value as a final argument to conditional functions.
5990 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5991 a define_expand and add an "else" operand. Assert for now that
5992 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5993 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5994 (*cond_<optab><mode>): New patterns.
5995 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5996 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5997 (UNSPEC_COND_EOR): Delete.
5998 (optab): Remove associated mappings.
5999 (SVE_INT_BINARY): New code iterator.
6000 (sve_int_op): Remove int attribute and add "minus" to the code
6002 (SVE_COND_INT_OP): Delete.
6003 (SVE_COND_FP_OP): Rename to...
6004 (SVE_COND_FP_BINARY): ...this.
6006 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
6008 * optabs.c (can_reuse_operands_p): New function.
6009 (maybe_legitimize_operands): Try to reuse the results for
6012 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6014 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
6015 Add {q} suffix to insn mnemonic.
6017 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
6019 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
6020 (msp430_warn_func_return): New.
6022 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
6024 * fold-const.c (tree_nonzero_bits): New function.
6025 * fold-const.h (tree_nonzero_bits): Likewise.
6026 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
6027 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
6029 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
6033 * varasm.c (assemble_alias): Check ifunc_resolver only on
6036 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6039 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6040 when memory input operand is handled.
6042 2018-05-24 Luis Machado <luis.machado@linaro.org>
6044 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6046 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6048 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6050 * match.pd: Delay FMA folds until after vectorization.
6052 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6055 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6056 address check not strict.
6058 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6060 * gimple-match.h (gimple_match_op): New class.
6061 (mprts_hook): Replace parameters with a gimple_match_op *.
6062 (maybe_build_generic_op): Likewise.
6063 (gimple_simplified_result_is_gimple_val): Replace parameters with
6064 a const gimple_match_op *.
6065 (gimple_simplify): Replace code_helper * and tree * parameters with
6066 a gimple_match_op * parameter.
6067 (gimple_resimplify1): Replace code_helper *, tree and tree *
6068 parameters with a gimple_match_op * parameter.
6069 (gimple_resimplify2): Likewise.
6070 (gimple_resimplify3): Likewise.
6071 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6072 parameters with a gimple_match_op * parameter.
6073 * gimple-match-head.c (gimple_simplify): Change prototypes of
6074 auto-generated functions to take a gimple_match_op * instead of
6075 separate code_helper * and tree * parameters. Make the same
6076 change in the top-level overload and update calls to the
6077 gimple_resimplify routines. Update calls to the auto-generated
6078 functions and to maybe_push_res_to_seq in the publicly-facing
6079 operation-specific gimple_simplify overloads.
6080 (gimple_match_op::MAX_NUM_OPS): Define.
6081 (gimple_resimplify1): Replace rcode and ops with a single res_op
6082 parameter. Update call to gimple_simplify.
6083 (gimple_resimplify2): Likewise.
6084 (gimple_resimplify3): Likewise.
6085 (mprts_hook): Replace parameters with a gimple_match_op *.
6086 (maybe_build_generic_op): Likewise.
6087 (build_call_internal): Replace type, nargs and ops with
6088 a gimple_match_op *.
6089 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6090 with a single gimple_match_op *. Update calls to mprts_hook,
6091 build_call_internal and gimple_simplified_result_is_gimple_val.
6092 Factor out code that is common to the tree_code and combined_fn cases.
6093 * genmatch.c (expr::gen_transform): Replace tem_code and
6094 tem_ops with a gimple_match_op called tem_op. Update calls
6095 to the gimple_resimplify functions and maybe_push_res_to_seq.
6096 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6097 res_ops. Update call to the gimple_resimplify functions.
6098 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6099 (decision_tree::gen): Make the functions take a gimple_match_op *
6100 called res_op instead of separate res_code and res_ops parameters.
6101 Update call accordingly.
6102 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6103 and ops with a single res_op parameter. Update calls to
6104 maybe_build_generic_op and maybe_push_res_to_seq.
6105 (fold_stmt_1): Update calls to gimple_simplify and
6106 replace_stmt_with_simplification.
6107 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6108 and gimple_simplified_result_is_gimple_val.
6109 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6111 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6112 with a gimple_match_op *.
6113 (vn_nary_build_or_lookup): Likewise. Update call to
6114 vn_nary_build_or_lookup_1.
6115 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6116 gimple_match_op *. Update calls to the gimple_resimplify routines
6117 and to gimple_simplified_result_is_gimple_val.
6118 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6119 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6120 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6121 (visit_nary_op): Likewise.
6122 (visit_reference_op_load): Likewise.
6124 2018-05-23 Luis Machado <luis.machado@linaro.org>
6126 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6127 modifier for printing the step amount.
6129 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6132 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6135 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6137 * doc/sourcebuild.texi (Endianness): New subsubsection.
6139 2018-05-23 Luis Machado <luis.machado@linaro.org>
6141 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6142 <prefetch_dynamic_strides>: New const bool field.
6143 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6144 prefetch_dynamic_strides.
6145 (exynosm1_prefetch_tune): Likewise.
6146 (thunderxt88_prefetch_tune): Likewise.
6147 (thunderx_prefetch_tune): Likewise.
6148 (thunderx2t99_prefetch_tune): Likewise.
6149 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6151 (aarch64_override_options_internal): Update to set
6152 PARAM_PREFETCH_DYNAMIC_STRIDES.
6153 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6154 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6155 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6156 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6157 prefetch-dynamic-strides setting.
6159 2018-05-23 Luis Machado <luis.machado@linaro.org>
6161 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6162 <minimum_stride>: New const int field.
6163 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6164 minimum_stride field defaulting to -1.
6165 (exynosm1_prefetch_tune): Likewise.
6166 (thunderxt88_prefetch_tune): Likewise.
6167 (thunderx_prefetch_tune): Likewise.
6168 (thunderx2t99_prefetch_tune): Likewise.
6169 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6170 <default_opt_level>: Set to 3.
6171 (aarch64_override_options_internal): Update to set
6172 PARAM_PREFETCH_MINIMUM_STRIDE.
6173 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6174 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6175 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6176 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6177 stride is constant and is below the minimum stride threshold.
6179 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6181 * config/arm/arm-cpus.in (mode26): Delete.
6182 (armv4): Delete mode26 reference.
6183 * config/arm/arm.c (arm_configure_build_target): Delete use of
6186 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6188 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6190 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6191 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6193 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6194 (floatunsdidf2): Ditto.
6196 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6198 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6199 (fixuns_trunc<mode>si2_avx512f): Ditto.
6200 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6201 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6202 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6204 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6206 PR rtl-optimization/79985
6207 * df-scan.c (df_insn_refs_collect): Remove special case for
6208 global registers and asm statements.
6210 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6212 * extend.texi (Global Register Variables): Rewrite the bullet list.
6213 Note that the register is available for allocation. Note that access
6214 via inline asm must use constraints. Add note about async-signal
6215 handlers. Remove paragraph about automagic register selection.
6217 2018-05-23 Richard Biener <rguenther@suse.de>
6219 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6220 of fixed offset from memset VN.
6222 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6224 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6226 (alloc_cand_and_find_basis): Initialize first_interp field.
6227 (slsr_process_mul): Modify first_interp field.
6228 (slsr_process_add): Likewise.
6229 (slsr_process_cast): Modify first_interp field for each new
6231 (slsr_process_copy): Likewise.
6232 (dump_candidate): Dump first_interp field.
6233 (replace_mult_candidate): Process all interpretations, not just
6235 (replace_rhs_if_not_dup): Likewise.
6236 (replace_one_candidate): Likewise.
6238 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6240 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6242 (aarch64_needs_frame_chain): New function.
6243 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6245 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6248 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6249 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6250 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6251 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6252 as true for strict-align.
6253 (aarch64_can_inline_p): Perform checks even when callee has no
6254 attributes to check for strict alignment.
6255 * doc/extend.texi (AArch64 Function Attributes): Document
6257 * doc/invoke.texi: (AArch64 Options): Likewise.
6259 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6261 PR tree-optimization/85853
6262 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6263 the handling of the root of the node to...
6264 (vect_slp_analyze_node_operations_1): ...this new function,
6265 and run the whole thing with the child nodes' def types
6266 set according to their SLP node's def type.
6268 2018-05-23 Richard Biener <rguenther@suse.de>
6271 * tree-data-ref.c (create_runtime_alias_checks): Defer
6272 and ignore overflow warnings.
6274 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6276 PR tree-optimization/85822
6277 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6280 2018-05-23 Richard Biener <rguenther@suse.de>
6282 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6283 memset constants via native_interpret_expr.
6285 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6288 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6290 (cgraph_node::create_alias): Likewise.
6291 (cgraph_node::get_availability): Check ifunc_resolver instead
6292 of looking up ifunc attribute.
6293 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6294 * varasm.c (do_assemble_alias): Likewise.
6295 (assemble_alias): Likewise.
6296 (default_binds_local_p_3): Likewise.
6297 * cgraph.h (cgraph_node): Add ifunc_resolver.
6298 (cgraph_node::only_called_directly_or_aliased_p): Return false
6300 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6302 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6303 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6304 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6305 instead of looking up ifunc attribute.
6307 2018-05-22 Luis Machado <luis.machado@linaro.org>
6309 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6311 2018-05-22 Martin Sebor <msebor@redhat.com>
6314 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6315 only when expasion succeeds.
6316 (expand_builtin_strcmp): Same.
6317 (expand_builtin_strncmp): Same.
6319 2018-05-22 Martin Sebor <msebor@redhat.com>
6321 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6323 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6324 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6326 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6327 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6329 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6330 Add check for legitimate_address.
6331 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6332 (aarch64_swap_ldrstr_operands): New.
6333 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6336 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6337 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6339 * config/aarch64/aarch64.md: New patterns to generate stp
6341 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6342 single words and double words.
6343 (load_pair_sw, load_pair_dw): Likewise.
6344 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6346 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6348 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6349 for different mode ldpstp and add peephole for merged zero stores.
6351 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6353 (aarch64_gen_store_pair): Rename calls to match new patterns.
6354 (aarch64_gen_load_pair): Rename calls to match new patterns.
6355 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6356 (load_pair<DREG:mode><DREG2:mode>): ... This.
6357 (store_pair<mode>): Rename to...
6358 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6359 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6361 (V_INT_EQUIV): Handle SImode.
6362 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6365 2018-05-22 Martin Sebor <msebor@redhat.com>
6368 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6369 or ajust the presumed bound on an operation to avoid unnecessary
6372 2018-05-22 Martin Sebor <msebor@redhat.com>
6374 PR tree-optimization/85826
6375 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6376 assuming that a DECL necesarily has a constant size.
6378 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6381 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6383 2018-05-22 Richard Biener <rguenther@suse.de>
6385 PR tree-optimization/85834
6386 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6387 non-constant and non-zero memset arguments.
6389 2018-05-22 Martin Liska <mliska@suse.cz>
6392 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6394 2018-05-22 Richard Biener <rguenther@suse.de>
6396 PR tree-optimization/85863
6397 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6398 comparisons when vectype is specified.
6399 (vectorizable_condition): Do not specify vectype for
6400 vect_is_simple_cond when SLP vectorizing.
6402 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6405 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6406 define __ibm128 as long double.
6407 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6408 as a distinct type when IEEE 128-bit support is enabled.
6409 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6410 128-bit types to use the correct functions.
6411 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6412 convert between 128-bit floating point types that have different
6413 modes but the same representation, instead of using gen_lowpart to
6415 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6417 (IFKF_reg): New attributes to give the register constraints for
6419 (extend<mode>tf2_internal): New insns to mark an explicit
6420 conversion between 128-bit floating point types that have a
6421 different mode but share the same representation.
6423 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6425 PR tree-optimization/85814
6426 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6427 a null return from get_strinfo when unsharing the next
6428 strinfo in the chain.
6430 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6433 * varasm.c (weak_finish): Clean up weak_decls.
6435 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6437 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6438 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6439 UNSPEC_UADALP values.
6440 * config/aarch64/iterators.md (ABAL): New int iterator.
6443 (sur): Add mappings for the above.
6444 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6446 (aarch64_<sur>abal<mode>_4): Likewise.
6447 (aarch64_<sur>adalp<mode>_3): Likewise.
6448 (<sur>sadv16qi): New define_expand.
6450 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6452 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6453 (*movdf_internal): Ditto.
6454 (*rcpsf2_sse): Ditto.
6455 (*rsqrtsf2_sse): Ditto.
6456 (*sqrt<mode>2_sse): Ditto.
6458 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6460 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6462 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6463 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6464 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6465 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6467 * config/aarch64/arm_neon.h: Likewise.
6468 * config/aarch64/iterators.md (VQ_I): New.
6470 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6472 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6474 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6477 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6479 * config/nds32/constraints.md (S): New constraint.
6480 * config/nds32/nds32.md (call_internal): Use constraint S.
6481 (call_value_internal): Likewise.
6482 (sibcall_internal): Likewise.
6483 (sibcall_value_internal): Likewise.
6485 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6486 Chung-Ju Wu <jasonwucj@gmail.com>
6488 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6491 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6492 Chung-Ju Wu <jasonwucj@gmail.com>
6494 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6495 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6496 (nds32_rtx_costs_impl): Simplify.
6497 (nds32_address_cost_impl): Simplify.
6498 (nds32_init_rtx_costs): New function.
6499 (nds32_rtx_costs_speed_prefer): Likewise.
6500 (nds32_rtx_costs_size_prefer): Likewise.
6501 (nds32_address_cost_speed_prefer): Likewise.
6502 (nds32_address_cost_speed_fwprop): Likewise.
6503 (nds32_address_cost_size_prefer): Likewise.
6504 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6505 * config/nds32/nds32.c (nds32_option_override): Use
6506 nds32_init_rtx_costs function.
6508 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6510 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6511 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6512 (TARGET_PIPELINE_N8): Likewise.
6513 (TARGET_PIPELINE_N10): Likewise.
6514 (TARGET_PIPELINE_N13): Likewise.
6515 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6517 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6519 * config/nds32/nds32-fpu.md: Update copyright year.
6521 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6523 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6525 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6527 * config/nds32/nds32.c
6528 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6529 * config/nds32/nds32.opt (minline-asm-r15): New option.
6531 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6533 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6535 * config/nds32/nds32.md (abssi2): New pattern.
6537 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6539 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6540 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6541 Merge insn pattern from sse_cvtsi2ss<round_name> and
6542 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6543 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6544 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6545 using SWI48 mode iterator.
6546 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6547 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6548 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6549 pattern from sse_cvttss2si<round_saeonly_name>
6550 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6551 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6552 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6553 using SWI48 mode iterator.
6554 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6555 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6556 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6557 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6558 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6559 using SWI48 mode iterator.
6560 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6561 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6562 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6563 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6564 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6565 SWI48 mode iterator.
6566 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6567 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6568 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6569 pattern from sse_cvttsd2si<round_saeonly_name>
6570 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6572 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6574 * config/nds32/nds32-md-auxiliary.c
6575 (nds32_valid_smw_lwm_base_p): Refine.
6576 (nds32_output_smw_single_word): Refine.
6577 (nds32_output_smw_double_word): New.
6578 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6580 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6582 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6583 (nds32_output_stack_pop): Refine.
6584 (nds32_expand_unaligned_load): Refine.
6585 (nds32_expand_unaligned_store): Refine.
6587 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6588 Chung-Ju Wu <jasonwucj@gmail.com>
6590 * config/nds32/constants.md: Add TP_REGNUM constant.
6591 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6592 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6594 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6595 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6596 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6597 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6598 and PIC code generation.
6599 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6601 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6603 * config/nds32/nds32.md: Support TLS and PIC.
6604 * config/nds32/nds32.c: Support TLS and PIC.
6605 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6606 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6609 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6611 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6612 mode with E_ prefix.
6614 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6615 Chung-Ju Wu <jasonwucj@gmail.com>
6617 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6618 * config/nds32/nds32-md-auxiliary.c
6619 (symbolic_reference_mentioned_p): New.
6620 (nds32_legitimize_ict_address): New.
6621 (nds32_expand_ict_move): New.
6622 (nds32_indirect_call_referenced_p): New.
6623 (nds32_symbol_binds_local_p): Delete.
6624 (nds32_long_call_p): Modify.
6625 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6626 * config/nds32/nds32-protos.h
6627 (symbolic_reference_mentioned_p): Declare.
6628 (nds32_legitimize_ict_address): Declare.
6629 (nds32_expand_ict_move): Declare.
6630 (nds32_indirect_call_referenced_p): Declare.
6631 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6632 (nds32_relax_group): Use nds32_ict_const_p as condition.
6633 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6634 (nds32_asm_file_start): Output ict_model directive in asm code.
6635 (nds32_legitimate_address_p): Consider indirect call.
6636 (nds32_print_operand): Consider indirect call.
6637 (nds32_print_operand_address): Consider indirect call.
6638 (nds32_insert_attributes): Handle "indirect_call" attribute.
6639 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6640 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6641 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6642 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6643 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6644 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6645 (TARGET_ICT_MODEL_SMALL): Define.
6646 (TARGET_ICT_MODEL_LARGE): Define.
6647 * config/nds32/nds32.md (movsi): Consider ict model.
6648 (call, call_value): Consider ict model.
6649 (sibcall, sibcall_value): Consider ict model.
6650 * config/nds32/nds32.opt (mict-model): New option.
6651 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6654 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6655 Monk Chiang <sh.chiang04@gmail.com>
6656 Jim Wilson <jimw@sifive.com>
6658 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6659 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6660 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6661 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6662 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6663 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6664 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6665 compute save_libcall_adjustment properly.
6666 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6667 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6668 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6669 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6670 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6671 (ABI_SPEC): Handle mabi=ilp32e.
6672 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6673 (RVE): Add RVE mask.
6674 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6675 <-march>: Add rv32e as an example.
6677 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6680 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6681 (intra_create_variable_infos): Handle C++ constructors.
6683 2018-05-18 Martin Liska <mliska@suse.cz>
6685 * passes.def: Remove a redundant pass.
6687 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6690 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6692 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6694 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6696 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6698 (armv2, armv2a, armv3, armv3m): Delete architectures.
6699 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6700 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6701 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6703 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6704 (*mulsidi3adddi): Likewise.
6705 (mulsidi3): Likewise.
6706 (*mulsidi3_nov6): Likewise.
6707 (umulsidi3): Likewise.
6708 (umulsidi3_nov6): Likewise.
6709 (umaddsidi4): Likewise.
6710 (*umulsidi3adddi): Likewise.
6711 (smulsi3_highpart): Likewise.
6712 (*smulsi3_highpart_nov6): Likewise.
6713 (umulsi3_highpart): Likewise.
6714 (*umulsi3_highpart_nov6): Likewise.
6715 * config/arm/arm.h (arm_arch3m): Delete.
6716 * config/arm/arm.c (arm_arch3m): Delete.
6717 (arm_option_override_internal): Update armv3-related comment.
6718 (arm_configure_build_target): Delete use of isa_bit_mode32.
6719 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6720 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6721 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6723 (mulusa3): Likewise.
6724 * config/arm/arm-protos.h (arm_arch3m): Delete.
6725 * config/arm/arm-tables.opt: Regenerate.
6726 * config/arm/arm-tune.md: Likewise.
6727 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6728 deleted architectures.
6730 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6732 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6733 (armv5t, armv5te): New features.
6734 (ARMv5, ARMv5e): Delete fgroups.
6735 (ARMv5t, ARMv5te): Adjust for above changes.
6737 (armv5, armv5e): Delete arches.
6738 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6740 (*call_reg_arm): Likewise.
6741 (*call_value_reg_armv5): Likewise.
6742 (*call_value_reg_arm): Likewise.
6743 (*call_symbol): Likewise.
6744 (*call_value_symbol): Likewise.
6745 (*sibcall_insn): Likewise.
6746 (*sibcall_value_insn): Likewise.
6748 (prefetch): Likewise.
6749 (define_split and define_peephole2 dependent on arm_arch5):
6751 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6753 (TARGET_ARM_QBIT): Likewise.
6754 (TARGET_DSP_MULTIPLY): Likewise.
6755 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6756 (arm_arch5, arm_arch5e): Delete.
6757 (arm_arch5t, arm_arch5te): Declare.
6758 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6759 (arm_arch5t): Declare.
6760 (arm_option_reconfigure_globals): Update for the above.
6761 (arm_options_perform_arch_sanity_checks): Update comment, replace
6762 use of arm_arch5 with arm_arch5t.
6763 (use_return_insn): Likewise.
6764 (arm_emit_call_insn): Likewise.
6765 (output_return_instruction): Likewise.
6766 (arm_final_prescan_insn): Likewise.
6767 (arm_coproc_builtin_available): Likewise.
6768 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6769 arm_arch5e with arm_arch5t and arm_arch5te.
6770 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6771 (arm_arch5t, arm_arch5te): Declare.
6772 * config/arm/arm-tables.opt: Regenerate.
6773 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6774 * config/arm/t-multilib: Likewise.
6775 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6776 instead of arm_arch5.
6777 (*call_reg_thumb1): Likewise.
6778 (*call_value_reg_thumb1_v5): Likewise.
6779 (*call_value_reg_thumb1): Likewise.
6780 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6782 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6784 2018-05-18 Martin Liska <mliska@suse.cz>
6786 PR gcov-profile/84846
6787 * doc/gcov.texi: Document -t option of gcov tool.
6789 2018-05-18 Martin Liska <mliska@suse.cz>
6791 PR gcov-profile/84846
6792 * gcov.c (print_usage): Add new -t option.
6793 (process_args): Handle the option.
6794 (generate_results): Use stdout as output when requested by
6797 2018-05-18 Martin Liska <mliska@suse.cz>
6799 PR gcov-profile/84846
6800 * coverage.c (coverage_init): Write PWD to .gcno file.
6801 * doc/gcov.texi: Document how working directory is printed.
6802 * gcov-dump.c (dump_gcov_file): Print PWD.
6803 * gcov.c (output_intermediate_file): Likewise.
6804 (read_graph_file): Read PWD string.
6805 (output_lines): Print PWD.
6807 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6810 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6811 for retval and return false if all args to phi are zero.
6813 2018-05-18 Richard Biener <rguenther@suse.de>
6815 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6817 (evrp_dom_walker::before_dom_children): Call it.
6819 2018-05-18 Richard Biener <rguenther@suse.de>
6821 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6822 results when processing array refs with variable index.
6824 2018-05-18 Toon Moene <toon@moene.org>
6826 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6827 directly after that of -floop-interchange. Indicate that both
6828 options are enabled by default when specifying -O3.
6830 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6832 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6833 iterator. Delete separate integer-mode vec_set<mode> expander.
6834 (aarch64_simd_vec_setv2di): Delete.
6835 (vec_setv2di): Delete.
6836 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6837 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6838 the "w, r" alternative.
6840 2018-05-18 Martin Liska <mliska@suse.cz>
6842 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6843 * tree-pass.h (make_pass_lower_switch_O0): New function.
6844 * tree-switch-conversion.c (node_has_low_bound): Remove.
6845 (node_has_high_bound): Likewise.
6846 (node_is_bounded): Likewise.
6847 (class pass_lower_switch): Make it a template type and create
6849 (pass_lower_switch::execute): Add template argument.
6850 (make_pass_lower_switch): New function.
6851 (make_pass_lower_switch_O0): New function.
6852 (do_jump_if_equal): Remove.
6853 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6854 all the hard work to tree optimization passes.
6856 2018-05-18 Martin Liska <mliska@suse.cz>
6858 * dbgcnt.c (limit_low): Renamed from limit.
6859 (limit_high): New variable.
6860 (dbg_cnt_is_enabled): Check for upper limit.
6861 (dbg_cnt): Adjust dumping.
6862 (dbg_cnt_set_limit_by_index): Add new argument for high
6864 (dbg_cnt_set_limit_by_name): Likewise.
6865 (dbg_cnt_process_single_pair): Parse new format.
6866 (dbg_cnt_process_opt): Use strtok.
6867 (dbg_cnt_list_all_counters): Remove 'value' and add
6869 * doc/invoke.texi: Document changes.
6871 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6873 * doc/sourcebuild.texi (scalar_all_fma): Document.
6874 * tree.def (FMA_EXPR): Delete.
6875 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6876 * internal-fn.c (ternary_direct): New macro.
6877 (expand_ternary_optab_fn): Likewise.
6878 (direct_ternary_optab_supported_p): Likewise.
6879 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6880 * builtins.c (fold_builtin_fma): Delete.
6881 (fold_builtin_3): Don't call it.
6882 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6883 * expr.c (expand_expr_real_2): Likewise.
6884 * fold-const.c (operand_equal_p): Likewise.
6885 (fold_ternary_loc): Likewise.
6886 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6887 * gimple.c (DEFTREECODE): Likewise.
6888 * gimplify.c (gimplify_expr): Likewise.
6889 * optabs-tree.c (optab_for_tree_code): Likewise.
6890 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6891 * tree-eh.c (operation_could_trap_p): Likewise.
6892 (stmt_could_throw_1_p): Likewise.
6893 * tree-inline.c (estimate_operator_cost): Likewise.
6894 * tree-pretty-print.c (dump_generic_node): Likewise.
6895 (op_code_prio): Likewise.
6896 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6897 * tree-ssa-operands.c (get_expr_operands): Likewise.
6898 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6899 * fold-const-call.h (fold_fma): Delete.
6900 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6901 CFN_FNMA and CFN_FNMS.
6903 * genmatch.c (combined_fn): New enum.
6904 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6905 (commutative_op): New function.
6906 (commutate): Use it. Handle more than 2 operands.
6907 (dt_operand::gen_gimple_expr): Use commutative_op.
6908 (parser::parse_expr): Allow :c to be used with non-binary
6909 operators if the commutative operand is known.
6910 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6911 CFN_FMS, CFN_FNMA and CFN_FNMS.
6912 (backprop::process_assign_use): Remove FMA_EXPR handling.
6913 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6914 (gen_hsa_fma): New function.
6915 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6916 IFN_FNMA and IFN_FNMS.
6917 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6918 * gimple-fold.h (follow_all_ssa_edges): Declare.
6919 * gimple-fold.c (follow_all_ssa_edges): New function.
6920 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6921 gimple_build interface and use follow_all_ssa_edges to fold the result.
6922 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6923 instead of checking for optabs directly.
6924 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6925 rather than FMA_EXPRs.
6926 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6927 call to IFN_FMA instead of an FMA_EXPR.
6929 2018-05-17 Jim Wilson <jimw@sifive.com>
6931 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6932 sign extended promoted subreg, and the range does not have the sign bit
6933 set, then do a sign extend.
6935 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6936 test, check for sign extended subreg and/or constant operands, and
6937 do a sign extend in that case.
6939 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6941 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6942 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6944 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6945 Change logics_shift_reg to logics_shift_imm.
6946 (thunderx2t99_fp_loadpair_basic): Delete.
6947 (thunderx2t99_fp_storepair_basic): Delete.
6948 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6949 (thunderx2t99_asimd_polynomial): Delete.
6950 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6951 and neon_fp_mul_d_scalar_q.
6952 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6953 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6954 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6955 (thunderx2t99_asimd_lut): Add missing tbl types.
6956 (thunderx2t99_asimd_ext): Delete.
6957 (thunderx2t99_asimd_load1_1_mult): Delete.
6958 (thunderx2t99_asimd_load1_2_mult): Delete.
6959 (thunderx2t99_asimd_load1_ldp): New.
6960 (thunderx2t99_asimd_load1): New.
6961 (thunderx2t99_asimd_load2): Add missing *load2* types.
6962 (thunderx2t99_asimd_load3): New.
6963 (thunderx2t99_asimd_load4): New.
6964 (thunderx2t99_asimd_store1_1_mult): Delete.
6965 (thunderx2t99_asimd_store1_2_mult): Delete.
6966 (thunderx2t99_asimd_store2_mult): Delete.
6967 (thunderx2t99_asimd_store2_onelane): Delete.
6968 (thunderx2t99_asimd_store_stp): New.
6969 (thunderx2t99_asimd_store1): New.
6970 (thunderx2t99_asimd_store2): New.
6971 (thunderx2t99_asimd_store3): New.
6972 (thunderx2t99_asimd_store4): New.
6974 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6976 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6977 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6979 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6980 Segher Boessenkool <segher@kernel.crashing.org>
6983 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6986 2018-05-17 Richard Biener <rguenther@suse.de>
6988 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6989 for pruning loop and prune defs feeding only already visited PHIs.
6991 2018-05-17 Richard Biener <rguenther@suse.de>
6993 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6995 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6996 Richard Biener <rguenther@suse.de>
6998 PR tree-optimization/85793
6999 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
7000 for VMAT_ELEMENTWISE.
7002 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7004 * internal-fn.h (lookup_internal_fn): Declare
7005 * internal-fn.c (lookup_internal_fn): New function.
7006 * gimple.c (gimple_build_call_from_tree): Handle calls to
7008 * gimple-pretty-print.c (dump_gimple_call): Print "." before
7009 internal function names.
7010 * tree-pretty-print.c (dump_generic_node): Likewise.
7011 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
7013 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
7015 * gimple-fold.h (gimple_build): Make the function forms take
7016 combined_fn rather than built_in_function.
7017 (gimple_simplify): Likewise.
7018 * gimple-match-head.c (gimple_simplify): Likewise.
7019 * gimple-fold.c (gimple_build): Likewise.
7020 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
7021 rather than gimple_build_call_internal.
7022 (get_initial_defs_for_reduction): Likewise.
7023 (vect_create_epilog_for_reduction): Likewise.
7024 (vectorizable_live_operation): Likewise.
7026 2018-05-17 Martin Liska <mliska@suse.cz>
7028 * gimple-ssa-sprintf.c (format_directive): Do not use
7029 space in between 'G_' and '('.
7031 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7034 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7035 even if the mask is not all ones.
7038 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7040 (ix86_gimple_fold_builtin): Likewise.
7043 * config/i386/i386.c: Include tree-vector-builder.h.
7044 (ix86_vector_shift_count): New function.
7045 (ix86_fold_builtin): Fold shift builtins by scalar count.
7046 (ix86_gimple_fold_builtin): Likewise.
7048 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7049 _mm512_setzero): New intrinsics.
7051 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7052 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7054 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7055 code generation for cases where splatting a value is not useful.
7056 * simplify-rtx.c (simplify_ternary_operation): Simplify
7057 vec_merge across a vec_duplicate and a paradoxical subreg forming
7058 a vector mode to a vec_concat.
7060 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7062 * config.gcc: Support "goldmont-plus".
7063 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7065 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7066 PROCESSOR_GOLDMONT_PLUS.
7067 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7068 (processor_target_table): Add "goldmont-plus".
7069 (PTA_GOLDMONT_PLUS): Define.
7070 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7071 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7072 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7073 (fold_builtin_cpu): Add "goldmont-plus".
7074 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7075 (ix86_option_override_internal): Add "goldmont-plus".
7076 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7077 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7078 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7079 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7081 2018-05-17 Richard Biener <rguenther@suse.de>
7083 PR tree-optimization/85757
7084 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7085 remove defs that only feed that PHI from further processing.
7087 2018-05-16 Jim Wilson <jimw@sifive.com>
7089 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7091 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7093 2018-05-16 Mark Wielaard <mark@klomp.org>
7095 * dwarf2out.c (count_index_strings): New function.
7096 (output_indirect_strings): Call count_index_strings and generate
7097 header for dwarf_version >= 5.
7099 2018-05-16 Mark Wielaard <mark@klomp.org>
7101 * dwarf2out.c (dwarf_FORM): New function.
7102 (set_indirect_string): Use dwarf_FORM.
7103 (reset_indirect_string): Likewise.
7104 (size_of_die): Likewise.
7105 (value_format): Likewise.
7106 (output_die): Likewise.
7107 (add_skeleton_AT_string): Likewise.
7108 (output_macinfo_op): Likewise.
7109 (index_string): Likewise.
7110 (output_index_string_offset): Likewise.
7111 (output_index_string): Likewise.
7112 (count_index_strings): Likewise.
7114 2018-05-16 Carl Love <cel@us.ibm.com>
7116 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7117 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7119 2018-05-16 Martin Jambor <mjambor@suse.cz>
7121 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7122 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7124 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7126 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7127 (fnma<mode>4): Likewise.
7128 (fms<mode>4): Likewise.
7129 (fnms<mode>4): Likewise.
7130 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7131 (aarch64_fnma<mode>4): Likewise.
7132 (aarch64_fms<mode>4): Likewise.
7133 (aarch64_fnms<mode>4): Likewise.
7134 (aarch64_fnmadd<mode>4): Likewise.
7136 2018-05-16 Jason Merrill <jason@redhat.com>
7138 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7140 2018-05-16 Richard Biener <rguenther@suse.de>
7142 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7143 (dump_stmt_cost): Declare.
7144 (add_stmt_cost): Dump cost we add.
7145 (add_stmt_costs): New function.
7146 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7148 (vect_analyze_stmt): Adjust prototype.
7149 (vectorizable_condition): Likewise.
7150 (vectorizable_live_operation): Likewise.
7151 (vectorizable_reduction): Likewise.
7152 (vectorizable_induction): Likewise.
7153 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7154 cost vector to pass to vectorizable_ and record afterwards.
7155 (vect_model_reduction_cost): Take cost vector argument and adjust.
7156 (vect_model_induction_cost): Likewise.
7157 (vectorizable_reduction): Likewise.
7158 (vectorizable_induction): Likewise.
7159 (vectorizable_live_operation): Likewise.
7160 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7161 SLP_TREE_NUMBER_OF_VEC_STMTS.
7162 (vect_analyze_slp_cost_1): Remove.
7163 (vect_analyze_slp_cost): Likewise.
7164 (vect_slp_analyze_node_operations): Take visited args and
7165 a target cost vector. Avoid processing already visited stmt sets.
7166 (vect_slp_analyze_operations): Use a local cost vector to gather
7167 costs and register those of non-discarded instances.
7168 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7169 (vect_schedule_slp_instance): Remove copying of
7170 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7172 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7173 adding cost. Record cost entry location.
7174 (vect_prologue_cost_for_slp_op): Function to compute cost of
7175 a constant or invariant generated for SLP vect in the prologue,
7176 split out from vect_analyze_slp_cost_1.
7177 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7178 (vect_model_promotion_demotion_cost): Likewise.
7179 (vect_model_store_cost): Likewise, make static.
7180 (vect_model_load_cost): Likewise.
7181 (vectorizable_bswap): Add cost vector arg and adjust.
7182 (vectorizable_call): Likewise.
7183 (vectorizable_simd_clone_call): Likewise.
7184 (vectorizable_conversion): Likewise.
7185 (vectorizable_assignment): Likewise.
7186 (vectorizable_shift): Likewise.
7187 (vectorizable_operation): Likewise.
7188 (vectorizable_store): Likewise.
7189 (vectorizable_load): Likewise.
7190 (vectorizable_condition): Likewise.
7191 (vectorizable_comparison): Likewise.
7192 (can_vectorize_live_stmts): Likewise.
7193 (vect_analyze_stmt): Likewise.
7194 (vect_transform_stmt): Adjust calls to vectorizable_*.
7195 * tree-vectorizer.c: Include gimple-pretty-print.h.
7196 (dump_stmt_cost): New function.
7198 2018-05-16 Richard Biener <rguenther@suse.de>
7200 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7201 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7202 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7203 (check_name): New callback.
7204 (dse_classify_store): Track cycles via a visited bitmap of PHI
7205 defs and simplify handling of in-loop and across loop dead stores
7206 and properly fail for loop-variant refs. Handle byte-tracking with
7207 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7210 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7212 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7213 (vect_get_mask_type_for_stmt): Likewise.
7214 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7216 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7217 to determine the statement's vector type and the vector type that
7218 should be used for calculating nunits. Deal with cases in which
7219 the type has to be deferred.
7220 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7221 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7222 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7223 (vect_determine_vf_for_stmt): New functions, split out from...
7224 (vect_determine_vectorization_factor): ...here.
7225 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7226 (vect_get_mask_type_for_stmt): New functions, split out from
7227 vect_determine_vectorization_factor.
7229 2018-05-16 Richard Biener <rguenther@suse.de>
7231 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7232 verify the [VEC_]COND_EXPR embedded comparison.
7234 2018-05-15 Martin Sebor <msebor@redhat.com>
7236 PR tree-optimization/85753
7237 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7238 RECORD_TYPE in addition to ARRAY_TYPE.
7240 2018-05-15 Martin Sebor <msebor@redhat.com>
7243 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7245 2018-05-15 Richard Biener <rguenther@suse.de>
7247 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7248 add by_clobber_p one. Change algorithm to collect all defs
7249 representing uses we need to walk and try reducing them to
7250 a single one before failing.
7251 (dse_dom_walker::dse_optimize_stmt): Adjust.
7253 2018-05-13 Mark Wielaard <mark@klomp.org>
7255 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7256 (size_of_loc_descr): Likewise.
7257 (output_loc_operands): Likewise.
7258 (output_loc_operands_raw): Likewise.
7259 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7260 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7261 (hash_loc_operands): Likewise.
7262 (compare_loc_operands): Likewise.
7264 2018-05-14 Mark Wielaard <mark@klomp.org>
7266 * dwarf2out.c (count_index_addrs): New function.
7267 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7269 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7271 PR tree-optimization/83648
7272 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7273 return value as malloc candidate.
7275 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7278 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7279 param as true in call to suggest_attribute.
7281 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7283 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7284 -mreadonly-in-sdata.
7286 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7288 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7290 (aarch64_crypto_aesd_fused): Likewise.
7292 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7294 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7295 (movsi_aarch64): Likewise.
7296 (load_pairsi): Likewise.
7297 (load_pairdi): Likewise.
7298 (store_pairsi): Likewise.
7299 (store_pairdi): Likewise.
7300 (load_pairsf): Likewise.
7301 (load_pairdf): Likewise.
7302 (store_pairsf): Likewise.
7303 (store_pairdf): Likewise.
7304 (zero_extend): Likewise.
7305 (trunc): Swap alternatives.
7306 (fcvt_target): Add '?' to prefer w over r.
7308 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7311 * config/i386/i386.md: Disallow non-commutative arithmetics in
7312 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7313 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7314 in the peephole2 before it.
7316 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7318 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7319 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7320 (ix86_handle_option): Handle -mcldemote.
7321 * config.gcc: New header.
7322 * config/i386/cldemoteintrin.h: New file.
7323 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7324 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7326 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7327 OPTION_MASK_ISA_CLDEMOTE.
7328 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7329 (ix86_valid_target_attribute_inner_p): Ditto.
7330 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7331 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7332 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7333 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7334 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7336 * config/i386/i386.opt: Add -mcldemote.
7337 * config/i386/x86intrin.h: New header.
7338 * doc/invoke.texi: Add -mcldemote.
7340 2018-05-14 Richard Biener <rguenther@suse.de>
7342 * doc/match-and-simplify.texi: Adjust :s documentation.
7344 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7346 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7347 intended memcpy size.
7348 (REORDER_45): Likewise.
7350 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7352 * sort.cc: New file.
7353 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7354 * vec.c (qsort_chk): Use gcc_qsort.
7355 * Makefile.in (OBJS-libcommon): Add sort.o.
7356 (build/sort.o): New target. Use it...
7357 (BUILD_RTL): ... here, and...
7358 (build/gencfn-macros): ... here, and...
7359 (build/genmatch): ... here.
7361 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7362 Chung-Ju Wu <jasonwucj@gmail.com>
7364 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7365 * config/nds32/nds32-graywolf.md: New file.
7366 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7367 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7369 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7370 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7371 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7372 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7373 * config/nds32/pipelines.md: Include n15 settings.
7375 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7376 Chung-Ju Wu <jasonwucj@gmail.com>
7378 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7379 * config/nds32/nds32-n13.md: New file.
7380 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7381 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7383 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7384 * config/nds32/nds32.md (pipeline_model): Add n13.
7385 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7386 * config/nds32/pipelines.md: Include n13 settings.
7388 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7389 Chung-Ju Wu <jasonwucj@gmail.com>
7391 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7392 * config/nds32/nds32-n10.md: New file.
7393 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7394 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7396 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7397 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7398 * config/nds32/nds32.md (pipeline_model): Add n10.
7399 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7400 * config/nds32/pipelines.md: Include n10 settings.
7402 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7403 Kito Cheng <kito.cheng@gmail.com>
7404 Chung-Ju Wu <jasonwucj@gmail.com>
7406 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7407 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7408 Add enum values for DSP extension instructions.
7409 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7411 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7412 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7414 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7415 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7416 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7417 * config/nds32/nds32-intrinsic.md: Likewise.
7418 * config/nds32/nds32_intrinsic.h: Likewise.
7419 * config/nds32/nds32-md-auxiliary.c: Likewise.
7420 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7421 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7422 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7423 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7424 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7425 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7426 TYPE_DMAC in switch statement.
7427 * config/nds32/nds32.c: New checking and implementation for DSP
7428 extension instructions.
7429 * config/nds32/nds32.h: Likewise.
7430 * config/nds32/nds32.md: Likewise.
7431 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7432 * config/nds32/predicates.md: Implement new predicates for DSP
7435 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7437 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7438 Reformat alternatives and attributes so it is easier to identify
7439 which constraints/attributes go with which instruction.
7440 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7441 (mov<mode>_softfloat32, FMOVE64): Likewise.
7442 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7443 (mov<mode>_softfloat64, FMOVE64): Likewise.
7445 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7447 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7449 (Basic PowerPC Built-in Functions): The new name of the
7450 subsection previously known as "PowerPC Built-in Functions".
7451 (Basic PowerPC Built-in Functions Available on all Configurations):
7453 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7454 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7455 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7456 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7458 2018-05-11 Martin Jambor <mjambor@suse.cz>
7461 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7464 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7467 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7469 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7471 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7472 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7473 (ix86_handle_option): Handle -mwaitpkg.
7474 * config.gcc: New header.
7475 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7476 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7477 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7479 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7480 OPTION_MASK_ISA_WAITPKG.
7481 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7482 (ix86_option_override_internal): Add PTA_WAITPKG.
7483 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7484 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7485 IX86_BUILTIN_TPAUSE.
7486 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7487 __builtin_ia32_umwait and __builtin_ia32_tpause.
7488 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7489 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7490 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7491 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7492 UNSPECV_TPAUSE): New.
7493 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7494 * config/i386/i386.opt: Add -mwaitpkg.
7495 * config/i386/waitpkgintrin.h: New file.
7496 * config/i386/x86intrin.h: New header.
7497 * doc/invoke.texi: Add -mwaitpkg.
7499 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7502 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7504 (cortex-m0): Use armv6s-m isa.
7505 (cortex-m0plus): Likewise.
7506 (cortex-m1): Likewise.
7507 (cortex-m0.small-multiply): Likewise.
7508 (cortex-m0plus.small-multiply): Likewise.
7509 (cortex-m1.small-multiply): Likewise.
7511 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7512 Jakub Jelinek <jakub@redhat.com>
7514 PR tree-optimization/85692
7515 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7516 source permute as well.
7518 2018-05-11 Martin Liska <mliska@suse.cz>
7521 * doc/extend.texi: Document LLVM style format for no_sanitize
7524 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7526 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7527 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7528 (mode_supports_vsx_dform_quad): Likewise.
7529 (mode_supports_vmx_dform): Move these functions to be next to the
7530 other mode_supports functions.
7531 (mode_supports_dq_form): Likewise.
7532 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7533 mode_supports_dq_form.
7534 (reg_offset_addressing_ok_p): Likewise.
7535 (offsettable_ok_by_alignment): Likewise.
7536 (rs6000_legitimate_offset_address_p): Likewise.
7537 (legitimate_lo_sum_address_p): Likewise.
7538 (rs6000_legitimize_address): Likewise.
7539 (rs6000_legitimize_reload_address): Likewise.
7540 (rs6000_secondary_reload_inner): Likewise.
7541 (rs6000_preferred_reload_class): Likewise.
7542 (rs6000_output_move_128bit): Likewise.
7544 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7546 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7547 Generate SImode target register for null target.
7548 <case IX86_BUILTIN_XGETBV>: Ditto.
7549 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7550 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7552 2018-05-10 Carl Love <cel@us.ibm.com>
7554 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7555 dcbtt and dcbtstt if operands[2] is 0.
7557 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7560 * config/i386/sse.md (usadv64qi): New expander.
7562 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7564 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7565 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7566 -maltivec=be support.
7567 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7568 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7569 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7570 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7571 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7572 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7573 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7574 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7575 altivec_vsumsws): Adjust.
7576 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7577 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7578 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7580 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7581 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7582 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7583 (altivec_lve<VI_char>x): Delete expand.
7584 (*altivec_lve<VI_char>x_internal): Rename to...
7585 (altivec_lve<VI_char>x): ... this.
7586 (altivec_lvxl_<mode>): Delete expand.
7587 (*altivec_lvxl_<mode>_internal): Rename to ...
7588 (altivec_lvxl_<mode>): ... this.
7589 (altivec_stvxl_<mode>): Delete expand.
7590 (*altivec_stvxl_<mode>_internal): Rename to ...
7591 (altivec_stvxl_<mode>): ... this.
7592 (altivec_stve<VI_char>x): Delete expand.
7593 (*altivec_stve<VI_char>x_internal): Rename to ...
7594 (altivec_stve<VI_char>x): ... this.
7595 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7596 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7597 reduc_plus_scal_<mode>): Adjust.
7598 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7600 (rs6000_cpu_cpp_builtins): Adjust.
7601 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7602 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7603 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7604 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7605 -maltivec=be support.
7606 (rs6000_split_vec_extract_var): Adjust.
7607 (rs6000_split_v4si_init): Adjust.
7608 (swap_selector_for_mode): Delete.
7609 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7610 altivec_expand_stvex_be): Delete.
7611 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7612 -maltivec=be support.
7613 (rs6000_gimple_fold_builtin): Ditto.
7614 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7616 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7617 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7618 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7619 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7620 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7621 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7622 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7623 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7624 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7625 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7626 anonymous split): Adjust.
7627 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7628 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7630 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7632 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7633 when --with-gxx-include-dir is also specified.
7634 * configure: Regenerate.
7636 2018-05-09 Jim Wilson <jimw@sifive.com>
7639 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7640 * config/riscv/t-withmultilib: New.
7641 * config/riscv/withmultilib.h: New.
7642 * doc/install.texi: Document RISC-V --with-multilib-list support.
7644 2018-05-09 Richard Biener <rguenther@suse.de>
7646 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7648 (vect_bb_vectorization_profitable_p): Adjust. Compute
7649 actual scalar cost using the cost vector and the add_stmt_cost
7652 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7654 PR rtl-optimization/85645
7655 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7656 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7658 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7660 PR rtl-optimization/85645
7661 * shrink-wrap.c (spread_components): Return a boolean saying if
7662 anything was changed.
7663 (try_shrink_wrapping_separate): Iterate spread_components until
7664 nothing changes anymore.
7666 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7668 PR rtl-optimization/85645
7669 * regrename.c (build_def_use): Also kill the chains that include the
7670 destination of a REG_CFA_REGISTER note.
7672 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7674 PR rtl-optimization/85645
7675 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7676 insn that has a REG_CFA_REGISTER note.
7678 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7680 * cfgexpand.c (expand_clobber): New function.
7681 (expand_gimple_stmt_1): Use it.
7682 * tree-vect-stmts.c (vect_clobber_variable): New function,
7684 (vectorizable_simd_clone_call): ...here.
7685 (vectorizable_store): Emit a clobber either side of an
7686 IFN_STORE_LANES sequence.
7687 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7689 2018-05-09 Tom de Vries <tom@codesourcery.com>
7692 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7693 (define_insn "trap_if_false"): Add exit after trap.
7695 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7697 PR rtl-optimization/85638
7698 * bb-reorder.c: Include common/common-target.h.
7699 (create_forwarder_block): New function extracted from...
7700 (fix_up_crossing_landing_pad): ...here. Rename into...
7701 (dw2_fix_up_crossing_landing_pad): ...this.
7702 (sjlj_fix_up_crossing_landing_pad): New function.
7703 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7704 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7705 from both partitions and exit the loop after one iteration.
7707 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7710 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7712 (Basic PowerPC Built-in Functions): The new name of the
7713 subsection previously known as "PowerPC Built-in Functions".
7714 (Basic PowerPC Built-in Functions Available on all Configurations):
7716 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7718 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7720 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7722 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7725 2018-05-08 Jim Wilson <jimw@sifive.com>
7727 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7728 (LD_EMUL_SUFFIX): New.
7729 (LINK_SPEC): Use it.
7731 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7733 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7735 (Basic PowerPC Built-in Functions): The new name of the
7736 subsection previously known as "PowerPC Built-in Functions".
7737 (Basic PowerPC Built-in Functions Available on all Configurations):
7739 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7741 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7743 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7745 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7748 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7751 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7752 after cmpelim optimization.
7754 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7756 * config.gcc: Support "goldmont".
7757 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7758 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7760 * config/i386/i386.c (m_GOLDMONT): Define.
7761 (processor_target_table): Add "goldmont".
7762 (PTA_GOLDMONT): Define.
7763 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7764 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7765 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7766 (fold_builtin_cpu): Add "goldmont".
7767 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7768 (ix86_option_override_internal): Add "goldmont".
7769 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7770 (processor_type): Add PROCESSOR_GOLDMONT.
7771 * config/i386/i386.md: Add CPU "glm".
7772 * config/i386/glm.md: New file.
7773 * config/i386/x86-tune.def: Add m_GOLDMONT.
7774 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7776 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7779 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7781 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7782 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7783 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7786 * config/i386/i386.c (ix86_fold_builtin): Handle
7787 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7790 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7791 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7793 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7796 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7797 (check_arch): Likewise.
7798 (check_fpu): Return the result rather than printing it.
7799 (end arch): Fix operator precedence.
7800 (end cpu): Likewise.
7801 (END): Print the result from check_fpu.
7803 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7804 Alan Hayward <alan.hayward@arm.com>
7805 David Sherwood <david.sherwood@arm.com>
7807 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7808 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7809 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7810 (*fcmuo<mode>_and): New patterns.
7812 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7814 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7815 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7816 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7817 (cmp_op, sve_imm_con): New code attributes.
7818 (SVE_COND_INT_CMP, imm_con): Delete.
7819 (cmp_op): Remove above unspecs from int attribute.
7820 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7822 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7823 comparison-specific unspecs.
7824 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7825 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7826 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7827 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7828 (*vec_fcm<cmp_op><mode>): Rename to...
7829 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7830 (*vec_fcmuo<mode>): Rename to...
7831 (*fcmuo<mode>): ...this and adjust likewise.
7832 (*pred_fcm<cmp_op><mode>): New pattern.
7833 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7834 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7836 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7838 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7839 (aarch64_emit_sve_predicated_cond): New function.
7840 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7841 (aarch64_emit_unspec_cond_or): Replace with...
7842 (aarch64_emit_sve_or_conds): ...this new function. Use
7843 aarch64_emit_sve_ptrue_op for the individual comparisons and
7844 aarch64_emit_binop to OR them together.
7845 (aarch64_emit_inverted_unspec_cond): Replace with...
7846 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7847 aarch64_emit_sve_ptrue_op for the comparison and
7848 aarch64_emit_unop to invert the result.
7849 (aarch64_expand_sve_vec_cmp_float): Update after the above
7850 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7852 2018-05-07 Nathan Sidwell <nathan@acm.org>
7854 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7855 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7856 (Backwards Compatibility): Likewise.
7858 2018-05-07 Luis Machado <luis.machado@linaro.org>
7862 2018-05-07 Luis Machado <luis.machado@linaro.org>
7864 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7865 <prefetch_dynamic_strides>: New const bool field.
7866 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7867 prefetch_dynamic_strides.
7868 (exynosm1_prefetch_tune): Likewise.
7869 (thunderxt88_prefetch_tune): Likewise.
7870 (thunderx_prefetch_tune): Likewise.
7871 (thunderx2t99_prefetch_tune): Likewise.
7872 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7874 (aarch64_override_options_internal): Update to set
7875 PARAM_PREFETCH_DYNAMIC_STRIDES.
7876 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7877 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7878 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7879 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7880 prefetch-dynamic-strides setting.
7882 2018-05-07 Luis Machado <luis.machado@linaro.org>
7884 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7885 <minimum_stride>: New const int field.
7886 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7887 minimum_stride field.
7888 (exynosm1_prefetch_tune): Likewise.
7889 (thunderxt88_prefetch_tune): Likewise.
7890 (thunderx_prefetch_tune): Likewise.
7891 (thunderx2t99_prefetch_tune): Likewise.
7892 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7893 (aarch64_override_options_internal): Update to set
7894 PARAM_PREFETCH_MINIMUM_STRIDE.
7895 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7896 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7897 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7898 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7899 stride is constant and is below the minimum stride threshold.
7901 2018-05-07 Luis Machado <luis.machado@linaro.org>
7903 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7906 2018-05-07 Luis Machado <luis.machado@linaro.org>
7908 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7909 <prefetch_dynamic_strides>: New const bool field.
7910 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7911 prefetch_dynamic_strides.
7912 (exynosm1_prefetch_tune): Likewise.
7913 (thunderxt88_prefetch_tune): Likewise.
7914 (thunderx_prefetch_tune): Likewise.
7915 (thunderx2t99_prefetch_tune): Likewise.
7916 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7918 (aarch64_override_options_internal): Update to set
7919 PARAM_PREFETCH_DYNAMIC_STRIDES.
7920 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7921 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7922 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7923 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7924 prefetch-dynamic-strides setting.
7926 2018-05-07 Luis Machado <luis.machado@linaro.org>
7928 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7929 <minimum_stride>: New const int field.
7930 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7931 minimum_stride field.
7932 (exynosm1_prefetch_tune): Likewise.
7933 (thunderxt88_prefetch_tune): Likewise.
7934 (thunderx_prefetch_tune): Likewise.
7935 (thunderx2t99_prefetch_tune): Likewise.
7936 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7937 (aarch64_override_options_internal): Update to set
7938 PARAM_PREFETCH_MINIMUM_STRIDE.
7939 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7940 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7941 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7942 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7943 stride is constant and is below the minimum stride threshold.
7945 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7948 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7949 the type is addressable. Don't force op into register if it has
7952 2018-05-05 Roland McGrath <mcgrathr@google.com>
7955 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7956 any section for which we don't know a specific type it should have,
7957 regardless of name. Previously this was done only for the exact
7958 names ".init_array", ".fini_array", and ".preinit_array".
7959 (default_elf_asm_named_section): Add comment about
7960 relationship with default_section_type_flags and SECTION_NOTYPE.
7961 (get_section): Don't consider it a type conflict if one side has
7962 SECTION_NOTYPE and the other doesn't, as long as neither has the
7963 SECTION_BSS et al used in the default_section_type_flags logic.
7965 2018-05-05 Tom de Vries <tom@codesourcery.com>
7968 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7969 (workaround_barsyncs): New function.
7970 (nvptx_reorg): Use workaround_barsyncs.
7971 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7972 (define_expand "nvptx_membar_cta"): New define_expand.
7973 (define_insn "*nvptx_membar_cta"): New insn.
7975 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7977 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7978 To improve optimization opportunities.
7979 * builtin-types.def: The new needed builtin types for the above.
7981 2018-05-04 Richard Biener <rguenther@suse.de>
7983 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7984 * gimple-ssa-store-merging.c
7985 (imm_store_chain_info::output_merged_store): Remove redundant create,
7986 release split_store vector contents on failure.
7987 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7988 scalar stmt vector on cache hit.
7990 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7992 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7994 * config.gcc (powerpc-xilinx-eabi*): Remove.
7995 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7997 (fusion_addis_mem_combo_load): Ditto.
7998 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
8000 (rs6000_cpu_cpp_builtins): Ditto.
8001 * config/rs6000/rs6000-linux.c
8002 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
8003 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
8004 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
8006 (rs6000_setup_reg_addr_masks): Ditto.
8007 (rs6000_init_hard_regno_mode_ok): Ditto.
8008 (rs6000_option_override_internal): Ditto.
8009 (legitimate_lo_sum_address_p): Ditto.
8010 (rs6000_legitimize_address): Ditto.
8011 (rs6000_legitimize_reload_address): Ditto.
8012 (rs6000_legitimate_address_p): Ditto.
8013 (abi_v4_pass_in_fpr): Ditto.
8014 (setup_incoming_varargs): Ditto.
8015 (rs6000_gimplify_va_arg): Ditto.
8016 (rs6000_split_multireg_move): Ditto.
8017 (rs6000_savres_strategy): Ditto.
8018 (rs6000_emit_prologue_components): Ditto.
8019 (rs6000_emit_epilogue_components): Ditto.
8020 (rs6000_emit_prologue): Ditto.
8021 (rs6000_emit_epilogue): Ditto.
8022 (rs6000_elf_file_end): Ditto.
8023 (rs6000_function_value): Ditto.
8024 (rs6000_libcall_value): Ditto.
8025 * config/rs6000/rs6000.h: Ditto.
8026 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
8027 (TARGET_MINMAX): ... this. New.
8028 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
8029 * config/rs6000/rs6000.md: Remove Xilinx FP support.
8030 (*movsi_internal1_single): Delete.
8031 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
8032 mfpu=, mxilinx-fpu): Delete.
8033 * config/rs6000/singlefp.h: Delete.
8034 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8035 * config/rs6000/t-rs6000: Ditto.
8036 * config/rs6000/t-xilinx: Delete.
8037 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
8038 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
8039 (VStype_simple): Delete.
8040 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8041 * config/rs6000/xfpu.h: Delete.
8042 * config/rs6000/xfpu.md: Delete.
8043 * config/rs6000/xilinx.h: Delete.
8044 * config/rs6000/xilinx.opt: Delete.
8045 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8046 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8048 2018-05-04 Tom de Vries <tom@codesourcery.com>
8051 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8054 2018-05-04 Richard Biener <rguenther@suse.de>
8057 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8058 (expand_complex_div_wide): Likewise.
8059 (expand_complex_operations_1): Likewise.
8060 (expand_complex_libcall): Preserve EH info of the original stmt.
8061 (tree_lower_complex): Handle removed blocks.
8062 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8063 on complex multiplication and division libcall builtins.
8065 2018-05-04 Richard Biener <rguenther@suse.de>
8068 * fold-const.c (negate_expr_p): Restrict negation of operand
8069 zero of a division to when we know that can happen without
8071 (fold_negate_expr_1): Likewise.
8073 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8076 * real.h (real_nextafter): Declare.
8077 * real.c (real_nextafter): New function.
8078 * fold-const-call.c (fold_const_nextafter): New function.
8079 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8080 CASE_CFN_NEXTTOWARD.
8081 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8082 even when arg1_mode is different from arg0_mode.
8084 2018-05-03 Nathan Sidwell <nathan@acm.org>
8086 * doc/extend.texi (Deprecated Features): Remove
8088 (Backwards Compatibility): Likewise.
8089 * doc/invoke.texi (C++ Language Options): Likewise.
8090 (C++ Dialect Options): Likewise.
8092 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8095 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8096 _mm512_mask_mullox_epi64): New intrinsics.
8098 2018-05-03 Tom de Vries <tom@codesourcery.com>
8101 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8102 dump files): Add offload-tree.
8104 2018-05-03 Richard Biener <rguenther@suse.de>
8106 PR tree-optimization/85615
8107 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8108 to loops not nested in BBs loop father to avoid creating multi-entry
8111 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8113 PR tree-optimization/70291
8114 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8115 arguments. Change return type to tree. Emit libcall as a new
8116 statement rather than replacing existing one when inplace_p is true.
8117 (expand_complex_multiplication_components): New function.
8118 (expand_complex_multiplication): Expand floating-point complex
8119 multiplication using the above.
8120 (expand_complex_division): Rename inner_type parameter to type.
8121 Update expand_complex_libcall call-site.
8122 (expand_complex_operations_1): Update expand_complex_multiplication
8123 and expand_complex_division call-sites.
8125 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8128 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8129 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8130 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8131 the highest significant bit of the shift count mask is clear. In
8132 check whether and[sq]i3 is needed verify that all significant bits
8133 of the shift count other than the highest are set.
8135 2018-05-02 Tom de Vries <tom@codesourcery.com>
8138 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8139 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8140 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8141 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8142 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8143 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8144 __builtin_goacc_parlevel_size.
8146 2018-05-02 Richard Biener <rguenther@suse.de>
8148 PR tree-optimization/85597
8149 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8150 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8153 2018-05-02 Tom de Vries <tom@codesourcery.com>
8156 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8157 dump files): Add ltrans-tree.
8159 2018-05-02 Tom de Vries <tom@codesourcery.com>
8162 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8163 dump files): Add wpa-ipa.
8165 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8167 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8168 powerpc*-*-linux*paired* target.
8169 * config/rs6000/750cl.h: Delete.
8170 * config/rs6000/paired.h: Delete.
8171 * config/rs6000/paired.md: Delete.
8172 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8174 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8175 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8176 comment. Remove paired float support.
8177 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8178 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8180 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8181 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8183 * config/rs6000/rs6000.c: Remove paired float support.
8184 (paired_expand_vector_init, paired_expand_vector_move,
8185 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8186 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8187 paired_expand_builtin, paired_expand_predicate_builtin,
8188 paired_init_builtins): Delete.
8189 * config/rs6000/rs6000.h: Remove paired float support.
8190 * config/rs6000/rs6000.md: Remove paired float support.
8191 (move_from_CR_ov_bit): Delete.
8192 * config/rs6000/rs6000.opt (mpaired): Delete.
8193 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8194 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8196 2018-05-02 Richard Biener <rguenther@suse.de>
8199 * gimplify.c (gimplify_save_expr): When in SSA form allow
8200 SAVE_EXPRs to compute to SSA vars.
8202 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8205 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8206 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8207 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8208 clobber operands[2], instead use a new pseudo. Formatting fixes.
8210 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8212 PR tree-optimization/85586
8213 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8214 exit early for statements in the same group if the accesses are
8217 2018-05-02 Tom de Vries <tom@codesourcery.com>
8220 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8223 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8225 PR tree-optimization/85143
8226 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8228 2018-05-01 Tom de Vries <tom@codesourcery.com>
8231 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8232 not found" error message.
8234 2018-05-01 Tom de Vries <tom@codesourcery.com>
8237 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8238 * vec.c (test_ordered_remove_if): New function.
8239 (vec_c_tests): Call test_ordered_remove_if.
8240 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8241 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8242 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8243 VEC_ORDERED_REMOVE_IF.
8245 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8247 PR tree-optimization/82665
8248 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8249 pointer subtraction where arguments come from a memchr call.
8251 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8253 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8254 --push-state --as-needed and --pop-state instead of --as-needed and
8255 --no-as-needed if ld supports it.
8256 * configure: Regenerated.
8259 * doc/install.texi2html: Replace _002d with - and _002a with * in
8260 generated html files using sed.
8262 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8265 * gcc-rich-location.c (blank_line_before_p): New function.
8266 (use_new_line): New function.
8267 (gcc_rich_location::add_fixit_insert_formatted): New function.
8268 * gcc-rich-location.h
8269 (gcc_rich_location::add_fixit_insert_formatted): New function.
8271 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8273 * selftest.c (assert_streq): Rename "expected" and "actual" to
8274 "val1" and "val2". Extend NULL-handling to cover both inputs
8275 symmetrically, while still requiring both to be non-NULL for a pass.
8276 * selftest.h (assert_streq): Rename "expected" and "actual" to
8278 (ASSERT_EQ): Likewise.
8279 (ASSERT_EQ_AT): Likewise.
8280 (ASSERT_KNOWN_EQ): Likewise.
8281 (ASSERT_KNOWN_EQ_AT): Likewise.
8282 (ASSERT_NE): Likewise.
8283 (ASSERT_MAYBE_NE): Likewise.
8284 (ASSERT_MAYBE_NE_AT): Likewise.
8285 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8286 the assertion to pass.
8287 (ASSERT_STREQ_AT): Likewise.
8289 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8291 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8292 interaction with -pie.
8294 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8296 * selftest.h: Fix alphabetization of per-source-file selftest
8299 2018-04-30 Jason Merrill <jason@redhat.com>
8301 PR c++/61982 - dead stores to destroyed objects.
8302 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8305 2018-04-30 Jason Merrill <jason@redhat.com>
8307 * tree.c (build_clobber): New.
8308 * tree.h: Declare it.
8309 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8311 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8313 * diagnostic-show-locus.c (layout::layout): Update for
8314 location_get_source_line returning a char_span.
8315 (struct char_span): Move to input.h.
8316 (struct correction): Update for fields in char_span becoming
8318 (struct source_line): Update for location_get_source_line
8319 returning a char_span.
8320 (layout::print_line): Likewise.
8321 * edit-context.c (edited_file::print_content): Likewise.
8322 (edited_file::print_diff_hunk): Likewise.
8323 (edited_file::print_run_of_changed_lines): Likewise.
8324 (edited_file::get_num_lines): Likewise.
8325 (edited_line::edited_line): Likewise.
8326 * final.c (asm_show_source): Likewise.
8327 * input.c (location_get_source_line): Convert return type
8328 from const char * to char_span, losing the final "line_len"
8330 (dump_location_info): Update for the above.
8331 (get_substring_ranges_for_loc): Likewise. Use a char_span
8332 when handling the literal within the line.
8333 (test_reading_source_line): Update for location_get_source_line
8334 returning a char_span.
8335 * input.h (class char_span): Move here from
8336 diagnostic-show-locus.c, converting from a struct to a class.
8337 Make data members private.
8338 (char_span::operator bool): New.
8339 (char_span::length): New.
8340 (char_span::get_buffer): New.
8341 (char_span::operator[]): New.
8342 (char_span::subspan): Make const.
8343 (char_span::xstrdup): New.
8344 (location_get_source_line): Convert return type from const char *
8345 to char_span, losing the final "line_size" param.
8347 2018-04-30 Jan Hubicka <jh@suse.cz>
8349 * lto-wrapper.c (ltrans_priorities): New static var.
8350 (cmp_priority): New.
8351 (run_gcc): Read priorities and if doing parallel build order
8352 the Makefile by them.
8354 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8356 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8358 2018-04-30 Richard Biener <rguenther@suse.de>
8360 * tree-cfg.c (verify_address): Remove base argument, add
8361 flag whether to check TREE_ADDRESSABLE and do that.
8362 (verify_expr): Remove.
8363 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8364 (verify_gimple_assign_single): Likewise.
8365 (verify_gimple_switch): Likewise.
8366 (verify_expr_location_1): Dereference tp once. Add (disabled)
8367 piece from verify_expr.
8368 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8370 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8372 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8374 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8376 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8377 (small_data_pattern): Likewise.
8378 (arc_rewrite_small_data): Likewise.
8379 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8380 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8381 (get_symbol_alignment): New function.
8382 (legitimate_small_data_address_p): Likewise.
8383 (legitimate_scaled_address): Update, call
8384 legitimate_small_data_address_p.
8385 (output_sdata): New static variable.
8386 (arc_print_operand): Update how we handle small data operands.
8387 (arc_print_operand_address): Likewise.
8388 (arc_legitimate_address_p): Update, use
8389 legitimate_small_data_address_p.
8390 (arc_rewrite_small_data_p): Remove.
8391 (arc_rewrite_small_data_1): Likewise.
8392 (arc_rewrite_small_data): Likewise.
8393 (small_data_pattern): Likewise.
8394 (compact_sda_memory_operand): Update to use
8395 legitimate_small_data_address_p and get_symbol_alignment.
8396 (prepare_move_operands): Don't rewite sdata pattern.
8397 (prepare_extend_operands): Remove.
8398 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8400 (zero_extendqisi2): Likewise.
8401 (zero_extendhisi2): Likewise.
8402 (extendqihi2): Likewise.
8403 (extendqisi2): Likewise.
8404 (extendhisi2): Likewise.
8408 * config/arc/constraints.md (Usd): Change it to memory constraint.
8410 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8412 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8413 as source of std instructions.
8414 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8415 allow 6-bit constants as source for store instructions.
8416 (movdi_insn): Update instruction pattern to allow 6-bit constants
8417 as source for store instructions.
8419 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8421 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8423 2018-04-30 Nathan Sidwell <nathan@acm.org>
8424 Sandra Loosemore <sandra@codesourcery.com>
8426 * dumpfile.c (dump_open): Allow '-' for stdout.
8427 * doc/invoke.texi (Developer Options): Document dump filename
8428 determination early. Document stdin/stdout selection.
8430 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8432 Microblaze Target: PIC data text relative
8434 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8435 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8437 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8438 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8439 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8440 New addressing mode for data-text relative position indepenedent code.
8441 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8442 'ADDRESS_SYMBOLIC_TXT_REL'.
8443 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8444 (microblaze_legitimate_pic_operand): Exclude function calls from
8445 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8446 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8448 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8449 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8450 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8451 for 'address + offset'.
8452 (microblaze_expand_prologue): Add new function prologue call for
8454 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8455 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8456 table in case of TARGET_PIC_DATA_TEXT_REL.
8457 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8458 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8459 Add new macros 'UNSPEC_TEXT',
8460 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8461 + exclude function calls from 'UNSPEC_PLT' in case of data text
8463 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8464 new target hook for generating address diff vector tables in case of
8466 * doc/tm.texi : Regenerate.
8467 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8468 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8469 of addr diff vector generation.
8470 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8471 target hook definition.
8472 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8473 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8474 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8475 option for data text relative.
8477 2018-04-30 Richard Biener <rguenther@suse.de>
8479 * tree-chrec.h (evolution_function_is_constant_p): Remove
8481 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8483 2018-04-30 Richard Biener <rguenther@suse.de>
8486 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8488 2018-04-30 Richard Biener <rguenther@suse.de>
8490 PR tree-optimization/28364
8491 PR tree-optimization/85275
8492 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8493 copying first exit test.
8495 2018-04-28 Mark Wielaard <mark@klomp.org>
8497 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8499 (dwarf_AT): Handle DW_AT_addr_base.
8500 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8502 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8505 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8506 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8507 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8508 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8510 2018-04-28 Richard Biener <rguenther@suse.de>
8512 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8513 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8514 to reflect use. Only add interesting stmts.
8516 2018-04-27 Martin Jambor <mjambor@suse.cz>
8519 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8520 the jump function allows for passing through aggregate values.
8522 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8524 * input.h (in_system_header_at): Convert from macro to inline
8526 (from_macro_expansion_at): Likewise.
8527 (from_macro_definition_at): Likewise.
8529 2018-04-27 Jeff Law <law@redhat.com>
8531 * config.gcc: Mark tile* targets as deprecated/obsolete.
8533 2018-04-27 Richard Biener <rguenther@suse.de>
8535 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8538 2018-04-27 Richard Biener <rguenther@suse.de>
8540 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8542 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8544 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8545 with Yd constraint. Set "preferred_for_speed" attribute from
8546 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8548 (*movdi_internal): Ditto.
8549 (movti_interunit splitters): Remove
8550 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8551 (movdi_interunit splitters): Ditto.
8552 * config/i386/constraints.md (Ye): Remove.
8553 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8555 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8558 * config/aarch64/constraints.md (Usg): Limit to 31.
8561 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8563 PR tree-optimization/85529
8564 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8565 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8566 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8567 zero extension or masking of the MSB bit.
8568 (optimize_range_tests): Add FIRST_BB argument, pass it through
8569 to optimize_range_tests_var_bound.
8570 (maybe_optimize_range_tests, reassociate_bb): Adjust
8571 optimize_range_tests callers.
8573 2018-04-26 Richard Biener <rguenther@suse.de>
8574 Jakub Jelinek <jakub@redhat.com>
8576 * cgraph.h (symbol_table): Just declare debug method here.
8577 * symtab.c (symbol_table::debug): Define.
8579 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8581 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8583 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8585 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8586 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8587 (*movdi_internal): Substitute Yi and Yj constraint with x
8588 and Ym and Yn constraint with y constraint. Update "isa"
8589 attribute and set "preferred_for_speed" attribute from
8590 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8591 (*movsi_internal): Ditto.
8592 (*movdf_internal): Ditto.
8593 (*movsf_internal): Ditto.
8594 (*zero_extendsidi2): Ditto.
8595 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8596 (sse2_loadld): Ditto.
8597 (*vec_extract<ssevecmodelower>_0): Ditto.
8598 (*vec_extractv4si_0_zext_sse4): Ditto.
8599 (vec_concatv2di): Ditto.
8600 (*vec_dup<mode>): Ditto.
8601 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8602 * config/i386/constraints.md (Yi): Remove.
8607 2018-04-26 Nathan Sidwell <nathan@acm.org>
8609 * dumpfile.c (dump_open): New.
8610 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8611 (dump_finish): Detect stdio/stderr by value not name.
8613 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8615 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8617 2018-04-26 Tom de Vries <tom@codesourcery.com>
8620 * config/nvptx/nvptx.c (verify_neutering_jumps)
8621 (verify_neutering_labels): New function
8622 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8624 2018-04-26 Tom de Vries <tom@codesourcery.com>
8627 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8628 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8631 2018-04-26 Richard Biener <rguenther@suse.de>
8632 Tom de Vries <tom@codesourcery.com>
8635 * lto-streamer-out.c (output_function): Fixup loops if required to match
8636 discovery done in the reader.
8638 2018-04-26 Richard Biener <rguenther@suse.de>
8640 PR tree-optimization/85116
8641 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8642 have a loop exit from the single latch predecessor. Remove
8643 case of header with just condition.
8644 (ch_base::copy_headers): Exclude infinite loops from any
8646 (pass_ch::execute): Record exits.
8648 2018-04-26 Richard Biener <rguenther@suse.de>
8650 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8651 prologue cost vector and pass it to vect_get_load_cost.
8652 (vect_get_peeling_costs_all_drs): Likewise.
8653 (vect_peeling_hash_get_lowest_cost): Likewise.
8654 (vect_enhance_data_refs_alignment): Likewise.
8656 2018-04-26 Richard Biener <rguenther@suse.de>
8659 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8660 checking of integer<->pointer conversions.
8661 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8662 sign-/zero-extending pointer types.
8663 (expand_omp_for_static_chunk): Likewise.
8665 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8666 Jean Lee <xiaoyur347@gmail.com>
8668 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8669 (TARGET_ASAN_SHADOW_OFFSET): Define.
8670 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8671 true for -fsanitize=address.
8673 2018-04-25 Mark Wielaard <mark@klomp.org>
8675 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8678 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8680 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8681 than "alu", remove explicit "memory" and "imm_disp" attributes.
8682 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8685 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8686 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8687 gen_lowpart_no_emit.
8689 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8692 * config/i386/i386.c (ix86_expand_builtin): Change memory
8693 operand to XI, extend p0 to Pmode.
8694 * config/i386/i386.md: Change unspec volatile and operand
8695 1 mode to XI, change operand 0 mode to P.
8697 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8699 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8700 GET_MODE_MASK before any checking.
8701 (nds32_can_use_bset_p): Likewise.
8702 (nds32_can_use_btgl_p): Likewise.
8704 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8706 * config/nds32/nds32-doubleword.md: New define_split pattern for
8707 illegal register number.
8709 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8711 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8713 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8715 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8717 2018-04-25 Richard Biener <rguenther@suse.de>
8719 * lto-streamer.h (LTO_major_version): Bump to 8.
8721 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8723 * BASE-VER: Set to 9.0.0.
8725 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8727 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8728 in __abskf2 and __powikf2.
8730 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8733 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8734 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8735 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8736 Use the above on operand 2. Reindent.
8737 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8739 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8742 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8744 * config/i386/i386.opt (mcet): Removed.
8745 * doc/install.texi: Remove -mcet documentation.
8746 * doc/invoke.texi: Likewise.
8748 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8751 * doc/install.texi: Remove -mcet from bootstrap-cet.
8753 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8756 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8757 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8761 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8762 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8763 containing a CONST_VECTOR.
8765 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8767 * doc/install.texi: Update newlib dependency for nvptx.
8769 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8772 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8773 instead of INTVAL when shifting x left.
8775 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8777 PR tree-optimization/85478
8778 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8779 vect_grouped_store_supported for single element vectors.
8781 2018-04-24 Richard Biener <rguenther@suse.de>
8784 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8785 load cost increase to the case of non-constant step.
8787 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8790 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8791 destination if any_malformed_asm.
8793 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8796 * expr.c (store_field): In the bitfield case, if the value comes from
8797 a function call and is returned in registers by means of a PARALLEL,
8798 do not change the mode of the temporary unless BLKmode and VOIDmode.
8800 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8802 PR rtl-optimization/85423
8803 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8804 dependencies to debug insns when the previous insn is non-debug.
8806 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8808 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8809 enums into a single definition.
8810 (fls): Fix predicates and printing.
8813 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8815 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8816 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8817 and short u6 immediate.
8818 (check_if_valid_sleep_operand): Remove.
8819 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8821 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8823 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8824 flag_always_save_lp condition.
8825 * config/nds32/nds32.opt (malways-save-lp): New option.
8827 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8829 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8830 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8831 * config/nds32/nds32.h
8832 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8833 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8835 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8837 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8838 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8840 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8841 Chung-Ju Wu <jasonwucj@gmail.com>
8843 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8845 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8846 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8848 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8850 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8852 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8854 * config/nds32/nds32-protos.h (nds32_data_alignment,
8855 nds32_local_alignment): Declare.
8856 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8857 nds32_local_alignment): New functions.
8858 (TARGET_CONSTANT_ALIGNMENT): Define.
8859 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8861 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8863 * config/nds32/nds32.c
8864 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8865 (TARGET_MODES_TIEABLE_P): Likewise.
8867 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8869 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8872 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8873 Chung-Ju Wu <jasonwucj@gmail.com>
8875 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8876 for unaligned access.
8877 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8879 * config/nds32/nds32-intrinsic.md: Likewise.
8880 * config/nds32/nds32_intrinsic.h: Likewise.
8881 * config/nds32/nds32.h (nds32_builtins): Likewise.
8882 * config/nds32/nds32.opt (munaligned-access): New option.
8883 * config/nds32/nds32.c (nds32_asm_file_start): Display
8884 flag_unaligned_access status.
8886 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8888 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8889 -mno-relax is present.
8890 * config/riscv/linux.h (LINK_SPEC): Ditto.
8892 2018-04-20 Martin Sebor <msebor@redhat.com>
8895 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8897 (gimple_fold_builtin_stxcpy_chk): Same.
8898 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8900 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8903 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8904 __powikf2 when long double is IEEE 128-bit.
8906 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8908 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8909 step to make sure stack always aligned.
8911 2018-04-20 Carl Love <cel@us.ibm.com>
8914 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8915 size check for arg0.
8917 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8918 Tom de Vries <tom@codesourcery.com>
8921 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8922 Emit insns for calls too.
8923 (nvptx_find_par): Always look for worker-level predecessor insn.
8924 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8926 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8927 (nvptx_process_pars): Propagate frames for calls.
8929 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8932 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8934 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8935 (ix86_handle_option): Don't handle OPT_mibt.
8936 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8938 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8940 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8941 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8942 (ix86_target_macros): Define __CET__ with flag_cf_protection
8943 for -fcf-protection.
8944 * config/i386/i386.c (isa2_opts): Remove -mibt.
8945 * config/i386/i386.h (TARGET_IBT): Removed.
8946 (TARGET_IBT_P): Likewise.
8947 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8948 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8949 * config/i386/i386.opt (mcet): Update help message.
8952 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8953 -mcet as an alias for -mshstk.
8955 2018-04-20 Richard Biener <rguenther@suse.de>
8958 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8959 complexity by forcing a single use of the multiply operand.
8961 2018-04-20 Martin Jambor <mjambor@suse.cz>
8964 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8965 recursion dependency to only apply to non-clones.
8967 2018-04-20 Martin Jambor <mjambor@suse.cz>
8970 * ipa-cp.c (create_specialized_node): Check that clones of
8971 self-recursive edges exist during IPA-CP.
8973 2018-04-19 Toon Moene <toon@moene.org>
8975 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8978 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8980 PR tree-optimization/85467
8981 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8982 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8983 VECTOR_CST element to type.
8985 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8988 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8989 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8990 (builtin_longjmp): Likewise.
8991 (save_stack_nonlocal): New pattern.
8992 (restore_stack_nonlocal): Likewise.
8994 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8997 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8998 Replace ASM_OUTPUT_LABEL with fprintf.
9000 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
9003 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
9004 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
9005 * config/i386/i386-c.c (ix86_target_macros_internal): Also
9006 define __IBT__ and __SHSTK__ for -fcf-protection.
9007 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
9009 (ix86_trampoline_init): Likewise.
9010 (x86_output_mi_thunk): Likewise.
9011 (ix86_notrack_prefixed_insn_p): Likewise.
9012 (ix86_option_override_internal): Don't disallow -fcf-protection.
9013 * config/i386/i386.md (rdssp<mode>): Also enable for
9015 (incssp<mode>): Likewise.
9016 (nop_endbr): Likewise.
9017 * config/i386/i386.opt (mcet): Change help message to built-in
9021 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
9022 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
9023 enable CET built-in functions.
9025 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
9027 * common/config/i386/i386-common.c
9028 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
9029 OPTION_MASK_ISA_MOVDIRI_UNSET,
9030 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
9031 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
9032 * config.gcc (movdirintrin.h): New header.
9033 * config/i386/cpuid.h (bit_MOVDIRI,
9034 bit_MOVDIR64B): New bits.
9035 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9037 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9038 (VOID, PVOID, PCVOID)): New function types.
9039 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9040 __builtin_ia32_directstoreu_u64,
9041 __builtin_ia32_movdir64b): New builtins.
9042 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9043 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9045 (ix86_valid_target_attribute_inner_p): Ditto.
9046 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9047 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9048 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9049 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9050 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9051 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9052 (movdiri<mode>, movdir64b_<mode>): New.
9053 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9054 * config/i386/immintrin.h: Include movdirintrin.h.
9055 * config/i386/movdirintrin.h: New file.
9056 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9058 2018-04-19 Richard Biener <rguenther@suse.de>
9061 * cfg.c (clear_bb_flags): When loop state says we have
9062 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9064 2018-04-19 Richard Biener <rguenther@suse.de>
9066 PR tree-optimization/84737
9067 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9068 copying restrict info.
9069 (vect_setup_realignment): Use it.
9070 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9071 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9072 the first DR to all generated stores.
9073 (vectorizable_load): Likewise for loads.
9075 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9077 PR tree-optimization/85446
9078 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9079 the integral and pointer types to have the same precision.
9081 * doc/install.texi: Document --disable-cet being the default and
9084 2018-04-18 Martin Liska <mliska@suse.cz>
9086 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9089 2018-04-18 Martin Liska <mliska@suse.cz>
9092 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9095 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9096 arguments if they are comparable.
9098 2018-04-18 Martin Liska <mliska@suse.cz>
9101 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9104 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9107 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9110 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9111 ENDBR after calling __morestack.
9113 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9116 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9117 by using gcc_base_ver to generate a gcc_driver_version, and use
9118 it when generating GCC_DRIVER_NAME.
9119 * configure: Regenerate.
9121 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9124 * config.gcc: Obsolete powerpc*-*-*spe*.
9126 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9129 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9130 (stabstr_D): Change type of unum from unsigned int to
9131 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9134 2018-04-17 Jim Wilson <jimw@sifive.com>
9137 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9138 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9139 Set arg_pointer_offset after using pretend_args_size.
9141 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9143 PR rtl-optimization/85431
9144 * dse.c (record_store): Ignore zero width stores.
9147 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9148 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9149 __builtin_stack_restore rather than after it.
9150 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9151 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9152 argument instead of virtual_dynamic_stack_rtx.
9154 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9156 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9158 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9159 Add note to error message to explain internal mapping of overloaded
9160 built-in function name to non-overloaded built-in function name.
9161 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9164 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9167 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9168 where the inputs overlap with the output.
9170 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9172 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9173 (=v, v) alternative and explicit "memory" attribute.
9174 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9175 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9177 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9178 "sselog1" type instead of "sselog".
9179 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9180 "sselog". Remove explicit "memory" attribute.
9181 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9182 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9184 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9185 "sselog1" type instead of "sselog". Remove explicit "memory"
9187 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9188 use "sselog1" type instead of "sselog". Remove explicit "memory"
9190 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9191 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9193 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9194 "sselog1" type instead of "sselog". Remove explicit "memory"
9196 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9197 use "sselog1" type instead of "sselog". Remove explicit "memory"
9201 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9204 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9207 2018-04-17 Martin Jambor <mjambor@suse.cz>
9210 * ipa-cp.c (create_specialized_node): Call
9211 expand_all_artificial_thunks if necessary.
9213 2018-04-17 Martin Liska <mliska@suse.cz>
9216 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9217 in message, remote space in between '_G' and '('.
9219 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9222 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9223 avx512f_vmcmp<mode>3<round_saeonly_name>,
9224 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9225 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9226 avx512f_rndscale<mode><round_saeonly_name>,
9227 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9228 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9229 Use %<iptr>2 instead of %2 for -masm=intel.
9230 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9231 avx512f_vcvttss2usi<round_saeonly_name>,
9232 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9234 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9235 avx512f_vcvttsd2usi<round_saeonly_name>,
9236 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9237 Use %q1 instead of %1 for -masm=intel.
9238 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9239 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9240 of %3 for -masm=intel.
9241 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9243 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9245 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9247 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9248 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9250 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9252 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9253 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9254 %g1 and one with %0 and %1.
9255 (avx512er_vmrcp28<mode><round_saeonly_name>,
9256 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9258 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9259 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9260 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9261 of %0 and %{%4%} for -masm=intel.
9262 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9263 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9264 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9265 order of %0 and %{%5%}%{z%} for -masm=intel.
9267 2018-04-17 Jan Hubicka <jh@suse.cz>
9270 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9272 2018-04-17 Martin Liska <mliska@suse.cz>
9275 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9276 for target_clone error message. Make default implementation
9277 clone to be a local declaration.
9278 (separate_attrs): Add new argument and check for an empty
9280 (expand_target_clones): Handle it.
9281 (ipa_target_clone): Make redirection just for target_clones
9284 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9285 Tom de Vries <tom@codesourcery.com>
9288 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9289 tiled basic blocks without omp continue statements.
9291 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9294 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9295 vec_extract expression as having side effects to make sure it gets
9298 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9301 * config/i386/i386.c (get_builtin_code_for_version): Check
9304 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9307 * gcc/config.gcc: Support "skylake".
9308 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9310 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9311 (processor_target_table): Add "skylake".
9312 (ix86_option_override_internal): Add "skylake".
9313 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9314 PROCESSOR_CANNONLAKE.
9315 (get_builtin_code_for_version): Fix priority for
9316 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9317 PROCESSOR_SKYLAKE-AVX512.
9318 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9319 (processor_type): Add PROCESSOR_SKYLAKE.
9321 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9322 Jason Merrill <jason@redhat.com>
9325 * convert.c (convert_to_integer_1): Use direct recursion for
9326 enumeral types and types with a precision less than the number
9327 of bits in their mode.
9329 2018-04-16 Julia Koval <julia.koval@intel.com>
9332 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9333 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9335 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9338 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9339 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9340 and -mno-direct-move.
9342 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9345 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9346 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9347 Ensure negative shifts result in {0}.
9349 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9351 PR rtl-optimization/79916
9352 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9353 regs (if any) to define how to gnerate SD moves when LRA is in
9356 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9358 PR rtl-optimization/85393
9359 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9360 * except.c (expand_dw2_landing_pad_for_region): Make static.
9361 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9362 a label and unconditional jump to old_bb, rather than
9363 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9366 PR rtl-optimization/85376
9367 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9368 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9369 instead of a specific value.
9371 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9372 Bin Cheng <bin.cheng@arm.com>
9374 PR tree-optimization/82965
9375 PR tree-optimization/83991
9376 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9377 by_profile_only parameter.
9378 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9379 information if the loop was predicted to iterate too many times.
9380 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9382 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9385 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9388 2018-04-13 Martin Liska <mliska@suse.cz>
9389 Jakub Jelinek <jakub@redhat.com>
9392 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9393 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9394 * builtins.c (expand_builtin_memory_copy_args): Use
9395 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9396 handle dest_addr == pc_rtx.
9398 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9401 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9402 asked to not generate direct moves.
9403 (fix_trunc<mode>si2_stfiwx): Similar.
9404 (fix_trunc<mode>si2_internal): Similar.
9406 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9409 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9410 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9411 lookup if dest in some wider mode is known to be const0_rtx and
9412 if so, record permanent equivalence for it to be ZERO_EXTEND of
9413 the narrower mode destination.
9415 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9417 * lto-streamer-out.c (output_function): Revert 259346.
9418 * omp-expand.c (expand_oacc_for): Likewise.
9420 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9422 PR rtl-optimization/85354
9423 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9424 * sel-sched.c (sel_global_init): ... here.
9426 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9429 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9430 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9431 mode for PE-COFF targets.
9432 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9433 (i386_pe_asm_lto_end): Likewise.
9434 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9435 (TARGET_ASM_LTO_END): Likewise.
9436 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9437 (i386_pe_asm_lto_start): New function.
9438 (i386_pe_asm_lto_end): Likewise.
9440 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9441 Richard Biener <rguenther@suse.de>
9444 * lto-streamer-out.c (output_function): Fix CFG loop state before
9446 * omp-expand.c (expand_oacc_for): Handle calls to internal
9447 functions like regular functions.
9449 2018-04-12 Richard Biener <rguenther@suse.de>
9452 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9453 for the early LTO debug to properly generate references to it
9454 during DIE emission. Do not re-use that for the skeleton for
9456 (dwarf2out_early_finish): Likewise.
9458 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9461 * config/i386/sse.md
9462 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9463 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9464 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9465 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9466 and output is a reg, avoid creating invalid lowpart subreg, but
9467 instead split into a 512-bit move. Don't split if not AVX512VL,
9468 input is xmm16+ reg and output is a mem.
9469 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9470 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9471 xmm16+ reg and output is a mem.
9473 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9475 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9476 also for flag_dwarf2_cfi_asm.
9478 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9480 PR rtl-optimization/85342
9481 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9482 a bool scalar var inside of the loop instead. Don't try to update
9483 recog_data.operand after failed apply_change_group.
9485 2018-04-12 Tom de Vries <tom@codesourcery.com>
9488 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9489 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9490 array with flexible array member as array without given dimension.
9491 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9492 argument for undefined param to true.
9494 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9497 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9498 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9499 from PowerPC section.
9500 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9501 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9502 help text that is too long.
9503 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9504 help text that is too long.
9505 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9506 help text that is too long.
9508 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9510 * config/alpha/alpha.md (stack_probe_internal): Rename
9511 from "probe_stack". Update all callers.
9513 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9515 PR rtl-optimization/84566
9516 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9517 sched_macro_fuse_insns.
9519 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9522 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9523 (compute_block_dependences): ... from here.
9525 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9527 PR tree-optimization/85331
9528 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9529 from int to HOST_WIDE_INT.
9531 2018-04-11 Martin Jambor <mjambor@suse.cz>
9534 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9535 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9536 not the same as the source val.
9537 (cgraph_edge_brings_value_p): New parameter.
9538 (gather_edges_for_value): Pass destination value to
9539 cgraph_edge_brings_value_p.
9540 (perhaps_add_new_callers): Likewise.
9541 (get_info_about_necessary_edges): Likewise and exclude values brought
9542 only by self-recursive edges.
9543 (create_specialized_node): Redirect only clones of self-calling edges.
9544 (+self_recursive_pass_through_p): New function.
9545 (find_more_scalar_values_for_callers_subset): Use it.
9546 (find_aggregate_values_for_callers_subset): Likewise.
9547 (known_aggs_to_agg_replacement_list): Removed.
9548 (decide_whether_version_node): Re-calculate known constants for all
9549 remaining context clones.
9551 2018-04-11 Richard Biener <rguenther@suse.de>
9554 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9555 from early DWARF output.
9556 (dwarf2out_early_finish): Output line info unconditionally into
9557 early DWARF and add reference to it.
9559 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9562 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9563 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9564 other than V2DFmode using iptr mode attribute.
9565 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9567 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9569 PR rtl-optimization/84659
9570 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9572 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9575 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9577 (output_loc_list): Pass address of a dummy size variable even in the
9578 locview handling loop.
9579 (index_location_lists): Add comment on why skip_loc_list_entry can't
9582 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9585 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9588 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9591 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9592 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9593 and -mstring-compare-inline-limit.
9595 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9598 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9599 for stack clash protection in a register whenever we need it to be in
9602 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9604 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9605 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9607 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9610 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9612 (mlong-double-): Ditto.
9613 * config/rs6000/sysv4.opt (msdata=): Ditto.
9614 (mtls-size=): Ditto.
9616 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9618 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9619 erroneous entries for
9620 "vector int vec_ldl (int, long int *)", and
9621 "vector unsigned int vec_ldl (int, unsigned long int *)".
9622 Add comments and entries for
9623 "vector bool char vec_ldl (int, bool char *)",
9624 "vector bool short vec_ldl (int, bool short *)",
9625 "vector bool int vec_ldl (int, bool int *)",
9626 "vector bool long long vec_ldl (int, bool long long *)",
9627 "vector pixel vec_ldl (int, pixel *)",
9628 "vector long long vec_ldl (int, long long *)",
9629 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9630 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9631 type tree bool_long_long_type_node and correct definition of
9632 bool_V2DI_type_node to make reference to this new type tree.
9633 (rs6000_mangle_type): Replace erroneous reference to
9634 bool_long_type_node with bool_long_long_type_node.
9635 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9636 comments to emphasize sign distinctions for char and int types and
9637 replace RS6000_BTI_bool_long constant with
9638 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9639 use of RS6000_BTI_pixel.
9640 (bool_long_type_node): Remove this macro definition.
9641 (bool_long_long_type_node): New macro definition
9643 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9645 PR rtl-optimization/85300
9646 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9647 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9648 simplify_unary_operation fails.
9650 2018-04-10 Martin Liska <mliska@suse.cz>
9652 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9653 cgraph_edge and ipa_ref.
9655 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9659 * config/i386/sse.md
9660 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9661 computation of the VEC_MERGE selector from mask.
9662 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9663 Fix decoding of the VEC_MERGE selector into mask.
9665 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9667 PR tree-optimization/85286
9668 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9670 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9672 * final.c (final_1): Set insn_last_address as well as
9673 insn_current_address.
9675 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9678 * explow.c (emit_stack_probe): Call validize_mem on memory location
9679 before passing it to gen_probe_stack. Create address operand and
9680 legitimize it for the probe_stack_address case.
9682 2018-04-09 Jan Hubicka <jh@suse.cz>
9685 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9686 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9687 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9689 (free_lang_data): Rebuild type inheritance graph.
9691 2018-04-09 Martin Sebor <msebor@redhat.com>
9693 * invoke.texi (-finline-small-functions): Mention other optimization
9695 (-findirect-inlining, -fpartial-inlining): Same.
9696 (-finline-functions-called-once): Same.
9697 (-freorder-blocks-and-partition): Same.
9699 2018-04-09 Jan Hubicka <jh@suse.cz>
9702 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9703 jumps; choose last target that matches the criteria (i.e.
9704 no partition changes for non-crossing jumps).
9705 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9706 support for redirecting crossing jumps to non-crossing.
9708 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9710 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9711 also for naked functions.
9713 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9715 * config/arc/arc.md (add_shift): New pattern.
9716 (add_shift2): Likewise.
9717 (sub_shift): Likewise.
9718 (sub_shift_cmp0_noout): Likewise.
9719 (compare_si_ashiftsi): Likewise.
9720 (xbfu_cmp0_noout): New combine pattern.
9721 (xbfu_cmp0"): Likewise.
9722 (movsi_set_cc_insn): Place the predicable variant first.
9723 (commutative_binary_cmp0_noout): Remove clobber.
9724 (commutative_binary_cmp0): New pattern.
9725 (noncommutative_binary_cmp0): Likewise.
9726 (noncommutative_binary_cmp0_noout): Likewise.
9727 (noncommutative_binary_comparison_result_used): Removed.
9728 (rsub_cmp0): New pattern.
9729 (rsub_cmp0_noout): Likewise.
9730 (extzvsi): Changed, keep only meaningful variants.
9731 (SQH, SEZ): New iterators.
9732 (SQH_postfix): New mode attribute.
9733 (SEZ_prefix): New code attribute.
9734 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9735 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9736 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9738 (noncommutative_operator): Check the availability of barrel
9741 2018-04-09 Richard Biener <rguenther@suse.de>
9743 PR tree-optimization/85284
9744 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9745 Only use the niter constraining form of simple_iv when the exit
9748 2018-04-09 Tom de Vries <tom@codesourcery.com>
9751 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9752 (define_expand "*memory_barrier"): New define_expand.
9753 (define_insn "memory_barrier"): New insn.
9755 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9757 PR rtl-optimization/80463
9758 PR rtl-optimization/83972
9759 PR rtl-optimization/83480
9761 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9762 correct producer for the insn.
9763 (tidy_control_flow): Fixup seqnos in case of debug insns.
9765 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9767 PR rtl-optimization/83913
9769 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9770 different sched-times when merging exprs.
9772 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9774 PR rtl-optimization/83962
9776 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9777 tidy_fallthru_edge and tidy_control_flow.
9779 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9781 PR rtl-optimization/83530
9783 * sel-sched.c (force_next_insn): New global variable.
9784 (remove_insn_for_debug): When force_next_insn is true, also leave only
9785 next insn in the ready list.
9786 (sel_sched_region): When the region wasn't scheduled, make another pass
9787 over it with force_next_insn set to 1.
9789 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9791 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9793 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9794 for interrupt control.
9795 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9796 functions for interrupt control.
9797 * config/nds32/nds32-intrinsic.md: Likewise.
9798 * config/nds32/nds32_intrinsic.h: Likewise.
9799 * config/nds32/nds32.h (nds32_builtins): Likewise.
9801 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9803 * config/nds32/nds32.c (nds32_init_machine_status,
9804 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9805 strict_aligned_p field.
9806 (nds32_expand_to_rtl_hook): New function.
9807 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9808 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9810 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9811 Chung-Ju Wu <jasonwucj@gmail.com>
9813 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9814 * config/nds32/nds32-n7.md: New file.
9815 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9816 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9818 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9819 * config/nds32/nds32.md (pipeline_model): Add n7.
9820 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9821 * config/nds32/pipelines.md: Include n7 settings.
9823 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9824 Chung-Ju Wu <jasonwucj@gmail.com>
9826 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9827 * config/nds32/nds32-e8.md: New file.
9828 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9829 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9831 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9832 * config/nds32/nds32.md (pipeline_model): Add e8.
9833 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9834 * config/nds32/pipelines.md: Include e8 settings.
9836 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9837 Chung-Ju Wu <jasonwucj@gmail.com>
9839 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9840 * config/nds32/nds32-n8.md: New file.
9841 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9842 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9844 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9845 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9846 * config/nds32/nds32.md (pipeline_model): Add n8.
9847 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9848 * config/nds32/pipelines.md: Include n8 settings.
9850 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9851 Chung-Ju Wu <jasonwucj@gmail.com>
9853 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9854 * config/nds32/nds32-n9-2r1w.md: New file.
9855 * config/nds32/nds32-n9-3r2w.md: New file.
9856 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9857 nds32_register_ports): New or modify for cpu n9.
9858 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9860 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9861 * config/nds32/nds32-utils.c: New file.
9862 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9863 TARGET_MUL_SLOW): Define.
9864 * config/nds32/nds32.md (pipeline_model): New attribute.
9865 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9866 New options that support cpu n9.
9867 * config/nds32/pipelines.md: Include n9 settings.
9868 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9870 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9872 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9873 information if necessary.
9874 (output_cond_branch_compare_zero): Likewise.
9875 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9876 (nds32_target_alignment): Refine for alignment.
9877 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9878 (FUNCTION_BOUNDARY): Modify.
9879 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9881 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9883 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9885 * config/nds32/constants.md (unspec_volatile_element): Add values for
9886 TLB operation and data prefetch.
9887 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9888 functions for TLB operation and data prefetch.
9889 * config/nds32/nds32-intrinsic.md: Likewise.
9890 * config/nds32/nds32_intrinsic.h: Likewise.
9891 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9892 (nds32_print_operand): Likewise.
9893 * config/nds32/nds32.h (nds32_builtins): Likewise.
9895 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9896 Andrew Pinski <pinsika@gcc.gnu.org>
9899 * match.pd: Use constant_boolean_node of correct type instead of
9900 boolean_true_node or boolean_false_node for simplifying
9901 pointer comparisons to zero.
9903 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9905 PR tree-optimization/80021
9906 * tree.c (verify_type_variant): Make error call in verify_variant_match
9907 translatable and remove final full stop.
9909 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9911 * config/nds32/constants.md (unspec_volatile_element): Add
9912 UNSPEC_VOLATILE_EH_RETURN.
9913 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9914 nds32_output_stack_pop): Support dwarf exception handling process.
9915 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9916 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9917 exception handling process.
9918 (nds32_compute_stack_frame): Likewise.
9919 (nds32_return_addr_rtx): Likewise.
9920 (nds32_initial_elimination_offset): Likewise.
9921 (nds32_expand_prologue): Likewise.
9922 (nds32_expand_epilogue): Likewise.
9923 (nds32_dynamic_chain_address): New function.
9924 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9926 (DYNAMIC_CHAIN_ADDRESS): Define.
9927 (EH_RETURN_DATA_REGNO): Define.
9928 (EH_RETURN_STACKADJ_RTX): Define.
9929 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9930 patterns for dwarf exception handling.
9932 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9934 * config/nds32/nds32.h: Clean up obsolete macros.
9936 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9938 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9939 Add enum values for particular instructions.
9940 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9941 particular intrinsic functions.
9942 * config/nds32/nds32-intrinsic.md: Likewise.
9943 * config/nds32/nds32_intrinsic.h: Likewise.
9944 * config/nds32/nds32.h (nds32_builtins): Likewise.
9945 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9946 (btst, ave): New patterns for particular instructions.
9948 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9950 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9951 Add enum values for atomic load/store and memory sync.
9952 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9954 * config/nds32/nds32-intrinsic.md: Likewise.
9955 * config/nds32/nds32_intrinsic.h: Likewise.
9956 * config/nds32/nds32.h (nds32_builtins): Likewise.
9958 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9960 PR tree-optimization/85257
9961 * fold-const.c (native_encode_vector): If not all elts could fit
9962 and off is -1, return 0 rather than offset.
9963 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9964 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9965 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9966 adjust buffer in native_interpret_expr call.
9968 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9970 * config/nds32/constants.md (unspec_volatile_element): Add cache
9971 control enum values.
9972 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9973 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9974 * config/nds32/nds32.c (nds32_cctl_names): New.
9975 (nds32_print_operand): Handle cache control register names.
9976 * config/nds32/nds32.h (nds32_builtins): New enum values.
9977 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9979 * config/nds32/nds32.md (type): Add mmu.
9980 * config/nds32/pipelines.md (simple_insn): Add mmu.
9982 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9984 * config/nds32/nds32.md (type): Remove call.
9985 * config/nds32/pipelines.md (simple_insn): Likewise.
9987 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9989 * config/nds32/constants.md (unspec_volatile_element): Add
9990 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9991 UNSPEC_VOLATILE_FMFCFG.
9992 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9993 description for fmfcfg and fmfcsr.
9994 (bdesc_1arg): Add fmtcsr.
9995 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9996 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9997 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9998 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9999 unspec_fmfcfg): New patterns.
10000 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
10001 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
10002 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
10003 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
10004 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
10005 __nds32__fmfcfg): Define.
10007 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
10009 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
10010 intrinsic register names.
10011 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
10012 intrinsic register enum values and macros.
10014 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
10016 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
10017 for load/store addressing form.
10018 (nds32_print_operand_address): Likewise.
10020 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
10023 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
10024 based on LABEL_REF. Remove useless assertion.
10025 (pic_address_needs_scratch): Fix formatting.
10026 (sparc_legitimize_pic_address): Minor tweaks.
10027 (sparc_delegitimize_address): Adjust assertion accordingly.
10028 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
10029 into symbolic_operand.
10030 (movsi_high_pic_label_ref): Likewise.
10031 (movsi_lo_sum_pic_label_ref): Likewise.
10032 (movdi_pic_label_ref): Likewise.
10033 (movdi_high_pic_label_ref): Likewise.
10034 (movdi_lo_sum_pic_label_ref): Likewise.
10036 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10038 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10039 custom LIB_SPEC setup.
10041 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10042 Kito Cheng <kito.cheng@gmail.com>
10044 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10045 * config/riscv/freebsd.h: New.
10047 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10049 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10050 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10053 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10054 Kito Cheng <kito.cheng@gmail.com>
10056 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10057 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10058 * config/nds32/nds32-protos.h (nds32_output_call,
10059 nds32_output_return): Declare.
10060 * config/nds32/nds32.md: Refine all the call and return patterns.
10062 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10065 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10066 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10068 PR rtl-optimization/84872
10069 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10070 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10071 EDGE_CROSSING edge.
10073 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10075 * expr.c (copy_blkmode_to_reg): Revert 254862.
10076 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10078 2018-04-06 Richard Biener <rguenther@suse.de>
10080 PR middle-end/85244
10081 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10082 after seeing a component reference with an adjacent field. Treat
10083 refs to arrays at struct end of external decls similar to
10084 refs to unconstrained commons.
10086 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10089 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10090 look through SAVE_EXPRs with non-side-effects argument. Adjust
10092 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10095 2018-04-06 Richard Biener <rguenther@suse.de>
10097 PR middle-end/85180
10098 * alias.c (find_base_term): New wrapper around find_base_term
10099 unwinding CSELIB_VAL_PTR changes.
10100 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10103 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10105 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10107 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10108 constant definitions.
10109 ("nop"): lr 0,0 -> nopr r0
10110 ("nop_lr0", "nop_lr1"): New insn definitions.
10112 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10114 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10115 NDS32_V3PUSH_AVAILABLE_P macro.
10117 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10118 Chung-Ju Wu <jasonwucj@gmail.com>
10120 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10121 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10122 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10123 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10124 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10125 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10126 * config/nds32/constraints.md: New constraints and checking for hard
10127 float configuration.
10128 * config/nds32/iterators.md: New mode iterator and attribute for hard
10129 float configuration.
10130 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10132 * config/nds32/nds32-fpu.md: New file.
10133 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10134 deal with hard float code generation.
10135 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10137 (abi_type, float_reg_number): New enum type.
10138 * config/nds32/nds32-predicates.c: New predicates for hard float.
10139 * config/nds32/nds32-protos.h: Declare functions for hard float.
10140 * config/nds32/nds32.c: Implementation for hard float configuration.
10141 * config/nds32/nds32.h: Definitions for hard float configuration.
10142 * config/nds32/nds32.md: Include hard float machine description and
10143 modify patterns for hard float configuration.
10144 * config/nds32/nds32.opt: New options for hard float configuration.
10145 * config/nds32/predicates.md: New predicates for hard float
10148 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10150 * common/config/nds32/nds32-common.c
10151 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10153 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10155 PR middle-end/85195
10156 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10157 CONSTRUCTOR_ELT (ctor, ...)->value.
10159 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10162 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10164 2018-04-05 Tom de Vries <tom@codesourcery.com>
10167 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10170 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10171 Kito Cheng <kito.cheng@gmail.com>
10173 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10174 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10175 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10176 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10178 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10179 Kito Cheng <kito.cheng@gmail.com>
10181 * config/nds32/constraints.md (Ufe): New memory constraint.
10182 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10183 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10184 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10186 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10187 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10189 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10191 * config/nds32/nds32.md: Use optimize_size in the condition for
10192 alu-shift instructions.
10194 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10196 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10198 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10200 * config/nds32/nds32.md (negsi2): Refine pattern.
10202 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10203 Chung-Ju Wu <jasonwucj@gmail.com>
10205 * config/nds32/iterators.md (shift_rotate): New code iterator.
10206 (shift): New code attribute.
10207 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10208 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10209 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10210 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10211 bit-wise operations.
10212 (andsi3, *andsi3): Ditto.
10213 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10214 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10215 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10216 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10217 nds32_ior_operand, nds32_xor_operand): New predicates.
10219 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10221 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10222 (addsi3, subsi3): ... this.
10224 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10226 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10228 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10230 * config/nds32/nds32.md: Adjust indention.
10232 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10234 * config/nds32/nds32.md (feature): New attribute.
10236 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10238 * config/nds32/nds32.md (subtype): New attribute.
10240 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10243 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10244 expansion to perform a bitwise AND of the argument followed by a
10245 boolean negation of the result.
10247 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10249 PR rtl-optimization/84878
10250 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10251 the basic block. Assert the use reference is not artificial and that
10252 it has an associated insn.
10254 2018-04-04 Michael Matz <matz@suse.de>
10256 * builtins.c (compute_objsize): Pass correct operand
10257 to array_at_struct_end_p.
10259 2018-04-04 Richard Biener <rguenther@suse.de>
10262 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10263 from contexts for DINFO_LEVEL_TERSE and below.
10265 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10267 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10268 resiter_operand condition.
10269 * config/nds32/nds32.md (*move<mode>): Ditto.
10271 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10272 Monk Chiang <sh.chiang04@gmail.com>
10274 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10276 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10278 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10280 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10281 Kito Cheng <kito.cheng@gmail.com>
10283 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10284 nds32_cond_code_str, output_cond_branch,
10285 output_cond_branch_compare_zero, nds32_expand_cbranch,
10286 nds32_expand_cstore, nds32_expand_movcc,
10287 nds32_output_cbranchsi4_equality_zero,
10288 nds32_output_cbranchsi4_equality_reg,
10289 nds32_output_cbranchsi4_equality_reg_or_const_int,
10290 nds32_output_cbranchsi4_greater_less_zero: New functions.
10291 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10292 nds32_expand_cstore, nds32_expand_movcc,
10293 nds32_output_cbranchsi4_equality_zero,
10294 nds32_output_cbranchsi4_equality_reg,
10295 nds32_output_cbranchsi4_equality_reg_or_const_int,
10296 nds32_output_cbranchsi4_greater_less_zero): Declare.
10297 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10298 nds32_rimm11s_operand): New predicates.
10299 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10300 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10303 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10305 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10306 * config/nds32/nds32.md: Ditto.
10307 * config/nds32/pipelines.md: Ditto.
10309 2018-04-04 Richard Biener <rguenther@suse.de>
10311 PR tree-optimization/85168
10312 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10313 propagating abnormals.
10315 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10317 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10319 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10320 Kito Cheng <kito.cheng@gmail.com>
10322 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10323 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10324 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10325 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10326 * config/nds32/nds32.md (sibcall_internal): New.
10327 (sibcall_register): Remove.
10328 (sibcall_immediate): Remove.
10329 (sibcall_value_internal): New.
10330 (sibcall_value_register): Remove.
10331 (sibcall_value_immediate): Remove.
10332 * config/nds32/predicates.md (nds32_general_register_operand): New.
10333 (nds32_call_address_operand): New.
10335 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10337 PR rtl-optimization/85167
10338 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10339 bb_defs if *split_p, instead preinitialize it to NULL.
10341 PR tree-optimization/85156
10342 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10343 evaluating the argument multiple times.
10345 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10347 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10349 (_mm_cvtpd_ps): Likewise.
10350 (_mm_cvttpd_epi32): Likewise.
10351 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10352 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10353 vector, pixel, and bool following altivec.h include.
10355 2018-04-03 Martin Sebor <msebor@redhat.com>
10357 * doc/extend.texi (Common Function Attributes): Clarify.
10358 (const attribute): Likewise.
10359 (pure attribute): Likewise.
10361 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10364 * config/i386/i386.c (ix86_expand_vector_set): Use
10365 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10367 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10369 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10370 instructions when changing rounding bits to preserve precision bits
10371 in the x87 control word.
10373 2018-04-03 Martin Liska <mliska@suse.cz>
10375 PR tree-optimization/82491
10376 * rtl.h (strip_offset_and_add): Replace += suboffset with
10377 poly_uint64 () + suboffset.
10379 2018-03-29 Martin Liska <mliska@suse.cz>
10380 Martin Jambor <mjambor@suse.cz>
10383 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10384 param_type is not an integral or pointer type.
10386 2018-04-03 Richard Biener <rguenther@suse.de>
10388 * sese.h (recompute_all_dominators): Remove.
10390 2018-04-02 Martin Sebor <msebor@redhat.com>
10392 * doc/invoke.texi (-Wrestrict): Fix typos.
10394 2018-04-02 Jim Wilson <jimw@sifive.com>
10396 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10397 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10398 (<optab>di3, <optab>si3_extend): Likewise.
10399 (<optab>si3_mask, <optab>si3_mask_1): New.
10400 (<optab>di3_mask, <optab>di3_mask_1): New.
10401 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10402 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10403 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10405 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10407 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10410 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10412 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10413 (nds32_canonicalize_comparison): New function.
10415 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10416 Kito Cheng <kito.cheng@gmail.com>
10417 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10419 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10420 * config/nds32/constants.md (unspec_volatile_element): Add
10421 UNSPEC_VOLATILE_RELAX_GROUP.
10422 * config/nds32/nds32-relax-opt.c: New file.
10423 * config/nds32/nds32-predicates.c
10424 (nds32_symbol_load_store_p): New function.
10425 * config/nds32/nds32-protos.h
10426 (nds32_symbol_load_store_p): Declare function.
10427 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10428 * config/nds32/nds32.c
10429 (nds32_register_pass): New function to register pass.
10430 (nds32_register_passes): New function to register passes.
10431 * config/nds32/nds32.md (relax_group): New pattern.
10432 * config/nds32/nds32.opt (mrelax-hint): New option.
10433 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10435 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10437 * config/nds32/t-nds32: Modify files dependency.
10439 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10441 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10442 (PROFILE_HOOK): Define its implementation.
10444 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10446 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10447 type and 32-bit size.
10449 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10451 PR middle-end/85090
10452 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10453 (V_128_256): New mode iterator.
10454 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10455 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10456 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10458 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10459 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10461 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10464 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10465 NaN inputs correctly.
10467 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10470 * config/rs6000/vsx.md (??r): New mode attribute.
10471 (*vsx_mov<mode>_64bit): Use it.
10472 (*vsx_mov<mode>_32bit): Likewise.
10474 2018-03-30 Martin Sebor <msebor@redhat.com>
10476 PR tree-optimization/84818
10477 * builtins.c (check_access): Use warning_n.
10479 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10482 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10484 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10487 2018-03-30 Julia Koval <julia.koval@intel.com>
10490 * x86-tune.def (movx, partial_reg_dependency): Enable for
10493 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10495 PR inline-asm/84985
10496 * lra-constraints.c (process_alt_operands): Move setting
10497 this_alternative_matches below.
10499 2018-03-29 Martin Liska <mliska@suse.cz>
10502 * doc/invoke.texi: Document how LTO works with debug info.
10503 Describe auto-load support of binutils. Mention 'x86-64'
10504 as valid option value of -march option.
10506 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10508 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10511 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10512 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10513 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10516 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10519 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10520 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10521 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10522 for RS6000_BTM_POWERPC64.
10523 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10524 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10525 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10528 (DIVDEU): Likewise.
10530 2018-03-28 Carl Love <cel@us.ibm.com>
10533 2017-09-27 Carl Love <cel@us.ibm.com>
10535 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10536 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10537 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10540 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10542 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10543 instead of __vector bool.
10544 (_mm_max_pu8): Likewise.
10545 (_mm_min_pi16): Likewise.
10547 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10550 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10551 (DIVWEUO): Likewise.
10552 (DIVDEO): Likewise.
10553 (DIVDEUO): Likewise.
10554 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10555 DIVWEUO and DIVDEUO.
10556 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10557 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10558 (div_extend): Likewise.
10559 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10561 (__builtin_divweuo): Likewise.
10562 (__builtin_divdeo): Likewise.
10563 (__builtin_divdeuo): Likewise.
10565 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10568 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10569 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10571 PR tree-optimization/82004
10572 * gimple-match-head.c (optimize_pow_to_exp): New function.
10573 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10574 Don't fold to exp if optimize_pow_to_exp is false.
10576 2018-03-28 Martin Liska <mliska@suse.cz>
10579 * calls.c (initialize_argument_information): Fix trailing space.
10580 * common.opt: Fix typo and provide better explanation for
10581 -fsanitize-coverage option.
10582 * config/i386/i386.opt: Fix typo.
10584 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10585 Martin Liska <mliska@suse.cz>
10588 * gimplify.c (asan_poison_variable): Don't do the check for
10589 gimplify_omp_ctxp here.
10590 (gimplify_decl_expr): Do it here.
10591 (gimplify_target_expr): Likewise.
10593 2018-03-28 Martin Liska <mliska@suse.cz>
10596 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10597 chkp_type_bounds_count if MPX is not enabled.
10599 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10601 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10603 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10606 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10607 function to create the function decl for complex long double
10608 multiply and divide for -mabi=ieeelongdouble.
10609 (init_float128_ieee): Call it.
10611 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10614 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10615 -fcf-protection=branch -mibt.
10616 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10618 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10621 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10622 arm_word_relocations.
10624 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10627 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10628 extern array declarations.
10630 2018-03-27 Richard Biener <rguenther@suse.de>
10632 PR middle-end/84067
10633 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10634 explicit single_use checks.
10636 2018-03-27 Richard Biener <rguenther@suse.de>
10638 PR tree-optimization/85082
10639 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10642 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10644 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10645 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10646 Turn on fasynchronous-unwind-tables and funwind-tables.
10648 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10651 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10652 (*bmi_blsr_<mode>_ccz): Ditto.
10654 2018-03-26 Tom de Vries <tom@codesourcery.com>
10656 PR tree-optimization/85063
10657 * omp-general.c (offloading_function_p): New function. Factor out
10659 * omp-offload.c (pass_omp_target_link::gate): ... here.
10660 * omp-general.h (offloading_function_p): Declare.
10661 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10662 with attribute omp declare target for offloading functions.
10664 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10666 PR tree-optimization/84005
10667 * tree-data-ref.h (get_base_for_alignment): Declare.
10668 * tree-data-ref.c (get_base_for_alignment_1): New function.
10669 (get_base_for_alignment): Likewise.
10670 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10671 get_base_for_alignment to find a suitable base object, instead
10672 of always using drb->base_address.
10674 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10676 PR inline-asm/85022
10677 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10678 known size by default.
10680 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10682 PR inline-asm/85030
10683 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10684 and non BLKmode operands.
10686 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10689 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10690 Clean up attributes.
10692 2018-03-23 Richard Biener <rguenther@suse.de>
10695 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10696 we are going to emit early debug for LTO.
10698 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10700 PR inline-asm/85034
10701 * function.c (match_asm_constraints_1): Don't optimize if input
10702 doesn't satisfy general_operand predicate for output's mode.
10704 PR inline-asm/85022
10705 * alias.c (write_dependence_p): Don't require for x_canonicalized
10706 non-VOIDmode if x has VOIDmode.
10709 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10710 just don't try to optimize it rather than assert it never happens.
10712 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10714 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10715 macro expansions for definition of ST_INTERNAL_<mode> and
10716 LD_INTERNAL_<mode> builtins.
10717 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10719 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10721 (altivec_expand_st_builtin): Likewise.
10722 (altivec_expand_builtin): Remove calls to deleted functions.
10723 (rs6000_address_for_altivec): Delete this function.
10724 * config/rs6000/vector.md: Remove expands for
10725 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10727 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10730 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10731 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10732 re-computing once computed.
10733 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10734 (arm_init_machine_status): Initialize
10735 machine->static_chain_stack_bytes.
10737 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10740 * doc/extend.texi: Add four new prototypes for vec_ld.
10741 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10742 definitions for more logical presentation.
10743 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10744 entries for V1TI variants of __builtin_altivec_ld builtin.
10745 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10746 handling of V1TI variant of LVX icode pattern.
10747 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10748 (rs6000_gimple_fold_builtin): Likewise.
10749 (altivec_init_builtins): Add code to define
10750 __builtin_altivec_lvx_v1ti function.
10752 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10754 PR inline-asm/84941
10755 * function.c (match_asm_constraints_1): Don't do the optimization
10756 if input isn't a REG, SUBREG, MEM or constant.
10758 2018-03-22 Tom de Vries <tom@codesourcery.com>
10760 PR tree-optimization/84956
10761 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10762 bb_has_abnormal_pred.
10764 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10767 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10768 DECL_INITIAL (decl) to decl at the end.
10769 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10770 adjust the comment.
10772 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10774 * doc/extend.texi (__builtin_tgmath): Document when complex
10775 integer types are treated as _Complex _Float64.
10777 2018-03-21 Tom de Vries <tom@codesourcery.com>
10779 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10781 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10783 PR tree-optimization/84960
10784 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10785 if it is ENTRY block, move them into single succ of ENTRY in that case.
10787 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10789 PR tree-optimization/84811
10790 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10791 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10792 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10793 (known_subrange_p): Update accordingly. Cast each value involved
10794 in the size comparison, rather than casting the result of the
10797 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10799 PR tree-optimization/84982
10800 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10801 by flipping the least significant bit rather than all bits from
10802 bitpos to bitpos + bitsize - 1.
10804 2018-03-21 Nathan Sidwell <nathan@acm.org>
10806 * doc/extend.texi (Deprecated Features): Remove mention of
10807 long-deleted deprecations.
10809 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10812 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10813 * configure: Regenerate.
10815 2018-03-21 Tom de Vries <tom@codesourcery.com>
10817 PR tree-optimization/83126
10818 * tree-parloops.c (num_phis): New function.
10819 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10821 2018-03-21 Nathan Sidwell <nathan@acm.org>
10823 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10824 mention anon-struct/union members and trailing attributes.
10826 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10828 PR tree-optimization/84969
10829 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10830 builtin memset partitions if they set different rhs values.
10832 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10834 PR rtl-optimization/84989
10835 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10836 VEC_DUPLICATE with scalar result mode.
10838 2018-03-21 Martin Liska <mliska@suse.cz>
10841 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10842 not intended return statement.
10844 2018-03-21 Martin Liska <mliska@suse.cz>
10847 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10848 (chkp_find_bound_slots_1): Limit number of iterations.
10850 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10853 * Minor grammar fixes for x86 options.
10855 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10858 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10859 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10861 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10864 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10865 (altivec_lvx_<mode>_1op): Likewise.
10866 (altivec_stvx_<mode>_2op): Likewise.
10867 (altivec_stvx_<mode>_1op): Likewise.
10868 (altivec_lvx_<VM2:mode>): New define_expand.
10869 (altivec_stvx_<VM2:mode>): Likewise.
10870 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10871 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10872 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10873 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10874 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10875 (rs6000_gen_lvx): Likewise.
10876 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10877 (altivec_expand_stv_builtin): Likewise.
10878 (altivec_expand_builtin): Likewise.
10879 * config/rs6000/vector.md: Likewise.
10881 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10884 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10887 2018-03-20 Richard Biener <rguenther@suse.de>
10890 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10891 sign-conversions as zero, fall back to standard scalar_stmt
10894 2018-03-20 Martin Liska <mliska@suse.cz>
10897 * predict.c (rebuild_frequencies): Handle case when we have
10898 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10900 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10903 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10904 flag_section_anchors.
10905 * varasm.c (use_blocks_for_decl_p): Remove hack for
10906 dw2_force_const_mem.
10909 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10911 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10912 be created, use lowpart_subreg of operands[0] rather than operands[0]
10914 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10915 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10916 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10917 and n constraint instead of aarch64_shift_imm_di and Usd.
10918 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10919 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10921 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10924 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10925 to favor GPR over NEON registers.
10926 (<shift>di3_neon): Likewise.
10928 2018-03-20 Tom de Vries <tom@codesourcery.com>
10931 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10932 (nvptx_process_pars): Emit bar.sync asap and alap.
10934 2018-03-20 Tom de Vries <tom@codesourcery.com>
10937 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10938 seen_label if seen_label is already set.
10940 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10943 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10944 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10945 Use 1U instead of 1. Formatting fixes.
10948 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10949 instead of TREE_TYPE (s1) for the return value.
10951 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10953 PR tree-optimization/84946
10954 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10955 bitsize + bitsize in poly_uint64 rather than poly_int64.
10958 * dwarf2asm.c: Include fold-const.c.
10959 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10960 of decl rather than decl itself.
10962 PR rtl-optimization/84643
10963 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10965 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10968 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10969 calling assemble_variable.
10971 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10974 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10975 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10977 2018-03-19 Jim Wilson <jimw@sifive.com>
10980 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10981 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10982 (riscv_first_stack_step): Likewise.
10983 (riscv_option_override): Use STACK_BOUNDARY instead of
10984 MIN_STACK_BOUNDARY.
10985 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10986 MIN_STACK_BOUNDARY.
10987 (BIGGEST_ALIGNMENT): Set to 128.
10988 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10989 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10992 2018-03-19 Richard Biener <rguenther@suse.de>
10994 PR tree-optimization/84933
10995 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10996 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10998 2018-03-19 Richard Biener <rguenther@suse.de>
11000 PR tree-optimization/84859
11001 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
11002 (cond_if_else_store_replacement): Perform sinking operation on
11003 single-store BBs regardless of MAX_STORES_TO_SINK setting.
11004 Generalize what a BB with a single eligible store is.
11006 2018-03-19 Richard Biener <rguenther@suse.de>
11008 PR tree-optimization/84929
11009 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
11010 chrec_is_positive against non-chrec arg.
11012 2018-03-19 Tamar Christina <tamar.christina@arm.com>
11015 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
11017 2018-03-18 Martin Liska <mliska@suse.cz>
11019 PR rtl-optimization/84635
11020 * regrename.c (build_def_use): Use matches_mode only when
11023 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
11025 PR tree-optimization/84913
11026 * tree-vect-loop.c (vectorizable_reduction): Don't try to
11027 vectorize chains of COND_EXPRs.
11029 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11031 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
11033 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11035 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11037 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11039 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11041 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11042 Kito Cheng <kito.cheng@gmail.com>
11044 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11045 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11046 (nds32_adjust_reg_alloc_order): New function.
11047 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11049 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11051 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11052 nds32_print_operand, nds32_print_operand_address): Use
11053 HOST_WIDE_INT_PRINT_DEC instead.
11055 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11057 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11059 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11062 * config/i386/i386.c (initial_ix86_tune_features,
11063 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11064 unsigned long long.
11065 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11066 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11067 rather than 1u << ix86_tune. Formatting fix.
11068 (ix86_option_override_internal): Change ix86_arch_mask from
11069 unsigned int to unsigned HOST_WIDE_INT, initialize to
11070 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11071 (ix86_function_specific_restore): Likewise.
11073 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11076 * postreload.c (reload_combine_recognize_pattern): Perform
11077 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11078 truncate_int_for_mode the result for the destination's mode.
11081 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11082 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11084 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11087 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11090 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11091 regno_allocno_class_array and sorted_pseudos.
11092 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11093 insns where regno is used.
11095 2018-03-16 Martin Liska <mliska@suse.cz>
11098 * multiple_target.c (create_dispatcher_calls): Redirect
11099 reference in the symbol table.
11101 2018-03-16 Martin Liska <mliska@suse.cz>
11104 * multiple_target.c (create_dispatcher_calls): Redirect also
11107 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11111 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11112 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11113 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11115 2018-03-16 Julia Koval <julia.koval@intel.com>
11117 * doc/invoke.texi (Skylake Server): Add CLWB.
11118 Cannonlake): Remove CLWB.
11120 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11122 PR tree-optimization/84841
11123 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11125 (FLOAT_ONE_CONST_TYPE): Define.
11126 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11127 (sort_by_operand_rank): Put entries with higher constant_type last
11128 rather than first to match comments.
11130 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11132 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11135 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11138 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11139 check and use address of target always.
11141 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11144 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11145 (indirect_thunk_bnd_needed): Likewise.
11146 (indirect_thunks_used): Likewise.
11147 (indirect_thunks_bnd_used): Likewise.
11148 (indirect_return_needed): New.
11149 (indirect_return_bnd_needed): Likewise.
11150 (output_indirect_thunk_function): Add a bool argument for
11152 (output_indirect_thunk_function): Don't generate alias for
11153 function return thunk.
11154 (ix86_code_end): Call output_indirect_thunk_function to generate
11155 function return thunks.
11156 (ix86_output_function_return): Set indirect_return_bnd_needed
11157 and indirect_return_needed instead of indirect_thunk_bnd_needed
11158 and indirect_thunk_needed.
11160 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11162 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11163 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11164 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11166 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11167 Paul Hua <paul.hua.gm@gmail.com>
11170 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11172 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11174 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11175 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11176 resp. SFmode cases.
11178 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11181 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11182 instead of GET_MODE_SIZE when comparing Units.
11184 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11187 * varasm.c (hash_section): Return an unchangeble hash value
11188 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11189 Return !aarch64_can_use_per_function_literal_pools_p ().
11191 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11194 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11195 rather than address of cmode as last argument to prepare_cmp_insn.
11197 2018-03-15 Julia Koval <julia.koval@intel.com>
11199 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11200 F_AVX512VNNI, F_AVX512BITALG): New.
11202 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11205 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11206 insn for floating-point loads and stores.
11208 2018-03-14 Carl Love <cel@us.ibm.com>
11210 * config/rs6000/rs6000-c.c: Add macro definitions for
11211 ALTIVEC_BUILTIN_VEC_PERMXOR.
11212 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11213 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11214 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11215 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11217 * config/doc/extend.texi: Add prototypes for vec_permxor.
11219 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11222 * diagnostic-show-locus.c (class layout_point): Convert m_line
11223 from int to linenum_type.
11224 (line_span::comparator): Use linenum "compare" function when
11225 comparing line numbers.
11226 (test_line_span): New function.
11227 (layout_range::contains_point): Convert param "row" from int to
11229 (layout_range::intersects_line_p): Likewise.
11230 (layout::will_show_line_p): Likewise.
11231 (layout::print_source_line): Likewise.
11232 (layout::should_print_annotation_line_p): Likewise.
11233 (layout::print_annotation_line): Likewise.
11234 (layout::print_leading_fixits): Likewise.
11235 (layout::annotation_line_showed_range_p): Likewise.
11236 (struct line_corrections): Likewise for field m_row.
11237 (line_corrections::line_corrections): Likewise for param "row".
11238 (layout::print_trailing_fixits): Likewise.
11239 (layout::get_state_at_point): Likewise.
11240 (layout::get_x_bound_for_row): Likewise.
11241 (layout::print_line): Likewise.
11242 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11243 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11244 * input.c (selftest::test_linenum_comparisons): New function.
11245 (selftest::input_c_tests): Call it.
11246 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11247 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11248 * selftest.h (ASSERT_GT): New macro.
11249 (ASSERT_GT_AT): New macro.
11250 (ASSERT_LT): New macro.
11251 (ASSERT_LT_AT): New macro.
11253 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11255 PR rtl-optimization/84780
11256 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11258 2018-03-14 Martin Liska <mliska@suse.cz>
11260 * tree.c (record_node_allocation_statistics): Use
11261 get_stats_node_kind.
11262 (get_stats_node_kind): New function extracted from
11263 record_node_allocation_statistics.
11264 (free_node): Use get_stats_node_kind.
11266 2018-03-14 Richard Biener <rguenther@suse.de>
11268 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11269 that the value-set of ANTIC_IN doesn't grow.
11272 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11274 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11275 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11277 2018-03-14 Julia Koval <julia.koval@intel.com>
11279 * config.gcc (icelake-client, icelake-server): New.
11281 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11282 (initial_ix86_arch_features): Ditto.
11283 (PTA_SKYLAKE): Add SGX.
11284 (PTA_ICELAKE): Remove.
11285 (PTA_ICELAKE_CLIENT): New.
11286 (PTA_ICELAKE_SERVER): New.
11287 (ix86_option_override_internal): Split up icelake on icelake client and
11289 (get_builtin_code_for_version): Ditto.
11290 (fold_builtin_cpu): Ditto.
11291 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11292 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11293 * config/i386/i386.h (processor_type): Ditto.
11294 * doc/invoke.texi: Ditto.
11296 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11299 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11300 INTEGER_CST offset, add it together with bitpos / 8 and
11301 sign extend based on POINTER_SIZE.
11305 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11308 * config/i386/constraints.md (Yc): New register constraint.
11309 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11310 Use Yc constraint for alternative 2 of operand 0. Remove
11311 preferred_for_speed attribute.
11313 2018-03-14 Richard Biener <rguenther@suse.de>
11315 PR tree-optimization/84830
11316 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11317 with the old one to avoid oscillations.
11319 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11322 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11324 (assign_by_spills): Return a flag of reload assignment failure.
11325 Do not process the reload assignment failures. Do not spill other
11326 reload pseudos if they has the same reg class. Update n if
11328 (lra_assign): Add a return arg. Set up from the result of
11329 assign_by_spills call.
11330 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11331 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11332 usage_insns if it is not NULL.
11333 (spill_hard_reg_in_range): New function.
11334 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11335 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11336 function prototypes.
11337 (lra_assign): Change prototype.
11338 * lra.c (lra): Add code to deal with fails by splitting hard reg
11341 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11343 * config/riscv/riscv.opt (mrelax): New option.
11344 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11345 "norelax" when riscv_mrelax is disabled.
11346 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11348 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11351 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11352 reassociation for int modes.
11354 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11356 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11357 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11359 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11360 * config/aarch64/aarch64-sve.md
11361 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11362 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11363 (*extend<mode><Vwide>2): Rename to...
11364 (aarch64_sve_extend<mode><Vwide>2): ...this.
11365 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11366 renaming the old pattern to...
11367 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11369 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11370 define_expand, renaming the old pattern to...
11371 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11372 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11373 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11374 account when deciding which SVE instruction the optab should use.
11375 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11377 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11379 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11380 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11381 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11382 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11383 (tlsdesc_small_<mode>): Turn a define_expand and use
11384 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11385 (tlsdesc_small_advsimd_<mode>): ...this.
11386 (tlsdesc_small_sve_<mode>): New pattern.
11388 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11390 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11391 (UNSPEC_UMUL_HIGHPART): New constants.
11392 (MUL_HIGHPART): New int iteraor.
11393 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11394 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11396 (*<su>mul<mode>3_highpart): New define_insn.
11398 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11401 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11404 2018-03-13 Martin Liska <mliska@suse.cz>
11407 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11409 (sem_item_optimizer::~sem_item_optimizer): Release it.
11410 (sem_item_optimizer::merge_classes): Register variable aliases.
11411 (sem_item_optimizer::fixup_pt_set): New function.
11412 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11413 * ipa-icf.h: Declare new variables and functions.
11415 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11417 PR middle-end/84834
11418 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11419 integer_pow2p@2 and test integer_pow2p in condition.
11420 (A < 0 ? C : 0): Similarly for @1.
11422 PR middle-end/84831
11423 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11424 characters starting at p contain '\0' character, don't look beyond
11428 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11429 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11432 * reg-stack.c (change_stack): Change update_end var from int to
11433 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11434 also call set_block_for_insn on the newly added insns and rescan.
11437 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11438 on the last operand.
11441 * tree.c (stabilize_reference_1): Return save_expr (e) for
11442 STATEMENT_LIST even if it doesn't have side-effects.
11444 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11446 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11448 2018-03-12 Renlin Li <renlin.li@arm.com>
11450 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11451 aarch64_output_scalar_simd_mov_immediate.
11453 2018-03-12 Martin Sebor <msebor@redhat.com>
11455 PR tree-optimization/83456
11456 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11457 for perfectly overlapping calls to memcpy.
11458 (gimple_fold_builtin_memory_chk): Same.
11459 (gimple_fold_builtin_strcpy): Handle no-warning.
11460 (gimple_fold_builtin_stxcpy_chk): Same.
11461 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11463 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11465 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11466 parameter. Use it for SFmode.
11467 (rs6000_function_arg_advance_1): Adjust.
11468 (rs6000_function_arg): Adjust.
11469 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11471 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11473 PR rtl-optimization/84169
11474 PR rtl-optimization/84780
11475 * combine.c (can_combine_p): Check for a 2-insn combination whether
11476 the destination register is used between the two insns, too.
11478 2018-03-12 Richard Biener <rguenther@suse.de>
11480 PR tree-optimization/84803
11481 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11482 for refs DR analysis didn't process.
11484 2018-03-12 Richard Biener <rguenther@suse.de>
11486 PR tree-optimization/84777
11487 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11488 force-vectorize loops ignore whether we are optimizing for size.
11490 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11492 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11493 (TARGET_MD_ASM_ADJUST): Define.
11495 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11496 Kito Cheng <kito.cheng@gmail.com>
11497 Chung-Ju Wu <jasonwucj@gmail.com>
11499 * config/nds32/nds32.c (nds32_compute_stack_frame,
11500 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11501 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11502 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11503 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11504 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11505 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11506 * config/nds32/nds32.md (prologue, epilogue): Use macro
11507 NDS32_V3PUSH_AVAILABLE_P to do checking.
11509 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11512 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11513 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11514 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11515 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11516 addition of most attributes on !orig_type_die or the attribute not
11517 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11519 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11520 Chung-Ju Wu <jasonwucj@gmail.com>
11522 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11523 __NDS32_VH__ macro.
11524 * config/nds32/nds32.opt (mvh): New option.
11526 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11527 Chung-Ju Wu <jasonwucj@gmail.com>
11529 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11531 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11532 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11535 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11536 Chung-Ju Wu <jasonwucj@gmail.com>
11538 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11540 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11541 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11543 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11544 Kito Cheng <kito.cheng@gmail.com>
11545 Chung-Ju Wu <jasonwucj@gmail.com>
11547 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11548 UNSPEC_FFMISM and UNSPEC_FLMISM.
11549 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11550 for ffb, ffmism and flmism.
11551 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11552 (unspec_ffmism): Ditto.
11553 (unspec_flmism): Ditto.
11554 (nds32_expand_builtin_impl): Check if string extension is available.
11555 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11556 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11558 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11561 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11564 * lra-assigns.c (assign_by_spills): Return a flag of reload
11565 assignment failure. Do not process the reload assignment
11566 failures. Do not spill other reload pseudos if they has the same
11568 (lra_assign): Add a return arg. Set up from the result of
11569 assign_by_spills call.
11570 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11571 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11572 usage_insns if it is not NULL.
11573 (spill_hard_reg_in_range): New function.
11574 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11575 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11576 function prototypes.
11577 (lra_assign): Change prototype.
11578 * lra.c (lra): Add code to deal with fails by splitting hard reg
11581 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11584 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11586 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11589 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11590 (dw_val_node): Add val_symbolic_view.
11591 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11592 (symview_upper_bound): New.
11593 (new_line_info_table): Initialize symviews_since_reset.
11594 (dwarf2out_source_line): Count symviews_since_reset and set
11595 symview_upper_bound.
11596 (dw_val_equal_p): Handle symview.
11597 (add_AT_symview): New.
11598 (print_dw_val): Handle symview.
11599 (attr_checksum, attr_checksum_ordered): Likewise.
11600 (same_dw_val_p, size_of_die): Likewise.
11601 (value_format, output_die): Likewise.
11602 (add_high_low_attributes): Use add_AT_symview for entry_view.
11603 (dwarf2out_finish): Reset symview_upper_bound, clear
11606 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11609 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11610 Add strict argument and use it.
11611 (rs6000_split_multireg_move): Update for new strict argument.
11612 (mem_operand_gpr): Disallow all non-offsettable addresses.
11613 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11615 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11618 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11619 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11620 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11623 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11624 decl, use remap_type if we want to use the type.
11626 2018-03-09 Martin Sebor <msebor@redhat.com>
11628 PR tree-optimization/84526
11629 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11631 (builtin_access::generic_overlap): Be prepared to handle non-array
11634 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11636 PR rtl-optimization/84682
11637 * lra-constraints.c (process_address_1): Check is_address flag
11638 for address constraints.
11639 (process_alt_operands): Likewise.
11640 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11641 preprocess_constraints.
11642 * recog.h (preprocess_constraints): Add oploc parameter.
11644 * recog.c (preprocess_constraints): Test address_operand for
11645 CT_ADDRESS constraints.
11647 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11650 * lra-assigns.c (assign_by_spills): Return a flag of reload
11651 assignment failure. Do not process the reload assignment
11652 failures. Do not spill other reload pseudos if they has the same
11654 (lra_assign): Add a return arg. Set up from the result of
11655 assign_by_spills call.
11656 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11657 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11658 usage_insns if it is not NULL.
11659 (spill_hard_reg_in_range): New function.
11660 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11661 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11662 function prototypes.
11663 (lra_assign): Change prototype.
11664 * lra.c (lra): Add code to deal with fails by splitting hard reg
11667 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11670 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11671 Accept complain bool parameter. Only emit errors if it is true.
11672 (arm_parse_cpu_option_name): Likewise.
11673 (arm_target_thumb_only): Adjust callers of the above.
11674 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11675 prototype to take a default true bool parameter.
11676 (arm_parse_arch_option_name): Likewise.
11678 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11679 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11683 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11684 * configure: Regenerate.
11685 * configure.ac ("linker --version-script option"): New.
11686 ("linker soname option"): New.
11688 2018-03-09 Richard Biener <rguenther@suse.de>
11690 PR tree-optimization/84775
11691 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11692 immediate uses of predicate stmts and mark them modified.
11695 PR tree-optimization/84178
11696 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11698 (version_loop_for_if_conversion): Delay update_ssa call.
11699 (tree_if_conversion): Delay update_ssa until after predicate
11702 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11705 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11706 when the function accesses prior frames.
11708 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11711 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11712 gen_llsym, otherwise call maybe_gen_llsym.
11714 PR inline-asm/84742
11715 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11716 has ',' character inside of it.
11718 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11721 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11722 as clobbering CC_REGNUM.
11724 2018-03-08 Richard Biener <rguenther@suse.de>
11726 PR middle-end/84552
11727 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11728 (follow_copies_to_constant): Do not follow SSA names registered
11731 2018-03-08 Richard Biener <rguenther@suse.de>
11733 PR tree-optimization/84178
11734 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11736 (version_loop_for_if_conversion): Delay update_ssa call.
11737 (tree_if_conversion): Delay update_ssa until after predicate
11740 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11742 PR tree-optimization/84178
11743 * tree-if-conv.c (release_bb_predicate): Remove the
11744 the assertion that the stmts have NULL use_ops.
11745 Discard the statements, asserting that they haven't
11746 yet been added to a BB.
11748 2018-03-08 Richard Biener <rguenther@suse.de>
11750 PR tree-optimization/84746
11751 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11752 (phi_translate): Pass in destination ANTIC_OUT set.
11753 (phi_translate_1): Likewise. For a simplified result lookup
11754 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11755 (phi_translate_set): Adjust.
11756 (do_pre_regular_insertion): Likewise.
11757 (do_pre_partial_partial_insertion): Likewise.
11759 2018-03-08 Martin Liska <mliska@suse.cz>
11761 PR gcov-profile/84735
11762 * doc/gcov.texi: Document usage of profile files.
11763 * gcov-io.h: Document changes in the format.
11765 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11769 * dwarf2out.c (struct dw_line_info_table): Update comments for
11771 (FORCE_RESET_NEXT_VIEW): New.
11772 (FORCE_RESETTING_VIEW_P): New.
11773 (RESETTING_VIEW_P): Check for -1 too.
11774 (ZERO_VIEW_P): Likewise.
11775 (new_line_info_table): Force-reset next view.
11776 (dwarf2out_begin_function): Likewise.
11777 (dwarf2out_source_line): Simplify zero_view_p initialization.
11778 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11779 view directly. Omit view when omitting .loc at line 0.
11781 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11783 PR tree-optimization/84740
11784 * tree-switch-conversion.c (process_switch): Call build_constructors
11785 only if info.phi_count is non-zero.
11787 PR tree-optimization/84739
11788 * tree-tailcall.c (find_tail_calls): Check call arguments against
11789 DECL_ARGUMENTS (current_function_decl) rather than
11790 DECL_ARGUMENTS (func) when checking for tail recursion.
11792 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11794 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11795 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11796 Volker Reichelt's entry and add entries for people that perform
11797 GCC fuzzy testing and report numerous bugs.
11799 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11802 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11803 readonly data in sdata, if that is disabled.
11804 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11805 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11806 -mreadonly-in-sdata option.
11808 2018-03-07 Martin Sebor <msebor@redhat.com>
11810 PR tree-optimization/84468
11811 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11812 basic block when looking for nul assignment.
11814 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11817 * except.h (output_function_exception_table): Adjust prototype.
11818 * except.c (output_function_exception_table): Remove FNNAME parameter
11819 and add SECTION parameter. Ouput one part of the table at a time.
11820 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11821 the first part of the exception table and emit unwind directives.
11822 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11823 (i386_pe_seh_cold_init): Likewise.
11824 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11825 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11826 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11827 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11828 * config/i386/winnt.c: Include except.h.
11829 (struct seh_frame_state): Add reg_offset, after_prologue and
11830 in_cold_section fields.
11831 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11832 (i386_pe_seh_cold_init): New function.
11833 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11834 to seh->in_cold_section.
11835 (seh_emit_push): Record the offset of the push.
11836 (seh_emit_save): Record the offet of the save.
11837 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11838 Test seh->after_prologue to disregard the epilogue.
11839 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11840 (i386_pe_end_cold_function): New function.
11842 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11845 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11846 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11849 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11851 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11852 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11854 PR middle-end/84723
11855 * multiple_target.c: Include tree-inline.h and intl.h.
11856 (expand_target_clones): Diagnose and fail if node->definition and
11857 !tree_versionable_function_p (node->decl).
11859 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11861 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11863 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11864 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11865 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11867 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11870 * combine.c (try_combine): Use reg_or_subregno instead of handling
11871 just paradoxical SUBREGs and REGs.
11873 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11875 * config/arc/arc.c (arc_finalize_pic): Remove function.
11876 (arc_must_save_register): We use single base PIC register, remove
11877 checks to save/restore the PIC register.
11878 (arc_expand_prologue): Likewise.
11879 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11881 (arc_verify_short): Likewise.
11882 (arc_attr_type): Likewise.
11883 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11884 (walk_stores): Likewise.
11885 (arc_address_cost): Make it static.
11886 (arc_verify_short): Likewise.
11887 (branch_dest): Likewise.
11888 (arc_attr_type): Likewise.
11889 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11890 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11891 (arc_final_prescan_insn): Remove inserting the nops due to
11892 hardware hazards. It is done in reorg step.
11893 (insn_length_variant_t): Remove.
11894 (insn_length_parameters_t): Likewise.
11895 (arc_insn_length_parameters): Likewise.
11896 (arc_get_insn_variants): Likewise.
11897 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11899 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11901 PR inline-asm/84683
11902 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11905 PR tree-optimization/84687
11906 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11908 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11910 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11912 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11913 Rename to ppc_speculation_barrier.
11914 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11915 __builtin_ppc_speculation_barrier.
11917 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11920 * combine.c (combine_simplify_rtx): Don't try to simplify if
11921 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11924 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11926 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11927 to 32 bytes when compiling for POWER9.
11929 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11932 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11933 regparm >= 3 with no arg reg available also for calls with
11934 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11937 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11939 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11941 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11944 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11946 2018-03-05 Richard Biener <rguenther@suse.de>
11948 PR tree-optimization/84486
11949 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11950 When inserting a __builtin_assume_aligned call set the LHS
11951 SSA name alignment info accordingly.
11953 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11955 PR tree-optimization/84114
11956 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11957 Avoid reassociation of FLOAT_MODE addition.
11959 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11961 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11962 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11963 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11964 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11965 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11966 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11967 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11969 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11970 __builtin_ia32_wbinvd): New builtins.
11971 (SPECIAL_ARGS2): New.
11972 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11973 (SPECIAL_ARGS2): New.
11974 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11975 (ix86_valid_target_attribute_inner_p): Ditto.
11976 (ix86_init_mmx_sse_builtins): Add special_args2.
11977 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11978 TARGET_WBNOINVD_P): New.
11979 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11980 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11981 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11982 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11983 * config/i386/pconfigintrin.h: New file.
11984 * config/i386/wbnoinvdintrin.h: Ditto.
11985 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11987 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11989 2018-03-05 Richard Biener <rguenther@suse.de>
11991 PR tree-optimization/84670
11992 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11994 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11995 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11996 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11997 doesn't grow if all successors (recursively) were visited at least
12000 2018-03-05 Richard Biener <rguenther@suse.de>
12002 PR tree-optimization/84650
12003 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
12004 if executed in the loop pipeline.
12006 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
12008 * doc/configfiles.texi (Configuration Files): Move info about
12009 conditionalizing $target-protos.h to...
12010 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
12011 differs from $target-protos.h.
12013 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
12014 Chung-Ju Wu <jasonwucj@gmail.com>
12016 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
12017 * config/nds32/nds32-multiple.md (setmemsi): Define.
12018 * config/nds32/nds32-memory-manipulation.c
12019 (nds32_gen_dup_4_byte_to_word_value): New.
12020 (emit_setmem_word_loop): New.
12021 (emit_setmem_byte_loop): New.
12022 (nds32_expand_setmem_loop): New.
12023 (nds32_expand_setmem_loop_v3m): New.
12024 (nds32_expand_setmem_unroll): New.
12025 (nds32_expand_setmem): New.
12027 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12028 Chung-Ju Wu <jasonwucj@gmail.com>
12030 * config/nds32/nds32-memory-manipulation.c
12031 (nds32_emit_load_store): New.
12032 (nds32_emit_post_inc_load_store): New.
12033 (nds32_emit_mem_move): New.
12034 (nds32_emit_mem_move_block): New.
12035 (nds32_expand_movmemsi_loop_unknown_size): New.
12036 (nds32_expand_movmemsi_loop_known_size): New.
12037 (nds32_expand_movmemsi_loop): New.
12038 (nds32_expand_movmemsi_unroll): New.
12039 (nds32_expand_movmemqi): Rename ...
12040 (nds32_expand_movmemsi): ... to this.
12041 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12042 (movmemsi): ... to this.
12043 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12044 (nds32_expand_movmemsi): ... to this.
12046 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12047 Monk Chiang <sh.chiang04@gmail.com>
12048 Chung-Ju Wu <jasonwucj@gmail.com>
12050 * config/nds32/nds32-protos.h
12051 (nds32_expand_load_multiple): New arguments.
12052 (nds32_expand_store_multiple): Ditto.
12053 (nds32_valid_multiple_load_store): Rename ...
12054 (nds32_valid_multiple_load_store_p): ... to this.
12055 * config/nds32/nds32-memory-manipulation.c
12056 (nds32_expand_load_multiple): Refine implementation.
12057 (nds32_expand_store_multiple): Ditto.
12058 * config/nds32/nds32-multiple.md
12059 (load_multiple): Update nds32_expand_load_multiple interface.
12060 (store_multiple): Update nds32_expand_store_multiple interface.
12061 * config/nds32/nds32-predicates.c
12062 (nds32_valid_multiple_load_store): Rename ...
12063 (nds32_valid_multiple_load_store_p): ... to this and refine
12065 * config/nds32/predicates.md
12066 (nds32_load_multiple_and_update_address_operation): New predicate.
12067 (nds32_store_multiple_and_update_address_operation): New predicate.
12069 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12070 Chung-Ju Wu <jasonwucj@gmail.com>
12072 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12073 (combo): New attribute.
12074 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12076 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12078 * config/nds32/nds32.opt: Change -mcmodel= default value.
12080 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12081 Monk Chiang <sh.chiang04@gmail.com>
12082 Chung-Ju Wu <jasonwucj@gmail.com>
12084 * config/nds32/constants.md (unspec_element): New enum.
12085 * config/nds32/constraints.md (Umw): New constraint.
12086 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12087 * config/nds32/nds32-intrinsic.md: Likewise.
12088 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12089 (nds32_valid_smw_lwm_base_p): New.
12090 (nds32_output_smw_single_word): New.
12091 (nds32_output_lmw_single_word): New.
12092 (nds32_expand_unaligned_load): New.
12093 (nds32_expand_unaligned_store): New.
12094 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12095 (nds32_output_smw_single_word): Declare.
12096 (nds32_output_lmw_single_word): Declare.
12097 (nds32_expand_unaligned_load): Declare.
12098 (nds32_expand_unaligned_store): Declare.
12099 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12100 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12101 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12102 NDS32_BUILTIN_UASTORE_DW.
12103 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12106 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12107 Kito Cheng <kito.cheng@gmail.com>
12108 Chung-Ju Wu <jasonwucj@gmail.com>
12110 * config/nds32/nds32-intrinsic.c
12111 (nds32_expand_builtin_null_ftype_reg): Delete.
12112 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12113 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12114 (nds32_read_argument): New.
12115 (nds32_legitimize_target): Ditto.
12116 (nds32_legitimize_argument): Ditto.
12117 (nds32_check_constant_argument): Ditto.
12118 (nds32_expand_unop_builtin): Ditto.
12119 (nds32_expand_unopimm_builtin): Ditto.
12120 (nds32_expand_binop_builtin): Ditto.
12121 (nds32_builtin_decl_impl): Ditto.
12122 (builtin_description): Ditto.
12123 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12124 (nds32_init_builtins_impl): Ditto.
12125 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12126 (nds32_builtin_decl): New.
12127 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12128 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12130 2018-03-02 Jeff Law <law@redhat.com>
12132 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12133 (redundant_insn, fill_simple_delay_slots): Likewise.
12134 (fill_slots_from_thread): Likewise.
12135 * resource.c (mark_referenced_resources): Likewise.
12136 (mark_set_resources, find_dead_or_set_registers): Likewise.
12138 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12140 * substring-locations.h (format_warning_va): Formatting fix for
12141 ATTRIBUTE_GCC_DIAG.
12142 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12144 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12145 * substring-locations.c: Include intl.h.
12146 (format_warning_va): Turned into small wrapper around
12147 format_warning_n_va, renamed to ...
12148 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12149 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12151 (format_warning_at_substring_n): New function.
12152 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12153 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12154 format_warning_at_substring with just a shorter name instead of
12155 const function pointer.
12156 (fmtwarn_n): New function.
12157 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12158 appropriate, get rid of all the fmtstr temporaries, move conditionals
12159 with G_() wrapped string literals directly into fmtwarn arguments,
12160 cast dir.len to (int), formatting fixes.
12162 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12164 * doc/invoke.texi: Remove "Cilk Plus" references.
12166 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12167 Richard Biener <rguenther@suse.de>
12170 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12171 for error or warning attributes if CALL_FROM_THUNK_P is set.
12174 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12177 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12178 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12181 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12182 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12184 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12185 instead of -1U in last predictors element's probability member.
12187 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12190 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12191 arguments if they are comparable.
12193 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12195 PR tree-optimization/84634
12196 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12197 masks and masked_loop_p with a single loop_masks, making sure it's
12198 null for bb vectorization.
12200 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12202 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12203 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12204 loop->force_vectorize to check whether there is no alias.
12206 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12209 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12211 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12213 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12214 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12215 instead of a loop around prev_real_insn.
12216 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12219 PR inline-asm/84625
12220 * config/i386/i386.c (ix86_print_operand): Use conditional
12221 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12224 2018-03-02 Richard Biener <rguenther@suse.de>
12226 PR tree-optimization/84427
12227 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12228 (bitmap_set_subtract_values): Rewrite to handle multiple
12231 (prune_clobbered_mems): Likewise.
12232 (phi_translate): Take edge instead of pred/phiblock.
12233 (phi_translate_1): Likewise.
12234 (phi_translate_set): Likewise. Insert all translated
12235 exprs for a value into the set, keeping possibly multiple
12236 expressions per value.
12237 (compute_antic_aux): Adjust for phi_translate changes.
12238 When intersecting union the expressions and prune those
12239 not in the final value set, keeping possibly multiple
12240 expressions per value. Do not use value-insertion
12241 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12242 all expressions. Add verification that the value-sets
12243 only shrink during iteration.
12244 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12245 (do_pre_regular_insertion): Likewise.
12246 (do_pre_partial_partial_insertion): Likewise.
12248 2018-03-02 Richard Biener <rguenther@suse.de>
12251 * config/darwin.c (saved_debug_info_level): New static global.
12252 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12253 (darwin_asm_lto_end): Restore debug info generation settings.
12255 2018-03-01 Martin Liska <mliska@suse.cz>
12258 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12259 volatile arguments.
12261 2018-03-01 Richard Biener <rguenther@suse.de>
12264 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12265 type annotation with LTO when debug was disabled at compile-time.
12267 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12269 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12271 (mips_final_postscan_insn): Likewise.
12273 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12275 PR rtl-optimization/84528
12276 * alias.c (init_alias_target): Add commentary.
12277 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12278 a unique base value if the frame pointer is not eliminated
12279 to the stack pointer.
12281 2018-03-01 Tom de Vries <tom@codesourcery.com>
12283 PR rtl-optimization/83327
12284 * lra-int.h (hard_regs_spilled_into): Declare.
12285 * lra.c (hard_regs_spilled_into): Define.
12286 (init_reg_info): Init hard_regs_spilled_into.
12287 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12288 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12289 (process_bb_lives): Handle hard_regs_spilled_into.
12290 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12291 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12293 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12295 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12296 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12297 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12298 * config/rs6000/aix72.h: New file.
12300 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12302 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12303 instead of warning_at with conditional singular and plural messages
12307 * stor-layout.c (update_alignment_for_field): For
12308 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12309 && !DECL_PACKED (field), do the alignment update, just use
12310 only desired_align instead of MAX (type_align, desired_align)
12312 (place_field): Don't do known_align < desired_align handling
12313 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12314 is non-NULL, instead do it after rli->prev_field handling and
12315 only if not within a bitfield word. For DECL_PACKED (field)
12316 use type_align of BITS_PER_UNIT.
12318 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12320 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12321 superfluous parentheses and trailing spaces.
12323 2018-02-28 Richard Biener <rguenther@suse.de>
12325 PR tree-optimization/84584
12326 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12327 SCoPs with fake exit edge.
12329 2018-02-28 Martin Liska <mliska@suse.cz>
12332 * timevar.c (timer::print): Fix format to properly print 100%
12335 2018-02-28 Richard Biener <rguenther@suse.de>
12337 PR middle-end/84607
12338 * genmatch.c (capture_info::walk_match): Do not mark
12339 captured expressions without operands as expr_p given
12340 they act more like predicates and should be subject to
12341 "lost tail" side-effect preserving.
12343 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12345 PR rtl-optimization/81611
12346 * auto-inc-dec.c (attempt_change): Move dead note from
12347 mem_insn if it's the next use of regno
12348 (find_address): Take address use of reg holding
12349 non-incremented value. Add parm to limit search to the named
12351 (merge_in_block): Attempt to use a mem insn that is the next
12352 use of the original regno.
12354 2018-02-27 Martin Sebor <msebor@redhat.com>
12357 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12358 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12360 2018-02-27 Martin Sebor <msebor@redhat.com>
12362 PR translation/84207
12363 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12364 n argument to unsigned HOST_WIDE_INT.
12365 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12366 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12367 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12368 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12370 2018-02-27 Richard Biener <rguenther@suse.de>
12372 PR tree-optimization/84512
12373 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12374 Do not use the estimate returned from record_stmt_cost for
12375 the scalar iteration cost but sum properly using add_stmt_cost.
12377 2018-02-27 Richard Biener <rguenther@suse.de>
12379 PR tree-optimization/84466
12380 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12381 Adjust last change to less strictly validate use operands.
12383 2018-02-27 Martin Liska <mliska@suse.cz>
12385 PR gcov-profile/84548
12386 * gcov.c (process_file): Allow partial overlap and consider it
12387 also as group functions.
12388 (output_lines): Properly calculate range of lines for a group.
12390 2018-02-27 Martin Liska <mliska@suse.cz>
12392 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12393 'ggc' suffixes. Change first column width.
12394 (timer::print): Fix formatting of the column.
12396 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12398 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12399 preserve inline entry blocks for the sake of debug inline
12400 entry point markers alone.
12401 (remove_unused_locals): Suggest in comments a better place to
12402 force the preservation of inline entry blocks that are
12403 otherwise unused, but do not preserve them.
12405 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12407 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12409 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12412 * config/i386/constraints.md (Bs): Replace
12413 ix86_indirect_branch_register with
12414 TARGET_INDIRECT_BRANCH_REGISTER.
12416 * config/i386/i386.md (indirect_jump): Likewise.
12417 (tablejump): Likewise.
12418 (*sibcall_memory): Likewise.
12419 (*sibcall_value_memory): Likewise.
12420 Peepholes of indirect call and jump via memory: Likewise.
12421 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12422 (*sibcall_value_GOT_32): Likewise.
12423 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12424 (GOT_memory_operand): Likewise.
12425 (call_insn_operand): Likewise.
12426 (sibcall_insn_operand): Likewise.
12427 (GOT32_symbol_operand): Likewise.
12428 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12430 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12432 PR rtl-optimization/83496
12433 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12434 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12435 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12436 redundant insn, if any.
12437 (relax_delay_slots): Likewise.
12438 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12440 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12442 PR tree-optimization/83965
12443 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12444 that grouped statements are part of a reduction chain. Return
12445 true if the statement is not marked as a reduction itself but
12446 is part of a group.
12447 (vect_recog_dot_prod_pattern): Don't check whether the statement
12448 is part of a group here.
12449 (vect_recog_sad_pattern): Likewise.
12450 (vect_recog_widen_sum_pattern): Likewise.
12452 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12455 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12457 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12460 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12462 (ix86_output_indirect_function_return): New prototype.
12463 (ix86_split_simple_return_pop_internal): Likewise.
12464 * config/i386/i386.c (indirect_return_via_cx): New.
12465 (indirect_return_via_cx_bnd): Likewise.
12466 (indirect_thunk_name): Handle return va CX_REG.
12467 (output_indirect_thunk_function): Create alias for
12468 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12469 (ix86_output_indirect_jmp): Remove the bool argument.
12470 (ix86_output_indirect_function_return): New function.
12471 (ix86_split_simple_return_pop_internal): Likewise.
12472 * config/i386/i386.md (*indirect_jump): Don't pass false
12473 to ix86_output_indirect_jmp.
12474 (*tablejump_1): Likewise.
12475 (simple_return_pop_internal): Change it to define_insn_and_split.
12476 Call ix86_split_simple_return_pop_internal to split it for
12477 -mfunction-return=.
12478 (simple_return_indirect_internal): Call
12479 ix86_output_indirect_function_return instead of
12480 ix86_output_indirect_jmp.
12482 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12485 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12486 memset and value initialization afterwards.
12488 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12490 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12492 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12495 * common/config/aarch64/aarch64-common.c
12496 (aarch_option_optimization_table[]): Switch
12497 off fomit-frame-pointer
12499 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12500 Chung-Ju Wu <jasonwucj@gmail.com>
12502 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12504 (store_multiple): Ditto.
12506 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12508 * config.gcc: Add --with-cpu support for nds32 target.
12509 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12510 * config/nds32/nds32.opt: Add -mcpu= option.
12512 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12514 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12515 isel=yes): Warn for these deprecated options.
12517 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12519 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12520 ISA_2_5_MASKS_EMBEDDED.
12522 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12524 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12525 p->max as pointers rather than using iterative_hash_expr.
12527 2018-02-23 Carl Love <cel@us.ibm.com>
12529 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12530 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12532 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12533 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12534 P8V_BUILTIN_VEC_VUNSIGNED2.
12536 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12539 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12540 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12542 * lra-constraints.c (curr_insn_transform): Set up
12543 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12545 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12546 * lra-eliminations.c (spill_pseudos): Ditto.
12547 (process_insn_for_elimination): Ditto.
12548 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12549 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12550 LRA_NON_CLOBBERED_ALT.
12552 2018-02-22 Martin Sebor <msebor@redhat.com>
12554 PR tree-optimization/84480
12555 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12556 to maybe_diag_stxncpy_trunc. Call it.
12557 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12558 from gimple_fold_builtin_strcpy. Print inlining stack.
12559 (handle_builtin_stxncpy): Print inlining stack.
12560 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12562 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12565 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12566 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12567 and -fcheck-pointer-bounds are used together.
12568 (indirect_thunk_prefix): New enum.
12569 (indirect_thunk_need_prefix): New function.
12570 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12571 "_nt" instead of "_bnd" for NOTRACK prefix.
12572 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12573 (output_indirect_thunk_function): Likewise.
12575 (ix86_code_end): Update output_indirect_thunk_function calls.
12576 (ix86_output_indirect_branch_via_reg): Replace
12577 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12578 (ix86_output_indirect_branch_via_push): Likewise.
12579 (ix86_output_function_return): Likewise.
12580 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12581 incompatible with -fcf-protection=branch and
12582 -fcheck-pointer-bounds.
12584 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12587 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12588 Change gcc_assert call to output_operand_lossage.
12590 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12592 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12594 2018-02-22 DJ Delorie <dj@redhat.com>
12595 Sebastian Perta <sebastian.perta@renesas.com>
12596 Oleg Endo <olegendo@gcc.gnu.org>
12598 * config/rx/rx.c (rx_rtx_costs): New function.
12599 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12601 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12603 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12605 2018-02-22 Martin Liska <mliska@suse.cz>
12608 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12609 Add "native" as a possible value.
12611 2018-02-22 Martin Liska <mliska@suse.cz>
12614 * config/i386/i386.c (ix86_option_override_internal):
12615 Add "native" as a possible value for -march and -mtune.
12617 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12620 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12621 to all type variants.
12623 PR tree-optimization/84503
12624 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12625 width as info->bitpos + info->bitsize - start.
12626 (merged_store_group::merge_overlapping): Simplify width computation.
12627 (check_no_overlap): New function.
12628 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12629 start + width and last_order of the group, fail if check_no_overlap
12631 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12632 to group if check_no_overlap fails.
12634 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12636 * config/rs6000/altivec.md: Delete contraint arguments to
12637 define_expand, define_split, and define_peephole2, and in
12638 define_insn_and_split if always unused.
12639 * config/rs6000/darwin.md: Ditto.
12640 * config/rs6000/dfp.md: Ditto.
12641 * config/rs6000/rs6000.md: Ditto.
12642 * config/rs6000/sync.md: Ditto.
12643 * config/rs6000/vector.md: Ditto.
12644 * config/rs6000/vsx.md: Ditto.
12646 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12648 * config/rs6000/altivec.md: Write output control strings as braced
12649 blocks instead of double-quoted strings.
12650 * config/rs6000/darwin.md: Ditto.
12651 * config/rs6000/rs6000.md: Ditto.
12652 * config/rs6000/vector.md: Ditto.
12653 * config/rs6000/vsx.md: Ditto.
12655 2018-02-21 Jason Merrill <jason@redhat.com>
12657 PR c++/84314 - ICE with templates and fastcall attribute.
12658 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12660 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12662 * ipa-cp.c (determine_versionability): Fix comment typos.
12664 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12667 * ipa-cp.c (determine_versionability): Do not version functions caling
12670 2018-02-21 Martin Liska <mliska@suse.cz>
12673 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12674 Add "native" as a possible value.
12675 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12676 the macro when native cpu detection is available.
12678 2018-02-21 Martin Liska <mliska@suse.cz>
12681 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12682 Add "native" as a possible value.
12683 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12684 when native cpu detection is available.
12686 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12687 Martin Sebor <msebor@redhat.com>
12689 PR tree-optimization/84478
12690 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12692 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12693 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12694 support which is conservatively correct, for 2 only stay conservative
12695 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12696 argument to the 2 argument get_range_strlen, adjust 6 arg
12697 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12699 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12700 (gimple_fold_builtin_strlen): Pass true as last argument to
12703 2018-02-20 Martin Sebor <msebor@redhat.com>
12705 PR middle-end/84095
12706 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12707 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12708 (builtin_memref::builtin_memref): Factor out parts into
12709 set_base_and_offset and call it.
12711 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12713 PR middle-end/84406
12714 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12715 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12716 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12717 search at the associated MODE_INT.
12719 2018-02-20 Jeff Law <law@redhat.com>
12721 PR middle-end/82123
12722 PR tree-optimization/81592
12723 PR middle-end/79257
12724 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12725 for range data rather than using global data.
12726 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12727 range data rather than using global data.
12728 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12729 pass it to children as needed.
12730 (struct directive::fmtresult): Similarly.
12731 (struct directive::set_width): Similarly.
12732 (struct directive::set_precision): Similarly.
12733 (format_integer, format_directive, parse_directive): Similarly.
12734 (format_none): Accept unnamed vr_values parameter.
12735 (format_percent, format_floating, format_character): Similarly.
12736 (format_string, format_plain): Similarly.
12737 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12738 the EVRP range analyzer for range data rather than using global data.
12739 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12740 gimple-ssa-evrp-analyze.h
12741 (class sprintf_dom_walker): Add after_dom_children member function.
12742 Add evrp_range_analyzer member.
12743 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12744 range analyzer as needed.
12745 (sprintf_dom_walker::after_dom_children): New member function.
12746 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12748 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12749 (evrp_range_analyzer::pop_to_marker): Likewise.
12751 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12753 PR tree-optimization/84419
12754 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12755 with the required type if its current type is compatible but
12758 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12760 PR middle-end/82004
12761 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12762 after vectorization.
12764 2018-02-20 Martin Liska <mliska@suse.cz>
12767 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12768 possible values if we don't have a hint.
12770 2018-02-20 Martin Liska <mliska@suse.cz>
12774 * final.c (shorten_branches): Build align_tab array with one
12776 * opts.c (finish_options): Add alignment option limit check.
12777 (MAX_CODE_ALIGN): Likewise.
12778 (MAX_CODE_ALIGN_VALUE): Likewise.
12779 * doc/invoke.texi: Document maximum allowed option value for
12780 all -falign-* options.
12782 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12785 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12786 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12787 * var-tracking.c (emit_note_insn_var_location): Remove all references
12788 to NOTE_INSN_CALL_ARG_LOCATION.
12789 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12790 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12791 Use copy_rtx_if_shared.
12792 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12793 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12794 (dwarf2out_var_location): Remove handling of
12795 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12797 * final.c (final_scan_insn): Remove all references to
12798 NOTE_INSN_CALL_ARG_LOCATION.
12799 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12800 before dumping final insns.
12801 * except.c (emit_note_eh_region_end): Remove all references to
12802 NOTE_INSN_CALL_ARG_LOCATION.
12803 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12804 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12805 * config/arc/arc.c (hwloop_optimize): Likewise.
12806 * config/arm/arm.c (create_fix_barrier): Likewise.
12807 * config/s390/s390.c (s390_chunkify_start): Likewise.
12808 * config/sh/sh.c (find_barrier): Likewise.
12809 * config/i386/i386.c (rest_of_insert_endbranch,
12810 ix86_seh_fixup_eh_fallthru): Likewise.
12811 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12812 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12813 * config/frv/frv.c (frv_function_prologue): Likewise.
12814 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12816 (note_outside_basic_block_p): Remove all references to
12817 NOTE_INSN_CALL_ARG_LOCATION.
12818 * gengtype.c (adjust_field_rtx_def): Likewise.
12819 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12821 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12822 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12825 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12828 PR tree-optimization/84452
12829 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12830 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12833 2018-02-19 Martin Liska <mliska@suse.cz>
12836 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12838 2018-02-19 Martin Liska <mliska@suse.cz>
12839 Richard Sandiford <richard.sandiford@linaro.org>
12841 PR tree-optimization/82491
12842 * gimple-fold.c (get_base_constructor): Make earlier bail out
12845 2018-02-19 Carl Love <cel@us.ibm.com>
12847 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12848 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12850 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12851 P8V_BUILTIN_VEC_NEG.
12853 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12855 * config/rl78/rl78.md (movdf): New define expand.
12857 2018-02-19 Martin Liska <mliska@suse.cz>
12860 * doc/invoke.texi: Fix typo.
12861 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12863 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12865 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12866 handle rs6000_single_float and rs6000_double_float specially for
12869 2018-02-16 Jeff Law <law@redhat.com>
12871 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12872 (pop_regs): Use it.
12874 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12877 * ipa-inline.c (inline_small_functions): Fix a typo.
12879 2018-02-16 Nathan Sidwell <nathan@acm.org>
12881 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12883 2018-02-16 Carl Love <cel@us.ibm.com>
12885 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12886 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12887 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12888 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12889 expansion to P8V_BUILTIN_VEC_FLOAT2.
12891 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12893 PR rtl-optimization/70023
12894 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12895 src_regno into account.
12897 2018-02-16 Carl Love <cel@us.ibm.com>
12899 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12900 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12901 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12902 * config/rs6000/rs6000.c: Remove case statements for
12903 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12904 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12905 and P9V_BUILTIN_VEC_VINSERT4B.
12906 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12907 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12908 * config/rs6000/vsx.md:
12909 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12912 2018-02-16 Carl Love <cel@us.ibm.com>
12914 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12916 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12918 * config/rs6000/rs6000-c.c: Add the definitions for
12919 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12920 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12921 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12922 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12923 definition for insert4b and define insn *insert3b_internal.
12924 * doc/extend.texi: Add documentation for vec_extract4b.
12926 2018-02-16 Nathan Sidwell <nathan@acm.org>
12928 * doc/extend.texi (Backwards Compatibility): Mention friend
12929 injection. Note for-scope is deprecated.
12930 * doc/invoke.texi (-ffriend-injection): Deprecate.
12932 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12934 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12935 that moved to I2, also allow destinations that are a paradoxical
12936 subreg (instead of a normal reg).
12938 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12941 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12944 2018-02-16 Richard Biener <rguenther@suse.de>
12946 PR tree-optimization/84037
12947 PR tree-optimization/84016
12949 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12950 Adjust vec_construct for the fact we need additional higher latency
12951 128bit inserts for AVX256 and AVX512 vector builds.
12952 (ix86_add_stmt_cost): Scale vector construction cost for
12955 2018-02-16 Richard Biener <rguenther@suse.de>
12957 PR tree-optimization/84417
12958 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12959 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12960 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12962 2018-02-16 Martin Liska <mliska@suse.cz>
12965 * internal-fn.def (ASAN_CHECK): Set proper flags.
12966 (ASAN_MARK): Likewise.
12968 2018-02-16 Julia Koval <julia.koval@intel.com>
12970 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12971 from PTA_CANNONLAKE.
12973 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12976 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12977 Use ++iter rather than iter++ for std::list iterators.
12978 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12979 defer deleting them until all nodes in the forest are processed. Do
12980 free even leaf nodes. Change to_process into auto_vec.
12983 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12984 * vec.h (vec_default_construct): Use memset instead of placement new
12985 if BROKEN_VALUE_INITIALIZATION is defined.
12986 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12987 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12990 PR rtl-optimization/83723
12991 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12992 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12993 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12995 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12997 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12999 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
13001 PR rtl-optimization/81443
13002 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
13003 from inner REGs to paradoxical SUBREGs.
13005 2018-02-16 Richard Biener <rguenther@suse.de>
13007 PR tree-optimization/84399
13008 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
13009 For operands we can analyze at their definition make sure we can
13010 analyze them at each use as well.
13012 2018-02-16 Richard Biener <rguenther@suse.de>
13014 PR tree-optimization/84190
13015 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
13016 volatile accesses if the decl isn't volatile.
13018 2018-02-15 Jason Merrill <jason@redhat.com>
13020 PR c++/84314 - ICE with templates and fastcall attribute.
13021 * attribs.c (build_type_attribute_qual_variant): Don't clobber
13022 TYPE_CANONICAL on an existing type.
13024 2018-02-15 Jakub Jelinek <jakub@redhat.com>
13026 PR tree-optimization/84383
13027 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
13028 dstoff nor call operand_equal_p if dstbase is NULL.
13030 PR tree-optimization/84334
13031 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
13032 also a CONSTANT_CLASS_P, punt.
13034 2018-02-14 Jim Wilson <jimw@sifive.com>
13036 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13037 first SMALL_OPERAND check. New local min_second_step. Move assert
13038 to where locals are set. Add TARGET_RVC support.
13039 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13041 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13043 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13045 2018-02-14 Martin Sebor <msebor@redhat.com>
13047 PR tree-optimization/83698
13048 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13049 arrays constrain the offset range to their bounds.
13050 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13051 (builtin_access::overlap): Avoid setting the size of overlap if it's
13053 (maybe_diag_overlap): Also consider arrays when deciding what values
13054 of offsets to include in diagnostics.
13056 2018-02-14 Martin Sebor <msebor@redhat.com>
13059 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13060 that correspond to the kind of a declaration.
13062 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13065 * config/pa/pa.md: Load address of PIC label using the linkage table
13066 if the label is nonlocal.
13068 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13070 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13071 warning message if user requests -maltivec=be.
13072 * doc/invoke.texi: Document deprecation of -maltivec=be.
13074 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13077 * config/rs6000/rs6000-c.c: Update definitions for
13078 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13079 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13081 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13084 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13085 add _get_ssp intrinsics. Remove argument from
13086 __builtin_ia32_rdssp[d|q].
13087 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13088 * config/i386/i386-builtin.def: Remove argument from
13089 __builtin_ia32_rdssp[d|q].
13090 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13091 ix86_expand_special_args_builtin for _rdssp[d|q].
13092 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13093 Clear register before usage.
13094 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13095 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13097 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13099 PR tree-optimization/84357
13100 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13101 operand 1 of an ARRAY_REF too.
13103 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13106 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13107 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13109 (set_of_reg): New struct.
13110 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13111 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13112 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13114 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13115 Split into bitclr, bitset, bitinvert patterns if appropriate.
13116 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13117 use rx_fuse_in_memory_bitop.
13118 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13119 to named insn, correct maximum insn length.
13121 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13124 * machmode.def: Define a complex mode for PARTIAL_INT.
13125 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13127 * doc/rtl.texi: Document CSPImode.
13128 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13130 (msp430_hard_regno_nregs_with_padding): Likewise.
13132 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13135 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13137 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13139 PR rtl-optimization/84169
13140 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13141 we generated a parallel as new i3 and we split that to new i2 and i3
13142 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13143 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13144 those to i2, not i1. Partially rewrite this scan code.
13146 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13149 * stor-layout.c (place_field): For variable length fields, adjust
13150 offset_align afterwards not just based on the field's alignment,
13151 but also on the size.
13153 PR middle-end/84309
13154 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13155 of exps and logs in the use_exp2 case.
13157 2018-02-13 Jeff Law <law@redhat.com>
13159 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13160 entry for "vector".
13162 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13165 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13169 * common.opt (gas-loc-support, gas-locview-support): New.
13170 (ginline-points, ginternal-reset-location-views): New.
13171 * doc/invoke.texi: Document them. Use @itemx where intended.
13172 (gvariable-location-views): Adjust.
13173 * target.def (reset_location_view): New.
13174 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13175 (TARGET_RESET_LOCATION_VIEW): New.
13176 * doc/tm.texi: Rebuilt.
13177 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13178 (dwarf2out_default_as_locview_support): New.
13179 (output_asm_line_debug_info): Use option variables.
13180 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13181 (output_loc_list): Likewise.
13182 (add_high_low_attributes): Check option variables.
13183 Don't output entry view attribute in strict mode.
13184 (gen_inlined_subroutine_die): Check option variables.
13185 (dwarf2out_inline_entry): Likewise.
13186 (init_sections_and_labels): Likewise.
13187 (dwarf2out_early_finish): Likewise.
13188 (maybe_reset_location_view): New, from...
13189 (dwarf2out_var_location): ... here. Call it.
13190 * debug.h (dwarf2out_default_as_loc_support): Declare.
13191 (dwarf2out_default_as_locview_support): Declare.
13192 * hooks.c (hook_int_rtx_insn_0): New.
13193 * hooks.h (hook_int_rtx_insn_0): Declare.
13194 * toplev.c (process_options): Take -gas-loc-support and
13195 -gas-locview-support from dwarf2out. Enable
13196 -gvariable-location-views by default only with locview
13197 assembler support. Enable -ginternal-reset-location-views by
13198 default only if the target defines the corresponding hook.
13199 Enable -ginline-points by default if location views are
13200 enabled; force it disabled if statement frontiers are
13202 * tree-inline.c (expand_call_inline): Check option variables.
13203 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13205 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13207 PR tree-optimization/84321
13208 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13209 handling. Also check whether the anti-range contains any values
13210 that satisfy the mask; switch to a VR_RANGE if not.
13212 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13215 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13217 2018-02-13 Martin Jambor <mjambor@suse.cz>
13220 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13221 of call statements, also set location of a load to a temporary.
13223 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13225 * config/rl78/rl78.c (add_vector_labels): New function.
13226 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13227 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13228 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13229 which checks that no arguments are passed.
13230 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13231 * doc/extend.texi: Documentation for the new attribute.
13233 2018-02-13 Andreas Schwab <schwab@suse.de>
13235 * config/riscv/linux.h (CPP_SPEC): Define.
13237 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13240 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13241 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13242 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13243 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13244 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13245 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13246 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13247 temporarily for AES and PCLMUL builtins.
13249 PR tree-optimization/84339
13250 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13251 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13254 PR middle-end/84309
13255 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13256 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13257 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13259 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13261 * omp-simd-clone.h: New file.
13262 * omp-simd-clone.c: Include omp-simd-clone.h.
13263 (expand_simd_clones): No longer static.
13264 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13265 cgraph.h and omp-simd-clone.h.
13266 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13267 (vect_recog_widen_shift_pattern): Formatting fix.
13268 (vect_pattern_recog_1): Don't check optab for calls.
13271 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13272 operands[2] into a REG before using gen_lowpart on it.
13274 2018-02-12 Jeff Law <law@redhat.com>
13277 * config/sh/sh.c (find_barrier): Consider a sibling call
13280 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13281 successfully back substituting a reg.
13283 2018-02-12 Richard Biener <rguenther@suse.de>
13285 PR tree-optimization/84037
13286 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13287 parameter, move visited init to caller.
13288 (vect_slp_analyze_operations): Separate cost from validity
13289 check, initialize visited once for all instances.
13290 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13292 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13294 (vect_model_promotion_demotion_cost): Likewise.
13295 (vectorizable_bswap): Guard cost modeling with !slp_node
13296 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13298 (vectorizable_call): Likewise.
13299 (vectorizable_conversion): Likewise.
13300 (vectorizable_assignment): Likewise.
13301 (vectorizable_shift): Likewise.
13302 (vectorizable_operation): Likewise.
13303 (vectorizable_store): Likewise.
13304 (vectorizable_load): Likewise.
13305 (vectorizable_condition): Likewise.
13306 (vectorizable_comparison): Likewise.
13308 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13311 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13312 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13313 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13315 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13317 PR middle-end/83665
13318 * params.def (inline-min-speedup): Increase from 8 to 15.
13319 (max-inline-insns-auto): Decrease from 40 to 30.
13320 * ipa-split.c (consider_split): Add some buffer for function to
13321 be considered inlining candidate.
13322 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13325 2018-02-12 Richard Biener <rguenther@suse.de>
13327 PR tree-optimization/84037
13328 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13329 matched stmts if we cannot swap the non-matched ones.
13331 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13333 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13334 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13335 _mm_maskz_scalef_round_ss): New intrinsics.
13336 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13337 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13338 __builtin_ia32_scalefss_round): Remove.
13339 (__builtin_ia32_scalefsd_mask_round,
13340 __builtin_ia32_scalefss_mask_round): New intrinsics.
13341 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13342 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13343 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13344 "<round_constraint>")): Changed to ...
13345 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13346 "<round_scalar_constraint>")): ... this.
13347 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13348 %0, %1, %2<round_op3>}"): Changed to ...
13349 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13350 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13351 %2<round_scalar_mask_op3>}"): ... this.
13352 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13354 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13356 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13357 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13358 (_mm_maskz_sqrt_round_ss): New intrinsics.
13359 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13360 (__builtin_ia32_sqrtsd_mask_round)
13361 (__builtin_ia32_sqrtss_mask_round): New builtins.
13362 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13363 (__builtin_ia32_sqrtss_round): Remove.
13364 (__builtin_ia32_sqrtsd_mask_round)
13365 (__builtin_ia32_sqrtss_mask_round): New builtins.
13366 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13367 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13368 ((match_operand:VF_128 1 "vector_operand"
13369 "xBm,<round_constraint>")): Changed to ...
13370 ((match_operand:VF_128 1 "vector_operand"
13371 "xBm,<round_scalar_constraint>")): ... this.
13372 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13373 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13374 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13375 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13376 %<iptr>1<round_scalar_mask_op3>}): ... this.
13377 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13378 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13380 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13383 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13384 Cast vec_cmpeq result to correct type.
13385 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13386 Cast vec_cmpgt result to correct type.
13388 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13390 * final.c (final_scan_insn_1): Renamed from...
13391 (final_scan_insn): ... this. New wrapper, to recover
13392 seen from the outermost call in recursive ones.
13393 * config/sparc/sparc.c (output_return): Drop seen from call.
13394 (output_sibcall): Likewise.
13395 * config/visium/visium.c (output_branch): Likewise.
13397 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13399 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13402 2018-02-10 Alan Modra <amodra@gmail.com>
13405 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13406 Specify LR as an input.
13408 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13411 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13412 remove_member_access_dummy_vars): New functions.
13413 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13414 lower_omp_1, execute_lower_omp): Use them.
13416 PR rtl-optimization/84308
13417 * shrink-wrap.c (spread_components): Release todo vector.
13419 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13421 PR rtl-optimization/57193
13422 * ira-color.c (struct allocno_color_data): Add member
13423 conflict_allocno_hard_prefs.
13424 (update_conflict_allocno_hard_prefs): New.
13425 (bucket_allocno_compare_func): Add a preference based on
13426 conflict_allocno_hard_prefs.
13427 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13428 (color_allocnos): Remove a dead code. Initiate
13429 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13431 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13434 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13435 constraint from =wa to wa. Avoid a subreg on the output operand,
13436 instead use a pseudo and subreg it in a move.
13437 (p9_xxbrd_<mode>): Changed to ...
13438 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13439 (p9_xxbrd_v2df): New expander.
13440 (p9_xxbrw_<mode>): Changed to ...
13441 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13442 (p9_xxbrw_v4sf): New expander.
13444 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13446 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13448 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13451 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13452 multiply in 32-bit mode.
13453 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13454 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13457 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13459 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13460 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13461 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13462 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13464 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13467 * dwarf2out.c (is_trivial_indirect_ref): New function.
13468 (dwarf2out_late_global_decl): Do not generate a location
13469 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13470 and that are not defined in the current unit.
13472 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13474 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13475 instead of a libcall for UNORDERED.
13477 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13480 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13481 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13483 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13486 * config/s390/s390.c (s390_set_current_function): Invoke
13487 s390_indirect_branch_settings also if fndecl didn't change.
13489 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13491 * config/rs6000/rs6000.md (blockage): Set length to zero.
13493 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13495 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13497 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13500 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13501 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13505 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13506 PARALLEL incoming that failed vt_get_decl_and_offset check.
13508 PR middle-end/84237
13509 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13510 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13512 (get_variable_section): For decls in named .bss* sections pass true as
13513 second argument to bss_initializer_p.
13515 2018-02-09 Marek Polacek <polacek@redhat.com>
13516 Jakub Jelinek <jakub@redhat.com>
13519 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13520 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13521 Sync some changes from cxx_fold_indirect_ref.
13523 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13525 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13527 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13528 (BLOCK_INLINE_ENTRY_LABEL): New.
13529 (dwarf2out_var_location): Disregard inline entry markers.
13530 (inline_entry_data): New struct.
13531 (inline_entry_data_hasher): New hashtable type.
13532 (inline_entry_data_hasher::hash): New.
13533 (inline_entry_data_hasher::equal): New.
13534 (inline_entry_data_table): New variable.
13535 (add_high_low_attributes): Add DW_AT_entry_pc and
13536 DW_AT_GNU_entry_view attributes if a pending entry is found
13537 in inline_entry_data_table. Add old entry_pc attribute only
13538 if debug nonbinding markers are disabled.
13539 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13540 markers are enabled.
13541 (block_within_block_p, dwarf2out_inline_entry): New.
13542 (dwarf2out_finish): Check that no entries remained in
13543 inline_entry_data_table.
13544 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13545 (final_scan_insn, notice_source_line): Likewise.
13546 (rest_of_clean_state): Skip inline entry markers.
13547 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13549 * gimple.c (gimple_build_debug_inline_entry): New.
13550 * gimple.h (enum gimple_debug_subcode): Add
13551 GIMPLE_DEBUG_INLINE_ENTRY.
13552 (gimple_build_debug_inline_entry): Declare.
13553 (gimple_debug_inline_entry_p): New.
13554 (gimple_debug_nonbind_marker_p): Adjust.
13555 * insn-notes.def (INLINE_ENTRY): New.
13556 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13557 inline entry marker notes.
13558 (print_insn): Likewise.
13559 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13560 (INSN_DEBUG_MARKER_KIND): Likewise.
13561 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13562 * tree-inline.c (expand_call_inline): Build and insert
13563 debug_inline_entry stmt.
13564 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13565 inline entry blocks early, if nonbind markers are enabled.
13566 (dump_scope_block): Dump fragment info.
13567 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13568 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13569 (gimple_build_debug_inline_entry): New.
13570 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13571 Enable/disable inline entry points too.
13572 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13573 (DEBUG_INSN): Describe inline entry markers.
13575 * common.opt (gvariable-location-views): New.
13576 (gvariable-location-views=incompat5): New.
13577 * config.in: Rebuilt.
13578 * configure: Rebuilt.
13579 * configure.ac: Test assembler for view support.
13580 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13581 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13582 * dwarf2out.c (var_loc_view): New typedef.
13583 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13584 (dwarf2out_locviews_in_attribute): New.
13585 (dwarf2out_locviews_in_loclist): New.
13586 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13587 (enum dw_line_info_opcode): Add LI_adv_address.
13588 (struct dw_line_info_table): Add view.
13589 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13590 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13591 (zero_view_p): New variable.
13592 (ZERO_VIEW_P): New macro.
13593 (output_asm_line_debug_info): New.
13594 (struct var_loc_node): Add view.
13595 (add_AT_view_list, AT_loc_list): New.
13596 (add_var_loc_to_decl): Add view param. Test it against last.
13597 (new_loc_list): Add view params. Record them.
13598 (AT_loc_list_ptr): Handle loc and view lists.
13599 (view_list_to_loc_list_val_node): New.
13600 (print_dw_val): Handle dw_val_class_view_list.
13601 (size_of_die): Likewise.
13602 (value_format): Likewise.
13603 (loc_list_has_views): New.
13604 (gen_llsym): Set vl_symbol too.
13605 (maybe_gen_llsym, skip_loc_list_entry): New.
13606 (dwarf2out_maybe_output_loclist_view_pair): New.
13607 (output_loc_list): Output view list or entries too.
13608 (output_view_list_offset): New.
13609 (output_die): Handle dw_val_class_view_list.
13610 (output_dwarf_version): New.
13611 (output_compilation_unit_header): Use it.
13612 (output_skeleton_debug_sections): Likewise.
13613 (output_rnglists, output_line_info): Likewise.
13614 (output_pubnames, output_aranges): Update version comments.
13615 (output_one_line_info_table): Output view numbers in asm comments.
13616 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13617 Call maybe_gen_llsym.
13618 (loc_list_from_tree_1): Adjust.
13619 (add_AT_location_description): Create view list attribute if
13620 needed, check it's absent otherwise.
13621 (convert_cfa_to_fb_loc_list): Adjust.
13622 (maybe_emit_file): Call output_asm_line_debug_info for test.
13623 (dwarf2out_var_location): Reset views as needed. Precompute
13624 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13625 attribute. Set view.
13626 (new_line_info_table): Reset next view.
13627 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13628 (dwarf2out_source_line): Likewise. Output view resets and labels to
13629 the assembler, or select appropriate line info opcodes.
13630 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13631 (optimize_string_length): Catch it. Adjust.
13632 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13633 dw_val_class_view_list, and remove it if no longer needed.
13634 (hash_loc_list): Hash view numbers.
13635 (loc_list_hasher::equal): Compare them.
13636 (optimize_location_lists): Check whether a view list symbol is
13637 needed, and whether the locview attribute is present, and
13638 whether they match. Remove the locview attribute if no longer
13640 (index_location_lists): Call skip_loc_list_entry for test.
13641 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13642 Use output_dwarf_version.
13643 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13644 (struct dw_val_node): Add val_view_list.
13645 * final.c (SEEN_NEXT_VIEW): New.
13646 (set_next_view_needed): New.
13647 (clear_next_view_needed): New.
13648 (maybe_output_next_view): New.
13649 (final_start_function): Rename to...
13650 (final_start_function_1): ... this. Take pointer to FIRST,
13651 add SEEN parameter. Emit param bindings in the initial view.
13652 (final_start_function): Reintroduce SEEN-less interface.
13653 (final): Rename to...
13654 (final_1): ... this. Take SEEN parameter. Output final pending
13655 next view at the end.
13656 (final): Reintroduce seen-less interface.
13657 (final_scan_insn): Output pending next view before switching
13658 sections or ending a block. Mark the next view as needed when
13659 outputting variable locations. Notify debug backend of section
13660 changes, and of location view changes.
13661 (rest_of_handle_final): Adjust.
13662 * toplev.c (process_options): Autodetect value for debug variable
13663 location views option. Warn on incompat5 without -gdwarf-5.
13664 * doc/invoke.texi (gvariable-location-views): New.
13665 (gvariable-location-views=incompat5): New.
13666 (gno-variable-location-views): New.
13668 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13670 PR tree-optimization/84136
13671 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13672 that the result of find_edge is non-NULL.
13674 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13677 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13678 storing integer register in SImode. Fix cost of 256 and 512
13679 byte aligned SSE register store.
13681 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13683 * config/i386/i386.c (ix86_multiplication_cost): Fix
13684 multiplication cost for TARGET_AVX512DQ.
13686 2018-02-08 Marek Polacek <polacek@redhat.com>
13688 PR tree-optimization/84238
13689 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13692 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13694 PR tree-optimization/84265
13695 * tree-vect-stmts.c (vectorizable_store): Don't treat
13696 VMAT_CONTIGUOUS accesses as grouped.
13697 (vectorizable_load): Likewise.
13699 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13701 PR tree-optimization/81635
13702 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13703 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13704 (test_round_for_mask): New functions.
13705 (wide_int_cc_tests): Call test_round_for_mask.
13706 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13707 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13708 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13709 range returned by get_range_info.
13711 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13714 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13715 * symtab.c: Include builtins.h
13716 (symtab_node::output_to_lto_symbol_table_p): Move here
13717 from lto-streamer-out.c:output_symbol_p.
13718 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13719 (output_symbol_p): Move all logic to symtab.c
13720 (produce_symtab): Update.
13722 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13724 * config/s390/s390-opts.h (enum indirect_branch): Define.
13725 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13726 (s390_indirect_branch_via_thunk)
13727 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13728 (enum s390_indirect_branch_type): Define.
13729 * config/s390/s390.c (struct s390_frame_layout, struct
13730 machine_function): Remove.
13731 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13732 (indirect_branch_table_label_no, indirect_branch_table_name):
13734 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13735 (enum s390_indirect_branch_option): Define.
13736 (s390_return_addr_from_memory): New function.
13737 (s390_handle_string_attribute): New function.
13738 (s390_attribute_table): Add new attribute handler.
13739 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13740 (s390_indirect_branch_via_thunk): New function.
13741 (s390_indirect_branch_via_inline_thunk): New function.
13742 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13743 sibling call optimization for non z10 compiles.
13744 (s390_emit_call): Force indirect branch target to be a single
13745 register. Add r1 clobber for non-z10 compiles.
13746 (s390_emit_epilogue): Emit return jump via return_use expander.
13747 (s390_reorg): Handle JUMP_INSNs as execute targets.
13748 (s390_option_override_internal): Perform validity checks for the
13749 new command line options.
13750 (s390_indirect_branch_attrvalue): New function.
13751 (s390_indirect_branch_settings): New function.
13752 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13753 (s390_output_indirect_thunk_function): New function.
13754 (s390_code_end): Implement target hook.
13755 (s390_case_values_threshold): Implement target hook.
13756 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13758 * config/s390/s390.h (struct s390_frame_layout)
13759 (struct machine_function): Move here from s390.c.
13760 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13761 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13762 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13763 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13764 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13765 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13766 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13767 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13768 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13769 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13770 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13771 (mnemonic attribute): Add values which aren't recognized
13773 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13774 pattern for branch conversion. Fix mnemonic attribute.
13775 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13776 indirect branch via thunk if requested.
13777 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13778 ("*indirect_jump"): Disable for branch conversion using out of
13780 ("indirect_jump_via_thunk<mode>_z10")
13781 ("indirect_jump_via_thunk<mode>")
13782 ("indirect_jump_via_inlinethunk<mode>_z10")
13783 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13784 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13785 ("casesi_jump_via_inlinethunk<mode>_z10")
13786 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13787 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13788 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13789 ("*indirect2_jump"): Disable for branch conversion.
13790 ("casesi_jump"): Turn into expander and expand patterns for branch
13792 ("return_use"): New expander.
13793 ("*return"): Emit return via thunk and rename it to ...
13794 ("*return<mode>"): ... this one.
13795 * config/s390/s390.opt: Add new options and and enum for the
13798 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13800 * lra-constraints.c (match_reload): Unconditionally use
13801 gen_lowpart_SUBREG, rather than selecting between that
13802 and equivalent gen_rtx_SUBREG code.
13804 2018-02-08 Richard Biener <rguenther@suse.de>
13806 PR tree-optimization/84233
13807 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13808 changed flag instead of boguously re-using phi_inserted.
13810 2018-02-08 Martin Jambor <mjambor@suse.cz>
13812 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13813 static local variables.
13815 2018-02-08 Richard Biener <rguenther@suse.de>
13817 PR tree-optimization/84278
13818 * tree-vect-stmts.c (vectorizable_store): When looking for
13819 smaller vector types to perform grouped strided loads/stores
13820 make sure the mode is supported by the target.
13821 (vectorizable_load): Likewise.
13823 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13825 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13826 Increase LDP/STP opportunities by adding adjacent callee-saves.
13828 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13830 PR rtl-optimization/84068
13831 PR rtl-optimization/83459
13832 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13834 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13836 PR tree-optimization/84224
13837 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13838 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13839 non-zero arguments.
13841 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13844 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13845 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13846 count, remove one USE.
13848 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13850 * doc/install.texi (Configuration): Document the
13851 --with-long-double-format={ibm,ieee} PowerPC configuration
13855 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13856 Convert from define_expand to be define_insn_and_split. Rework
13857 float/double/_Float128 conversions to QI/HI/SImode to work with
13858 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13859 conversions to QI/HImode types did a store and then a load to
13860 truncate the value. For conversions to VSX registers, don't split
13861 the insn, instead emit the code directly. Use the code iterator
13862 any_fix to combine signed and unsigned conversions.
13863 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13864 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13865 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13866 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13867 (fix_<mode>di2_hw): Likewise.
13868 (fixuns_<mode>di2_hw): Likewise.
13869 (fix_<mode>si2_hw): Likewise.
13870 (fixuns_<mode>si2_hw): Likewise.
13871 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13872 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13873 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13874 fix<uns>_trunc<SFDF:mode>si2_p8.
13875 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13876 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13877 (fix<uns>_<mode>_mem): Likewise.
13878 (fctiw<u>z_<mode>_mem): Likewise.
13879 (fix<uns>_<mode>_mem): Likewise.
13880 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13881 the register allocator from doing a direct move to the GPRs to do
13882 a store, and instead use the ISA 3.0 store byte/half-word from
13883 vector register instruction. For IEEE 128-bit floating point,
13884 also optimize stores of 32-bit ints.
13885 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13887 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13889 * genextract.c (push_pathstr_operand): New function to support
13891 (walk_rtx): Call push_pathstr_operand.
13892 (print_path): Support [a-zA-Z].
13894 2018-02-07 Richard Biener <rguenther@suse.de>
13896 PR tree-optimization/84037
13897 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13898 (cse_and_gimplify_to_preheader): Declare.
13899 (vect_get_place_in_interleaving_chain): Likewise.
13900 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13902 (_loop_vec_info::~_loop_vec_info): Delete it.
13903 (cse_and_gimplify_to_preheader): New function.
13904 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13905 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13906 (vectorizable_load): Likewise. For grouped stores always base
13907 the IV on the first element.
13908 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13909 condition before gimplifying.
13911 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13913 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13914 *DIV_EXPR and *MOD_EXPR.
13916 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13919 * config/i386/i386.c (ix86_option_override_internal): Mask out
13920 the CF_SET bit when checking -fcf-protection.
13922 2018-02-07 Tom de Vries <tom@codesourcery.com>
13925 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13928 2018-02-07 Richard Biener <rguenther@suse.de>
13930 PR tree-optimization/84204
13931 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13934 PR tree-optimization/84205
13935 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13936 special-case isl_ast_op_zdiv_r.
13938 PR tree-optimization/84223
13939 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13940 Only add conditions from within the region.
13941 (gather_bbs::after_dom_children): Adjust.
13943 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13946 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13947 * config/avr/avr.md: Only post-reload split REG-REG moves if
13948 either register is GENERAL_REG_P.
13950 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13952 PR tree-optimization/84235
13953 * tree-ssa-scopedtables.c
13954 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13955 if the subtraction is performed in floating point type where NaNs are
13956 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13957 build 1. Formatting fix.
13959 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13962 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13963 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13964 and skip it regardless of bb boundaries. Use CALL_P macro,
13965 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13966 unnecessarily, formatting fix.
13968 2018-02-06 Michael Collison <michael.collison@arm.com>
13970 * config/arm/thumb2.md:
13971 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13972 (*thumb_mov_notscc): Ditto.
13974 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13977 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13978 unsigned_fix, not "s".
13980 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13982 * configure.ac (gcc_fn_eh_frame_ro): New function.
13983 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13984 correct .eh_frame permissions.
13985 * configure: Regenerate.
13987 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13989 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13990 irrelevant options.
13992 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13994 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13995 Display warning message for -mno-speculate-indirect-jumps.
13997 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13999 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
14001 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
14003 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
14005 PR tree-optimization/84225
14006 * tree-eh.c (find_trapping_overflow): Only call
14007 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
14009 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14012 * config/i386/i386.c: Reimplement the check of possible options
14013 -mibt/-mshstk conbination. Change error messages.
14014 * doc/invoke.texi: Fix a typo: remove extra '='.
14016 2018-02-06 Marek Polacek <polacek@redhat.com>
14018 PR tree-optimization/84228
14019 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
14021 2018-02-06 Tamar Christina <tamar.christina@arm.com>
14024 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
14025 emitted arch directives.
14026 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
14027 __ARM_FEATURE_COPROC before changing architectures.
14029 2018-02-06 Richard Biener <rguenther@suse.de>
14031 * config/i386/i386.c (print_reg): Fix typo.
14032 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14034 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14036 * configure: Regenerate.
14038 2018-02-05 Martin Sebor <msebor@redhat.com>
14040 PR tree-optimization/83369
14041 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14044 2018-02-05 Martin Liska <mliska@suse.cz>
14046 * doc/invoke.texi: Cherry-pick upstream r323995.
14048 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14050 * ira.c (ira_init_register_move_cost): Adjust comment.
14052 2018-02-05 Martin Liska <mliska@suse.cz>
14054 PR gcov-profile/84137
14055 * doc/gcov.texi: Fix typo in documentation.
14057 2018-02-05 Martin Liska <mliska@suse.cz>
14059 PR gcov-profile/83879
14060 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14061 using dlopen functionality.
14063 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14065 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14066 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14067 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14068 _mm_maskz_range_round_ss): New intrinsics.
14069 (__builtin_ia32_rangesd128_round)
14070 (__builtin_ia32_rangess128_round): Remove.
14071 (__builtin_ia32_rangesd128_mask_round,
14072 __builtin_ia32_rangess128_mask_round): New builtins.
14073 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14074 __builtin_ia32_rangess128_round): Remove.
14075 (__builtin_ia32_rangesd128_mask_round,
14076 __builtin_ia32_rangess128_mask_round): New builtins.
14077 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14078 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14079 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14080 "<round_saeonly_constraint>")): Changed to ...
14081 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14082 "<round_saeonly_scalar_constraint>")): ... this.
14083 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14084 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14085 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14086 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14087 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14089 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14091 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14093 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14094 Remove all values except native, 8540 and 8548.
14096 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14098 * config/i386/i386.c (ix86_output_function_return): Pass
14099 INVALID_REGNUM, instead of -1, as invalid register number to
14100 indirect_thunk_name and output_indirect_thunk.
14102 2018-02-02 Julia Koval <julia.koval@intel.com>
14104 * config.gcc: Add -march=icelake.
14105 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14106 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14107 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14108 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14109 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14110 (processor_target_table): Add icelake.
14111 (ix86_option_override_internal): Handle new PTAs.
14112 (get_builtin_code_for_version): Handle icelake.
14113 (M_INTEL_COREI7_ICELAKE): New.
14114 (fold_builtin_cpu): Handle icelake.
14115 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14116 * doc/invoke.texi: Add -march=icelake.
14118 2018-02-02 Julia Koval <julia.koval@intel.com>
14120 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14121 to wide_int_bitmask.
14122 * wide-int-bitmask.h: New.
14124 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14127 * config/i386/i386.md: Replace Pmode with word_mode in
14128 builtin_setjmp_setup and builtin_longjmp to support x32.
14130 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14134 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14136 (rs6000_supported_cpu_names): New static variable.
14137 (linux_cpu_translation_table): Likewise.
14138 (elf_platform) <cpu>: Define new static variable and use it.
14139 Translate kernel AT_PLATFORM name to canonical name if needed.
14140 Error if platform name is unknown.
14142 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14145 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14147 2018-02-01 Jeff Law <law@redhat.com>
14150 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14151 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14152 the scratch if RELEASE_VIA_POP is false.
14153 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14154 If we have to save a temporary register, decrement SIZE appropriately.
14155 Pass new arguments to release_scratch_register_on_entry.
14156 (ix86_adjust_stack_and_probe): Likewise.
14157 (ix86_emit_probe_stack_range): Pass new arguments to
14158 release_scratch_register_on_entry.
14160 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14162 PR rtl-optimization/84157
14163 * combine.c (change_zero_ext): Use REG_P predicate in
14164 front of HARD_REGISTER_P predicate.
14166 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14168 * config/avr/avr.c (avr_option_override): Move disabling of
14169 -fdelete-null-pointer-checks to...
14170 * common/config/avr/avr-common.c (avr_option_optimization_table):
14173 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14175 PR tree-optimization/81635
14176 * tree-data-ref.c (split_constant_offset_1): For types that
14177 wrap on overflow, try to use range info to prove that wrapping
14180 2018-02-01 Renlin Li <renlin.li@arm.com>
14183 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14184 TAILCALL_ADDR_REGS.
14185 (aarch64_register_move_cost): Likewise.
14186 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14187 TAILCALL_ADDR_REGS.
14188 (REG_CLASS_NAMES): Likewise.
14189 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14190 TAILCALL_ADDR_REGS. Remove IP registers.
14191 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14193 2018-02-01 Richard Biener <rguenther@suse.de>
14195 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14196 for specifying RPO order and allow NULL for that.
14197 * domwalk.c (dom_walker::dom_walker): Likewise.
14198 (dom_walker::walk): Handle NULL RPO order.
14199 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14201 (rewrite_update_dom_walker): Likewise.
14202 (mark_def_dom_walker): Likewise.
14204 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14206 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14207 (aarch64_maybe_expand_sve_subreg_move): Declare.
14208 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14209 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14211 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14212 that are semantically a reverse operation.
14213 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14214 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14215 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14217 (aarch64_can_change_mode_class): For big-endian, forbid changes
14218 between two SVE modes if they have different element sizes.
14220 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14222 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14223 the TImode handling for big-endian targets.
14225 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14227 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14228 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14230 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14231 Remove BSWAP handing for big-endian targets and use the form of
14232 LD1RQ appropariate for the mode.
14234 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14236 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14237 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14238 duplicated element.
14240 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14243 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14244 check for operands that need to go through aarch64_sve_reload_be.
14246 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14248 PR tree-optimization/81661
14249 PR tree-optimization/84117
14250 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14251 * tree-eh.c: Include gimplify.h.
14252 (find_trapping_overflow, replace_trapping_overflow,
14253 rewrite_to_non_trapping_overflow): New functions.
14254 * tree-vect-loop.c: Include tree-eh.h.
14255 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14256 * tree-data-ref.c: Include tree-eh.h.
14257 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14259 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14261 PR rtl-optimization/84123
14262 * combine.c (change_zero_ext): Check if hard register satisfies
14263 can_change_dest_mode before calling gen_lowpart_SUBREG.
14265 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14268 * ira.c (ira_init_register_move_cost): Remove assert.
14270 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14272 PR rtl-optimization/84071
14273 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14274 * doc/tm.texi: Regenerate.
14276 2018-01-31 Richard Biener <rguenther@suse.de>
14278 PR tree-optimization/84132
14279 * tree-data-ref.c (analyze_miv_subscript): Properly
14280 check whether evolution_function_is_affine_multivariate_p
14281 before calling gcd_of_steps_may_divide_p.
14283 2018-01-31 Julia Koval <julia.koval@intel.com>
14286 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14287 * config/i386/i386.md (rdpid_rex64) New.
14288 (rdpid): Make 32bit only.
14290 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14293 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14294 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14296 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14299 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14301 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14303 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14305 PR rtl-optimization/84071
14306 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14307 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14309 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14311 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14312 (arc_attribute_table): Add 'aux' attribute.
14313 (arc_in_small_data_p): Consider aux like variables.
14314 (arc_is_aux_reg_p): New function.
14315 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14316 (arc_get_aux_arg): New function.
14317 (prepare_move_operands): Handle aux-register access.
14318 (arc_handle_aux_attribute): New function.
14319 * doc/extend.texi (ARC Variable attributes): Add subsection.
14321 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14323 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14324 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14325 (arc_attribute_table): Add 'uncached' attribute.
14326 (arc_print_operand): Print '.di' flag for uncached memory
14328 (arc_in_small_data_p): Do not consider for small data the uncached
14330 (arc_is_uncached_mem_p): New function.
14331 * config/arc/predicates.md (compact_store_memory_operand): Check
14332 for uncached memory accesses.
14333 (nonvol_nonimm_operand): Likewise.
14334 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14336 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14339 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14340 falign-loops=): Add Optimization flag.
14342 2018-01-30 Jeff Law <law@redhat.com>
14345 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14346 INT_REGISTERS_SAVED. Check it prior to calling
14347 get_scratch_register_on_entry.
14348 (ix86_adjust_stack_and_probe): Similarly.
14349 (ix86_emit_probe_stack_range): Similarly.
14350 (ix86_expand_prologue): Corresponding changes.
14352 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14355 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14356 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14358 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14361 * lra-constraints.c (curr_insn_transform): Process AND in the
14364 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14366 PR rtl-optimization/83986
14367 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14368 dependence against last_pending_memory_flush in addition to
14369 pending_jump_insns.
14371 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14373 PR tree-optimization/81611
14374 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14377 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14380 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14383 2018-01-30 Richard Biener <rguenther@suse.de>
14384 Jakub Jelinek <jakub@redhat.com>
14386 PR tree-optimization/84111
14387 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14388 inner loops added during recursion, as they don't have up-to-date
14391 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14394 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14395 (can_inline_edge_by_limits_p): ... here.
14396 (can_early_inline_edge_p, check_callers,
14397 update_caller_keys, update_callee_keys, recursive_inlining,
14398 add_new_edges_to_heap, speculation_useful_p,
14399 inline_small_functions,
14400 inline_small_functions, flatten_function,
14401 inline_to_all_callers_1): Update.
14403 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14405 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14408 2018-01-30 Richard Biener <rguenther@suse.de>
14410 PR tree-optimization/83008
14411 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14412 invariant and constant vector uses in stmts when they need
14413 more than one stmt.
14415 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14418 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14419 * configure: Regenerate.
14421 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14423 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14425 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14426 Use gen_rtx_REG rather than gen_lowpart.
14428 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14430 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14431 rather than 0 when creating partial subregs.
14433 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14435 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14438 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14441 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14442 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14443 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14444 flags. This restores the settings used before the 2017-07-24.
14445 Turning off pre increment/decrement/modify allows IVOPTS to
14446 optimize DF/SF loops where the index is an int.
14448 2018-01-29 Richard Biener <rguenther@suse.de>
14449 Kelvin Nilsen <kelvin@gcc.gnu.org>
14452 * tree-vect-stmts.c (vectorizable_call): Don't call
14453 targetm.vectorize_builtin_md_vectorized_function if callee is
14456 2018-01-22 Carl Love <cel@us.ibm.com>
14458 * doc/extend.tex: Fix typo in second arg in
14459 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14461 2018-01-29 Richard Biener <rguenther@suse.de>
14463 PR tree-optimization/84086
14464 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14465 (flush_ssaname_freelist): When SSA names were released reset
14466 the SCEV hash table.
14468 2018-01-29 Richard Biener <rguenther@suse.de>
14470 PR tree-optimization/84057
14471 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14472 removed paths when removing edges.
14474 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14476 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14477 -mfunction-return=@var{choice}.
14479 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14481 PR diagnostic/84034
14482 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14483 Handle CR like TAB.
14484 (layout::print_source_line): Likewise.
14485 (test_get_line_width_without_trailing_whitespace): Add test cases.
14487 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14489 PR middle-end/84040
14490 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14493 2018-01-26 Jim Wilson <jimw@sifive.com>
14495 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14497 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14500 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14502 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14503 and CMP + SUB-immediate -> SUBS.
14505 2018-01-26 Martin Sebor <msebor@redhat.com>
14507 PR tree-optimization/83896
14508 * tree-ssa-strlen.c (get_string_len): Rename...
14509 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14510 Avoid assuming length is constant.
14511 (handle_char_store): Use HOST_WIDE_INT for string length.
14513 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14516 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14517 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14519 2018-01-26 Richard Biener <rguenther@suse.de>
14521 PR rtl-optimization/84003
14522 * dse.c (record_store): Only record redundant stores when
14523 the earlier store aliases at least all accesses the later one does.
14525 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14527 PR rtl-optimization/83985
14528 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14529 REG_CFA_RESTORE insns.
14530 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14531 REG_CFA_RESTORE insns here.
14534 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14535 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14537 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14539 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14540 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14541 (arc_init): Likewise.
14542 (arc_override_options): Likewise.
14543 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14545 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14547 * config/arc/arc.h (TARGET_DBNZ): Define.
14548 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14549 properly set the tune attribute.
14550 (dbnz): Use TARGET_DBNZ guard.
14551 * config/arc/arc.opt (mtune): Add core3 option.
14553 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14555 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14556 recognize new pic like addresses.
14557 (arc_delegitimize_address): Clean up.
14559 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14561 * config/arc/arc-arches.def: Option mrf16 valid for all
14563 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14564 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14565 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14566 * config/arc/arc-tables.opt: Regenerate.
14567 * config/arc/arc.c (arc_conditional_register_usage): Handle
14568 reduced register file case.
14569 (arc_file_start): Set must have build attributes.
14570 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14571 mrf16 option value.
14572 * config/arc/arc.opt (mrf16): Add new option.
14573 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14574 * config/arc/genmultilib.awk: Handle new mrf16 option.
14575 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14576 * config/arc/t-multilib: Regenerate.
14577 * doc/invoke.texi (ARC Options): Document mrf16 option.
14579 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14581 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14582 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14583 (arc_attribute_table): Add 'secure_call' attribute.
14584 (arc_print_operand): Print secure call operand.
14585 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14587 (arc_is_secure_call_p): New function. * config/arc/arc.md
14588 (call_i): Add support for sjli instruction.
14589 (call_value_i): Likewise.
14590 * config/arc/constraints.md (Csc): New constraint.
14592 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14593 John Eric Martin <John.Martin@emmicro-us.com>
14595 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14596 * config/arc/arc.c (_arc_jli_section): New struct.
14597 (arc_jli_section): New type.
14598 (rc_jli_sections): New static variable.
14599 (arc_handle_jli_attribute): New function.
14600 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14601 (arc_file_end): New function.
14602 (TARGET_ASM_FILE_END): Define.
14603 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14604 (arc_add_jli_section): New function.
14605 (jli_call_scan): Likewise.
14606 (arc_reorg): Call jli_call_scan.
14607 (arc_output_addsi): Remove 'S' from printing asm operand.
14608 (arc_is_jli_call_p): New function.
14609 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14611 (movhi_insn): Likewise.
14612 (movsi_insn): Likewise.
14613 (movsi_set_cc_insn): Likewise.
14614 (loadqi_update): Likewise.
14615 (load_zeroextendqisi_update): Likewise.
14616 (load_signextendqisi_update): Likewise.
14617 (loadhi_update): Likewise.
14618 (load_zeroextendhisi_update): Likewise.
14619 (load_signextendhisi_update): Likewise.
14620 (loadsi_update): Likewise.
14621 (loadsf_update): Likewise.
14622 (movsicc_insn): Likewise.
14623 (bset_insn): Likewise.
14624 (bxor_insn): Likewise.
14625 (bclr_insn): Likewise.
14626 (bmsk_insn): Likewise.
14627 (bicsi3_insn): Likewise.
14628 (cmpsi_cc_c_insn): Likewise.
14629 (movsi_ne): Likewise.
14630 (movsi_cond_exec): Likewise.
14631 (clrsbsi2): Likewise.
14632 (norm_f): Likewise.
14642 (call_i): Remove 'S' asm letter, add jli instruction.
14643 (call_value_i): Likewise.
14644 * config/arc/arc.op (mjli-always): New option.
14645 * config/arc/constraints.md (Cji): New constraint.
14646 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14648 (subsf3_fpx): Likewise.
14649 (mulsf3_fpx): Likewise.
14650 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14652 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14653 function attrbutes.
14654 * doc/invoke.texi (ARC): Document mjli-always option.
14656 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14658 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14659 avoid addition with 0 and use incw and decw where possible.
14661 2018-01-26 Richard Biener <rguenther@suse.de>
14663 PR tree-optimization/81082
14664 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14665 association if it requires casting to unsigned.
14666 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14667 from fold_plusminus_mult_expr to catch important cases late when
14668 range info is available.
14670 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14672 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14673 * configure.ac (hidden_linkonce): New test.
14674 * configure: Regenerate.
14675 * config.in: Regenerate.
14677 2018-01-26 Julia Koval <julia.koval@intel.com>
14679 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14680 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14681 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14682 _mm_mask_bitshuffle_epi64_mask): Fix type.
14683 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14684 USI_FTYPE_V4DI_V4DI_USI): Remove.
14685 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14686 __builtin_ia32_vpshufbitqmb256_mask,
14687 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14688 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14689 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14691 2018-01-26 Alan Modra <amodra@gmail.com>
14694 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14695 UNSPEC_VBPERMQ. Sort other unspecs.
14697 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14699 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14701 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14703 PR middle-end/83055
14704 * predict.c (drop_profile): Do not push/pop cfun; update also
14706 (handle_missing_profiles): Fix logic looking for zero profiles.
14708 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14710 PR middle-end/83977
14711 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14712 on functions with #pragma omp declare simd or functions with simd
14714 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14715 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14716 Remove trailing \n from warning_at calls.
14718 2018-01-25 Tom de Vries <tom@codesourcery.com>
14721 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14722 for neutered workers.
14724 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14727 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14728 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14730 2018-01-24 Jeff Law <law@redhat.com>
14733 * i386.c (get_probe_interval): Move to earlier point.
14734 (ix86_compute_frame_layout): If -fstack-clash-protection and
14735 the frame is larger than the probe interval, then use pushes
14736 to save registers rather than reg->mem moves.
14737 (ix86_expand_prologue): Remove conditional for int_registers_saved
14740 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14743 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14744 min/max for never referenced object.
14746 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14748 PR middle-end/83977
14749 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14751 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14752 attributes from DECL_ATTRIBUTES (decl) without affecting
14753 DECL_ATTRIBUTES (current_function_decl).
14754 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14755 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14757 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14759 PR tree-optimization/83979
14760 * fold-const.c (fold_comparison): Use constant_boolean_node
14761 instead of boolean_{true,false}_node.
14763 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14765 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14768 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14770 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14771 Simplify the clause that sets the length attribute.
14772 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14773 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14774 clause that sets the length attribute.
14775 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14777 2018-01-24 Tom de Vries <tom@codesourcery.com>
14780 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14781 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14782 Add strict parameter.
14783 (prevent_branch_around_nothing): Insert dummy insn between branch to
14784 label and label with no ptx insn inbetween.
14785 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14787 2018-01-24 Tom de Vries <tom@codesourcery.com>
14790 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14791 for neutered threads in warp.
14792 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14794 2018-01-24 Richard Biener <rguenther@suse.de>
14796 PR tree-optimization/83176
14797 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14800 2018-01-24 Richard Biener <rguenther@suse.de>
14802 PR tree-optimization/82819
14803 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14804 code generating pluses that are no-ops in the target precision.
14806 2018-01-24 Richard Biener <rguenther@suse.de>
14808 PR middle-end/84000
14809 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14811 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14813 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14814 to merge probabilities.
14815 * predict.c (probably_never_executed): Also mark as cold functions
14816 with global 0 profile and guessed local profile.
14817 * profile-count.c (profile_probability::combine_with_count): New
14819 * profile-count.h (profile_probability::operator*,
14820 profile_probability::operator*=, profile_probability::operator/,
14821 profile_probability::operator/=): Reduce precision to adjusted
14822 and set value to guessed on contradictory divisions.
14823 (profile_probability::combine_with_freq): Remove.
14824 (profile_probability::combine_wiht_count): Declare.
14825 (profile_count::force_nonzero):: Set to adjusted.
14826 (profile_count::probability_in):: Set quality to adjusted.
14827 * tree-ssa-tail-merge.c (replace_block_by): Use
14828 combine_with_count.
14830 2018-01-23 Andrew Waterman <andrew@sifive.com>
14831 Jim Wilson <jimw@sifive.com>
14833 * config/riscv/riscv.c (riscv_stack_boundary): New.
14834 (riscv_option_override): Set riscv_stack_boundary. Handle
14835 riscv_preferred_stack_boundary_arg.
14836 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14837 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14838 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14839 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14840 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14841 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14843 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14846 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14847 of struct ix86_frame.
14848 (ix86_expand_epilogue): Likewise. Add a local variable for
14849 the reg_save_offset field in struct ix86_frame.
14851 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14853 PR tree-optimization/82604
14854 * tree-loop-distribution.c (enum partition_kind): New enum item
14855 PKIND_PARTIAL_MEMSET.
14856 (partition_builtin_p): Support above new enum item.
14857 (generate_code_for_partition): Ditto.
14858 (compute_access_range): Differentiate cases that equality can be
14859 proven at all loops, the innermost loops or no loops.
14860 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14861 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14862 (finalize_partitions, distribute_loop): Don't fuse partition of
14863 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14864 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14865 parloop is enabled.
14867 2018-01-23 Martin Liska <mliska@suse.cz>
14869 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14870 order to ignore the predictor.
14871 (PRED_POLYMORPHIC_CALL): Likewise.
14872 (PRED_RECURSIVE_CALL): Likewise.
14874 2018-01-23 Martin Liska <mliska@suse.cz>
14876 * tree-profile.c (tree_profiling): Print function header to
14877 aware reader which function we are working on.
14878 * value-prof.c (gimple_find_values_to_profile): Do not print
14879 not interesting value histograms.
14881 2018-01-23 Martin Liska <mliska@suse.cz>
14883 * profile-count.h (enum profile_quality): Add
14884 profile_uninitialized as the first value. Do not number values
14885 as they are zero based.
14886 (profile_count::verify): Update sanity check.
14887 (profile_probability::verify): Likewise.
14889 2018-01-23 Nathan Sidwell <nathan@acm.org>
14891 * doc/invoke.texi (ffor-scope): Deprecate.
14893 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14895 PR tree-optimization/83510
14896 * domwalk.c (set_all_edges_as_executable): New function.
14897 (dom_walker::dom_walker): Convert bool param
14898 "skip_unreachable_blocks" to enum reachability. Move setup of
14899 edge flags to set_all_edges_as_executable and only do it when
14900 reachability is REACHABLE_BLOCKS.
14901 * domwalk.h (enum dom_walker::reachability): New enum.
14902 (dom_walker::dom_walker): Convert bool param
14903 "skip_unreachable_blocks" to enum reachability.
14904 (set_all_edges_as_executable): New decl.
14905 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14906 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14908 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14909 but converting true to REACHABLE_BLOCKS.
14910 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14912 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14913 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14914 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14916 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14917 if check_all_array_refs will be called.
14919 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14921 * tree.c (selftest::test_location_wrappers): Add more test
14924 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14926 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14927 (selftest::test_bit_in_range): Likewise.
14929 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14932 * doc/sourcebuild.texi (vect_float): Say that the selector
14933 only describes the situation when -funsafe-math-optimizations is on.
14934 (vect_float_strict): Document.
14936 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14938 PR tree-optimization/83965
14939 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14940 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14941 instead of checking only for a reduction.
14942 (vect_recog_widen_sum_pattern): Likewise.
14944 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14946 * predict.c (probably_never_executed): Only use precise profile info.
14947 (compute_function_frequency): Skip after inlining hack since we now
14948 have quality checking.
14950 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14952 * profile-count.h (profile_probability::very_unlikely,
14953 profile_probability::unlikely, profile_probability::even): Set
14954 precision to guessed.
14956 2018-01-23 Richard Biener <rguenther@suse.de>
14958 PR tree-optimization/83963
14959 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14960 Properly terminate dominator walk when crossing the exit edge not
14961 when visiting its source block.
14963 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14966 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14967 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14969 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14971 PR tree-optimization/83957
14972 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14973 semicolon after for body surrounded by braces.
14975 PR tree-optimization/83081
14976 * profile-count.h (profile_probability::split): New method.
14977 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14978 Use profile_probability::split.
14979 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14980 when splitting a single conditional jump into 2.
14982 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14984 PR tree-optimization/69452
14985 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14988 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14990 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14991 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14992 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14994 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14996 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14998 * config/rl78/rl78.md (movdi): New define_expand.
14999 * config/rl78/rl78.c (rl78_split_movdi): New function.
15001 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
15004 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
15006 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
15007 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
15008 128-bit to produce an UNSPEC move to get the double word with the
15009 signbit and then a shift directly to do signbit.
15010 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
15011 implementation with a new version that just does either a direct
15012 move or a regular move. Move memory interface to separate insns.
15013 Move insns so they are next to the expander.
15014 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
15015 with signbit move. Split big and little endian case.
15016 (signbit<mode>2_dm_mem_le): Likewise.
15017 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
15018 (signbit<mode>2_dm2): Likewise.
15020 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15022 * config/rl78/rl78.md (anddi3): New define_expand.
15024 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15026 * config/rl78/rl78.md (umindi3): New define_expand.
15028 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15030 * config/rl78/rl78.md (smindi3): New define_expand.
15032 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15034 * config/rl78/rl78.md (smaxdi3): New define_expand.
15036 2018-01-22 Carl Love <cel@us.ibm.com>
15038 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15039 LVX_V1TI): Add macro expansion.
15040 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15041 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15042 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15043 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15044 Change check to determine if the instruction is a byte reversing
15045 entry. Fix typo in comment.
15046 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15047 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15048 Add def_builtin calls for new builtins.
15049 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15050 Add define_insn expansion.
15052 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15054 * config/rl78/rl78.md (umaxdi3): New define_expand.
15056 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15058 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15059 for non-QImode registers.
15061 2018-01-22 Richard Biener <rguenther@suse.de>
15063 PR tree-optimization/83963
15064 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15065 including the loop exit block.
15066 (scop_detection::merge_sese): Likewise.
15067 (scop_detection::add_scop): Do it here instead.
15069 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15071 * doc/sourcebuild.texi (arm_softfloat): Document.
15073 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15076 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15077 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15078 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15080 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15081 David Edelsohn <dje.gcc@gmail.com>
15084 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15085 Change "crset eq" to "crset 2".
15086 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15087 (*call_indirect_aix<mode>_nospec): Likewise.
15088 (*call_value_indirect_aix<mode>_nospec): Likewise.
15089 (*call_indirect_elfv2<mode>_nospec): Likewise.
15090 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15091 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15092 change assembly output from . to $.
15093 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15094 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15095 (*tablejump<mode>_internal1_nospec): Likewise.
15097 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15100 * config/sh/sh_optimize_sett_clrt.cc:
15101 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15103 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15105 PR tree-optimization/83940
15106 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15107 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15108 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15109 use it to pass back the definition type.
15110 (vect_check_store_rhs): Likewise rhs_dt_out.
15111 (vect_build_gather_load_calls): Add a mask_dt argument and use
15112 it instead of a call to vect_is_simple_use.
15113 (vectorizable_store): Update calls to vect_check_load_store_mask
15114 and vect_check_store_rhs. Use the dt returned by the latter instead
15115 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15116 instead of calls to vect_is_simple_use. Pass the scalar rather
15117 than the vector operand to vect_is_simple_use when handling
15118 second and subsequent copies of an rhs value.
15119 (vectorizable_load): Update calls to vect_check_load_store_mask
15120 and vect_build_gather_load_calls. Use the cached mask_dt and
15121 gs_info.offset_dt instead of calls to vect_is_simple_use.
15123 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15125 PR middle-end/83945
15126 * tree-emutls.c: Include gimplify.h.
15127 (lower_emutls_2): New function.
15128 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15129 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15130 it before further processing.
15133 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15134 UINTVAL (trueop1) instead of INTVAL (op1).
15136 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15140 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15141 INCOMING_FRAME_SP_OFFSET if not defined.
15142 (scan_trace): Add ENTRY argument. If true and
15143 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15144 emit a note to adjust the CFA offset.
15145 (create_cfi_notes): Adjust scan_trace callers.
15146 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15147 INCOMING_FRAME_SP_OFFSET in the CIE.
15148 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15149 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15151 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15152 * doc/tm.texi: Regenerated.
15154 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15156 PR rtl-optimization/83147
15157 * lra-constraints.c (remove_inheritance_pseudos): Use
15158 lra_substitute_pseudo_within_insn.
15160 2018-01-19 Tom de Vries <tom@codesourcery.com>
15161 Cesar Philippidis <cesar@codesourcery.com>
15164 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15166 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15169 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15170 spaces for function labels.
15172 2018-01-19 Martin Liska <mliska@suse.cz>
15174 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15175 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15176 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15177 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15178 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15179 (PRED_CONST_RETURN): Change from 69 to 65.
15180 (PRED_NULL_RETURN): Change from 91 to 71.
15181 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15182 (PRED_LOOP_GUARD): Change from 66 to 73.
15184 2018-01-19 Martin Liska <mliska@suse.cz>
15186 * predict.c (predict_insn_def): Add new assert.
15187 (struct branch_predictor): Change type to signed integer.
15188 (test_prediction_value_range): Amend test to cover
15189 PROB_UNINITIALIZED.
15190 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15191 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15192 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15193 (PRED_LOOP_IV_COMPARE): Likewise.
15194 * predict.h (PROB_UNINITIALIZED): Define new constant.
15196 2018-01-19 Martin Liska <mliska@suse.cz>
15198 * predict.c (dump_prediction): Add new format for
15199 analyze_brprob.py script which is enabled with -details
15201 * profile-count.h (precise_p): New function.
15203 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15205 PR tree-optimization/83922
15206 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15207 there are no statements that need masking.
15208 (vect_active_double_reduction_p): New function.
15209 (vect_analyze_loop_operations): Use it when handling phis that
15210 are not in the loop header.
15212 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15214 PR tree-optimization/83914
15215 * tree-vect-loop.c (vectorizable_induction): Don't convert
15216 init_expr or apply the peeling adjustment for inductions
15217 that are nested within the vectorized loop.
15219 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15221 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15224 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15228 * function.h (gimplify_parameters): Add gimple_seq * argument.
15229 * function.c: Include gimple.h and options.h.
15230 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15231 for the added local temporaries if needed.
15232 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15233 if there are any parameter cleanups, wrap whole body into a
15234 try/finally with the cleanups.
15236 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15239 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15240 Use GET_MODE_CLASS for scalar floating point.
15242 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15246 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15247 Fix call of call_cgraph_insertion_hooks.
15249 2018-01-18 Martin Sebor <msebor@redhat.com>
15251 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15253 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15256 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15259 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15262 * common.opt: (-ffile-prefix-map): New option.
15263 * opts.c (common_handle_option): Defer it.
15264 * opts-global.c (handle_common_deferred_options): Handle it.
15265 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15266 * file-prefix-map.h: New file.
15267 (remap_debug_filename, add_debug_prefix_map): ...here.
15268 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15269 * final.c (debug_prefix_map, add_debug_prefix_map
15270 remap_debug_filename): Move to...
15271 * file-prefix-map.c: New file.
15272 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15273 generalize, get rid of alloca(), use strrchr() instead of strchr().
15274 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15275 Implement in terms of add_prefix_map().
15276 (remap_macro_filename, remap_debug_filename): Implement in term of
15278 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15279 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15280 * dbxout.c: Include file-prefix-map.h.
15281 * varasm.c: Likewise.
15282 * vmsdbgout.c: Likewise.
15283 * xcoffout.c: Likewise.
15284 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15285 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15286 * doc/invoke.texi (-ffile-prefix-map): Document.
15287 (-fdebug-prefix-map): Update description.
15289 2018-01-18 Martin Liska <mliska@suse.cz>
15291 * config/i386/i386.c (indirect_thunk_name): Document that also
15293 (output_indirect_thunk): Document why both instructions
15294 (pause and lfence) are generated.
15296 2018-01-18 Richard Biener <rguenther@suse.de>
15298 PR tree-optimization/83887
15299 * graphite-scop-detection.c
15300 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15301 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15302 (scop_detection::merge_sese): Re-implement with a flood-fill
15303 algorithm that properly finds a SESE region if it exists.
15305 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15308 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15309 pointer_diff optimizations use view_convert instead of convert.
15311 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15313 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15314 Generate different code for -mno-speculate-indirect-jumps.
15315 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15316 (*call_indirect_aix<mode>): Disable for
15317 -mno-speculate-indirect-jumps.
15318 (*call_indirect_aix<mode>_nospec): New define_insn.
15319 (*call_value_indirect_aix<mode>): Disable for
15320 -mno-speculate-indirect-jumps.
15321 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15322 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15323 -mno-speculate-indirect-jumps.
15324 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15326 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15328 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15329 long double type, set the flags for noting the default long double
15330 type, even if we don't pass or return a long double type.
15332 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15335 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15338 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15340 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15341 support for merge[hl].
15342 (fold_mergehl_helper): New helper function.
15343 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15344 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15345 (altivec_vmrglw_direct): Add xxmrglw insn.
15347 2018-01-17 Andrew Waterman <andrew@sifive.com>
15349 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15350 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15352 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15355 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15356 call the lto_location_cache before reading the
15357 DECL_SOURCE_LOCATION of the types.
15359 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15360 Richard Sandiford <richard.sandiford@linaro.org>
15362 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15363 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15364 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15365 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15366 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15368 * config/aarch64/constraints.md (aarch64_movti_operand):
15370 * config/aarch64/predicates.md (Uti): Add new constraint.
15372 2018-01-17 Carl Love <cel@us.ibm.com>
15374 * config/rs6000/vsx.md (define_expand xl_len_r,
15375 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15376 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15378 (define_expand, define_insn): Move the shift left from the
15379 define_insn to the define_expand for lxvl and stxvl instructions.
15380 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15381 and XL_LEN_R definitions to PURE.
15383 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15385 * config/i386/i386.c (indirect_thunk_name): Declare regno
15386 as unsigned int. Compare regno with INVALID_REGNUM.
15387 (output_indirect_thunk): Ditto.
15388 (output_indirect_thunk_function): Ditto.
15389 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15390 in the call to output_indirect_thunk_function.
15392 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15394 PR middle-end/83884
15395 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15396 rather than the size of inner_type to determine the stack slot size
15397 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15399 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15402 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15405 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15407 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15408 endian Linux systems to optionally enable multilibs for selecting
15409 the long double type if the user configured an explicit type.
15410 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15411 have no long double multilibs if not defined.
15412 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15413 warn if the user used -mabi={ieee,ibm}longdouble and we built
15414 multilibs for long double.
15415 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15416 appropriate multilib option.
15417 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15419 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15420 for building long double multilibs.
15421 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15423 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15425 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15428 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15430 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15433 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15436 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15439 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15441 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15442 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15444 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15446 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15447 different rtl trees depending on TARGET_64BIT.
15448 (rs6000_gen_lvx): Likewise.
15450 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15452 * config/visium/visium.md (nop): Tweak comment.
15453 (hazard_nop): Likewise.
15455 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15457 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15458 -mspeculate-indirect-jumps.
15459 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15460 for -mno-speculate-indirect-jumps.
15461 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15462 (*call_value_indirect_elfv2<mode>): Disable for
15463 -mno-speculate-indirect-jumps.
15464 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15465 (indirect_jump): Emit different RTL for
15466 -mno-speculate-indirect-jumps.
15467 (*indirect_jump<mode>): Disable for
15468 -mno-speculate-indirect-jumps.
15469 (*indirect_jump<mode>_nospec): New define_insn.
15470 (tablejump): Emit different RTL for
15471 -mno-speculate-indirect-jumps.
15472 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15473 (tablejumpsi_nospec): New define_expand.
15474 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15475 (tablejumpdi_nospec): New define_expand.
15476 (*tablejump<mode>_internal1): Disable for
15477 -mno-speculate-indirect-jumps.
15478 (*tablejump<mode>_internal1_nospec): New define_insn.
15479 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15482 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15484 * caller-save.c (insert_save): Drop unnecessary parameter. All
15487 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15488 Richard Biener <rguenth@suse.de>
15491 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15492 return early, inline manually is_gimple_sizepos. Make sure if we
15493 call gimplify_expr we don't end up with a gimple constant.
15494 * tree.c (variably_modified_type_p): Don't return true for
15495 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15496 * gimplify.h (is_gimple_sizepos): Remove.
15498 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15500 PR tree-optimization/83857
15501 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15502 vectorizable_live_operation for pure SLP statements.
15503 (vectorizable_live_operation): Handle PHIs.
15505 2018-01-16 Richard Biener <rguenther@suse.de>
15507 PR tree-optimization/83867
15508 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15509 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15511 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15514 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15515 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15516 If off is not INTEGER_CST, issue a may not be aligned warning
15517 rather than isn't aligned. Use isn%'t rather than isn't.
15518 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15520 <case MULT_EXPR>: Improve the case when bottom and one of the
15521 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15522 operand, in that case check if the other operand is multiple of
15523 bottom divided by the INTEGER_CST operand.
15525 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15528 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15529 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15530 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15531 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15532 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15533 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15534 (pa_function_arg_size): New function.
15536 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15538 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15539 in a separate statement.
15541 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15543 PR tree-optimization/83847
15544 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15545 group gathers and scatters.
15547 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15549 PR rtl-optimization/86620
15550 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15552 PR rtl-optimization/83213
15553 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15554 to last if both are JUMP_INSNs.
15556 PR tree-optimization/83843
15557 * gimple-ssa-store-merging.c
15558 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15559 store_immediate_info for bswap/nop orig_stores.
15561 2018-01-15 Andrew Waterman <andrew@sifive.com>
15563 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15565 <UDIV>: Increase cost if !TARGET_DIV.
15567 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15569 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15570 (define_attr "cr_logical_3op"): New.
15571 (cceq_ior_compare): Adjust.
15572 (cceq_ior_compare_complement): Adjust.
15573 (*cceq_rev_compare): Adjust.
15574 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15575 (is_cracked_insn): Adjust.
15576 (insn_must_be_first_in_group): Adjust.
15577 * config/rs6000/40x.md: Adjust.
15578 * config/rs6000/440.md: Adjust.
15579 * config/rs6000/476.md: Adjust.
15580 * config/rs6000/601.md: Adjust.
15581 * config/rs6000/603.md: Adjust.
15582 * config/rs6000/6xx.md: Adjust.
15583 * config/rs6000/7450.md: Adjust.
15584 * config/rs6000/7xx.md: Adjust.
15585 * config/rs6000/8540.md: Adjust.
15586 * config/rs6000/cell.md: Adjust.
15587 * config/rs6000/e300c2c3.md: Adjust.
15588 * config/rs6000/e500mc.md: Adjust.
15589 * config/rs6000/e500mc64.md: Adjust.
15590 * config/rs6000/e5500.md: Adjust.
15591 * config/rs6000/e6500.md: Adjust.
15592 * config/rs6000/mpc.md: Adjust.
15593 * config/rs6000/power4.md: Adjust.
15594 * config/rs6000/power5.md: Adjust.
15595 * config/rs6000/power6.md: Adjust.
15596 * config/rs6000/power7.md: Adjust.
15597 * config/rs6000/power8.md: Adjust.
15598 * config/rs6000/power9.md: Adjust.
15599 * config/rs6000/rs64.md: Adjust.
15600 * config/rs6000/titan.md: Adjust.
15602 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15604 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15605 ix86_indirect_branch_register logic.
15607 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15609 * config/i386/constraints.md (Bs): Update
15610 ix86_indirect_branch_register check. Don't check
15611 ix86_indirect_branch_register with GOT_memory_operand.
15613 * config/i386/predicates.md (GOT_memory_operand): Don't check
15614 ix86_indirect_branch_register here.
15615 (GOT32_symbol_operand): Likewise.
15617 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15619 * config/i386/predicates.md (constant_call_address_operand):
15620 Rewrite ix86_indirect_branch_register logic.
15621 (sibcall_insn_operand): Likewise.
15623 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15625 * config/i386/constraints.md (Bs): Replace
15626 ix86_indirect_branch_thunk_register with
15627 ix86_indirect_branch_register.
15629 * config/i386/i386.md (indirect_jump): Likewise.
15630 (tablejump): Likewise.
15631 (*sibcall_memory): Likewise.
15632 (*sibcall_value_memory): Likewise.
15633 Peepholes of indirect call and jump via memory: Likewise.
15634 * config/i386/i386.opt: Likewise.
15635 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15636 (GOT_memory_operand): Likewise.
15637 (call_insn_operand): Likewise.
15638 (sibcall_insn_operand): Likewise.
15639 (GOT32_symbol_operand): Likewise.
15641 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15643 PR middle-end/83837
15644 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15645 type rather than type addr's type points to.
15646 (expand_omp_atomic_mutex): Likewise.
15647 (expand_omp_atomic): Likewise.
15649 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15652 * config/i386/i386.c (output_indirect_thunk_function): Use
15653 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15654 for __x86_return_thunk.
15656 2018-01-15 Richard Biener <rguenther@suse.de>
15658 PR middle-end/83850
15659 * expmed.c (extract_bit_field_1): Fix typo.
15661 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15664 * config/arm/iterators.md (VF): New mode iterator.
15665 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15666 Remove integer-related logic from pattern.
15667 (neon_vabd<mode>_3): Likewise.
15669 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15671 PR middle-end/82694
15672 * common.opt (fstrict-overflow): No longer an alias.
15673 (fwrapv-pointer): New option.
15674 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15675 also for pointer types based on flag_wrapv_pointer.
15676 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15677 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15678 opts->x_flag_wrapv got set.
15679 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15680 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15681 POINTER_TYPE_OVERFLOW_UNDEFINED.
15682 * match.pd: Likewise in address comparison pattern.
15683 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15685 2018-01-15 Richard Biener <rguenther@suse.de>
15688 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15689 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15690 Reset type names to their identifier if their TYPE_DECL doesn't
15691 have linkage (and thus is used for ODR and devirt).
15692 (save_debug_info_for_decl): Remove.
15693 (save_debug_info_for_type): Likewise.
15694 (add_tree_to_fld_list): Adjust.
15695 * tree-pretty-print.c (dump_generic_node): Make dumping of
15696 type names more robust.
15698 2018-01-15 Richard Biener <rguenther@suse.de>
15700 * BASE-VER: Bump to 8.0.1.
15702 2018-01-14 Martin Sebor <msebor@redhat.com>
15705 * builtins.c (check_access): Avoid warning when the no-warning bit
15708 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15710 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15711 * ira-color (allocno_hard_regs_compare): Likewise.
15713 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15716 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15717 Use .pushsection/.popsection.
15719 2018-01-14 Martin Sebor <msebor@redhat.com>
15722 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15724 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15726 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15727 entry from extra_headers.
15728 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15729 extra_headers, make the list bitwise identical to the i?86-*-* one.
15731 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15733 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15734 -mcmodel=large with -mindirect-branch=thunk,
15735 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15736 -mfunction-return=thunk-extern.
15737 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15738 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15739 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15741 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15743 * config/i386/i386.c (print_reg): Print the name of the full
15744 integer register without '%'.
15745 (ix86_print_operand): Handle 'V'.
15746 * doc/extend.texi: Document 'V' modifier.
15748 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15750 * config/i386/constraints.md (Bs): Disallow memory operand for
15751 -mindirect-branch-register.
15753 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15754 (GOT_memory_operand): Likewise.
15755 (call_insn_operand): Likewise.
15756 (sibcall_insn_operand): Likewise.
15757 (GOT32_symbol_operand): Likewise.
15758 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15759 for -mindirect-branch-register.
15760 (tablejump): Likewise.
15761 (*sibcall_memory): Likewise.
15762 (*sibcall_value_memory): Likewise.
15763 Disallow peepholes of indirect call and jump via memory for
15764 -mindirect-branch-register.
15765 (*call_pop): Replace m with Bw.
15766 (*call_value_pop): Likewise.
15767 (*sibcall_pop_memory): Replace m with Bs.
15768 * config/i386/i386.opt (mindirect-branch-register): New option.
15769 * doc/invoke.texi: Document -mindirect-branch-register option.
15771 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15773 * config/i386/i386-protos.h (ix86_output_function_return): New.
15774 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15775 set function_return_type.
15776 (indirect_thunk_name): Add ret_p to indicate thunk for function
15778 (output_indirect_thunk_function): Pass false to
15779 indirect_thunk_name.
15780 (ix86_output_indirect_branch_via_reg): Likewise.
15781 (ix86_output_indirect_branch_via_push): Likewise.
15782 (output_indirect_thunk_function): Create alias for function
15783 return thunk if regno < 0.
15784 (ix86_output_function_return): New function.
15785 (ix86_handle_fndecl_attribute): Handle function_return.
15786 (ix86_attribute_table): Add function_return.
15787 * config/i386/i386.h (machine_function): Add
15788 function_return_type.
15789 * config/i386/i386.md (simple_return_internal): Use
15790 ix86_output_function_return.
15791 (simple_return_internal_long): Likewise.
15792 * config/i386/i386.opt (mfunction-return=): New option.
15793 (indirect_branch): Mention -mfunction-return=.
15794 * doc/extend.texi: Document function_return function attribute.
15795 * doc/invoke.texi: Document -mfunction-return= option.
15797 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15799 * config/i386/i386-opts.h (indirect_branch): New.
15800 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15801 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15802 with local indirect jump when converting indirect call and jump.
15803 (ix86_set_indirect_branch_type): New.
15804 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15805 (indirectlabelno): New.
15806 (indirect_thunk_needed): Likewise.
15807 (indirect_thunk_bnd_needed): Likewise.
15808 (indirect_thunks_used): Likewise.
15809 (indirect_thunks_bnd_used): Likewise.
15810 (INDIRECT_LABEL): Likewise.
15811 (indirect_thunk_name): Likewise.
15812 (output_indirect_thunk): Likewise.
15813 (output_indirect_thunk_function): Likewise.
15814 (ix86_output_indirect_branch_via_reg): Likewise.
15815 (ix86_output_indirect_branch_via_push): Likewise.
15816 (ix86_output_indirect_branch): Likewise.
15817 (ix86_output_indirect_jmp): Likewise.
15818 (ix86_code_end): Call output_indirect_thunk_function if needed.
15819 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15821 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15822 (ix86_attribute_table): Add indirect_branch.
15823 * config/i386/i386.h (machine_function): Add indirect_branch_type
15824 and has_local_indirect_jump.
15825 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15827 (tablejump): Likewise.
15828 (*indirect_jump): Use ix86_output_indirect_jmp.
15829 (*tablejump_1): Likewise.
15830 (simple_return_indirect_internal): Likewise.
15831 * config/i386/i386.opt (mindirect-branch=): New option.
15832 (indirect_branch): New.
15835 (thunk-inline): Likewise.
15836 (thunk-extern): Likewise.
15837 * doc/extend.texi: Document indirect_branch function attribute.
15838 * doc/invoke.texi: Document -mindirect-branch= option.
15840 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15843 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15845 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15847 * ipa-inline.c (want_inline_small_function_p): Return false if
15848 inlining has already failed with CIF_FINAL_ERROR.
15849 (update_caller_keys): Call want_inline_small_function_p before
15851 (update_callee_keys): Likewise.
15853 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15855 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15857 (rs6000_quadword_masked_address_p): Likewise.
15858 (quad_aligned_load_p): Likewise.
15859 (quad_aligned_store_p): Likewise.
15860 (const_load_sequence_p): Add comment to describe the outer-most loop.
15861 (mimic_memory_attributes_and_flags): New function.
15862 (rs6000_gen_stvx): Likewise.
15863 (replace_swapped_aligned_store): Likewise.
15864 (rs6000_gen_lvx): Likewise.
15865 (replace_swapped_aligned_load): Likewise.
15866 (replace_swapped_load_constant): Capitalize argument name in
15867 comment describing this function.
15868 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15869 and stores that access quad-word aligned addresses and replace
15870 with stvx or lvx instructions when appropriate.
15871 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15872 New function prototype.
15873 (rs6000_quadword_masked_address_p): Likewise.
15874 (rs6000_gen_lvx): Likewise.
15875 (rs6000_gen_stvx): Likewise.
15876 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15877 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15878 when memory address is aligned.
15879 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15880 this split to select lvx instruction when memory address is aligned.
15881 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15882 instruction when memory address is aligned.
15883 (*vsx_le_perm_load_v16qi): Likewise.
15884 (four unnamed splitters): Modify to select the stvx instruction
15885 when memory is aligned.
15887 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15889 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15890 which appears in the queue multiple times.
15892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15893 Alan Hayward <alan.hayward@arm.com>
15894 David Sherwood <david.sherwood@arm.com>
15896 * tree-vectorizer.h (vec_lower_bound): New structure.
15897 (_loop_vec_info): Add check_nonzero and lower_bounds.
15898 (LOOP_VINFO_CHECK_NONZERO): New macro.
15899 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15900 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15901 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15902 fields. Make seg_len the distance travelled, not including the
15904 (dr_direction_indicator): Declare.
15905 (dr_zero_step_indicator): Likewise.
15906 (dr_known_forward_stride_p): Likewise.
15907 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15909 (runtime_alias_check_p): Allow runtime alias checks with
15911 (operator ==): Compare access_size and align.
15912 (prune_runtime_alias_test_list): Rework for new distinction between
15913 the access_size and seg_len.
15914 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15916 (get_segment_min_max): New function.
15917 (create_intersect_range_checks): Use it.
15918 (dr_step_indicator): New function.
15919 (dr_direction_indicator): Likewise.
15920 (dr_zero_step_indicator): Likewise.
15921 (dr_known_forward_stride_p): Likewise.
15922 * tree-loop-distribution.c (data_ref_segment_size): Return
15923 DR_STEP * (niters - 1).
15924 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15926 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15927 (vect_preserves_scalar_order_p): New function, split out from...
15928 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15929 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15930 (vect_vfa_access_size): New function.
15931 (vect_vfa_align): Likewise.
15932 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15933 (dump_lower_bound): New function.
15934 (vect_check_lower_bound): Likewise.
15935 (vect_small_gap_p): Likewise.
15936 (vectorizable_with_step_bound_p): Likewise.
15937 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15938 depencies if the vectorization factor is 1. Convert the checks
15939 for nonzero steps into checks on the bounds of DR_STEP. Try using
15940 a bunds check for variable steps if the minimum required step is
15941 relatively small. Update calls to the dr_with_seg_len
15942 constructor and to vect_compile_time_alias.
15943 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15945 (vect_loop_versioning): Call it.
15946 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15948 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15950 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15951 Alan Hayward <alan.hayward@arm.com>
15952 David Sherwood <david.sherwood@arm.com>
15954 * doc/sourcebuild.texi (vect_scatter_store): Document.
15955 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15957 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15959 * genopinit.c (main): Add supports_vec_scatter_store and
15960 supports_vec_scatter_store_cached to target_optabs.
15961 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15962 IFN_MASK_SCATTER_STORE.
15963 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15965 * internal-fn.h (internal_store_fn_p): Declare.
15966 (internal_fn_stored_value_index): Likewise.
15967 * internal-fn.c (scatter_store_direct): New macro.
15968 (expand_scatter_store_optab_fn): New function.
15969 (direct_scatter_store_optab_supported_p): New macro.
15970 (internal_store_fn_p): New function.
15971 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15972 IFN_MASK_SCATTER_STORE.
15973 (internal_fn_mask_index): Likewise.
15974 (internal_fn_stored_value_index): New function.
15975 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15976 for scatter stores.
15977 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15978 * optabs-query.c (supports_vec_scatter_store_p): New function.
15979 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15980 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15981 true for scatter stores.
15982 (vect_gather_scatter_fn_p): Handle scatter stores too.
15983 (vect_check_gather_scatter): Consider using scatter stores if
15984 supports_vec_scatter_store_p.
15985 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15986 scatter stores too.
15987 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15988 internal_fn_stored_value_index.
15989 (check_load_store_masking): Handle scatter stores too.
15990 (vect_get_store_rhs): Make public.
15991 (vectorizable_call): Use internal_store_fn_p.
15992 (vectorizable_store): Handle scatter store internal functions.
15993 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15994 when deciding whether the end of the group has been reached.
15995 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15996 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15997 (mask_scatter_store<mode>): New insns.
15999 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16000 Alan Hayward <alan.hayward@arm.com>
16001 David Sherwood <david.sherwood@arm.com>
16003 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
16004 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
16005 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
16007 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
16008 Use vect_truncate_gather_scatter_offset if we can't treat the
16009 operation as a normal gather load or scatter store.
16010 (get_group_load_store_type): Take the gather_scatter_info
16011 as argument. Try using a gather load or scatter store for
16012 single-element groups.
16013 (get_load_store_type): Update calls to get_group_load_store_type
16014 and vect_use_strided_gather_scatters_p.
16016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16017 Alan Hayward <alan.hayward@arm.com>
16018 David Sherwood <david.sherwood@arm.com>
16020 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
16021 optional tree argument.
16022 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
16024 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
16025 but continue to use the current value as a fallback.
16026 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
16027 to compare the updates.
16028 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
16029 (get_load_store_type): Use it when handling a strided access.
16030 (vect_get_strided_load_store_ops): New function.
16031 (vect_get_data_ptr_increment): Likewise.
16032 (vectorizable_load): Handle strided gather loads. Always pass
16033 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
16040 (mask_gather_load@var{m}): Likewise.
16041 * genopinit.c (main): Add supports_vec_gather_load and
16042 supports_vec_gather_load_cached to target_optabs.
16043 * optabs-tree.c (init_tree_optimization_optabs): Use
16044 ggc_cleared_alloc to allocate target_optabs.
16045 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16046 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16048 * internal-fn.h (internal_load_fn_p): Declare.
16049 (internal_gather_scatter_fn_p): Likewise.
16050 (internal_fn_mask_index): Likewise.
16051 (internal_gather_scatter_fn_supported_p): Likewise.
16052 * internal-fn.c (gather_load_direct): New macro.
16053 (expand_gather_load_optab_fn): New function.
16054 (direct_gather_load_optab_supported_p): New macro.
16055 (direct_internal_fn_optab): New function.
16056 (internal_load_fn_p): Likewise.
16057 (internal_gather_scatter_fn_p): Likewise.
16058 (internal_fn_mask_index): Likewise.
16059 (internal_gather_scatter_fn_supported_p): Likewise.
16060 * optabs-query.c (supports_at_least_one_mode_p): New function.
16061 (supports_vec_gather_load_p): Likewise.
16062 * optabs-query.h (supports_vec_gather_load_p): Declare.
16063 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16064 and memory_type field.
16065 (NUM_PATTERNS): Bump to 15.
16066 * tree-vect-data-refs.c: Include internal-fn.h.
16067 (vect_gather_scatter_fn_p): New function.
16068 (vect_describe_gather_scatter_call): Likewise.
16069 (vect_check_gather_scatter): Try using internal functions for
16070 gather loads. Recognize existing calls to a gather load function.
16071 (vect_analyze_data_refs): Consider using gather loads if
16072 supports_vec_gather_load_p.
16073 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16074 (vect_get_gather_scatter_offset_type): Likewise.
16075 (vect_convert_mask_for_vectype): Likewise.
16076 (vect_add_conversion_to_patterm): Likewise.
16077 (vect_try_gather_scatter_pattern): Likewise.
16078 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16079 (vect_vect_recog_func_ptrs): Add it.
16080 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16081 internal_fn_mask_index and internal_gather_scatter_fn_p.
16082 (check_load_store_masking): Take the gather_scatter_info as an
16083 argument and handle gather loads.
16084 (vect_get_gather_scatter_ops): New function.
16085 (vectorizable_call): Check internal_load_fn_p.
16086 (vectorizable_load): Likewise. Handle gather load internal
16088 (vectorizable_store): Update call to check_load_store_masking.
16089 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16090 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16091 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16092 (aarch64_gather_scale_operand_d): New predicates.
16093 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16094 (mask_gather_load<mode>): New insns.
16096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16097 Alan Hayward <alan.hayward@arm.com>
16098 David Sherwood <david.sherwood@arm.com>
16100 * optabs.def (fold_left_plus_optab): New optab.
16101 * doc/md.texi (fold_left_plus_@var{m}): Document.
16102 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16103 * internal-fn.c (fold_left_direct): Define.
16104 (expand_fold_left_optab_fn): Likewise.
16105 (direct_fold_left_optab_supported_p): Likewise.
16106 * fold-const-call.c (fold_const_fold_left): New function.
16107 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16108 * tree-parloops.c (valid_reduction_p): New function.
16109 (gather_scalar_reductions): Use it.
16110 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16111 (vect_finish_replace_stmt): Declare.
16112 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16113 (needs_fold_left_reduction_p): New function, split out from...
16114 (vect_is_simple_reduction): ...here. Accept reductions that
16115 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16116 (vect_force_simple_reduction): Also store the reduction type in
16117 the assignment's STMT_VINFO_REDUC_TYPE.
16118 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16119 (merge_with_identity): New function.
16120 (vect_expand_fold_left): Likewise.
16121 (vectorize_fold_left_reduction): Likewise.
16122 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16123 scalar phi in place for it. Check for target support and reject
16124 cases that would reassociate the operation. Defer the transform
16125 phase to vectorize_fold_left_reduction.
16126 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16127 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16128 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16132 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16133 call to ifc_temp_var.
16135 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16136 Alan Hayward <alan.hayward@arm.com>
16137 David Sherwood <david.sherwood@arm.com>
16139 * target.def (legitimize_address_displacement): Take the original
16140 offset as a poly_int.
16141 * targhooks.h (default_legitimize_address_displacement): Update
16143 * targhooks.c (default_legitimize_address_displacement): Likewise.
16144 * doc/tm.texi: Regenerate.
16145 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16146 as an argument, moving assert of ad->disp == ad->disp_term to...
16147 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16148 Try calling targetm.legitimize_address_displacement before expanding
16149 the address rather than afterwards, and adjust for the new interface.
16150 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16151 Match the new hook interface. Handle SVE addresses.
16152 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16153 new hook interface.
16155 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16157 * Makefile.in (OBJS): Add early-remat.o.
16158 * target.def (select_early_remat_modes): New hook.
16159 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16160 * doc/tm.texi: Regenerate.
16161 * targhooks.h (default_select_early_remat_modes): Declare.
16162 * targhooks.c (default_select_early_remat_modes): New function.
16163 * timevar.def (TV_EARLY_REMAT): New timevar.
16164 * passes.def (pass_early_remat): New pass.
16165 * tree-pass.h (make_pass_early_remat): Declare.
16166 * early-remat.c: New file.
16167 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16169 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16172 Alan Hayward <alan.hayward@arm.com>
16173 David Sherwood <david.sherwood@arm.com>
16175 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16176 vfm1 with a bound_epilog parameter.
16177 (vect_do_peeling): Update calls accordingly, and move the prologue
16178 call earlier in the function. Treat the base bound_epilog as 0 for
16179 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16180 this base when peeling for gaps.
16181 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16182 with fully-masked loops.
16183 (vect_estimate_min_profitable_iters): Handle the single peeled
16184 iteration in that case.
16186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16187 Alan Hayward <alan.hayward@arm.com>
16188 David Sherwood <david.sherwood@arm.com>
16190 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16191 single-element interleaving even if the size is not a power of 2.
16192 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16193 accesses for single-element interleaving if the group size is
16196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16197 Alan Hayward <alan.hayward@arm.com>
16198 David Sherwood <david.sherwood@arm.com>
16200 * doc/md.texi (fold_extract_last_@var{m}): Document.
16201 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16202 * optabs.def (fold_extract_last_optab): New optab.
16203 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16204 * internal-fn.c (fold_extract_direct): New macro.
16205 (expand_fold_extract_optab_fn): Likewise.
16206 (direct_fold_extract_optab_supported_p): Likewise.
16207 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16208 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16209 EXTRACT_LAST_REDUCTION.
16210 (get_initial_def_for_reduction): Do not create an initial vector
16211 for EXTRACT_LAST_REDUCTION reductions.
16212 (vectorizable_reduction): Leave the scalar phi in place for
16213 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16214 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16215 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16216 transform phase to vectorizable_condition.
16217 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16219 (vect_finish_stmt_generation): ...here.
16220 (vect_finish_replace_stmt): New function.
16221 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16222 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16224 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16226 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16227 Alan Hayward <alan.hayward@arm.com>
16228 David Sherwood <david.sherwood@arm.com>
16230 * doc/md.texi (extract_last_@var{m}): Document.
16231 * optabs.def (extract_last_optab): New optab.
16232 * internal-fn.def (EXTRACT_LAST): New internal function.
16233 * internal-fn.c (cond_unary_direct): New macro.
16234 (expand_cond_unary_optab_fn): Likewise.
16235 (direct_cond_unary_optab_supported_p): Likewise.
16236 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16237 loops using EXTRACT_LAST.
16238 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16239 (extract_last_<mode>): ...this optab.
16240 (vec_extract<mode><Vel>): Update accordingly.
16242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16243 Alan Hayward <alan.hayward@arm.com>
16244 David Sherwood <david.sherwood@arm.com>
16246 * target.def (empty_mask_is_expensive): New hook.
16247 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16248 * doc/tm.texi: Regenerate.
16249 * targhooks.h (default_empty_mask_is_expensive): Declare.
16250 * targhooks.c (default_empty_mask_is_expensive): New function.
16251 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16252 if the target says that empty masks are expensive.
16253 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16255 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16257 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16258 Alan Hayward <alan.hayward@arm.com>
16259 David Sherwood <david.sherwood@arm.com>
16261 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16262 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16263 (vect_use_loop_mask_for_alignment_p): New function.
16264 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16265 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16266 niters_skip argument. Make sure that the first niters_skip elements
16267 of the first iteration are inactive.
16268 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16269 Update call to vect_set_loop_masks_directly.
16270 (get_misalign_in_elems): New function, split out from...
16271 (vect_gen_prolog_loop_niters): ...here.
16272 (vect_update_init_of_dr): Take a code argument that specifies whether
16273 the adjustment should be added or subtracted.
16274 (vect_update_init_of_drs): Likewise.
16275 (vect_prepare_for_masked_peels): New function.
16276 (vect_do_peeling): Skip prologue peeling if we're using a mask
16277 instead. Update call to vect_update_inits_of_drs.
16278 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16280 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16281 alignment. Do not include the number of peeled iterations in
16282 the minimum threshold in that case.
16283 (vectorizable_induction): Adjust the start value down by
16284 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16285 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16286 Take the number of skipped iterations into account when calculating
16288 * tree-vect-stmts.c (vect_gen_while_not): New function.
16290 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16291 Alan Hayward <alan.hayward@arm.com>
16292 David Sherwood <david.sherwood@arm.com>
16294 * doc/sourcebuild.texi (vect_fully_masked): Document.
16295 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16296 default value to 0.
16297 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16299 (vect_analyze_loop_2): ...here. Don't check the vectorization
16300 factor against the number of loop iterations if the loop is
16303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16304 Alan Hayward <alan.hayward@arm.com>
16305 David Sherwood <david.sherwood@arm.com>
16307 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16308 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16309 (dump_groups): Update accordingly.
16310 (iv_use::mem_type): New member variable.
16311 (address_p): New function.
16312 (record_use): Add a mem_type argument and initialize the new
16314 (record_group_use): Add a mem_type argument. Use address_p.
16315 Remove obsolete null checks of base_object. Update call to record_use.
16316 (find_interesting_uses_op): Update call to record_group_use.
16317 (find_interesting_uses_cond): Likewise.
16318 (find_interesting_uses_address): Likewise.
16319 (get_mem_type_for_internal_fn): New function.
16320 (find_address_like_use): Likewise.
16321 (find_interesting_uses_stmt): Try find_address_like_use before
16322 calling find_interesting_uses_op.
16323 (addr_offset_valid_p): Use the iv mem_type field as the type
16324 of the addressed memory.
16325 (add_autoinc_candidates): Likewise.
16326 (get_address_cost): Likewise.
16327 (split_small_address_groups_p): Use address_p.
16328 (split_address_groups): Likewise.
16329 (add_iv_candidate_for_use): Likewise.
16330 (autoinc_possible_for_pair): Likewise.
16331 (rewrite_groups): Likewise.
16332 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16333 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16334 (get_alias_ptr_type_for_ptr_address): New function.
16335 (rewrite_use_address): Rewrite address uses in calls that were
16336 identified by find_address_like_use.
16338 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16339 Alan Hayward <alan.hayward@arm.com>
16340 David Sherwood <david.sherwood@arm.com>
16342 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16344 * gimple-expr.h (is_gimple_addressable: Likewise.
16345 * gimple-expr.c (is_gimple_address): Likewise.
16346 * internal-fn.c (expand_call_mem_ref): New function.
16347 (expand_mask_load_optab_fn): Use it.
16348 (expand_mask_store_optab_fn): Likewise.
16350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16351 Alan Hayward <alan.hayward@arm.com>
16352 David Sherwood <david.sherwood@arm.com>
16354 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16355 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16356 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16357 (cond_umax@var{mode}): Document.
16358 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16359 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16360 (cond_umin_optab, cond_umax_optab): New optabs.
16361 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16362 (COND_IOR, COND_XOR): New internal functions.
16363 * internal-fn.h (get_conditional_internal_fn): Declare.
16364 * internal-fn.c (cond_binary_direct): New macro.
16365 (expand_cond_binary_optab_fn): Likewise.
16366 (direct_cond_binary_optab_supported_p): Likewise.
16367 (get_conditional_internal_fn): New function.
16368 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16369 Cope with reduction statements that are vectorized as calls rather
16371 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16372 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16373 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16374 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16375 (UNSPEC_COND_EOR): New unspecs.
16376 (optab): Add mappings for them.
16377 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16378 (sve_int_op, sve_fp_op): New int attributes.
16380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16381 Alan Hayward <alan.hayward@arm.com>
16382 David Sherwood <david.sherwood@arm.com>
16384 * optabs.def (while_ult_optab): New optab.
16385 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16386 * internal-fn.def (WHILE_ULT): New internal function.
16387 * internal-fn.h (direct_internal_fn_supported_p): New override
16388 that takes two types as argument.
16389 * internal-fn.c (while_direct): New macro.
16390 (expand_while_optab_fn): New function.
16391 (convert_optab_supported_p): Likewise.
16392 (direct_while_optab_supported_p): New macro.
16393 * wide-int.h (wi::udiv_ceil): New function.
16394 * tree-vectorizer.h (rgroup_masks): New structure.
16395 (vec_loop_masks): New typedef.
16396 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16397 and fully_masked_p.
16398 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16399 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16400 (vect_max_vf): New function.
16401 (slpeel_make_loop_iterate_ntimes): Delete.
16402 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16403 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16404 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16405 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16406 internal-fn.h, stor-layout.h and optabs-query.h.
16407 (vect_set_loop_mask): New function.
16408 (add_preheader_seq): Likewise.
16409 (add_header_seq): Likewise.
16410 (interleave_supported_p): Likewise.
16411 (vect_maybe_permute_loop_masks): Likewise.
16412 (vect_set_loop_masks_directly): Likewise.
16413 (vect_set_loop_condition_masked): Likewise.
16414 (vect_set_loop_condition_unmasked): New function, split out from
16415 slpeel_make_loop_iterate_ntimes.
16416 (slpeel_make_loop_iterate_ntimes): Rename to..
16417 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16418 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16419 (vect_do_peeling): Update call accordingly.
16420 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16422 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16423 mask_compare_type, can_fully_mask_p and fully_masked_p.
16424 (release_vec_loop_masks): New function.
16425 (_loop_vec_info): Use it to free the loop masks.
16426 (can_produce_all_loop_masks_p): New function.
16427 (vect_get_max_nscalars_per_iter): Likewise.
16428 (vect_verify_full_masking): Likewise.
16429 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16430 retries, and free the mask rgroups before retrying. Check loop-wide
16431 reasons for disallowing fully-masked loops. Make the final decision
16432 about whether use a fully-masked loop or not.
16433 (vect_estimate_min_profitable_iters): Do not assume that peeling
16434 for the number of iterations will be needed for fully-masked loops.
16435 (vectorizable_reduction): Disable fully-masked loops.
16436 (vectorizable_live_operation): Likewise.
16437 (vect_halve_mask_nunits): New function.
16438 (vect_double_mask_nunits): Likewise.
16439 (vect_record_loop_mask): Likewise.
16440 (vect_get_loop_mask): Likewise.
16441 (vect_transform_loop): Handle the case in which the final loop
16442 iteration might handle a partial vector. Call vect_set_loop_condition
16443 instead of slpeel_make_loop_iterate_ntimes.
16444 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16445 (check_load_store_masking): New function.
16446 (prepare_load_store_mask): Likewise.
16447 (vectorizable_store): Handle fully-masked loops.
16448 (vectorizable_load): Likewise.
16449 (supportable_widening_operation): Use vect_halve_mask_nunits for
16451 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16452 (vect_gen_while): New function.
16453 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16454 (aarch64_uqdec<mode>): New insn.
16456 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16457 Alan Hayward <alan.hayward@arm.com>
16458 David Sherwood <david.sherwood@arm.com>
16460 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16461 (reduc_xor_scal_optab): New optabs.
16462 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16463 (reduc_xor_scal_@var{m}): Document.
16464 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16465 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16466 internal functions.
16467 * fold-const-call.c (fold_const_call): Handle them.
16468 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16469 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16470 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16471 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16472 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16473 (UNSPEC_XORV): New unspecs.
16474 (optab): Add entries for them.
16475 (BITWISEV): New int iterator.
16476 (bit_reduc_op): New int attributes.
16478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16479 Alan Hayward <alan.hayward@arm.com>
16480 David Sherwood <david.sherwood@arm.com>
16482 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16483 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16484 * optabs.def (vec_shl_insert_optab): New optab.
16485 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16486 (duplicate_and_interleave): Likewise.
16487 * tree-vect-loop.c: Include internal-fn.h.
16488 (neutral_op_for_slp_reduction): New function, split out from
16489 get_initial_defs_for_reduction.
16490 (get_initial_def_for_reduction): Handle option 2 for variable-length
16491 vectors by loading the neutral value into a vector and then shifting
16492 the initial value into element 0.
16493 (get_initial_defs_for_reduction): Replace the code argument with
16494 the neutral value calculated by neutral_op_for_slp_reduction.
16495 Use gimple_build_vector for constant-length vectors.
16496 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16497 but the first group_size elements have a neutral value.
16498 Use duplicate_and_interleave otherwise.
16499 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16500 Update call to get_initial_defs_for_reduction. Handle SLP
16501 reductions for variable-length vectors by creating one vector
16502 result for each scalar result, with the elements associated
16503 with other scalar results stubbed out with the neutral value.
16504 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16505 Require IFN_VEC_SHL_INSERT for double reductions on
16506 variable-length vectors, or SLP reductions that have
16507 a neutral value. Require can_duplicate_and_interleave_p
16508 support for variable-length unchained SLP reductions if there
16509 is no neutral value, such as for MIN/MAX reductions. Also require
16510 the number of vector elements to be a multiple of the number of
16511 SLP statements when doing variable-length unchained SLP reductions.
16512 Update call to vect_create_epilog_for_reduction.
16513 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16514 and remove initial values.
16515 (duplicate_and_interleave): Make public.
16516 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16517 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16519 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16520 Alan Hayward <alan.hayward@arm.com>
16521 David Sherwood <david.sherwood@arm.com>
16523 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16524 (can_duplicate_and_interleave_p): New function.
16525 (vect_get_and_check_slp_defs): Take the vector of statements
16526 rather than just the current one. Remove excess parentheses.
16527 Restriction rejectinon of vect_constant_def and vect_external_def
16528 for variable-length vectors to boolean types, or types for which
16529 can_duplicate_and_interleave_p is false.
16530 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16531 (duplicate_and_interleave): New function.
16532 (vect_get_constant_vectors): Use gimple_build_vector for
16533 constant-length vectors and suitable variable-length constant
16534 vectors. Use duplicate_and_interleave for other variable-length
16535 vectors. Don't defer the update when inserting new statements.
16537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16538 Alan Hayward <alan.hayward@arm.com>
16539 David Sherwood <david.sherwood@arm.com>
16541 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16542 min_profitable_iters doesn't go negative.
16544 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16545 Alan Hayward <alan.hayward@arm.com>
16546 David Sherwood <david.sherwood@arm.com>
16548 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16549 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16550 * optabs.def (vec_mask_load_lanes_optab): New optab.
16551 (vec_mask_store_lanes_optab): Likewise.
16552 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16553 (MASK_STORE_LANES): Likewise.
16554 * internal-fn.c (mask_load_lanes_direct): New macro.
16555 (mask_store_lanes_direct): Likewise.
16556 (expand_mask_load_optab_fn): Handle masked operations.
16557 (expand_mask_load_lanes_optab_fn): New macro.
16558 (expand_mask_store_optab_fn): Handle masked operations.
16559 (expand_mask_store_lanes_optab_fn): New macro.
16560 (direct_mask_load_lanes_optab_supported_p): Likewise.
16561 (direct_mask_store_lanes_optab_supported_p): Likewise.
16562 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16564 (vect_load_lanes_supported): Likewise.
16565 * tree-vect-data-refs.c (strip_conversion): New function.
16566 (can_group_stmts_p): Likewise.
16567 (vect_analyze_data_ref_accesses): Use it instead of checking
16568 for a pair of assignments.
16569 (vect_store_lanes_supported): Take a masked_p parameter.
16570 (vect_load_lanes_supported): Likewise.
16571 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16572 vect_store_lanes_supported and vect_load_lanes_supported.
16573 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16574 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16575 parameter. Don't allow gaps for masked accesses.
16576 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16577 and vect_load_lanes_supported.
16578 (get_load_store_type): Take a masked_p parameter and update
16579 call to get_group_load_store_type.
16580 (vectorizable_store): Update call to get_load_store_type.
16581 Handle IFN_MASK_STORE_LANES.
16582 (vectorizable_load): Update call to get_load_store_type.
16583 Handle IFN_MASK_LOAD_LANES.
16585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16586 Alan Hayward <alan.hayward@arm.com>
16587 David Sherwood <david.sherwood@arm.com>
16589 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16591 * config/aarch64/aarch64-protos.h
16592 (aarch64_sve_struct_memory_operand_p): Declare.
16593 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16594 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16595 (VPRED, vpred): Handle SVE structure modes.
16596 * config/aarch64/constraints.md (Utx): New constraint.
16597 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16598 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16599 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16600 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16601 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16602 structure modes. Split into pieces after RA.
16603 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16604 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16606 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16607 SVE structure modes.
16608 (aarch64_classify_address): Likewise.
16609 (sizetochar): Move earlier in file.
16610 (aarch64_print_operand): Handle SVE register lists.
16611 (aarch64_array_mode): New function.
16612 (aarch64_sve_struct_memory_operand_p): Likewise.
16613 (TARGET_ARRAY_MODE): Redefine.
16615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16616 Alan Hayward <alan.hayward@arm.com>
16617 David Sherwood <david.sherwood@arm.com>
16619 * target.def (array_mode): New target hook.
16620 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16621 * doc/tm.texi: Regenerate.
16622 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16623 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16624 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16625 targetm.array_mode.
16626 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16629 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16630 Alan Hayward <alan.hayward@arm.com>
16631 David Sherwood <david.sherwood@arm.com>
16633 * fold-const.c (fold_binary_loc): Check the argument types
16634 rather than the result type when testing for a vector operation.
16636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16638 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16639 * doc/tm.texi: Regenerate.
16641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16642 Alan Hayward <alan.hayward@arm.com>
16643 David Sherwood <david.sherwood@arm.com>
16645 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16646 (sve): Document new AArch64 extension.
16647 * doc/md.texi (w): Extend the description of the AArch64
16648 constraint to include SVE vectors.
16649 (Upl, Upa): Document new AArch64 predicate constraints.
16650 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16652 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16653 (msve-vector-bits=): New option.
16654 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16655 SVE when these are disabled.
16656 (sve): New extension.
16657 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16658 modes. Adjust their number of units based on aarch64_sve_vg.
16659 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16660 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16661 aarch64_addr_query_type.
16662 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16663 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16664 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16665 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16666 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16667 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16668 (aarch64_simd_imm_zero_p): Delete.
16669 (aarch64_check_zero_based_sve_index_immediate): Declare.
16670 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16671 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16672 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16673 (aarch64_sve_float_mul_immediate_p): Likewise.
16674 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16675 rather than an rtx.
16676 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16677 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16678 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16679 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16680 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16681 (aarch64_regmode_natural_size): Likewise.
16682 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16683 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16685 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16686 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16687 for VG and the SVE predicate registers.
16688 (V_ALIASES): Add a "z"-prefixed alias.
16689 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16690 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16691 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16692 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16693 (REG_CLASS_NAMES): Add entries for them.
16694 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16695 and the predicate registers.
16696 (aarch64_sve_vg): Declare.
16697 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16698 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16699 (REGMODE_NATURAL_SIZE): Define.
16700 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16702 * config/aarch64/aarch64.c: Include cfgrtl.h.
16703 (simd_immediate_info): Add a constructor for series vectors,
16704 and an associated step field.
16705 (aarch64_sve_vg): New variable.
16706 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16707 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16708 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16709 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16710 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16711 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16712 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16713 (aarch64_get_mask_mode): New functions.
16714 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16715 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16716 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16717 predicate modes and predicate registers. Explicitly restrict
16718 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16719 to store a vector mode if it is recognized by
16720 aarch64_classify_vector_mode.
16721 (aarch64_regmode_natural_size): New function.
16722 (aarch64_hard_regno_caller_save_mode): Return the original mode
16724 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16725 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16726 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16727 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16729 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16730 does not overlap dest if the function is frame-related. Handle
16732 (aarch64_split_add_offset): New function.
16733 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16734 them aarch64_add_offset.
16735 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16736 and update call to aarch64_sub_sp.
16737 (aarch64_add_cfa_expression): New function.
16738 (aarch64_expand_prologue): Pass extra temporary registers to the
16739 functions above. Handle the case in which we need to emit new
16740 DW_CFA_expressions for registers that were originally saved
16741 relative to the stack pointer, but now have to be expressed
16742 relative to the frame pointer.
16743 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16745 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16746 IP0 and IP1 values for SVE frames.
16747 (aarch64_expand_vec_series): New function.
16748 (aarch64_expand_sve_widened_duplicate): Likewise.
16749 (aarch64_expand_sve_const_vector): Likewise.
16750 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16751 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16752 into the register, rather than emitting a SET directly.
16753 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16754 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16755 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16756 (offset_9bit_signed_scaled_p): New functions.
16757 (aarch64_replicate_bitmask_imm): New function.
16758 (aarch64_bitmask_imm): Use it.
16759 (aarch64_cannot_force_const_mem): Reject expressions involving
16760 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16761 (aarch64_classify_index): Handle SVE indices, by requiring
16762 a plain register index with a scale that matches the element size.
16763 (aarch64_classify_address): Handle SVE addresses. Assert that
16764 the mode of the address is VOIDmode or an integer mode.
16765 Update call to aarch64_classify_symbol.
16766 (aarch64_classify_symbolic_expression): Update call to
16767 aarch64_classify_symbol.
16768 (aarch64_const_vec_all_in_range_p): New function.
16769 (aarch64_print_vector_float_operand): Likewise.
16770 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16771 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16772 and the FP immediates 1.0 and 0.5.
16773 (aarch64_print_address_internal): Handle SVE addresses.
16774 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16775 (aarch64_regno_regclass): Handle predicate registers.
16776 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16778 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16779 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16780 (aarch64_convert_sve_vector_bits): New function.
16781 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16782 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16783 rather than an rtx.
16784 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16785 Handle SVE vector and predicate modes. Accept VL-based constants
16786 that need only one temporary register, and VL offsets that require
16787 no temporary registers.
16788 (aarch64_conditional_register_usage): Mark the predicate registers
16789 as fixed if SVE isn't available.
16790 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16791 Return true for SVE vector and predicate modes.
16792 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16793 rather than an unsigned int. Handle SVE modes.
16794 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16796 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16798 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16799 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16800 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16801 (aarch64_sve_float_mul_immediate_p): New functions.
16802 (aarch64_sve_valid_immediate): New function.
16803 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16804 Explicitly reject structure modes. Check for INDEX constants.
16805 Handle PTRUE and PFALSE constants.
16806 (aarch64_check_zero_based_sve_index_immediate): New function.
16807 (aarch64_simd_imm_zero_p): Delete.
16808 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16809 vector modes. Accept constants in the range of CNT[BHWD].
16810 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16811 ask for an Advanced SIMD mode.
16812 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16813 (aarch64_simd_vector_alignment): Handle SVE predicates.
16814 (aarch64_vectorize_preferred_vector_alignment): New function.
16815 (aarch64_simd_vector_alignment_reachable): Use it instead of
16817 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16818 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16820 (MAX_VECT_LEN): Delete.
16821 (expand_vec_perm_d): Add a vec_flags field.
16822 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16823 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16824 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16826 (aarch64_evpc_rev): Rename to...
16827 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16828 (aarch64_evpc_rev_global): New function.
16829 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16830 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16832 (aarch64_evpc_sve_tbl): New function.
16833 (aarch64_expand_vec_perm_const_1): Update after rename of
16834 aarch64_evpc_rev. Handle SVE permutes too, trying
16835 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16836 than aarch64_evpc_tbl.
16837 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16838 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16839 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16840 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16841 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16842 (aarch64_expand_sve_vcond): New functions.
16843 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16844 of aarch64_vector_mode_p.
16845 (aarch64_dwarf_poly_indeterminate_value): New function.
16846 (aarch64_compute_pressure_classes): Likewise.
16847 (aarch64_can_change_mode_class): Likewise.
16848 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16849 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16850 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16851 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16852 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16853 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16854 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16855 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16857 (Dn, Dl, Dr): Accept const as well as const_vector.
16858 (Dz): Likewise. Compare against CONST0_RTX.
16859 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16860 of "vector" where appropriate.
16861 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16862 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16863 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16864 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16865 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16866 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16867 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16868 (v_int_equiv): Extend to SVE modes.
16869 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16871 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16872 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16873 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16874 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16875 (SVE_COND_FP_CMP): New int iterators.
16876 (perm_hilo): Handle the new unpack unspecs.
16877 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16879 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16880 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16881 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16882 (aarch64_equality_operator, aarch64_constant_vector_operand)
16883 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16884 (aarch64_sve_nonimmediate_operand): Likewise.
16885 (aarch64_sve_general_operand): Likewise.
16886 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16887 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16888 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16889 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16890 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16891 (aarch64_sve_float_arith_immediate): Likewise.
16892 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16893 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16894 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16895 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16896 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16897 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16898 (aarch64_sve_float_arith_operand): Likewise.
16899 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16900 (aarch64_sve_float_mul_operand): Likewise.
16901 (aarch64_sve_vec_perm_operand): Likewise.
16902 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16903 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16904 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16905 as well as const_vector.
16906 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16907 in file. Use CONST0_RTX and CONSTM1_RTX.
16908 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16909 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16910 Use aarch64_simd_imm_zero.
16911 * config/aarch64/aarch64-sve.md: New file.
16912 * config/aarch64/aarch64.md: Include it.
16913 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16914 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16915 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16916 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16917 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16918 (sve): New attribute.
16919 (enabled): Disable instructions with the sve attribute unless
16921 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16922 aarch64_expand_mov_immediate.
16923 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16924 CNT[BHSD] immediates.
16925 (movti): Split CONST_POLY_INT moves into two halves.
16926 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16927 Split additions that need a temporary here if the destination
16928 is the stack pointer.
16929 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16930 (*add<mode>3_poly_1): New instruction.
16931 (set_clobber_cc): New expander.
16933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16935 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16936 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16937 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16938 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16939 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16940 Change innermode from fixed_mode_size to machine_mode.
16941 (simplify_subreg): Update call accordingly. Handle a constant-sized
16942 subreg of a variable-length CONST_VECTOR.
16944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16945 Alan Hayward <alan.hayward@arm.com>
16946 David Sherwood <david.sherwood@arm.com>
16948 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16949 (add_offset_to_base): New function, split out from...
16950 (create_mem_ref): ...here. When handling a scale other than 1,
16951 check first whether the address is valid without the offset.
16952 Add it into the base if so, leaving the index and scale as-is.
16954 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16957 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16958 fold_for_warn before checking if arg2 is INTEGER_CST.
16960 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16962 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16963 (store_multiple_operation): Delete.
16964 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16965 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16966 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16967 guarded by TARGET_STRING.
16968 (rs6000_output_load_multiple): Delete.
16969 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16970 OPTION_MASK_STRING / TARGET_STRING handling.
16971 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16972 (const rs6000_opt_masks) <"string">: Change mask to 0.
16973 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16974 (MASK_STRING): Delete.
16975 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16977 (load_multiple): Delete.
16984 (store_multiple): Delete.
16991 (movmemsi_8reg): Delete.
16992 (corresponding unnamed define_insn): Delete.
16993 (movmemsi_6reg): Delete.
16994 (corresponding unnamed define_insn): Delete.
16995 (movmemsi_4reg): Delete.
16996 (corresponding unnamed define_insn): Delete.
16997 (movmemsi_2reg): Delete.
16998 (corresponding unnamed define_insn): Delete.
16999 (movmemsi_1reg): Delete.
17000 (corresponding unnamed define_insn): Delete.
17001 * config/rs6000/rs6000.opt (mno-string): New.
17002 (mstring): Replace by deprecation warning stub.
17003 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
17005 2018-01-12 Jakub Jelinek <jakub@redhat.com>
17007 * regrename.c (regrename_do_replace): If replacing the same
17008 reg multiple times, try to reuse last created gen_raw_REG.
17011 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
17012 main to workaround a bug in GDB.
17014 2018-01-12 Tom de Vries <tom@codesourcery.com>
17017 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
17019 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
17021 PR rtl-optimization/80481
17022 * ira-color.c (get_cap_member): New function.
17023 (allocnos_conflict_by_live_ranges_p): Use it.
17024 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
17025 (setup_slot_coalesced_allocno_live_ranges): Ditto.
17027 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
17030 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
17031 (*saddl_se_1): Ditto.
17032 (*ssubsi_1): Ditto.
17033 (*ssubl_se_1): Ditto.
17035 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17037 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17038 rather than wi::to_widest for DR_INITs.
17039 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17040 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17041 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17043 (vect_analyze_group_access_1): Note that here.
17045 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17047 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17048 polynomial type sizes.
17050 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17052 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17053 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17054 (gimple_add_tmp_var): Likewise.
17056 2018-01-12 Martin Liska <mliska@suse.cz>
17058 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17059 (gimple_alloc_sizes): Likewise.
17060 (dump_gimple_statistics): Use PRIu64 in printf format.
17061 * gimple.h: Change uint64_t to int.
17063 2018-01-12 Martin Liska <mliska@suse.cz>
17065 * tree-core.h: Use uint64_t instead of int.
17066 * tree.c (tree_node_counts): Likewise.
17067 (tree_node_sizes): Likewise.
17068 (dump_tree_statistics): Use PRIu64 in printf format.
17070 2018-01-12 Martin Liska <mliska@suse.cz>
17072 * Makefile.in: As qsort_chk is implemented in vec.c, add
17073 vec.o to linkage of gencfn-macros.
17074 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17075 passing the info to record_node_allocation_statistics.
17076 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17078 * ggc-common.c (struct ggc_usage): Add operator== and use
17079 it in operator< and compare function.
17080 * mem-stats.h (struct mem_usage): Likewise.
17081 * vec.c (struct vec_usage): Remove operator< and compare
17082 function. Can be simply inherited.
17084 2018-01-12 Martin Jambor <mjambor@suse.cz>
17087 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17088 * tree-ssa-math-opts.c: Include domwalk.h.
17089 (convert_mult_to_fma_1): New function.
17090 (fma_transformation_info): New type.
17091 (fma_deferring_state): Likewise.
17092 (cancel_fma_deferring): New function.
17093 (result_of_phi): Likewise.
17094 (last_fma_candidate_feeds_initial_phi): Likewise.
17095 (convert_mult_to_fma): Added deferring logic, split actual
17096 transformation to convert_mult_to_fma_1.
17097 (math_opts_dom_walker): New type.
17098 (math_opts_dom_walker::after_dom_children): New method, body moved
17099 here from pass_optimize_widening_mul::execute, added deferring logic
17101 (pass_optimize_widening_mul::execute): Moved most of code to
17102 math_opts_dom_walker::after_dom_children.
17103 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17104 * config/i386/i386.c (ix86_option_override_internal): Added
17105 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17107 2018-01-12 Richard Biener <rguenther@suse.de>
17110 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17111 inline instance vars.
17113 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17116 * config/rx/rx.c (rx_is_restricted_memory_address):
17117 Handle SUBREG case.
17119 2018-01-12 Richard Biener <rguenther@suse.de>
17121 PR tree-optimization/80846
17122 * target.def (split_reduction): New target hook.
17123 * targhooks.c (default_split_reduction): New function.
17124 * targhooks.h (default_split_reduction): Declare.
17125 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17126 target requests first reduce vectors by combining low and high
17128 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17129 (get_vectype_for_scalar_type_and_size): Export.
17130 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17131 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17132 * doc/tm.texi: Regenerate.
17133 * config/i386/i386.c (ix86_split_reduction): Implement
17134 TARGET_VECTORIZE_SPLIT_REDUCTION.
17136 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17139 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17140 in PIC mode except for TARGET_VXWORKS_RTP.
17141 * config/sparc/sparc.c: Include cfgrtl.h.
17142 (TARGET_INIT_PIC_REG): Define.
17143 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17144 (sparc_pic_register_p): New predicate.
17145 (sparc_legitimate_address_p): Use it.
17146 (sparc_legitimize_pic_address): Likewise.
17147 (sparc_delegitimize_address): Likewise.
17148 (sparc_mode_dependent_address_p): Likewise.
17149 (gen_load_pcrel_sym): Remove 4th parameter.
17150 (load_got_register): Adjust call to above. Remove obsolete stuff.
17151 (sparc_expand_prologue): Do not call load_got_register here.
17152 (sparc_flat_expand_prologue): Likewise.
17153 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17154 (sparc_use_pseudo_pic_reg): New function.
17155 (sparc_init_pic_reg): Likewise.
17156 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17157 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17159 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17161 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17162 Add item for branch_cost.
17164 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17166 PR rtl-optimization/83565
17167 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17168 not extend the result to a larger mode for rotate operations.
17169 (num_sign_bit_copies1): Likewise.
17171 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17174 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17176 Use values-Xc.o for -pedantic.
17177 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17179 2018-01-12 Martin Liska <mliska@suse.cz>
17182 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17184 (possible_polymorphic_call_targets): Use it.
17185 (ipa_devirt): Likewise.
17187 2018-01-12 Martin Liska <mliska@suse.cz>
17189 * profile-count.h (enum profile_quality): Use 0 as invalid
17190 enum value of profile_quality.
17192 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17194 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17195 -mext-string options.
17197 2018-01-12 Richard Biener <rguenther@suse.de>
17199 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17200 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17201 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17203 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17205 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17207 * configure.ac (--with-long-double-format): Add support for the
17208 configuration option to change the default long double format on
17210 * config.gcc (powerpc*-linux*-*): Likewise.
17211 * configure: Regenerate.
17212 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17213 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17214 used without modification.
17216 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17218 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17219 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17220 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17221 MISC_BUILTIN_SPEC_BARRIER.
17222 (rs6000_init_builtins): Likewise.
17223 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17225 (speculation_barrier): New define_insn.
17226 * doc/extend.texi: Document __builtin_speculation_barrier.
17228 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17231 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17232 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17233 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17235 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17236 integral modes instead of "ss" and "sd".
17237 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17238 vectors with 32-bit and 64-bit elements.
17239 (vecdupssescalarmodesuffix): New mode attribute.
17240 (vec_dup<mode>): Use it.
17242 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17245 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17246 frame if argument is passed on stack.
17248 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17251 * ree.c (combine_reaching_defs): Optimize also
17252 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17253 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17255 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17257 PR middle-end/83189
17258 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17260 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17262 PR middle-end/83718
17263 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17264 after they are computed.
17266 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17268 PR tree-optimization/83695
17269 * gimple-loop-linterchange.cc
17270 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17271 reset cached scev information after interchange.
17272 (pass_linterchange::execute): Remove call to scev_reset_htab.
17274 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17276 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17277 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17278 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17279 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17280 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17281 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17282 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17283 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17284 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17285 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17286 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17287 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17288 (V_lane_reg): Likewise.
17289 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17291 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17292 (vfmal_lane_low<mode>_intrinsic,
17293 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17294 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17295 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17296 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17297 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17298 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17300 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17302 * config/arm/arm-cpus.in (fp16fml): New feature.
17303 (ALL_SIMD): Add fp16fml.
17304 (armv8.2-a): Add fp16fml as an option.
17305 (armv8.3-a): Likewise.
17306 (armv8.4-a): Add fp16fml as part of fp16.
17307 * config/arm/arm.h (TARGET_FP16FML): Define.
17308 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17310 * config/arm/arm-modes.def (V2HF): Define.
17311 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17312 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17313 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17314 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17315 vfmsl_low, vfmsl_high): New set of builtins.
17316 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17317 (vfml_op): New code attribute.
17318 (VFMLHALVES): New int iterator.
17319 (VFML, VFMLSEL): New mode attributes.
17320 (V_reg): Define mapping for V2HF.
17321 (V_hi, V_lo): New mode attributes.
17322 (VF_constraint): Likewise.
17323 (vfml_half, vfml_half_selector): New int attributes.
17324 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17326 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17327 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17329 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17330 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17331 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17332 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17334 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17335 Document new effective target and option set.
17337 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17339 * config/arm/arm-cpus.in (armv8_4): New feature.
17340 (ARMv8_4a): New fgroup.
17341 (armv8.4-a): New arch.
17342 * config/arm/arm-tables.opt: Regenerate.
17343 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17344 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17345 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17346 Add matching rules for -march=armv8.4-a and extensions.
17347 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17349 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17352 * config/rx/rx.md (BW): New mode attribute.
17353 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17355 2018-01-11 Richard Biener <rguenther@suse.de>
17357 PR tree-optimization/83435
17358 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17359 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17360 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17362 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17363 Alan Hayward <alan.hayward@arm.com>
17364 David Sherwood <david.sherwood@arm.com>
17366 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17368 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17369 (aarch64_print_address_internal): Use it to check for a zero offset.
17371 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17372 Alan Hayward <alan.hayward@arm.com>
17373 David Sherwood <david.sherwood@arm.com>
17375 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17376 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17377 Return a poly_int64 rather than a HOST_WIDE_INT.
17378 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17379 rather than a HOST_WIDE_INT.
17380 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17381 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17382 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17383 final_offset from HOST_WIDE_INT to poly_int64.
17384 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17385 to_constant when getting the number of units in an Advanced SIMD
17387 (aarch64_builtin_vectorized_function): Check for a constant number
17389 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17391 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17392 attribute instead of GET_MODE_NUNITS.
17393 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17394 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17395 GET_MODE_SIZE for fixed-size registers.
17396 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17397 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17398 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17399 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17400 (aarch64_print_operand, aarch64_print_address_internal)
17401 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17402 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17403 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17404 Handle polynomial GET_MODE_SIZE.
17405 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17406 wider than SImode without modification.
17407 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17408 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17409 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17410 passing and returning SVE modes.
17411 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17412 rather than GEN_INT.
17413 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17414 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17415 (aarch64_allocate_and_probe_stack_space): Likewise.
17416 (aarch64_layout_frame): Cope with polynomial offsets.
17417 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17418 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17419 polynomial offsets.
17420 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17421 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17422 poly_int64 rather than a HOST_WIDE_INT.
17423 (aarch64_get_separate_components, aarch64_process_components)
17424 (aarch64_expand_prologue, aarch64_expand_epilogue)
17425 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17426 (aarch64_anchor_offset): New function, split out from...
17427 (aarch64_legitimize_address): ...here.
17428 (aarch64_builtin_vectorization_cost): Handle polynomial
17429 TYPE_VECTOR_SUBPARTS.
17430 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17432 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17433 number of elements from the PARALLEL rather than the mode.
17434 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17435 rather than GET_MODE_BITSIZE.
17436 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17437 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17438 (aarch64_expand_vec_perm_const_1): Handle polynomial
17439 d->perm.length () and d->perm elements.
17440 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17441 Apply to_constant to d->perm elements.
17442 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17443 polynomial CONST_VECTOR_NUNITS.
17444 (aarch64_move_pointer): Take amount as a poly_int64 rather
17446 (aarch64_progress_pointer): Avoid temporary variable.
17447 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17448 the mode attribute instead of GET_MODE.
17450 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17451 Alan Hayward <alan.hayward@arm.com>
17452 David Sherwood <david.sherwood@arm.com>
17454 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17455 x exists before using it.
17456 (aarch64_add_constant_internal): Rename to...
17457 (aarch64_add_offset_1): ...this. Replace regnum with separate
17458 src and dest rtxes. Handle the case in which they're different,
17459 including when the offset is zero. Replace scratchreg with an rtx.
17460 Use 2 additions if there is no spare register into which we can
17461 move a 16-bit constant.
17462 (aarch64_add_constant): Delete.
17463 (aarch64_add_offset): Replace reg with separate src and dest
17464 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17465 Use aarch64_add_offset_1.
17466 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17467 an rtx rather than an int. Take the delta as a poly_int64
17468 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17469 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17470 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17471 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17472 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17473 and aarch64_add_sp.
17474 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17475 aarch64_add_constant.
17477 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17479 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17480 Use scalar_float_mode.
17482 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17484 * config/aarch64/aarch64-simd.md
17485 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17486 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17487 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17488 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17489 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17490 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17491 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17492 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17493 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17494 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17496 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17499 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17500 targ_options->x_arm_arch_string is non NULL.
17502 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17504 * config/aarch64/aarch64.h
17505 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17507 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17510 * expmed.c (emit_store_flag_force): Swap if const op0
17511 and change VOIDmode to mode of op0.
17513 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17515 PR rtl-optimization/83761
17516 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17517 than bytes to mode_for_size.
17519 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17521 PR middle-end/83189
17522 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17523 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17526 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17528 PR middle-end/83575
17529 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17530 when in layout mode.
17531 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17532 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17535 2018-01-10 Michael Collison <michael.collison@arm.com>
17537 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17538 * config/aarch64/aarch64-option-extension.def: Add
17539 AARCH64_OPT_EXTENSION of 'fp16fml'.
17540 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17541 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17542 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17543 * config/aarch64/constraints.md (Ui7): New constraint.
17544 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17545 (VFMLA_SEL_W): Ditto.
17548 (VFMLA16_LOW): New int iterator.
17549 (VFMLA16_HIGH): Ditto.
17550 (UNSPEC_FMLAL): New unspec.
17551 (UNSPEC_FMLSL): Ditto.
17552 (UNSPEC_FMLAL2): Ditto.
17553 (UNSPEC_FMLSL2): Ditto.
17554 (f16mac): New code attribute.
17555 * config/aarch64/aarch64-simd-builtins.def
17556 (aarch64_fmlal_lowv2sf): Ditto.
17557 (aarch64_fmlsl_lowv2sf): Ditto.
17558 (aarch64_fmlalq_lowv4sf): Ditto.
17559 (aarch64_fmlslq_lowv4sf): Ditto.
17560 (aarch64_fmlal_highv2sf): Ditto.
17561 (aarch64_fmlsl_highv2sf): Ditto.
17562 (aarch64_fmlalq_highv4sf): Ditto.
17563 (aarch64_fmlslq_highv4sf): Ditto.
17564 (aarch64_fmlal_lane_lowv2sf): Ditto.
17565 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17566 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17567 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17568 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17569 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17570 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17571 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17572 (aarch64_fmlal_lane_highv2sf): Ditto.
17573 (aarch64_fmlsl_lane_highv2sf): Ditto.
17574 (aarch64_fmlal_laneq_highv2sf): Ditto.
17575 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17576 (aarch64_fmlalq_lane_highv4sf): Ditto.
17577 (aarch64_fmlsl_lane_highv4sf): Ditto.
17578 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17579 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17580 * config/aarch64/aarch64-simd.md:
17581 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17582 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17583 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17584 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17585 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17586 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17587 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17588 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17589 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17590 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17591 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17592 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17593 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17594 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17595 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17596 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17597 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17598 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17599 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17600 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17601 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17602 (vfmlsl_low_u32): Ditto.
17603 (vfmlalq_low_u32): Ditto.
17604 (vfmlslq_low_u32): Ditto.
17605 (vfmlal_high_u32): Ditto.
17606 (vfmlsl_high_u32): Ditto.
17607 (vfmlalq_high_u32): Ditto.
17608 (vfmlslq_high_u32): Ditto.
17609 (vfmlal_lane_low_u32): Ditto.
17610 (vfmlsl_lane_low_u32): Ditto.
17611 (vfmlal_laneq_low_u32): Ditto.
17612 (vfmlsl_laneq_low_u32): Ditto.
17613 (vfmlalq_lane_low_u32): Ditto.
17614 (vfmlslq_lane_low_u32): Ditto.
17615 (vfmlalq_laneq_low_u32): Ditto.
17616 (vfmlslq_laneq_low_u32): Ditto.
17617 (vfmlal_lane_high_u32): Ditto.
17618 (vfmlsl_lane_high_u32): Ditto.
17619 (vfmlal_laneq_high_u32): Ditto.
17620 (vfmlsl_laneq_high_u32): Ditto.
17621 (vfmlalq_lane_high_u32): Ditto.
17622 (vfmlslq_lane_high_u32): Ditto.
17623 (vfmlalq_laneq_high_u32): Ditto.
17624 (vfmlslq_laneq_high_u32): Ditto.
17625 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17626 (AARCH64_FL_FOR_ARCH8_4): New.
17627 (AARCH64_ISA_F16FML): New ISA flag.
17628 (TARGET_F16FML): New feature flag for fp16fml.
17629 (doc/invoke.texi): Document new fp16fml option.
17631 2018-01-10 Michael Collison <michael.collison@arm.com>
17633 * config/aarch64/aarch64-builtins.c:
17634 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17635 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17636 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17637 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17638 (AARCH64_ISA_SHA3): New ISA flag.
17639 (TARGET_SHA3): New feature flag for sha3.
17640 * config/aarch64/iterators.md (sha512_op): New int attribute.
17641 (CRYPTO_SHA512): New int iterator.
17642 (UNSPEC_SHA512H): New unspec.
17643 (UNSPEC_SHA512H2): Ditto.
17644 (UNSPEC_SHA512SU0): Ditto.
17645 (UNSPEC_SHA512SU1): Ditto.
17646 * config/aarch64/aarch64-simd-builtins.def
17647 (aarch64_crypto_sha512hqv2di): New builtin.
17648 (aarch64_crypto_sha512h2qv2di): Ditto.
17649 (aarch64_crypto_sha512su0qv2di): Ditto.
17650 (aarch64_crypto_sha512su1qv2di): Ditto.
17651 (aarch64_eor3qv8hi): Ditto.
17652 (aarch64_rax1qv2di): Ditto.
17653 (aarch64_xarqv2di): Ditto.
17654 (aarch64_bcaxqv8hi): Ditto.
17655 * config/aarch64/aarch64-simd.md:
17656 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17657 (aarch64_crypto_sha512su0qv2di): Ditto.
17658 (aarch64_crypto_sha512su1qv2di): Ditto.
17659 (aarch64_eor3qv8hi): Ditto.
17660 (aarch64_rax1qv2di): Ditto.
17661 (aarch64_xarqv2di): Ditto.
17662 (aarch64_bcaxqv8hi): Ditto.
17663 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17664 (vsha512h2q_u64): Ditto.
17665 (vsha512su0q_u64): Ditto.
17666 (vsha512su1q_u64): Ditto.
17667 (veor3q_u16): Ditto.
17668 (vrax1q_u64): Ditto.
17669 (vxarq_u64): Ditto.
17670 (vbcaxq_u16): Ditto.
17671 * config/arm/types.md (crypto_sha512): New type attribute.
17672 (crypto_sha3): Ditto.
17673 (doc/invoke.texi): Document new sha3 option.
17675 2018-01-10 Michael Collison <michael.collison@arm.com>
17677 * config/aarch64/aarch64-builtins.c:
17678 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17679 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17680 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17681 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17682 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17683 (AARCH64_ISA_SM4): New ISA flag.
17684 (TARGET_SM4): New feature flag for sm4.
17685 * config/aarch64/aarch64-simd-builtins.def
17686 (aarch64_sm3ss1qv4si): Ditto.
17687 (aarch64_sm3tt1aq4si): Ditto.
17688 (aarch64_sm3tt1bq4si): Ditto.
17689 (aarch64_sm3tt2aq4si): Ditto.
17690 (aarch64_sm3tt2bq4si): Ditto.
17691 (aarch64_sm3partw1qv4si): Ditto.
17692 (aarch64_sm3partw2qv4si): Ditto.
17693 (aarch64_sm4eqv4si): Ditto.
17694 (aarch64_sm4ekeyqv4si): Ditto.
17695 * config/aarch64/aarch64-simd.md:
17696 (aarch64_sm3ss1qv4si): Ditto.
17697 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17698 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17699 (aarch64_sm4eqv4si): Ditto.
17700 (aarch64_sm4ekeyqv4si): Ditto.
17701 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17702 (sm3part_op): Ditto.
17703 (CRYPTO_SM3TT): Ditto.
17704 (CRYPTO_SM3PART): Ditto.
17705 (UNSPEC_SM3SS1): New unspec.
17706 (UNSPEC_SM3TT1A): Ditto.
17707 (UNSPEC_SM3TT1B): Ditto.
17708 (UNSPEC_SM3TT2A): Ditto.
17709 (UNSPEC_SM3TT2B): Ditto.
17710 (UNSPEC_SM3PARTW1): Ditto.
17711 (UNSPEC_SM3PARTW2): Ditto.
17712 (UNSPEC_SM4E): Ditto.
17713 (UNSPEC_SM4EKEY): Ditto.
17714 * config/aarch64/constraints.md (Ui2): New constraint.
17715 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17716 * config/arm/types.md (crypto_sm3): New type attribute.
17717 (crypto_sm4): Ditto.
17718 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17719 (vsm3tt1aq_u32): Ditto.
17720 (vsm3tt1bq_u32): Ditto.
17721 (vsm3tt2aq_u32): Ditto.
17722 (vsm3tt2bq_u32): Ditto.
17723 (vsm3partw1q_u32): Ditto.
17724 (vsm3partw2q_u32): Ditto.
17725 (vsm4eq_u32): Ditto.
17726 (vsm4ekeyq_u32): Ditto.
17727 (doc/invoke.texi): Document new sm4 option.
17729 2018-01-10 Michael Collison <michael.collison@arm.com>
17731 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17732 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17733 (AARCH64_FL_FOR_ARCH8_4): New.
17734 (AARCH64_FL_V8_4): New flag.
17735 (doc/invoke.texi): Document new armv8.4-a option.
17737 2018-01-10 Michael Collison <michael.collison@arm.com>
17739 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17740 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17741 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17742 * config/aarch64/aarch64-option-extension.def: Add
17743 AARCH64_OPT_EXTENSION of 'sha2'.
17744 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17745 (crypto): Disable sha2 and aes if crypto disabled.
17746 (crypto): Enable aes and sha2 if enabled.
17747 (simd): Disable sha2 and aes if simd disabled.
17748 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17750 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17751 (TARGET_SHA2): New feature flag for sha2.
17752 (TARGET_AES): New feature flag for aes.
17753 * config/aarch64/aarch64-simd.md:
17754 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17755 conditional on TARGET_AES.
17756 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17757 (aarch64_crypto_sha1hsi): Make pattern conditional
17759 (aarch64_crypto_sha1hv4si): Ditto.
17760 (aarch64_be_crypto_sha1hv4si): Ditto.
17761 (aarch64_crypto_sha1su1v4si): Ditto.
17762 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17763 (aarch64_crypto_sha1su0v4si): Ditto.
17764 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17765 (aarch64_crypto_sha256su0v4si): Ditto.
17766 (aarch64_crypto_sha256su1v4si): Ditto.
17767 (doc/invoke.texi): Document new aes and sha2 options.
17769 2018-01-10 Martin Sebor <msebor@redhat.com>
17771 PR tree-optimization/83781
17772 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17775 2018-01-11 Martin Sebor <msebor@gmail.com>
17776 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17778 PR tree-optimization/83501
17779 PR tree-optimization/81703
17781 * tree-ssa-strlen.c (get_string_cst): Rename...
17782 (get_string_len): ...to this. Handle global constants.
17783 (handle_char_store): Adjust.
17785 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17786 Jim Wilson <jimw@sifive.com>
17788 * config/riscv/riscv-protos.h (riscv_output_return): New.
17789 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17790 (riscv_attribute_table, riscv_output_return),
17791 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17792 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17793 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17795 (riscv_expand_prologue): Add early return for naked function.
17796 (riscv_expand_epilogue): Likewise.
17797 (riscv_function_ok_for_sibcall): Return false for naked function.
17798 (riscv_set_current_function): New.
17799 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17800 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17801 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17802 * doc/extend.texi (RISC-V Function Attributes): New.
17804 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17806 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17807 check for 128-bit long double before checking TCmode.
17808 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17809 128-bit long doubles before checking TFmode or TCmode.
17810 (FLOAT128_IBM_P): Likewise.
17812 2018-01-10 Martin Sebor <msebor@redhat.com>
17814 PR tree-optimization/83671
17815 * builtins.c (c_strlen): Unconditionally return zero for the empty
17817 Use -Warray-bounds for warnings.
17818 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17819 for non-constant array indices with COMPONENT_REF, arrays of
17820 arrays, and pointers to arrays.
17821 (gimple_fold_builtin_strlen): Determine and set length range for
17822 non-constant character arrays.
17824 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17826 PR middle-end/81897
17827 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17830 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17832 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17834 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17837 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17838 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17839 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17840 indexed_or_indirect_operand predicate.
17841 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17842 (*vsx_le_perm_load_v8hi): Likewise.
17843 (*vsx_le_perm_load_v16qi): Likewise.
17844 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17845 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17846 (*vsx_le_perm_store_v8hi): Likewise.
17847 (*vsx_le_perm_store_v16qi): Likewise.
17848 (eight unnamed splitters): Likewise.
17850 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17852 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17853 * config/rs6000/emmintrin.h: Likewise.
17854 * config/rs6000/mmintrin.h: Likewise.
17855 * config/rs6000/xmmintrin.h: Likewise.
17857 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17860 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17862 * tree.c (tree_nop_conversion): Return true for location wrapper
17864 (maybe_wrap_with_location): New function.
17865 (selftest::check_strip_nops): New function.
17866 (selftest::test_location_wrappers): New function.
17867 (selftest::tree_c_tests): Call it.
17868 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17869 (maybe_wrap_with_location): New decl.
17870 (EXPR_LOCATION_WRAPPER_P): New macro.
17871 (location_wrapper_p): New inline function.
17872 (tree_strip_any_location_wrapper): New inline function.
17874 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17877 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17878 stack_realign_offset for the largest alignment of stack slot
17880 (ix86_find_max_used_stack_alignment): New function.
17881 (ix86_finalize_stack_frame_flags): Use it. Set
17882 max_used_stack_alignment if we don't realign stack.
17883 * config/i386/i386.h (machine_function): Add
17884 max_used_stack_alignment.
17886 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17888 * config/arm/arm.opt (-mbranch-cost): New option.
17889 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17892 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17895 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17896 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17898 2018-01-10 Richard Biener <rguenther@suse.de>
17901 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17902 early out so it also covers the case where we have a non-NULL
17905 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17907 PR tree-optimization/83753
17908 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17909 for non-strided grouped accesses if the number of elements is 1.
17911 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17914 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17915 * i386.h (TARGET_USE_GATHER): Define.
17916 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17918 2018-01-10 Martin Liska <mliska@suse.cz>
17921 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17922 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17924 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17925 CLEANUP_NO_PARTITIONING is not set.
17927 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17929 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17930 for vectors, as a partial revert of r254296.
17931 * rtl.h (const_vec_p): Delete.
17932 (const_vec_duplicate_p): Don't test for vector CONSTs.
17933 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17934 * expmed.c (make_tree): Likewise.
17937 * common.md (E, F): Use CONSTANT_P instead of checking for
17939 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17940 checking for CONST_VECTOR.
17942 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17944 PR middle-end/83575
17945 * predict.c (force_edge_cold): Handle in more sane way edges
17946 with no prediction.
17948 2018-01-09 Carl Love <cel@us.ibm.com>
17950 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17952 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17953 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17954 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17955 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17956 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17957 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17958 * config/rs6000/rs6000-protos.h: Add extern defition for
17959 rs6000_generate_float2_double_code.
17960 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17962 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17963 (float2_v2df): Add define_expand.
17965 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17968 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17969 op_mode in the force_to_mode call.
17971 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17973 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17974 instead of checking each element individually.
17975 (aarch64_evpc_uzp): Likewise.
17976 (aarch64_evpc_zip): Likewise.
17977 (aarch64_evpc_ext): Likewise.
17978 (aarch64_evpc_rev): Likewise.
17979 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17980 instead of checking each element individually. Return true without
17982 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17983 whether all selected elements come from the same input, instead of
17984 checking each element individually. Remove calls to gen_rtx_REG,
17985 start_sequence and end_sequence and instead assert that no rtl is
17988 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17990 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17991 order of HIGH and CONST checks.
17993 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17995 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17996 if the destination isn't an SSA_NAME.
17998 2018-01-09 Richard Biener <rguenther@suse.de>
18000 PR tree-optimization/83668
18001 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
18003 (canonicalize_loop_form): ... here, renamed from ...
18004 (canonicalize_loop_closed_ssa_form): ... this and amended to
18005 swap successor edges for loop exit blocks to make us use
18006 the RPO order we need for initial schedule generation.
18008 2018-01-09 Joseph Myers <joseph@codesourcery.com>
18010 PR tree-optimization/64811
18011 * match.pd: When optimizing comparisons with Inf, avoid
18012 introducing or losing exceptions from comparisons with NaN.
18014 2018-01-09 Martin Liska <mliska@suse.cz>
18017 * asan.c (shadow_mem_size): Add gcc_assert.
18019 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
18021 Don't save registers in main().
18024 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
18025 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
18026 * config/avr/avr.c (avr_set_current_function): Don't error if
18027 naked, OS_task or OS_main are specified at the same time.
18028 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
18030 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
18032 * common/config/avr/avr-common.c (avr_option_optimization_table):
18033 Switch on -mmain-is-OS_task for optimizing compilations.
18035 2018-01-09 Richard Biener <rguenther@suse.de>
18037 PR tree-optimization/83572
18038 * graphite.c: Include cfganal.h.
18039 (graphite_transform_loops): Connect infinite loops to exit
18040 and remove fake edges at the end.
18042 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18044 * ipa-inline.c (edge_badness): Revert accidental checkin.
18046 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18049 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18050 symbols; not inline clones.
18052 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18055 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18056 hard registers. Formatting fixes.
18058 PR preprocessor/83722
18059 * gcc.c (try_generate_repro): Pass
18060 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18061 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18064 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18065 Kito Cheng <kito.cheng@gmail.com>
18067 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18068 (riscv_leaf_function_p): Delete.
18069 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18071 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18073 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18075 (do_ifelse): New function.
18076 (do_isel): New function.
18077 (do_sub3): New function.
18078 (do_add3): New function.
18079 (do_load_mask_compare): New function.
18080 (do_overlap_load_compare): New function.
18081 (expand_compare_loop): New function.
18082 (expand_block_compare): Call expand_compare_loop() when appropriate.
18083 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18084 option description.
18085 (-mblock-compare-inline-loop-limit): New option.
18087 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18090 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18091 Reverse order of second and third operands in first alternative.
18092 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18093 of first and second elements in UNSPEC_VPERMR vector.
18094 (altivec_expand_vec_perm_le): Likewise.
18096 2018-01-08 Jeff Law <law@redhat.com>
18098 PR rtl-optimizatin/81308
18099 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18100 (process_switch): If group_case_labels makes a change, then set
18102 (pass_convert_switch::execute): If a switch is converted, then
18103 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18105 PR rtl-optimization/81308
18106 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18109 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18111 PR target/83663 - Revert r255946
18112 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18113 generation for cases where splatting a value is not useful.
18114 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18115 across a vec_duplicate and a paradoxical subreg forming a vector
18116 mode to a vec_concat.
18118 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18120 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18121 -march=armv8.3-a variants.
18122 * config/arm/t-multilib: Likewise.
18123 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18125 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18127 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18129 (cceq_ior_compare_complement): Give it a name so I can use it, and
18130 change boolean_or_operator predicate to boolean_operator so it can
18131 be used to generate a crand.
18132 (eqne): New code iterator.
18133 (bd/bd_neg): New code_attrs.
18134 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18135 a single define_insn.
18136 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18137 decrement (bdnzt/bdnzf/bdzt/bdzf).
18138 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18139 with the new names of the branch decrement patterns, and added the
18140 names of the branch decrement conditional patterns.
18142 2018-01-08 Richard Biener <rguenther@suse.de>
18144 PR tree-optimization/83563
18145 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18148 2018-01-08 Richard Biener <rguenther@suse.de>
18150 PR middle-end/83713
18151 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18153 2018-01-08 Richard Biener <rguenther@suse.de>
18155 PR tree-optimization/83685
18156 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18157 references to abnormals.
18159 2018-01-08 Richard Biener <rguenther@suse.de>
18162 * dwarf2out.c (output_indirect_strings): Handle empty
18163 skeleton_debug_str_hash.
18164 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18166 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18168 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18169 (emit_store_direct): Likewise.
18170 (arc_trampoline_adjust_address): Likewise.
18171 (arc_asm_trampoline_template): New function.
18172 (arc_initialize_trampoline): Use asm_trampoline_template.
18173 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18174 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18175 * config/arc/arc.md (flush_icache): Delete pattern.
18177 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18179 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18180 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18183 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18186 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18187 by not USED_FOR_TARGET.
18188 (make_pass_resolve_sw_modes): Likewise.
18190 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18192 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18195 2018-01-08 Richard Biener <rguenther@suse.de>
18197 PR middle-end/83580
18198 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18200 2018-01-08 Richard Biener <rguenther@suse.de>
18202 PR middle-end/83517
18203 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18205 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18207 PR middle-end/81897
18208 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18209 basic blocks with a small number of successors.
18210 (convert_control_dep_chain_into_preds): Improve handling of
18212 (dump_predicates): Split apart into...
18213 (dump_pred_chain): ...here...
18214 (dump_pred_info): ...and here.
18215 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18216 (can_chain_union_be_invalidated_p): Improve check for invalidation
18218 (uninit_uses_cannot_happen): Avoid unnecessary if
18219 convert_control_dep_chain_into_preds yielded nothing.
18221 2018-01-06 Martin Sebor <msebor@redhat.com>
18223 PR tree-optimization/83640
18224 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18225 subtracting negative offset from size.
18226 (builtin_access::overlap): Adjust offset bounds of the access to fall
18227 within the size of the object if possible.
18229 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18231 PR rtl-optimization/83699
18232 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18233 extract_bit_field_as_subreg to cases in which the extracted
18234 value is also a vector.
18236 * lra-constraints.c (process_alt_operands): Test for the equivalence
18237 substitutions when detecting a possible reload cycle.
18239 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18242 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18243 by default if flag_selective_schedling{,2}. Formatting fixes.
18245 PR rtl-optimization/83682
18246 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18247 if it has non-VECTOR_MODE element mode.
18248 (vec_duplicate_p): Likewise.
18250 PR middle-end/83694
18251 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18252 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18254 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18257 * config/i386/i386-builtin.def
18258 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18259 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18260 Require also OPTION_MASK_ISA_AVX512F in addition to
18261 OPTION_MASK_ISA_GFNI.
18262 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18263 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18264 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18265 to OPTION_MASK_ISA_GFNI.
18266 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18267 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18268 OPTION_MASK_ISA_AVX512BW.
18269 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18270 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18271 addition to OPTION_MASK_ISA_GFNI.
18272 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18273 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18274 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18275 to OPTION_MASK_ISA_GFNI.
18276 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18277 a requirement for all ISAs rather than any of them with a few
18279 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18281 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18282 bitmasks to be enabled with 3 exceptions, instead of requiring any
18283 enabled ISA with lots of exceptions.
18284 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18285 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18286 Change avx512bw in isa attribute to avx512f.
18287 * config/i386/sgxintrin.h: Add license boilerplate.
18288 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18289 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18290 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18291 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18293 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18294 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18295 temporarily sse2 rather than sse if not enabled already.
18298 * config/i386/sse.md (VI248_VLBW): Rename to ...
18299 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18300 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18301 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18302 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18303 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18304 mode iterator instead of VI248_VLBW.
18306 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18308 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18309 (record_modified): Skip clobbers; add debug output.
18310 (param_change_prob): Use sreal frequencies.
18312 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18314 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18315 punt for user-aligned variables.
18317 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18319 * tree-chrec.c (chrec_contains_symbols): Return true for
18322 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18325 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18326 of (x|y) == x for BICS pattern.
18328 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18330 PR tree-optimization/83605
18331 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18332 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18335 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18337 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18338 * config/epiphany/rtems.h: New file.
18340 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18341 Uros Bizjak <ubizjak@gmail.com>
18344 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18345 QIreg_operand instead of register_operand predicate.
18346 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18347 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18348 comments instead of -fmitigate[-_]rop.
18350 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18353 * cgraphunit.c (symbol_table::compile): Switch to text_section
18354 before calling assembly_start debug hook.
18355 * run-rtl-passes.c (run_rtl_passes): Likewise.
18358 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18360 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18361 range_int_cst_p rather than !symbolic_range_p before calling
18362 extract_range_from_multiplicative_op_1.
18364 2018-01-04 Jeff Law <law@redhat.com>
18366 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18367 redundant test in assertion.
18369 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18371 * doc/rtl.texi: Document machine_mode wrapper classes.
18373 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18375 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18376 using tree_to_uhwi.
18378 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18380 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18381 the VEC_PERM_EXPR fold to fail.
18383 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18386 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18387 to switched_sections.
18389 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18392 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18393 test for d.testing.
18395 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18398 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18399 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18401 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18404 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18405 is BLKmode and bitpos not zero or mode change is needed.
18407 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18410 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18413 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18416 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18417 instead of MULT rtx. Update all corresponding splitters.
18418 (*saddl_se): Ditto.
18419 (*ssub<modesuffix>): Ditto.
18420 (*ssubl_se): Ditto.
18421 (*cmp_sadd_di): Update split patterns.
18422 (*cmp_sadd_si): Ditto.
18423 (*cmp_sadd_sidi): Ditto.
18424 (*cmp_ssub_di): Ditto.
18425 (*cmp_ssub_si): Ditto.
18426 (*cmp_ssub_sidi): Ditto.
18427 * config/alpha/predicates.md (const23_operand): New predicate.
18428 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18429 Look for ASHIFT, not MULT inner operand.
18430 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18432 2018-01-04 Martin Liska <mliska@suse.cz>
18434 PR gcov-profile/83669
18435 * gcov.c (output_intermediate_file): Add version to intermediate
18437 * doc/gcov.texi: Document new field 'version' in intermediate
18438 file format. Fix location of '-k' option of gcov command.
18440 2018-01-04 Martin Liska <mliska@suse.cz>
18443 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18445 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18447 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18449 2018-01-03 Martin Sebor <msebor@redhat.com>
18451 PR tree-optimization/83655
18452 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18453 checking calls with invalid arguments.
18455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18457 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18458 (vectorizable_mask_load_store): Delete.
18459 (vectorizable_call): Return false for masked loads and stores.
18460 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18461 instead of gimple_assign_rhs1.
18462 (vectorizable_load): Handle IFN_MASK_LOAD.
18463 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18465 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18467 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18469 (vectorizable_mask_load_store): ...here.
18470 (vectorizable_load): ...and here.
18472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18474 * tree-vect-stmts.c (vect_build_all_ones_mask)
18475 (vect_build_zero_merge_argument): New functions, split out from...
18476 (vectorizable_load): ...here.
18478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18480 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18482 (vectorizable_mask_load_store): ...here.
18483 (vectorizable_store): ...and here.
18485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18487 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18489 (vectorizable_mask_load_store): ...here.
18491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18493 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18494 (vect_model_store_cost): Take a vec_load_store_type instead of a
18496 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18497 (vect_model_store_cost): Take a vec_load_store_type instead of a
18499 (vectorizable_mask_load_store): Update accordingly.
18500 (vectorizable_store): Likewise.
18501 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18505 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18506 IFN_MASK_LOAD calls here rather than...
18507 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18510 Alan Hayward <alan.hayward@arm.com>
18511 David Sherwood <david.sherwood@arm.com>
18513 * expmed.c (extract_bit_field_1): For vector extracts,
18514 fall back to extract_bit_field_as_subreg if vec_extract
18517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18518 Alan Hayward <alan.hayward@arm.com>
18519 David Sherwood <david.sherwood@arm.com>
18521 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18522 they are variable or constant sized.
18523 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18524 slots for constant-sized data.
18526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18527 Alan Hayward <alan.hayward@arm.com>
18528 David Sherwood <david.sherwood@arm.com>
18530 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18531 handling COND_EXPRs with boolean comparisons, try to find a better
18532 basis for the mask type than the boolean itself.
18534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18536 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18537 is calculated and how it can be overridden.
18538 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18539 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18541 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18545 Alan Hayward <alan.hayward@arm.com>
18546 David Sherwood <david.sherwood@arm.com>
18548 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18549 Remove the mode argument.
18550 (aarch64_simd_valid_immediate): Remove the mode and inverse
18552 * config/aarch64/iterators.md (bitsize): New iterator.
18553 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18554 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18555 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18556 aarch64_simd_valid_immediate.
18557 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18558 (aarch64_reg_or_bic_imm): Likewise.
18559 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18560 with an insn_type enum and msl with a modifier_type enum.
18561 Replace element_width with a scalar_mode. Change the shift
18562 to unsigned int. Add constructors for scalar_float_mode and
18563 scalar_int_mode elements.
18564 (aarch64_vect_float_const_representable_p): Delete.
18565 (aarch64_can_const_movi_rtx_p)
18566 (aarch64_simd_scalar_immediate_valid_for_move)
18567 (aarch64_simd_make_constant): Update call to
18568 aarch64_simd_valid_immediate.
18569 (aarch64_advsimd_valid_immediate_hs): New function.
18570 (aarch64_advsimd_valid_immediate): Likewise.
18571 (aarch64_simd_valid_immediate): Remove mode and inverse
18572 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18573 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18574 and aarch64_float_const_representable_p on the result.
18575 (aarch64_output_simd_mov_immediate): Remove mode argument.
18576 Update call to aarch64_simd_valid_immediate and use of
18577 simd_immediate_info.
18578 (aarch64_output_scalar_simd_mov_immediate): Update call
18581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18582 Alan Hayward <alan.hayward@arm.com>
18583 David Sherwood <david.sherwood@arm.com>
18585 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18586 (mode_nunits): Likewise CONST_MODE_NUNITS.
18587 * machmode.def (ADJUST_NUNITS): Document.
18588 * genmodes.c (mode_data::need_nunits_adj): New field.
18589 (blank_mode): Update accordingly.
18590 (adj_nunits): New variable.
18591 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18593 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18594 listed in adj_nunits.
18595 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18596 listed in adj_nunits. Don't emit case statements for such modes.
18597 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18598 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18599 nothing if adj_nunits is nonnull.
18600 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18601 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18602 (emit_mode_fbit): Update use of print_maybe_const_decl.
18603 (emit_move_size): Likewise. Treat the array as non-const
18605 (emit_mode_adjustments): Handle adj_nunits.
18607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18609 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18610 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18611 (VECTOR_MODES): Use it.
18612 (make_vector_modes): Take the prefix as an argument.
18614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18615 Alan Hayward <alan.hayward@arm.com>
18616 David Sherwood <david.sherwood@arm.com>
18618 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18619 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18620 for MODE_VECTOR_BOOL.
18621 * machmode.def (VECTOR_BOOL_MODE): Document.
18622 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18623 (make_vector_bool_mode): New function.
18624 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18626 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18627 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18629 * stor-layout.c (int_mode_for_mode): Likewise.
18630 * tree.c (build_vector_type_for_mode): Likewise.
18631 * varasm.c (output_constant_pool_2): Likewise.
18632 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18633 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18634 for MODE_VECTOR_BOOL.
18635 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18636 of mode class checks.
18637 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18638 instead of a list of mode class checks.
18639 (expand_vector_scalar_condition): Likewise.
18640 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18643 Alan Hayward <alan.hayward@arm.com>
18644 David Sherwood <david.sherwood@arm.com>
18646 * machmode.h (mode_size): Change from unsigned short to
18648 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18649 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18650 or if measurement_type is not polynomial.
18651 (fixed_size_mode::includes_p): Check for constant-sized modes.
18652 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18653 return a poly_uint16 rather than an unsigned short.
18654 (emit_mode_size): Change the type of mode_size from unsigned short
18655 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18656 (emit_mode_adjustments): Cope with polynomial vector sizes.
18657 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18659 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18661 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18662 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18663 * caller-save.c (setup_save_areas): Likewise.
18664 (replace_reg_with_saved_mem): Likewise.
18665 * calls.c (emit_library_call_value_1): Likewise.
18666 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18667 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18668 (gen_lowpart_for_combine): Likewise.
18669 * convert.c (convert_to_integer_1): Likewise.
18670 * cse.c (equiv_constant, cse_insn): Likewise.
18671 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18672 (cselib_subst_to_values): Likewise.
18673 * dce.c (word_dce_process_block): Likewise.
18674 * df-problems.c (df_word_lr_mark_ref): Likewise.
18675 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18676 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18677 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18678 (rtl_for_decl_location): Likewise.
18679 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18680 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18681 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18682 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18683 (expand_expr_real_1): Likewise.
18684 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18685 (pad_below): Likewise.
18686 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18687 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18688 * ira.c (get_subreg_tracking_sizes): Likewise.
18689 * ira-build.c (ira_create_allocno_objects): Likewise.
18690 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18691 (ira_sort_regnos_for_alter_reg): Likewise.
18692 * ira-costs.c (record_operand_costs): Likewise.
18693 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18694 (resolve_simple_move): Likewise.
18695 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18696 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18697 (lra_constraints): Likewise.
18698 (CONST_POOL_OK_P): Reject variable-sized modes.
18699 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18700 (add_pseudo_to_slot, lra_spill): Likewise.
18701 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18702 * optabs-query.c (get_best_extraction_insn): Likewise.
18703 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18704 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18705 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18706 * recog.c (offsettable_address_addr_space_p): Likewise.
18707 * regcprop.c (maybe_mode_change): Likewise.
18708 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18709 * regrename.c (build_def_use): Likewise.
18710 * regstat.c (dump_reg_info): Likewise.
18711 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18712 (find_reloads, find_reloads_subreg_address): Likewise.
18713 * reload1.c (eliminate_regs_1): Likewise.
18714 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18715 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18716 (simplify_binary_operation_1, simplify_subreg): Likewise.
18717 * targhooks.c (default_function_arg_padding): Likewise.
18718 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18719 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18720 (verify_gimple_assign_ternary): Likewise.
18721 * tree-inline.c (estimate_move_cost): Likewise.
18722 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18723 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18724 (get_address_cost_ainc): Likewise.
18725 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18726 (vect_supportable_dr_alignment): Likewise.
18727 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18728 (vectorizable_reduction): Likewise.
18729 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18730 (vectorizable_operation, vectorizable_load): Likewise.
18731 * tree.c (build_same_sized_truth_vector_type): Likewise.
18732 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18733 * var-tracking.c (emit_note_insn_var_location): Likewise.
18734 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18735 (ADDR_VEC_ALIGN): Likewise.
18737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18738 Alan Hayward <alan.hayward@arm.com>
18739 David Sherwood <david.sherwood@arm.com>
18741 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18743 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18744 or if measurement_type is polynomial.
18745 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18746 * combine.c (make_extraction): Likewise.
18747 * dse.c (find_shift_sequence): Likewise.
18748 * dwarf2out.c (mem_loc_descriptor): Likewise.
18749 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18750 (extract_bit_field, extract_low_bits): Likewise.
18751 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18752 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18753 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18754 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18755 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18756 * reload.c (find_reloads): Likewise.
18757 * reload1.c (alter_reg): Likewise.
18758 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18759 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18760 * tree-if-conv.c (predicate_mem_writes): Likewise.
18761 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18762 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18763 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18764 * valtrack.c (dead_debug_insert_temp): Likewise.
18765 * varasm.c (mergeable_constant_section): Likewise.
18766 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18769 Alan Hayward <alan.hayward@arm.com>
18770 David Sherwood <david.sherwood@arm.com>
18772 * expr.c (expand_assignment): Cope with polynomial mode sizes
18773 when assigning to a CONCAT.
18775 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18776 Alan Hayward <alan.hayward@arm.com>
18777 David Sherwood <david.sherwood@arm.com>
18779 * machmode.h (mode_precision): Change from unsigned short to
18781 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18783 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18784 or if measurement_type is not polynomial.
18785 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18786 in which the mode is already known to be a scalar_int_mode.
18787 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18788 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18790 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18791 for GET_MODE_PRECISION.
18792 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18793 for GET_MODE_PRECISION.
18794 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18796 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18797 (expand_field_assignment, make_extraction): Likewise.
18798 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18799 (get_last_value): Likewise.
18800 * convert.c (convert_to_integer_1): Likewise.
18801 * cse.c (cse_insn): Likewise.
18802 * expr.c (expand_expr_real_1): Likewise.
18803 * lra-constraints.c (simplify_operand_subreg): Likewise.
18804 * optabs-query.c (can_atomic_load_p): Likewise.
18805 * optabs.c (expand_atomic_load): Likewise.
18806 (expand_atomic_store): Likewise.
18807 * ree.c (combine_reaching_defs): Likewise.
18808 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18809 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18810 * tree.h (type_has_mode_precision_p): Likewise.
18811 * ubsan.c (instrument_si_overflow): Likewise.
18813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18814 Alan Hayward <alan.hayward@arm.com>
18815 David Sherwood <david.sherwood@arm.com>
18817 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18818 polynomial numbers of units.
18819 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18820 (valid_vector_subparts_p): New function.
18821 (build_vector_type): Remove temporary shim and take the number
18822 of units as a poly_uint64 rather than an int.
18823 (build_opaque_vector_type): Take the number of units as a
18824 poly_uint64 rather than an int.
18825 * tree.c (build_vector_from_ctor): Handle polynomial
18826 TYPE_VECTOR_SUBPARTS.
18827 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18828 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18829 (build_vector_from_val): If the number of units is variable,
18830 use build_vec_duplicate_cst for constant operands and
18831 VEC_DUPLICATE_EXPR otherwise.
18832 (make_vector_type): Remove temporary is_constant ().
18833 (build_vector_type, build_opaque_vector_type): Take the number of
18834 units as a poly_uint64 rather than an int.
18835 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18837 * cfgexpand.c (expand_debug_expr): Likewise.
18838 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18839 (store_constructor, expand_expr_real_1): Likewise.
18840 (const_scalar_mask_from_tree): Likewise.
18841 * fold-const-call.c (fold_const_reduction): Likewise.
18842 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18843 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18844 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18845 (fold_relational_const): Likewise.
18846 (native_interpret_vector): Likewise. Change the size from an
18847 int to an unsigned int.
18848 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18849 TYPE_VECTOR_SUBPARTS.
18850 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18851 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18852 duplicating a non-constant operand into a variable-length vector.
18853 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18854 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18855 * ipa-icf.c (sem_variable::equals): Likewise.
18856 * match.pd: Likewise.
18857 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18858 * print-tree.c (print_node): Likewise.
18859 * stor-layout.c (layout_type): Likewise.
18860 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18861 * tree-cfg.c (verify_gimple_comparison): Likewise.
18862 (verify_gimple_assign_binary): Likewise.
18863 (verify_gimple_assign_ternary): Likewise.
18864 (verify_gimple_assign_single): Likewise.
18865 * tree-pretty-print.c (dump_generic_node): Likewise.
18866 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18867 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18868 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18869 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18870 (vect_shift_permute_load_chain): Likewise.
18871 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18872 (expand_vector_condition, optimize_vector_constructor): Likewise.
18873 (lower_vec_perm, get_compute_type): Likewise.
18874 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18875 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18876 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18877 (vect_recog_mask_conversion_pattern): Likewise.
18878 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18879 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18880 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18881 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18882 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18883 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18884 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18885 (supportable_widening_operation): Likewise.
18886 (supportable_narrowing_operation): Likewise.
18887 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18889 * varasm.c (output_constant): Likewise.
18891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18892 Alan Hayward <alan.hayward@arm.com>
18893 David Sherwood <david.sherwood@arm.com>
18895 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18896 so that both the length == 3 and length != 3 cases set up their
18897 own permute vectors. Add comments explaining why we know the
18898 number of elements is constant.
18899 (vect_permute_load_chain): Likewise.
18901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18902 Alan Hayward <alan.hayward@arm.com>
18903 David Sherwood <david.sherwood@arm.com>
18905 * machmode.h (mode_nunits): Change from unsigned char to
18907 (ONLY_FIXED_SIZE_MODES): New macro.
18908 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18909 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18910 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18912 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18913 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18914 or if measurement_type is not polynomial.
18915 * genmodes.c (ZERO_COEFFS): New macro.
18916 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18918 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18919 Use ZERO_COEFFS when emitting initializers.
18920 * data-streamer.h (bp_pack_poly_value): New function.
18921 (bp_unpack_poly_value): Likewise.
18922 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18923 for GET_MODE_NUNITS.
18924 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18925 for GET_MODE_NUNITS.
18926 * tree.c (make_vector_type): Remove temporary shim and make
18927 the real function take the number of units as a poly_uint64
18928 rather than an int.
18929 (build_vector_type_for_mode): Handle polynomial nunits.
18930 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18931 * emit-rtl.c (const_vec_series_p_1): Likewise.
18932 (gen_rtx_CONST_VECTOR): Likewise.
18933 * fold-const.c (test_vec_duplicate_folding): Likewise.
18934 * genrecog.c (validate_pattern): Likewise.
18935 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18936 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18937 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18938 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18939 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18940 * rtlanal.c (subreg_get_info): Likewise.
18941 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18942 (vect_grouped_load_supported): Likewise.
18943 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18944 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18945 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18946 (simplify_const_unary_operation, simplify_binary_operation_1)
18947 (simplify_const_binary_operation, simplify_ternary_operation)
18948 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18949 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18950 instead of CONST_VECTOR_NUNITS.
18951 * varasm.c (output_constant_pool_2): Likewise.
18952 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18953 explicit-encoded elements in the XVEC for variable-length vectors.
18955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18957 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18959 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18960 Alan Hayward <alan.hayward@arm.com>
18961 David Sherwood <david.sherwood@arm.com>
18963 * coretypes.h (fixed_size_mode): Declare.
18964 (fixed_size_mode_pod): New typedef.
18965 * builtins.h (target_builtins::x_apply_args_mode)
18966 (target_builtins::x_apply_result_mode): Change type to
18967 fixed_size_mode_pod.
18968 * builtins.c (apply_args_size, apply_result_size, result_vector)
18969 (expand_builtin_apply_args_1, expand_builtin_apply)
18970 (expand_builtin_return): Update accordingly.
18972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18974 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18975 * cselib.c (cselib_hash_rtx): Likewise.
18976 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18977 CONST_VECTOR encoding.
18979 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18980 Jeff Law <law@redhat.com>
18983 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18984 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18985 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18986 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18989 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18990 explicitly probe *sp in a noreturn function if there were any callee
18991 register saves or frame pointer is needed.
18993 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18996 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18997 BLKmode for ternary, binary or unary expressions.
19000 * var-tracking.c (delete_vta_debug_insn): New inline function.
19001 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
19002 insns from get_insns () to NULL instead of each bb separately.
19003 Use delete_vta_debug_insn. No longer static.
19004 (vt_debug_insns_local, variable_tracking_main_1): Adjust
19005 delete_vta_debug_insns callers.
19006 * rtl.h (delete_vta_debug_insns): Declare.
19007 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
19008 instead of variable_tracking_main.
19010 2018-01-03 Martin Sebor <msebor@redhat.com>
19012 PR tree-optimization/83603
19013 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
19014 arguments past the endof the argument list in functions declared
19015 without a prototype.
19016 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
19017 Avoid checking when arguments are null.
19019 2018-01-03 Martin Sebor <msebor@redhat.com>
19022 * doc/extend.texi (attribute const): Fix a typo.
19023 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
19024 issuing -Wsuggest-attribute for void functions.
19026 2018-01-03 Martin Sebor <msebor@redhat.com>
19028 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
19029 offset_int::from instead of wide_int::to_shwi.
19030 (maybe_diag_overlap): Remove assertion.
19031 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
19032 * gimple-ssa-sprintf.c (format_directive): Same.
19033 (parse_directive): Same.
19034 (sprintf_dom_walker::compute_format_length): Same.
19035 (try_substitute_return_value): Same.
19037 2018-01-03 Jeff Law <law@redhat.com>
19039 PR middle-end/83654
19040 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19041 non-constant residual for zero at runtime and avoid probing in
19042 that case. Reorganize code for trailing problem to mirror handling
19045 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19047 PR tree-optimization/83501
19048 * tree-ssa-strlen.c (get_string_cst): New.
19049 (handle_char_store): Call get_string_cst.
19051 2018-01-03 Martin Liska <mliska@suse.cz>
19053 PR tree-optimization/83593
19054 * tree-ssa-strlen.c: Include tree-cfg.h.
19055 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19056 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19057 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19059 (strlen_dom_walker::before_dom_children): Call
19060 gimple_purge_dead_eh_edges. Dump tranformation with details
19062 (strlen_dom_walker::before_dom_children): Update call by adding
19063 new argument cleanup_eh.
19064 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19066 2018-01-03 Martin Liska <mliska@suse.cz>
19069 * cif-code.def (VARIADIC_THUNK): New enum value.
19070 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19073 2018-01-03 Jan Beulich <jbeulich@suse.com>
19075 * sse.md (mov<mode>_internal): Tighten condition for when to use
19076 vmovdqu<ssescalarsize> for TI and OI modes.
19078 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19080 Update copyright years.
19082 2018-01-03 Martin Liska <mliska@suse.cz>
19085 * ipa-visibility.c (function_and_variable_visibility): Skip
19086 functions with noipa attribure.
19088 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19090 * gcc.c (process_command): Update copyright notice dates.
19091 * gcov-dump.c (print_version): Ditto.
19092 * gcov.c (print_version): Ditto.
19093 * gcov-tool.c (print_version): Ditto.
19094 * gengtype.c (create_file): Ditto.
19095 * doc/cpp.texi: Bump @copying's copyright year.
19096 * doc/cppinternals.texi: Ditto.
19097 * doc/gcc.texi: Ditto.
19098 * doc/gccint.texi: Ditto.
19099 * doc/gcov.texi: Ditto.
19100 * doc/install.texi: Ditto.
19101 * doc/invoke.texi: Ditto.
19103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19105 * vector-builder.h (vector_builder::m_full_nelts): Change from
19106 unsigned int to poly_uint64.
19107 (vector_builder::full_nelts): Update prototype accordingly.
19108 (vector_builder::new_vector): Likewise.
19109 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19110 (vector_builder::operator ==): Likewise.
19111 (vector_builder::finalize): Likewise.
19112 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19113 Take the number of elements as a poly_uint64 rather than an
19115 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19116 from unsigned int to poly_uint64.
19117 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19118 (vec_perm_indices::new_vector): Likewise.
19119 (vec_perm_indices::length): Likewise.
19120 (vec_perm_indices::nelts_per_input): Likewise.
19121 (vec_perm_indices::input_nelts): Likewise.
19122 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19123 number of elements per input as a poly_uint64 rather than an
19124 unsigned int. Use the original encoding for variable-length
19125 vectors, rather than clamping each individual element.
19126 For the second and subsequent elements in each pattern,
19127 clamp the step and base before clamping their sum.
19128 (vec_perm_indices::series_p): Handle polynomial element counts.
19129 (vec_perm_indices::all_in_range_p): Likewise.
19130 (vec_perm_indices_to_tree): Likewise.
19131 (vec_perm_indices_to_rtx): Likewise.
19132 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19133 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19134 (tree_vector_builder::new_binary_operation): Handle polynomial
19135 element counts. Return false if we need to know the number
19136 of elements at compile time.
19137 * fold-const.c (fold_vec_perm): Punt if the number of elements
19138 isn't known at compile time.
19140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19142 * vec-perm-indices.h (vec_perm_builder): Change element type
19143 from HOST_WIDE_INT to poly_int64.
19144 (vec_perm_indices::element_type): Update accordingly.
19145 (vec_perm_indices::clamp): Handle polynomial element_types.
19146 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19147 (vec_perm_indices::all_in_range_p): Likewise.
19148 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19150 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19151 polynomial vec_perm_indices element types.
19152 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19153 * fold-const.c (fold_vec_perm): Likewise.
19154 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19155 * tree-vect-generic.c (lower_vec_perm): Likewise.
19156 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19157 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19158 element type to HOST_WIDE_INT.
19160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19161 Alan Hayward <alan.hayward@arm.com>
19162 David Sherwood <david.sherwood@arm.com>
19164 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19165 rather than an int. Use plus_constant.
19166 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19167 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19170 Alan Hayward <alan.hayward@arm.com>
19171 David Sherwood <david.sherwood@arm.com>
19173 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19174 a HOST_WIDE_INT to a poly_int64.
19176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19177 Alan Hayward <alan.hayward@arm.com>
19178 David Sherwood <david.sherwood@arm.com>
19180 * calls.c (load_register_parameters): Cope with polynomial
19181 mode sizes. Require a constant size for BLKmode parameters
19182 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19183 forces a parameter to be padded at the lsb end in order to
19184 fill a complete number of words, require the parameter size
19185 to be ordered wrt UNITS_PER_WORD.
19187 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19188 Alan Hayward <alan.hayward@arm.com>
19189 David Sherwood <david.sherwood@arm.com>
19191 * reload1.c (spill_stack_slot_width): Change element type
19192 from unsigned int to poly_uint64_pod.
19193 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
19200 (reload_inner_reg_of_subreg, push_reload): Use it.
19202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19203 Alan Hayward <alan.hayward@arm.com>
19204 David Sherwood <david.sherwood@arm.com>
19206 * lra-constraints.c (process_alt_operands): Reject matched
19207 operands whose sizes aren't ordered.
19208 (match_reload): Refer to this check here.
19210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19211 Alan Hayward <alan.hayward@arm.com>
19212 David Sherwood <david.sherwood@arm.com>
19214 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19215 that the mode size is in the set {1, 2, 4, 8, 16}.
19217 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19218 Alan Hayward <alan.hayward@arm.com>
19219 David Sherwood <david.sherwood@arm.com>
19221 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19222 Use plus_constant instead of gen_rtx_PLUS.
19224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19225 Alan Hayward <alan.hayward@arm.com>
19226 David Sherwood <david.sherwood@arm.com>
19228 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19229 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19230 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19231 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19232 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19233 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19234 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19235 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19236 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19237 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19239 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19240 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19241 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19242 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19243 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19244 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19245 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19246 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19247 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19248 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19250 * expr.c (emit_move_resolve_push): Treat the input and result
19251 of PUSH_ROUNDING as a poly_int64.
19252 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19253 (emit_push_insn): Likewise.
19254 * lra-eliminations.c (mark_not_eliminable): Likewise.
19255 * recog.c (push_operand): Likewise.
19256 * reload1.c (elimination_effects): Likewise.
19257 * rtlanal.c (nonzero_bits1): Likewise.
19258 * calls.c (store_one_arg): Likewise. Require the padding to be
19259 known at compile time.
19261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19262 Alan Hayward <alan.hayward@arm.com>
19263 David Sherwood <david.sherwood@arm.com>
19265 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19266 Use plus_constant instead of gen_rtx_PLUS.
19268 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19269 Alan Hayward <alan.hayward@arm.com>
19270 David Sherwood <david.sherwood@arm.com>
19272 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19273 rather than an int.
19275 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19276 Alan Hayward <alan.hayward@arm.com>
19277 David Sherwood <david.sherwood@arm.com>
19279 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19280 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19281 via stack temporaries. Treat the mode size as polynomial too.
19283 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19284 Alan Hayward <alan.hayward@arm.com>
19285 David Sherwood <david.sherwood@arm.com>
19287 * expr.c (expand_expr_real_2): When handling conversions involving
19288 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19289 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19290 as a poly_uint64 too.
19292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19293 Alan Hayward <alan.hayward@arm.com>
19294 David Sherwood <david.sherwood@arm.com>
19296 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19298 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19299 Alan Hayward <alan.hayward@arm.com>
19300 David Sherwood <david.sherwood@arm.com>
19302 * combine.c (can_change_dest_mode): Handle polynomial
19303 REGMODE_NATURAL_SIZE.
19304 * expmed.c (store_bit_field_1): Likewise.
19305 * expr.c (store_constructor): Likewise.
19306 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19307 and polynomial REGMODE_NATURAL_SIZE.
19308 (gen_lowpart_common): Likewise.
19309 * reginfo.c (record_subregs_of_mode): Likewise.
19310 * rtlanal.c (read_modify_subreg_p): Likewise.
19312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19313 Alan Hayward <alan.hayward@arm.com>
19314 David Sherwood <david.sherwood@arm.com>
19316 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19317 numbers of elements.
19319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19320 Alan Hayward <alan.hayward@arm.com>
19321 David Sherwood <david.sherwood@arm.com>
19323 * match.pd: Cope with polynomial numbers of vector elements.
19325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19326 Alan Hayward <alan.hayward@arm.com>
19327 David Sherwood <david.sherwood@arm.com>
19329 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19330 in a POINTER_PLUS_EXPR.
19332 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19333 Alan Hayward <alan.hayward@arm.com>
19334 David Sherwood <david.sherwood@arm.com>
19336 * omp-simd-clone.c (simd_clone_subparts): New function.
19337 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19338 (ipa_simd_modify_function_body): Likewise.
19340 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19341 Alan Hayward <alan.hayward@arm.com>
19342 David Sherwood <david.sherwood@arm.com>
19344 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19345 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19346 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19347 (expand_vector_condition, vector_element): Likewise.
19348 (subparts_gt): New function.
19349 (get_compute_type): Use subparts_gt.
19350 (count_type_subparts): Delete.
19351 (expand_vector_operations_1): Use subparts_gt instead of
19352 count_type_subparts.
19354 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19355 Alan Hayward <alan.hayward@arm.com>
19356 David Sherwood <david.sherwood@arm.com>
19358 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19359 (vect_compile_time_alias): ...this new function. Do the calculation
19360 on poly_ints rather than trees.
19361 (vect_prune_runtime_alias_test_list): Update call accordingly.
19363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19364 Alan Hayward <alan.hayward@arm.com>
19365 David Sherwood <david.sherwood@arm.com>
19367 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19369 (vect_schedule_slp_instance): Likewise.
19371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19372 Alan Hayward <alan.hayward@arm.com>
19373 David Sherwood <david.sherwood@arm.com>
19375 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19376 constant and extern definitions for variable-length vectors.
19377 (vect_get_constant_vectors): Note that the number of units
19378 is known to be constant.
19380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19381 Alan Hayward <alan.hayward@arm.com>
19382 David Sherwood <david.sherwood@arm.com>
19384 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19385 of units as polynomial. Choose between WIDE and NARROW based
19388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19389 Alan Hayward <alan.hayward@arm.com>
19390 David Sherwood <david.sherwood@arm.com>
19392 * tree-vect-stmts.c (simd_clone_subparts): New function.
19393 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19396 Alan Hayward <alan.hayward@arm.com>
19397 David Sherwood <david.sherwood@arm.com>
19399 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19400 vectors as polynomial. Use build_index_vector for
19401 IFN_GOMP_SIMD_LANE.
19403 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19404 Alan Hayward <alan.hayward@arm.com>
19405 David Sherwood <david.sherwood@arm.com>
19407 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19408 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19409 for variable-length vectors.
19410 (vectorizable_mask_load_store): Treat the number of units as
19411 polynomial, asserting that it is constant if the condition has
19412 already been enforced.
19413 (vectorizable_store, vectorizable_load): Likewise.
19415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19416 Alan Hayward <alan.hayward@arm.com>
19417 David Sherwood <david.sherwood@arm.com>
19419 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19420 of units as polynomial. Punt if we can't tell at compile time
19421 which vector contains the final result.
19423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19424 Alan Hayward <alan.hayward@arm.com>
19425 David Sherwood <david.sherwood@arm.com>
19427 * tree-vect-loop.c (vectorizable_induction): Treat the number
19428 of units as polynomial. Punt on SLP inductions. Use an integer
19429 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19430 cast of such a series for variable-length floating-point
19433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19434 Alan Hayward <alan.hayward@arm.com>
19435 David Sherwood <david.sherwood@arm.com>
19437 * tree.h (build_index_vector): Declare.
19438 * tree.c (build_index_vector): New function.
19439 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19440 of units as polynomial, forcibly converting it to a constant if
19441 vectorizable_reduction has already enforced the condition.
19442 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19443 to create a {1,2,3,...} vector.
19444 (vectorizable_reduction): Treat the number of units as polynomial.
19445 Choose vectype_in based on the largest scalar element size rather
19446 than the smallest number of units. Enforce the restrictions
19449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19450 Alan Hayward <alan.hayward@arm.com>
19451 David Sherwood <david.sherwood@arm.com>
19453 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
19454 number of units as polynomial.
19456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19457 Alan Hayward <alan.hayward@arm.com>
19458 David Sherwood <david.sherwood@arm.com>
19460 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19461 * target.def (autovectorize_vector_sizes): Return the vector sizes
19462 by pointer, using vector_sizes rather than a bitmask.
19463 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19464 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19465 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19467 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19468 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19469 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19470 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19471 * omp-general.c (omp_max_vf): Likewise.
19472 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19473 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19474 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19475 * tree-vect-slp.c (vect_slp_bb): Likewise.
19476 * doc/tm.texi: Regenerate.
19477 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19479 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19480 the vector size as a poly_uint64 rather than an unsigned int.
19481 (current_vector_size): Change from an unsigned int to a poly_uint64.
19482 (get_vectype_for_scalar_type): Update accordingly.
19483 * tree.h (build_truth_vector_type): Take the size and number of
19484 units as a poly_uint64 rather than an unsigned int.
19485 (build_vector_type): Add a temporary overload that takes
19486 the number of units as a poly_uint64 rather than an unsigned int.
19487 * tree.c (make_vector_type): Likewise.
19488 (build_truth_vector_type): Take the number of units as a poly_uint64
19489 rather than an unsigned int.
19491 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19492 Alan Hayward <alan.hayward@arm.com>
19493 David Sherwood <david.sherwood@arm.com>
19495 * target.def (get_mask_mode): Take the number of units and length
19496 as poly_uint64s rather than unsigned ints.
19497 * targhooks.h (default_get_mask_mode): Update accordingly.
19498 * targhooks.c (default_get_mask_mode): Likewise.
19499 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19500 * doc/tm.texi: Regenerate.
19502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19503 Alan Hayward <alan.hayward@arm.com>
19504 David Sherwood <david.sherwood@arm.com>
19506 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19507 * omp-general.c (omp_max_vf): Likewise.
19508 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19509 (expand_omp_simd): Handle polynomial safelen.
19510 * omp-low.c (omplow_simd_context): Add a default constructor.
19511 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19512 (lower_rec_simd_input_clauses): Update accordingly.
19513 (lower_rec_input_clauses): Likewise.
19515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19516 Alan Hayward <alan.hayward@arm.com>
19517 David Sherwood <david.sherwood@arm.com>
19519 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19520 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19521 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19522 (vect_analyze_slp_cost): Likewise.
19523 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19524 (vect_model_load_cost): Likewise.
19526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19527 Alan Hayward <alan.hayward@arm.com>
19528 David Sherwood <david.sherwood@arm.com>
19530 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19531 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19532 from an unsigned int * to a poly_uint64_pod *.
19533 (calculate_unrolling_factor): New function.
19534 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
19536 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19537 Alan Hayward <alan.hayward@arm.com>
19538 David Sherwood <david.sherwood@arm.com>
19540 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19541 from an unsigned int to a poly_uint64.
19542 (_loop_vec_info::slp_unrolling_factor): Likewise.
19543 (_loop_vec_info::vectorization_factor): Change from an int
19545 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19546 (vect_get_num_vectors): New function.
19547 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19548 (vect_get_num_copies): Use vect_get_num_vectors.
19549 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19550 to an unsigned int *.
19551 (vect_analyze_data_refs): Change min_vf from an int * to a
19553 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19554 than an unsigned HOST_WIDE_INT.
19555 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19556 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19557 to an unsigned int *.
19558 (vect_analyze_data_ref_dependences): Likewise.
19559 (vect_compute_data_ref_alignment): Handle polynomial vf.
19560 (vect_enhance_data_refs_alignment): Likewise.
19561 (vect_prune_runtime_alias_test_list): Likewise.
19562 (vect_shift_permute_load_chain): Likewise.
19563 (vect_supportable_dr_alignment): Likewise.
19564 (dependence_distance_ge_vf): Take the vectorization factor as a
19565 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19566 (vect_analyze_data_refs): Change min_vf from an int * to a
19568 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19569 vfm1 as a poly_uint64 rather than an int. Make the same change
19570 for the returned bound_scalar.
19571 (vect_gen_vector_loop_niters): Handle polynomial vf.
19572 (vect_do_peeling): Likewise. Update call to
19573 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19574 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19576 * tree-vect-loop.c (vect_determine_vectorization_factor)
19577 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19578 (vect_get_known_peeling_cost): Likewise.
19579 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19580 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19581 (vect_transform_loop): Likewise. Use the lowest possible VF when
19582 updating the upper bounds of the loop.
19583 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19584 rather than an int.
19585 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19586 polynomial unroll factors.
19587 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19588 (vect_make_slp_decision): Likewise.
19589 (vect_supported_load_permutation_p): Likewise, and polynomial
19591 (vect_analyze_slp_cost): Handle polynomial vf.
19592 (vect_slp_analyze_node_operations): Likewise.
19593 (vect_slp_analyze_bb_1): Likewise.
19594 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19595 than an unsigned HOST_WIDE_INT.
19596 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19597 (vectorizable_load): Handle polynomial vf.
19598 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19600 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19603 Alan Hayward <alan.hayward@arm.com>
19604 David Sherwood <david.sherwood@arm.com>
19606 * match.pd: Handle bit operations involving three constants
19607 and try to fold one pair.
19609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19611 * tree-vect-loop-manip.c: Include gimple-fold.h.
19612 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19613 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19614 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19615 Add a path that uses a step of VF instead of 1, but disable it
19617 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19618 and niters_no_overflow parameters. Update calls to
19619 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19620 Create a new SSA name if the latter choses to use a ste other
19621 than zero, and return it via niters_vector_mult_vf_var.
19622 * tree-vect-loop.c (vect_transform_loop): Update calls to
19623 vect_do_peeling, vect_gen_vector_loop_niters and
19624 slpeel_make_loop_iterate_ntimes.
19625 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19626 (vect_gen_vector_loop_niters): Update declarations after above changes.
19628 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19630 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19631 128-bit round to integer instructions.
19632 (ceil<mode>2): Likewise.
19633 (btrunc<mode>2): Likewise.
19634 (round<mode>2): Likewise.
19636 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19638 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19639 unaligned VSX load/store on P8/P9.
19640 (expand_block_clear): Allow the use of unaligned VSX
19641 load/store on P8/P9.
19643 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19645 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19647 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19648 swap associated with both a load and a store.
19650 2018-01-02 Andrew Waterman <andrew@sifive.com>
19652 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19653 * config/riscv/riscv.md (clear_cache): Use it.
19655 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19657 * web.c: Remove out-of-date comment.
19659 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19661 * expr.c (fixup_args_size_notes): Check that any existing
19662 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19663 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19664 (emit_single_push_insn): ...here.
19666 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19668 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19669 (const_vector_encoded_nelts): New function.
19670 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19671 (const_vector_int_elt, const_vector_elt): Declare.
19672 * emit-rtl.c (const_vector_int_elt_1): New function.
19673 (const_vector_elt): Likewise.
19674 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19675 of CONST_VECTOR_ELT.
19677 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19679 * expr.c: Include rtx-vector-builder.h.
19680 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19681 directly on the tree encoding.
19682 (const_vector_from_tree): Likewise.
19683 * optabs.c: Include rtx-vector-builder.h.
19684 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19685 sequence of "u" values.
19686 * vec-perm-indices.c: Include rtx-vector-builder.h.
19687 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19688 directly on the vec_perm_indices encoding.
19690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19692 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19693 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19694 * rtx-vector-builder.h: New file.
19695 * rtx-vector-builder.c: Likewise.
19696 * rtl.h (rtx_def::u2): Add a const_vector field.
19697 (CONST_VECTOR_NPATTERNS): New macro.
19698 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19699 (CONST_VECTOR_DUPLICATE_P): Likewise.
19700 (CONST_VECTOR_STEPPED_P): Likewise.
19701 (CONST_VECTOR_ENCODED_ELT): Likewise.
19702 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19703 (unwrap_const_vec_duplicate): Likewise.
19704 (const_vec_series_p): Check for a non-duplicated vector encoding.
19705 Say that the function only returns true for integer vectors.
19706 * emit-rtl.c: Include rtx-vector-builder.h.
19707 (gen_const_vec_duplicate_1): Delete.
19708 (gen_const_vector): Call gen_const_vec_duplicate instead of
19709 gen_const_vec_duplicate_1.
19710 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19711 (gen_const_vec_duplicate): Use rtx_vector_builder.
19712 (gen_const_vec_series): Likewise.
19713 (gen_rtx_CONST_VECTOR): Likewise.
19714 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19715 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19716 Build a new vector rather than modifying a CONST_VECTOR in-place.
19717 (handle_special_swappables): Update call accordingly.
19718 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19719 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19720 Build a new vector rather than modifying a CONST_VECTOR in-place.
19721 (handle_special_swappables): Update call accordingly.
19723 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19725 * simplify-rtx.c (simplify_const_binary_operation): Use
19726 CONST_VECTOR_ELT instead of XVECEXP.
19728 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19730 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19731 the selector elements to be different from the data elements
19732 if the selector is a VECTOR_CST.
19733 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19734 ssizetype for the selector.
19736 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19738 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19739 before testing each element individually.
19740 * tree-vect-generic.c (lower_vec_perm): Likewise.
19742 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19744 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19745 * selftest-run-tests.c (selftest::run_tests): Call it.
19746 * vector-builder.h (vector_builder::operator ==): New function.
19747 (vector_builder::operator !=): Likewise.
19748 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19749 (vec_perm_indices::all_from_input_p): New function.
19750 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19751 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19752 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19753 instead of reading the VECTOR_CST directly. Detect whether both
19754 vector inputs are the same before constructing the vec_perm_indices,
19755 and update the number of inputs argument accordingly. Use the
19756 utility functions added above. Only construct sel2 if we need to.
19758 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19760 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19761 the broadcast of the low byte.
19762 (expand_mult_highpart): Use an explicit encoding for the permutes.
19763 * optabs-query.c (can_mult_highpart_p): Likewise.
19764 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19765 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19766 (vectorizable_bswap): Likewise.
19767 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19768 explicit encoding for the power-of-2 permutes.
19769 (vect_permute_store_chain): Likewise.
19770 (vect_grouped_load_supported): Likewise.
19771 (vect_permute_load_chain): Likewise.
19773 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19775 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19776 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19777 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19778 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19779 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19780 (vect_gen_perm_mask_any): Likewise.
19782 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19784 * int-vector-builder.h: New file.
19785 * vec-perm-indices.h: Include int-vector-builder.h.
19786 (vec_perm_indices): Redefine as an int_vector_builder.
19787 (auto_vec_perm_indices): Delete.
19788 (vec_perm_builder): Redefine as a stand-alone class.
19789 (vec_perm_indices::vec_perm_indices): New function.
19790 (vec_perm_indices::clamp): Likewise.
19791 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19792 (vec_perm_indices::new_vector): New function.
19793 (vec_perm_indices::new_expanded_vector): Update for new
19794 vec_perm_indices class.
19795 (vec_perm_indices::rotate_inputs): New function.
19796 (vec_perm_indices::all_in_range_p): Operate directly on the
19797 encoded form, without computing elided elements.
19798 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19799 encoding. Update for new vec_perm_indices class.
19800 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19801 the given vec_perm_builder.
19802 (expand_vec_perm_var): Update vec_perm_builder constructor.
19803 (expand_mult_highpart): Use vec_perm_builder instead of
19804 auto_vec_perm_indices.
19805 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19806 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19807 or double series encoding as appropriate.
19808 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19809 vec_perm_indices instead of auto_vec_perm_indices.
19810 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19811 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19812 (vect_permute_store_chain): Likewise.
19813 (vect_grouped_load_supported): Likewise.
19814 (vect_permute_load_chain): Likewise.
19815 (vect_shift_permute_load_chain): Likewise.
19816 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19817 (vect_transform_slp_perm_load): Likewise.
19818 (vect_schedule_slp_instance): Likewise.
19819 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19820 (vectorizable_mask_load_store): Likewise.
19821 (vectorizable_bswap): Likewise.
19822 (vectorizable_store): Likewise.
19823 (vectorizable_load): Likewise.
19824 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19825 vec_perm_indices instead of auto_vec_perm_indices. Use
19826 tree_to_vec_perm_builder to read the vector from a tree.
19827 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19828 vec_perm_builder instead of a vec_perm_indices.
19829 (have_whole_vector_shift): Use vec_perm_builder and
19830 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19831 truncation to calc_vec_perm_mask_for_shift.
19832 (vect_create_epilog_for_reduction): Likewise.
19833 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19834 from auto_vec_perm_indices to vec_perm_indices.
19835 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19836 instead of changing individual elements.
19837 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19838 the vector in d.perm.
19839 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19840 from auto_vec_perm_indices to vec_perm_indices.
19841 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19842 instead of changing individual elements.
19843 (arm_vectorize_vec_perm_const): Use new_vector to install
19844 the vector in d.perm.
19845 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19846 Update vec_perm_builder constructor.
19847 (rs6000_expand_interleave): Likewise.
19848 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19849 (rs6000_expand_interleave): Likewise.
19851 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19853 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19854 to qimode could truncate the indices.
19855 * optabs.c (expand_vec_perm_var): Likewise.
19857 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19859 * Makefile.in (OBJS): Add vec-perm-indices.o.
19860 * vec-perm-indices.h: New file.
19861 * vec-perm-indices.c: Likewise.
19862 * target.h (vec_perm_indices): Replace with a forward class
19864 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19865 * optabs.h: Include vec-perm-indices.h.
19866 (expand_vec_perm): Delete.
19867 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19868 (expand_vec_perm_const): Declare.
19869 * target.def (vec_perm_const_ok): Replace with...
19870 (vec_perm_const): ...this new hook.
19871 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19872 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19873 * doc/tm.texi: Regenerate.
19874 * optabs.def (vec_perm_const): Delete.
19875 * doc/md.texi (vec_perm_const): Likewise.
19876 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19877 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19878 expand_vec_perm for constant permutation vectors. Assert that
19879 the mode of variable permutation vectors is the integer equivalent
19880 of the mode that is being permuted.
19881 * optabs-query.h (selector_fits_mode_p): Declare.
19882 * optabs-query.c: Include vec-perm-indices.h.
19883 (selector_fits_mode_p): New function.
19884 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19885 is defined, instead of checking whether the vec_perm_const_optab
19886 exists. Use targetm.vectorize.vec_perm_const instead of
19887 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19888 fit in the vector mode before using a variable permute.
19889 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19890 vec_perm_indices instead of an rtx.
19891 (expand_vec_perm): Replace with...
19892 (expand_vec_perm_const): ...this new function. Take the selector
19893 as a vec_perm_indices rather than an rtx. Also take the mode of
19894 the selector. Update call to shift_amt_for_vec_perm_mask.
19895 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19896 Use vec_perm_indices::new_expanded_vector to expand the original
19897 selector into bytes. Check whether the indices fit in the vector
19898 mode before using a variable permute.
19899 (expand_vec_perm_var): Make global.
19900 (expand_mult_highpart): Use expand_vec_perm_const.
19901 * fold-const.c: Includes vec-perm-indices.h.
19902 * tree-ssa-forwprop.c: Likewise.
19903 * tree-vect-data-refs.c: Likewise.
19904 * tree-vect-generic.c: Likewise.
19905 * tree-vect-loop.c: Likewise.
19906 * tree-vect-slp.c: Likewise.
19907 * tree-vect-stmts.c: Likewise.
19908 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19910 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19911 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19912 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19913 (aarch64_vectorize_vec_perm_const): ...this new function.
19914 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19915 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19916 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19917 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19918 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19919 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19920 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19922 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19923 check for NEON modes.
19924 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19925 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19926 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19927 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19929 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19930 the old VEC_PERM_CONST conditions.
19931 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19932 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19933 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19934 (ia64_vectorize_vec_perm_const_ok): Merge into...
19935 (ia64_vectorize_vec_perm_const): ...this new function.
19936 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19937 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19938 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19939 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19940 * config/mips/mips.c (mips_expand_vec_perm_const)
19941 (mips_vectorize_vec_perm_const_ok): Merge into...
19942 (mips_vectorize_vec_perm_const): ...this new function.
19943 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19944 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19945 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19946 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19947 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19948 (rs6000_expand_vec_perm_const): Delete.
19949 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19951 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19952 (altivec_expand_vec_perm_const_le): Take each operand individually.
19953 Operate on constant selectors rather than rtxes.
19954 (altivec_expand_vec_perm_const): Likewise. Update call to
19955 altivec_expand_vec_perm_const_le.
19956 (rs6000_expand_vec_perm_const): Delete.
19957 (rs6000_vectorize_vec_perm_const_ok): Delete.
19958 (rs6000_vectorize_vec_perm_const): New function.
19959 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19960 an element count and rtx array.
19961 (rs6000_expand_extract_even): Update call accordingly.
19962 (rs6000_expand_interleave): Likewise.
19963 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19964 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19965 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19966 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19967 (rs6000_expand_vec_perm_const): Delete.
19968 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19969 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19970 (altivec_expand_vec_perm_const_le): Take each operand individually.
19971 Operate on constant selectors rather than rtxes.
19972 (altivec_expand_vec_perm_const): Likewise. Update call to
19973 altivec_expand_vec_perm_const_le.
19974 (rs6000_expand_vec_perm_const): Delete.
19975 (rs6000_vectorize_vec_perm_const_ok): Delete.
19976 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19977 reference to the SPE evmerge intructions.
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/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19983 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19985 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19987 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19989 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19990 vector mode and that that mode matches the mode of the data
19992 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19993 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19994 directly using expand_vec_perm_1 when forcing selectors into
19996 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19998 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20000 * optabs-query.h (can_vec_perm_p): Delete.
20001 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
20002 * optabs-query.c (can_vec_perm_p): Split into...
20003 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
20004 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
20005 particular selector is valid.
20006 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
20007 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
20008 (vect_grouped_load_supported): Likewise.
20009 (vect_shift_permute_load_chain): Likewise.
20010 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
20011 (vect_transform_slp_perm_load): Likewise.
20012 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
20013 (vectorizable_bswap): Likewise.
20014 (vect_gen_perm_mask_checked): Likewise.
20015 * fold-const.c (fold_ternary_loc): Likewise. Don't take
20016 implementations of variable permutation vectors into account
20017 when deciding which selector to use.
20018 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
20019 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
20020 with a false third argument.
20021 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
20022 to test whether the constant selector is valid and can_vec_perm_var_p
20023 to test whether a variable selector is valid.
20025 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20027 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
20028 * optabs-query.c (can_vec_perm_p): Likewise.
20029 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
20030 instead of vec_perm_indices.
20031 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
20032 (vect_gen_perm_mask_checked): Likewise,
20033 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20034 (vect_gen_perm_mask_checked): Likewise,
20036 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20038 * optabs-query.h (qimode_for_vec_perm): Declare.
20039 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20040 (qimode_for_vec_perm): ...this new function.
20041 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20043 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20045 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20046 does not have a conditional at the top.
20048 2018-01-02 Richard Biener <rguenther@suse.de>
20050 * ipa-inline.c (big_speedup_p): Fix expression.
20052 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20055 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20058 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20062 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20063 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20064 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20065 cond_taken_branch_cost 3->4.
20067 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20069 PR tree-optimization/83581
20070 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20071 TODO_cleanup_cfg if any changes have been made.
20073 PR middle-end/83608
20074 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20075 convert_modes if target mode has the right side, but different mode
20078 PR middle-end/83609
20079 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20080 last argument when extracting from CONCAT. If either from_real or
20081 from_imag is NULL, use expansion through memory. If result is not
20082 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20083 the parts directly to inner mode, if even that fails, use expansion
20086 PR middle-end/83623
20087 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20088 check for bswap in mode rather than HImode and use that in expand_unop
20091 Copyright (C) 2018 Free Software Foundation, Inc.
20093 Copying and distribution of this file, with or without modification,
20094 are permitted in any medium without royalty provided the copyright
20095 notice and this notice are preserved.