1 2018-07-17 Martin Liska <mliska@suse.cz>
3 * opts.c: Do not enable OPT_falign_* for -Os.
5 2018-07-17 Martin Liska <mliska@suse.cz>
7 * align.h (MAX_CODE_ALIGN): New.
8 (MAX_CODE_ALIGN_VALUE): New.
9 * common/config/i386/i386-common.c (ix86_handle_option):
10 (MAX_CODE_ALIGN): Moved to align.h.
11 * final.c (MAX_CODE_ALIGN): Likewise.
12 * opts.c (parse_and_check_align_values):
13 (MAX_CODE_ALIGN): Likewise.
14 (MAX_CODE_ALIGN_VALUE): Likewise.
16 2018-07-17 Martin Liska <mliska@suse.cz>
18 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
19 in order to fulfil coding style.
20 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
21 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
22 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
23 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
24 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
25 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
26 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
27 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
28 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
30 2018-07-17 Martin Liska <mliska@suse.cz>
33 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions directly.
34 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
35 align_flags of label_to_alignment.
36 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into align_flags
38 * config/m68k/m68k.c: Do not use removed align_labels_value and
40 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
41 (LOOP_ALIGN): Likewise.
42 (LABEL_ALIGN): Likewise.
43 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
44 Remove not used macro.
45 (rs6000_loop_align): Change return type to align_flags.
46 (rs6000_loop_align_max_skip): Remove.
47 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
48 Change return type to align_flags.
49 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
50 Remove not used macro.
51 (rs6000_loop_align): Change return type to align_flags.
52 (rs6000_loop_align_max_skip): Remove.
53 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
54 * config/rx/rx-protos.h (rx_align_for_label): Make it
56 * config/rx/rx.c (rx_align_for_label): Change return type
58 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
60 into align_flags class.
61 (LABEL_ALIGN): Likewise.
62 (LOOP_ALIGN): Likewise.
63 * config/s390/s390.c (s390_label_align): Use align_flags
65 (s390_asm_output_function_label): Likewise.
66 * config/sh/sh.c (sh_override_options_after_change):
67 Use align_flags class directly without macros.
68 (find_barrier): Likewise.
69 (barrier_align): Likewise.
70 (sh_loop_align): Likewise.
71 * config/spu/spu.c (spu_option_override):
72 Use align_flags_tuple::get_value instead of removed macros.
73 (spu_sched_init): Likewise.
74 * config/spu/spu.h (GTY): Likewise.
75 * config/visium/visium.c (visium_option_override):
76 Set "8" as default secondary alignment.
77 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
78 in order to guarantee secondary alignment of 8.
79 * coretypes.h: Include align.h header file.
80 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
81 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
82 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
83 * doc/tm.texi.in: Likewise.
84 * final.c (struct label_alignment): Remove not used structure.
85 (LABEL_ALIGN): Change type to align_flags.
86 (LOOP_ALIGN): Likewise.
87 (JUMP_ALIGN): Likewise.
88 (default_loop_align_max_skip): Remove.
89 (default_label_align_max_skip): Likewise.
90 (default_jump_align_max_skip): Likewise.
91 (default_label_align_after_barrier_max_skip):
92 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
93 (LABEL_TO_MAX_SKIP): Remove.
94 (label_to_alignment): Return align_flags type instead of integer.
95 (label_to_max_skip): Remove.
96 (align_fuzz): Use align_flags type.
97 (compute_alignments): Use align_flags type and use align_flags::max
98 to combine multiple alignments.
99 (grow_label_align): Grow vec instead of C array.
100 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
101 (shorten_branches): Use align_flags type and use align_flags::max
102 to combine multiple alignments.
103 (final_scan_insn_1): Remove usage of secondary alignment that comes
104 from label alignment, but instead use proper secondary alignment
105 which is computed in grow_label_align.
106 * flags.h (struct align_flags_tuple): Move to align.h.
107 (struct align_flags): Likewise.
108 (state_align_loops): Rename to align_loops.
109 (state_align_jumps): Rename to align_jumps.
110 (state_align_labels): Rename to align_labels.
111 (state_align_functions): Rename to align_functions.
112 (align_loops_log): Remove.
113 (align_jumps_log): Remove.
114 (align_labels_log): Remove.
115 (align_functions_log): Remove.
116 (align_loops_max_skip): Remove.
117 (align_jumps_max_skip): Remove.
118 (align_labels_max_skip): Remove.
119 (align_functions_max_skip): Remove.
120 (align_loops_value): Remove.
121 (align_jumps_value): Remove.
122 (align_labels_value): Remove.
123 (align_functions_value): Remove.
124 * output.h (label_to_alignment): Change return type to align_flags.
125 (label_to_max_skip): Remove.
126 * target.def: Remove loop_align_max_skip, label_align_max_skip,
127 jump_align_max_skip macros.
128 * targhooks.h (default_loop_align_max_skip): Remove.
129 (default_label_align_max_skip): Likewise.
130 (default_jump_align_max_skip): Likewise.
131 (default_label_align_after_barrier_max_skip): Remove.
132 * toplev.c (read_log_maxskip): Use ::normalize function.
133 (parse_N_M): Remove not used argument and also call ::normalize.
134 (parse_alignment_opts): Do not pass unused arguments.
135 * varasm.c (assemble_start_function): Use directly align_functions
136 instead of removed macros.
137 * system.h: Do not poison removed macros.
139 2018-07-17 Jakub Jelinek <jakub@redhat.com>
142 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
143 and cond temporaries don't have reference type if iterator has
144 pointer type. For init use &for_pre_body instead of pre_p if
145 for_pre_body is non-empty.
147 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
149 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
150 double-double modes to SFmode directly directly.
151 (trunc<mode>sf2_fprs): Delete.
153 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
155 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
156 for conversions between IFmode and the decimal floating point modes.
157 (init_float128_ieee): Use the correct names for conversions between
158 KFmode and the decimal floating point modes.
160 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
162 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
163 for the conversions between TDmode and IFmode.
164 (init_float128_ieee): Use more correct names for the conversions
165 between TDmode and KFmode.
167 2018-07-16 Jakub Jelinek <jakub@redhat.com>
169 PR tree-optimization/86526
170 * builtins.c (expand_builtin_memcmp): Formatting fixes.
171 (inline_expand_builtin_string_cmp): Likewise.
172 (inline_string_cmp): Likewise. Use c_readstr instead of
173 builtin_memcpy_read_str. Add unit_mode temporary.
175 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
178 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
179 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
181 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
183 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
184 Alphabetize prototypes of built-in functions, separating out
185 built-in functions that are listed in this section but should be
188 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
191 * expmed.c (emit_store_flag): Do not emit setcc followed by a
192 conditional move when trapping comparison was split to a
193 non-trapping one (and vice versa).
195 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
197 * config/s390/s390.c (s390_function_profiler): Generate nops
198 instead of profiler call sequences.
199 * config/s390/s390.opt: Add the new option.
201 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
203 * config/s390/s390.c (s390_function_profiler): Generate
204 __mcount_loc section.
205 * config/s390/s390.opt: Add the new option.
207 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
209 * common.opt: Add the new warning.
210 * config/s390/s390.c (s390_function_profiler): Emit "brasl
211 %r0,__fentry__" when -mfentry is specified.
212 (s390_option_override_internal): Disallow -mfentry for 31-bit
214 * config/s390/s390.opt: Add the new option.
216 2018-07-16 Richard Biener <rguenther@suse.de>
219 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
220 for function-local FUNCTION_DECL and RESULT_DECL immediately.
222 2018-07-16 Martin Liska <mliska@suse.cz>
225 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
228 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
230 * config/arc/arcHS.md: Update ARCHS scheduling rules.
232 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
234 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
236 * config/arc/arc-cpus.def (hs4x): New cpu.
238 * config/arc/arc-tables.opt: Regenerate.
239 * config/arc/arc.c (arc_sched_issue_rate): New function.
240 (TARGET_SCHED_ISSUE_RATE): Define.
241 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
242 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
244 (attr tune): Add ARCHS4x tune values.
245 (attr tune_dspmpy): Define.
246 (*tst): Correct instruction type.
247 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
248 * config/arc/arcHS4x.md: New file.
249 * config/arc/fpu.md: Update instruction type attributes.
250 * config/arc/t-multilib: Regenerate.
252 2018-07-16 Tom de Vries <tdevries@suse.de>
255 * var-tracking.c (vt_initialize): Fix pre_dec handling.
257 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
259 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
262 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
264 PR tree-optimization/86514
265 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
266 conversion to a boolean type from a type with greater precision.
268 2018-07-16 Tom de Vries <tdevries@suse.de>
270 * var-tracking.c (vt_initialize): Print adjusted insn slim if
271 dump_flags request TDF_SLIM.
273 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
275 * fold-const.c (int_const_binop_1): Abstract...
276 (wide_int_binop): ...wide int code here.
277 (poly_int_binop): ...poly int code here.
278 Abstract the rest of int_const_binop_1 into int_const_binop.
279 * fold-const.h (wide_int_binop): New.
280 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
281 Remove useless PLUS/MINUS_EXPR case.
282 (zero_nonzero_bits_from_vr): Move wide int code...
283 (zero_nonzero_bits_from_bounds): ...here.
284 (extract_range_from_binary_expr_1): Move mask optimization code...
285 (range_easy_mask_min_max): ...here.
286 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
287 (range_easy_mask_min_max): New.
289 2018-07-15 Jeff Law <law@redhat.com>
292 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
295 2018-07-14 Jim Wilson <jimw@sifive.com>
297 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
299 2018-07-14 Paul Koning <ni1d@arrl.net>
301 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
303 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
305 * lto-streamer-out.c (copy_function_or_variable): Dump info about
308 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
309 Steve Munroe <munroesj52@gmail.com>
311 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
312 (_mm_andnot_si128): Likewise.
313 (_mm_or_si128): Likewise.
314 (_mm_xor_si128): Likewise.
316 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
319 * builtins.c (expand_builtin_memcmp): Inline the calls first
320 when result_eq is false.
321 (expand_builtin_strcmp): Inline the calls first.
322 (expand_builtin_strncmp): Likewise.
323 (inline_string_cmp): New routine. Expand a string compare
324 call by using a sequence of char comparison.
325 (inline_expand_builtin_string_cmp): New routine. Inline expansion
326 a call to str(n)cmp/memcmp.
327 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
329 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
331 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
333 * config/arm/driver-arm.c: Include arm-native.h.
334 (host_detect_local_cpu): Use auto-generated data tables.
335 (vendors, arm_cpu_table): Delete. Move part information to ...
336 * config/arm/arm-cpus.in: ... here.
337 * config/arm/parsecpu.awk (gen_native): New function.
338 (vendor, part): New CPU fields.
339 (END): Add support for building the native CPU detection tables.
340 * config/arm/t-arm (arm-native.h): Add build rule.
341 (driver-arm.o): Add dependency on arm-native.h.
343 2018-07-13 Richard Biener <rguenther@suse.de>
346 * match.pd (addr1 - addr2): Allow either of the operand to
349 2018-07-13 Tom de Vries <tdevries@suse.de>
351 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
354 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
356 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
357 arrays instead of numbered variables.
359 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
361 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
362 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
364 2018-07-13 Richard Biener <rguenther@suse.de>
367 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
368 instead of get_context_die.
370 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
371 Richard Biener <rguenther@suse.de>
374 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
375 that the loop latch destination where phi is defined.
377 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
379 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
380 (riscv_expand_epilogue): Add assertion to check interrupt mode.
381 (riscv_set_current_function): Extract getting interrupt type to new
383 (riscv_get_interrupt_type): New function.
384 (riscv_merge_decl_attributes): New function, checking interrupt type is
386 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
388 2018-07-12 Paul Koning <ni1d@arrl.net>
390 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
393 2018-07-12 Paul Koning <ni1d@arrl.net>
395 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
396 zero reference, add doloop_end instead.
397 * doc/md.texi (decrement_and_branch_until_zero): Remove.
398 (Looping patterns): Remove decrement_and_branch_until_zero. Add
399 detail for doloop_end.
401 2018-07-12 Martin Sebor <msebor@redhat.com>
404 * attribs.c (decl_attributes): Reject conflicting attributes before
405 calling attribute handlers.
407 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
409 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
411 (gcc::dump_manager::get_dump_file_name): likewise.
412 (dump_begin): Likewise.
413 * dumpfile.h (dump_begin): Update prototype.
414 (gcc::dump_manager::get_dump_file_name,
415 gcc::dump_manager::get_dump_file_name): Update prototype.
417 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
419 * internal-fn.h (vectorizable_internal_fn_p): New function.
420 * tree-vect-slp.c (compatible_calls_p): Likewise.
421 (vect_build_slp_tree_1): Remove nops argument. Handle calls
422 to internal functions.
423 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
425 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
427 * fold-const.h (inverse_conditions_p): Declare.
428 * fold-const.c (inverse_conditions_p): New function.
429 * match.pd: Use inverse_conditions_p. Add folds of view_converts
430 that test the inverse condition of a conditional internal function.
431 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
432 * internal-fn.c (internal_fn_mask_index): Handle conditional
434 (vectorized_internal_fn_supported_p): New function.
435 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
436 (any_pred_load_store): Replace with...
437 (need_to_predicate): ...this new variable.
438 (redundant_ssa_names): New variable.
439 (ifcvt_can_use_mask_load_store): Move initial checks to...
440 (ifcvt_can_predicate): ...this new function. Handle tree codes
441 for which a conditional internal function exists.
442 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
443 instead of ifcvt_can_use_mask_load_store. Update after variable
445 (predicate_load_or_store): New function, split out from
446 predicate_mem_writes.
447 (check_redundant_cond_expr): New function.
448 (value_available_p): Likewise.
449 (predicate_rhs_code): Likewise.
450 (predicate_mem_writes): Rename to...
451 (predicate_statements): ...this. Use predicate_load_or_store
452 and predicate_rhs_code.
453 (combine_blocks, tree_if_conversion): Update after above name changes.
454 (ifcvt_local_dce): Handle redundant_ssa_names.
455 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
456 general conditional functions.
457 * tree-vect-stmts.c (vectorizable_call): Likewise.
459 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
460 Alan Hayward <alan.hayward@arm.com>
461 David Sherwood <david.sherwood@arm.com>
463 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
464 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
465 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
466 plus and minus and convert them into IFN_COND_FMA-based sequences.
467 (convert_mult_to_fma): Handle conditional plus and minus.
469 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
471 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
472 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
473 (cond_fnms_optab): New optabs.
474 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
476 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
477 * internal-fn.h (get_conditional_internal_fn): Declare.
478 (get_unconditional_internal_fn): Likewise.
479 * internal-fn.c (cond_ternary_direct): New macro.
480 (expand_cond_ternary_optab_fn): Likewise.
481 (direct_cond_ternary_optab_supported_p): Likewise.
482 (FOR_EACH_COND_FN_PAIR): Likewise.
483 (get_conditional_internal_fn): New function.
484 (get_unconditional_internal_fn): Likewise.
485 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
486 (gimple_match_op::gimple_match_op): Add a new overload for 5
488 (gimple_match_op::set_op): Likewise.
489 (gimple_resimplify5): Declare.
490 * genmatch.c (decision_tree::gen): Generate simplifications for
492 * gimple-match-head.c (gimple_simplify): Define an overload for
493 5 operands. Handle calls with 5 arguments in the top-level overload.
494 (convert_conditional_op): Handle conversions from unconditional
495 internal functions to conditional ones.
496 (gimple_resimplify5): New function.
497 (build_call_internal): Pass a fifth operand.
498 (maybe_push_res_to_seq): Likewise.
499 (try_conditional_simplification): Try converting conditional
500 internal functions to unconditional internal functions.
501 Handle 3-operand unconditional forms.
502 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
503 Define ternary equivalents of the current rules for binary conditional
505 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
507 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
508 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
509 (optab): Handle them.
510 (SVE_COND_FP_TERNARY): New int iterator.
511 (sve_fmla_op, sve_fmad_op): New int attributes.
512 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
513 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
514 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
516 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
518 * target.def (preferred_else_value): New target hook.
519 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
520 * doc/tm.texi: Regenerate.
521 * targhooks.h (default_preferred_else_value): Declare.
522 * targhooks.c (default_preferred_else_value): New function.
523 * internal-fn.h (conditional_internal_fn_code): Declare.
524 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
525 (get_conditional_internal_fn): Use it.
526 (conditional_internal_fn_code): New function.
527 * gimple-match.h (gimple_match_cond): New struct.
528 (gimple_match_op): Add a cond member function.
529 (gimple_match_op::gimple_match_op): Update all forms to take a
531 * genmatch.c (expr::gen_transform): Use the same condition as res_op
532 for the suboperation, but don't specify a particular else_value.
533 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
534 (visit_nary_op, visit_reference_op_load): Pass
535 gimple_match_cond::UNCOND to the gimple_match_op constructor.
536 * gimple-match-head.c: Include tree-eh.h
537 (convert_conditional_op): New function.
538 (maybe_resimplify_conditional_op): Likewise.
539 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
540 (gimple_resimplify2): Likewise.
541 (gimple_resimplify3): Likewise.
542 (gimple_resimplify4): Likewise.
543 (maybe_push_res_to_seq): Return null for conditional operations.
544 (try_conditional_simplification): New function.
545 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
547 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
549 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
551 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
553 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
555 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
557 (hash_tree): Do not hash DECL_FCONTEXT
558 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
559 Do not stream DECL_FCONTEXT.
560 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
561 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
563 2018-07-12 Richard Biener <rguenther@suse.de>
566 * dwarf2out.c (gen_block_die): Only output blocks when they have
567 at least one !DECL_IGNORED_P variable.
569 2018-07-12 Richard Biener <rguenther@suse.de>
572 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
573 Remove -mieee-fp handling.
575 2018-07-12 Richard Biener <rguenther@suse.de>
577 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
578 left-over from last patch.
580 2018-07-12 Jakub Jelinek <jakub@redhat.com>
582 PR tree-optimization/86492
583 * gimple-ssa-store-merging.c
584 (imm_store_chain_info::coalesce_immediate_stores): Call
585 check_no_overlap even for the merge_overlapping case. Formatting fix.
587 2018-07-12 Richard Biener <rguenther@suse.de>
590 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
591 move possibly trapping operations into the conditional.
593 2018-07-12 Richard Biener <rguenther@suse.de>
595 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
596 (vn_lookup_simplify_result): Remove recursion limit applied
598 (vn_nary_build_or_lookup_1): Adjust.
599 (try_to_simplify): Likewise.
600 * gimple-match-head.c (gimple_resimplify1): Instead apply one
602 (gimple_resimplify2): Likewise.
603 (gimple_resimplify3): Likewise.
604 (gimple_resimplify4): Likewise.
606 2018-07-11 Jakub Jelinek <jakub@redhat.com>
608 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
609 Use __mmask64 type instead of __mmask8 for __M argument.
610 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
611 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
613 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
614 __mmask16 for __M argument.
615 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
616 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
617 to __mmask16 instead of __mmask8.
618 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
619 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
620 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
621 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
622 instead of __mmask16 for __U argument.
623 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
624 __mmask16 instead of __mmask8 for __U argument.
625 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
627 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
629 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
631 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
633 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
635 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
637 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
639 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
641 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
643 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
645 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
646 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
647 return type as well as __M argument type and all casts from __mmask8
649 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
650 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
651 return type as well as __M argument type and all casts from __mmask8
653 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
654 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
655 return type as well as __M argument type and all casts from __mmask8
657 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
658 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
659 return type as well as __M argument type and all casts from __mmask8
661 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
662 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
665 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
667 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
668 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
671 2018-07-11 Paul Koning <ni1d@arrl.net>
673 * doc/md.texi (define_subst): Document how multiple occurrences of
674 the same argument in the replacement pattern are handled.
676 2018-07-11 Paul Koning <ni1d@arrl.net>
678 * doc/extend.texi (Common Variable Attributes): Move "mode" into
680 (Common Type Attributes): Add "mode" attribute.
682 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
684 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
685 stream DECL_ORIGINAL_TYPE.
686 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
687 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
688 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
689 Do not walk original type.
690 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
692 (write_ts_decl_non_common_tree_pointers): Do not stream
694 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
695 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
697 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
699 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
700 thread twice from the same starting edge.
702 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
704 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
705 * gimple.c (gimple_call_nonnull_result_p): ...here...
706 (gimple_call_nonnull_arg): ...and here.
707 * gimple.h (gimple_call_nonnull_result_p): New.
708 (gimple_call_nonnull_arg): New.
710 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
712 * config/arm/arm-cpus.in: Move information from fpu field of each
713 cpu definition to the isa field.
714 * config/arm/parsecpu.awk (fpu): Delete match rule.
715 (gen_comm_data): Don't add bits from the CPU's FPU entry.
717 2018-07-11 Richard Biener <rguenther@suse.de>
720 * dwarf2out.c (init_sections_and_labels): Use
721 output_asm_line_debug_info consistently.
722 (dwarf2out_early_finish): Likewise.
723 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
726 2018-07-11 Richard Biener <rguenther@suse.de>
728 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
729 Rework father_bb setting in a way to avoid propagating constants
730 multiple times on a loop body.
732 2018-07-10 Mark Wielaard <mark@klomp.org>
735 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
737 2018-07-10 Richard Biener <rguenther@suse.de>
739 * hash-map.h (hash_map::iterator::operator*): Return
740 references to key and value.
742 2018-07-10 Jakub Jelinek <jakub@redhat.com>
745 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
746 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
747 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
748 (gimplify_omp_for): For composite loops, move outer
749 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
750 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
751 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
752 TREE_LIST for both the original class iterator and the "last" helper
753 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
754 loop, remember has_decl_expr from outer composite loops for the
755 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
757 2018-07-09 Martin Sebor <msebor@redhat.com>
761 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
762 accessing implicitly initialized array elements.
763 * expr.c (string_constant): Handle string initializers of
764 character arrays within aggregates.
765 * gimple-fold.c (fold_array_ctor_reference): Add argument.
766 Store element offset. As a special case, handle zero size.
767 (fold_nonarray_ctor_reference): Same.
768 (fold_ctor_reference): Add argument. Store subobject offset.
769 * gimple-fold.h (fold_ctor_reference): Add argument.
771 2018-07-09 Paul Koning <ni1d@arrl.net>
773 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
774 (pdp11_insn_cost): New function.
775 (pdp11_md_asm_adjust): New function.
776 (TARGET_INVALID_WITHIN_DOLOOP): Define.
777 (pdp11_rtx_costs): Update to match machine better.
778 (output_addr_const_pdp11): Correct format mismatch warnings.
779 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
780 * config/pdp11/pdp11.md: General change to add base_cost and/or
781 length attributes for use by new pdp11_insn_cost function.
782 (MIN_BRANCH): Correct definition.
784 (doloop_end): Use standard pattern name for looping pattern.
785 (doloop_end_nocc): New.
786 (movsf): Add another constraint alternative.
787 (zero_extendqihi2): Add constraint alternatives for not in place
789 (zero_extendhisi2): Remove.
790 (shift patterns): Add CC handling variants.
794 (define_peephole2): New peephole to recognize mov that sets CC for
797 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
799 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
800 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
803 2018-07-09 Jakub Jelinek <jakub@redhat.com>
806 * real.c (real_nextafter): Return true if result is denormal.
808 2018-07-09 Martin Liska <mliska@suse.cz>
810 * common.opt: Add back wrongly removed attribute.
812 2018-07-09 Richard Biener <rguenther@suse.de>
815 * dwarf2out.c (gen_block_die): For an early generated DIE
816 always output high/low PC attributes.
818 2018-07-09 Tom de Vries <tdevries@suse.de>
820 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
821 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
822 onto VAR_DECL with abstract origin.
824 2018-07-07 Jim Wilson <jimw@sifive.com>
826 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
828 2018-07-07 Tom de Vries <tdevries@suse.de>
830 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
832 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
834 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
836 (combine_bound): Use wide-int overflow calculation instead of
838 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
840 * fold-const.c (int_const_binop_2): Same.
841 (extract_muldiv_1): Same.
842 (fold_div_compare): Same.
843 (fold_abs_const): Same.
845 * poly-int.h (add): Same.
849 * predict.c (predict_iv_comparison): Same.
850 * profile-count.c (slow_safe_scale_64bit): Same.
851 * simplify-rtx.c (simplify_const_binary_operation): Same.
852 * tree-chrec.c (tree_fold_binomial): Same.
853 * tree-data-ref.c (split_constant_offset_1): Same.
854 * tree-if-conv.c (idx_within_array_bound): Same.
855 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
856 * tree-ssa-phiopt.c (minmax_replacement): Same.
857 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
858 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
859 * vr-values.c (vr_values::adjust_range_with_scev): Same.
860 * wide-int.cc (wi::add_large): Same.
861 (wi::mul_internal): Same.
862 (wi::sub_large): Same.
863 (wi::divmod_internal): Same.
864 * wide-int.h: Change overflow type to overflow_type for neg, add,
865 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
866 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
867 mul_internal, divmod_internal.
868 (overflow_type): New enum.
869 (accumulate_overflow): New.
871 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
873 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
874 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
876 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
878 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
879 argument is checked for zero before entering loop, avoid checking again.
881 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
883 * gimplify.h (generic_expr_could_trap_p): Set as global function.
884 * gimplify.h (generic_expr_could_trap_p): Likwise.
885 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
887 2018-07-06 Jakub Jelinek <jakub@redhat.com>
889 PR tree-optimization/86401
890 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
891 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
892 (fold_bit_and_mask): ... here. New helper function for match.pd.
893 * fold-const.h (fold_bit_and_mask): Declare.
894 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
896 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
899 * target.def (translate_mode_attribute): New hook.
900 * targhooks.h (default_translate_mode_attribute): Declare.
901 * targhooks.c (default_translate_mode_attribute): New function.
902 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
903 * doc/tm.texi: Regenerate.
904 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
905 (rs6000_translate_mode_attribute): New function.
907 2018-07-06 Paul Koning <ni1d@arrl.net>
909 * doc/md.texi (define_split): Document DONE and FAIL.
910 (define_peephole2): Ditto.
912 2018-07-05 Jeff Law <law@redhat.com>
914 PR tree-optimization/86010
915 * tree-ssa-dse.c (compute_trims): More aggressively trim at
916 both the head and tail of mem* and str* calls.
918 2018-07-05 Jim Wilson <jimw@sifive.com>
920 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
922 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
924 * config/aarch64/aarch64-simd.md: correct flags text for
925 MIN_EXPR replacement.
927 2018-07-05 James Clarke <jrtc27@jrtc27.com>
929 * configure: Regenerated.
931 2018-07-05 Carl Love <cel@us.ibm.com>
933 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
934 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
935 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
936 VSX_BUILTIN_DOUBLEL_V4SF.
938 2018-07-05 Martin Sebor <msebor@redhat.com>
941 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
942 than its domain to compute its the upper bound of a char array.
944 2018-07-05 Nathan Sidwell <nathan@acm.org>
946 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
947 * doc/cpp.texi: Update comment.
948 * doc/tm.texi: Rebuilt.
949 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
950 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
951 * doc/extend.texi (Backwards Compatibility): Clarify it is system
952 headers affected by extern "C".
953 * system.h: Poison NO_IMPLICIT_EXTERN_C.
954 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
955 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
956 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
957 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
958 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
959 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
960 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
961 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
962 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
963 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
964 config/sparc/sp64-elf.h, config/spu/spu.h,
965 config/stormy16/stormy16.h, config/v850/v850.h,
966 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
967 define NO_IMPLICIT_EXTERN_C.
968 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
970 2018-07-05 Tamar Christina <tamar.christina@arm.com>
973 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
974 instead of GET_MODE_SIZE when comparing Units.
976 2018-07-05 Tamar Christina <tamar.christina@arm.com>
979 * rtlanal.c (set_noop_p): Constrain on mode change,
980 include hard-reg-set.h
982 2018-07-05 Tamar Christina <tamar.christina@arm.com>
984 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
986 2018-07-05 Jakub Jelinek <jakub@redhat.com>
989 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
992 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
994 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
996 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
999 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
1001 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
1003 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
1006 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
1008 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
1010 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
1011 neg_*_op* variables.
1013 2018-07-04 Martin Liska <mliska@suse.cz>
1015 * tree-switch-conversion.c: Define
1016 max_ratio_for_speed and max_ratio_for_size constants.
1018 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1019 Martin Liska <mliska@suse.cz>
1024 * common.opt: Rename align options with 'str_' prefix.
1025 * common/config/i386/i386-common.c (set_malign_value): New
1027 (ix86_handle_option): Use it to set -falign-* options/
1028 * config/aarch64/aarch64-protos.h (struct tune_params): Change
1029 type from int to string.
1030 * config/aarch64/aarch64.c: Update default values from int
1032 * config/alpha/alpha.c (alpha_override_options_after_change):
1034 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
1035 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1036 max skip conditionally.
1037 * config/i386/freebsd.h (SUBALIGN_LOG): New.
1038 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1039 max skip conditionally.
1040 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1041 max skip conditionally.
1042 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
1043 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1044 max skip conditionally.
1045 * config/i386/i386.c (struct ptt): Change type from int to
1047 (ix86_default_align): Set default values.
1048 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
1049 max skip conditionally.
1050 * config/i386/iamcu.h (SUBALIGN_LOG): New.
1051 (ASM_OUTPUT_MAX_SKIP_ALIGN):
1052 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
1053 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1054 max skip conditionally.
1055 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
1056 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
1057 * config/i386/x86-64.h (SUBALIGN_LOG): New.
1058 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
1059 max skip conditionally.
1060 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1061 * config/ia64/ia64.c (ia64_option_override): Set default values
1062 for alignment options.
1063 * config/m68k/m68k.c: Handle new str_align_* options.
1064 * config/mips/mips.c (mips_set_compression_mode): Change
1066 (mips_option_override): Set default values for options.
1067 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
1069 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1071 * config/rx/rx.c (rx_option_override): Likewise.
1072 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
1073 (LABEL_ALIGN): Use align_labels_log.
1074 (LOOP_ALIGN): Use align_loops_align.
1075 * config/s390/s390.c (s390_asm_output_function_label): Use new
1077 * config/sh/sh.c (sh_override_options_after_change):
1078 Change type of constants.
1079 * config/spu/spu.c (spu_sched_init): Likewise.
1080 * config/sparc/sparc.c (sparc_option_override): Set default
1082 * config/visium/visium.c (visium_option_override): Likewise.
1083 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
1084 emit p2align format with last argument if it's not needed.
1085 * doc/invoke.texi: Document extended format of -falign-*.
1086 * final.c: Use align_labels alignment.
1087 * flags.h (struct target_flag_state): Change type to use
1089 (struct align_flags_tuple): New.
1090 (struct align_flags): Likewise.
1091 (align_loops_log): Redefine macro to use new types.
1092 (align_loops_max_skip): Redefine macro to use new types.
1093 (align_jumps_log): Redefine macro to use new types.
1094 (align_jumps_max_skip): Redefine macro to use new types.
1095 (align_labels_log): Redefine macro to use new types.
1096 (align_labels_max_skip): Redefine macro to use new types.
1097 (align_functions_log): Redefine macro to use new types.
1098 (align_loops): Redefine macro to use new types.
1099 (align_jumps): Redefine macro to use new types.
1100 (align_labels): Redefine macro to use new types.
1101 (align_functions): Redefine macro to use new types.
1102 (align_functions_max_skip): Redefine macro to use new types.
1103 (align_loops_value): New macro.
1104 (align_jumps_value): New macro.
1105 (align_labels_value): New macro.
1106 (align_functions_value): New macro.
1107 * function.c (invoke_set_current_function_hook): Propagate
1108 alignment values from flags to global variables default in
1110 * ipa-icf.c (sem_function::equals_wpa): Use
1111 cl_optimization_option_eq instead of memcmp.
1112 * lto-streamer.h (cl_optimization_stream_out): Support streaming
1114 (cl_optimization_stream_in): Likewise.
1115 * optc-save-gen.awk: Support strings in cl_optimization.
1116 * opth-gen.awk: Likewise.
1117 * opts.c (finish_options): Remove error checking of invalid
1119 (MAX_CODE_ALIGN): Remove.
1120 (MAX_CODE_ALIGN_VALUE): Likewise.
1121 (parse_and_check_align_values): New function.
1122 (check_alignment_argument): Likewise.
1123 (common_handle_option): Use check_alignment_argument.
1124 * opts.h (parse_and_check_align_values): Declare.
1125 * toplev.c (init_alignments): Remove.
1126 (read_log_maxskip): New.
1127 (parse_N_M): Likewise.
1128 (parse_alignment_opts): Likewise.
1129 (backend_init_target): Remove usage of init_alignments.
1130 * toplev.h (parse_alignment_opts): Declare.
1131 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
1133 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
1134 * tree.c (cl_option_hasher::equal): New.
1135 * varasm.c: Use new global macros.
1137 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
1139 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
1140 Use a simpler align directive also if MAXSKIP = ALIGN-1.
1141 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1142 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1143 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1144 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
1145 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
1146 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
1147 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
1148 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1149 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1150 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1151 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
1153 2018-07-04 Martin Liska <mliska@suse.cz>
1154 Jonathan Wakely <jwakely@redhat.com>
1156 * coverage.c: Use correct type.
1157 * doc/invoke.texi: Language correction.
1159 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
1162 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
1163 ENDBRANCH for non-tail call which may return via indirect branch.
1164 * doc/extend.texi: Document indirect_return attribute.
1166 2018-07-03 Martin Sebor <msebor@redhat.com>
1168 PR tree-optimization/86274
1169 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
1171 (format_floating): Correct handling of infinities and NaNs.
1173 2018-07-03 Martin Sebor <msebor@redhat.com>
1175 * print-tree.c (print_real_cst): New function.
1176 (print_node_brief): Call it.
1177 (print_node): Ditto.
1179 2018-07-03 Jeff Law <law@redhat.com>
1181 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
1182 into a single pattern.
1184 * config/h8300/h8300.md (ors code_iterator): New.
1185 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
1186 a single pattern and single splitter.
1187 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
1188 (iorqi3_1, xorqi3_1): Likewise.
1189 (iorqi3, xorqi3 expanders): Similarly.
1191 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
1192 (movmd_internal) into a single pattern using the P mode iterator.
1193 (movmd splitters): Similarly.
1194 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
1195 (movsd splitters): Similarly.
1197 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
1198 ADDB, ADDW and ADDL into a single ADD attribute which selects the
1199 right table based on the size of the operand.
1200 * config/h8300/h8300.md (length_table): Corresponding changes. All
1201 references to "addb", "addw" and "addl" changed to "add".
1202 (btst patterns): Merge two variants into a single pattern.
1203 (tstqi, tsthi): Likewise.
1204 (addhi3_incdec, addsi3_incdec): Likewise.
1205 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
1206 (mulhi3, mulsi3): Likewise.
1207 (udivhi3, udivsi3): Likewise.
1208 (divhi3, divsi3): Likewise.
1209 (andorqi3, andorhi3, andorsi3): Likewise.
1211 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
1214 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
1215 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
1217 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1219 PR tree-optimization/85694
1220 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
1221 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
1223 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
1224 (<u>avg<mode>3_ceil): New patterns.
1226 2018-07-03 David Malcolm <dmalcolm@redhat.com>
1228 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
1229 scan-tree-dump directive.
1230 * gcc.dg/vect/slp-perm-2.c: Likewise.
1231 * gcc.dg/vect/slp-perm-3.c: Likewise.
1232 * gcc.dg/vect/slp-perm-5.c: Likewise.
1233 * gcc.dg/vect/slp-perm-6.c: Likewise.
1234 * gcc.dg/vect/slp-perm-7.c: Likewise.
1235 * gcc.dg/vect/slp-perm-8.c: Likewise.
1237 2018-07-03 Marek Polacek <polacek@redhat.com>
1240 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
1242 2018-07-03 Richard Biener <rguenther@suse.de>
1245 * tree-ssa-structalias.c (find_func_clobbers): Properly
1246 handle indirect calls.
1248 2018-07-03 Jeff Law <law@redhat.com>
1250 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
1251 (shifts): New code iterator.
1252 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
1253 expander. Fix HImode handling on H8/SX.
1254 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
1255 (subqi3, subhi3, subsi3 expanders): Likewise.
1256 (andqi3, andhi3, andsi3 expanders): Likewise.
1257 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
1258 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
1259 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
1260 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
1261 (zero_extendqihi2, zero_extendqisi2): Likewise.
1262 (extendqihi2, extendqisi2): Likewise.
1263 (rotlqi3, rotlhi3, rotlsi3): Likewise.
1264 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
1265 (rotlqi3_1, rotlhi3_1): Likewise.
1266 (logicalhi3_sn, logicalsi3_sn): Likewise.
1267 (logicalhi3, logicalsi3): Likewise.
1269 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1271 * tree-vect-patterns.c (vect_recog_rotate_pattern)
1272 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1273 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
1274 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
1275 type to append_pattern_def_seq instead of creating a stmt_vec_info
1277 (build_mask_conversion): Likewise. Remove vinfo argument.
1278 (vect_add_conversion_to_patterm): Likewise, renaming to...
1279 (vect_add_conversion_to_pattern): ...this.
1280 (vect_recog_mask_conversion_pattern): Update call to
1281 build_mask_conversion. Pass the vector type to
1282 append_pattern_def_seq here too.
1283 (vect_recog_gather_scatter_pattern): Update call to
1284 vect_add_conversion_to_pattern.
1286 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1288 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
1289 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1290 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
1291 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
1292 STMT_VINFO_PATTERN_DEF_SEQ to null here.
1293 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
1294 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
1295 append_pattern_def_seq instead of new_pattern_def_seq.
1296 (vect_recog_divmod_pattern): Do both of the above.
1297 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
1300 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1302 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
1303 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
1304 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
1305 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
1306 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
1307 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1308 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
1309 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
1310 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
1311 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
1312 parameter with a single stmt_vec_info.
1313 (vect_recog_func_ptr): Likewise.
1314 (vect_recog_gather_scatter_pattern): Likewise, folding in...
1315 (vect_try_gather_scatter_pattern): ...this.
1316 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
1317 the stmt_vec_info of the statement to be matched. Don't clear
1318 STMT_VINFO_RELATED_STMT.
1319 (vect_pattern_recog): Update call accordingly.
1321 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1323 PR tree-optimization/85694
1324 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
1325 (uavgM3_ceil): Document new optabs.
1326 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
1327 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
1329 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
1330 (savg_ceil_optab): New optabs.
1331 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
1332 (vect_vect_recog_func_ptrs): Add it.
1333 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
1334 constant directly from the associated lhs.
1336 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1338 * tree-vect-patterns.c (vect_split_statement): New function.
1339 (vect_convert_input): Use it to try to split an existing cast.
1341 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1343 * poly-int.h (print_hex): New function.
1344 * dumpfile.h (dump_dec, dump_hex): Declare.
1345 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
1346 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
1347 min_input_precision, operation_precision and operation_sign.
1348 * tree-vect-patterns.c (vect_get_range_info): New function.
1349 (vect_same_loop_or_bb_p, vect_single_imm_use)
1350 (vect_operation_fits_smaller_type): Delete.
1351 (vect_look_through_possible_promotion): Add an optional
1352 single_use_p parameter.
1353 (vect_recog_over_widening_pattern): Rewrite to use new
1354 stmt_vec_info infomration. Handle one operation at a time.
1355 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
1356 (vect_truncatable_operation_p, vect_set_operation_type)
1357 (vect_set_min_input_precision): New functions.
1358 (vect_determine_min_output_precision_1): Likewise.
1359 (vect_determine_min_output_precision): Likewise.
1360 (vect_determine_precisions_from_range): Likewise.
1361 (vect_determine_precisions_from_users): Likewise.
1362 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
1363 (vect_vect_recog_func_ptrs): Put over_widening first.
1365 (vect_pattern_recog): Call vect_determine_precisions.
1367 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
1369 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
1370 statements that have been replaced by further pattern statements.
1371 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
1373 2018-07-03 Richard Biener <rguenther@suse.de>
1375 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
1376 always set *dt. Dump vectype in vectype overload.
1377 * dumpfile.h (dump_gimple_expr): New function.
1378 (dump_gimple_expr_loc): Likewise.
1379 * dumpfile.c (dump_gimple_expr): New function.
1380 (dump_gimple_expr_loc): Likewise.
1382 2018-07-02 Jeff Law <law@redhat.com>
1384 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
1385 the H8/300, H8/300H and H8/S variants into a single pattern.
1386 (movhi_h8300, movqi_h8300hs): Similarly.
1387 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
1388 (QHI mode iterator): New.
1390 * config/h8300/h8300.md: Remove trailing whitespace.
1392 2018-07-02 Jim Wilson <jimw@sifive.com>
1394 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
1395 instead of emit_insn for interrupt returns.
1396 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
1397 (riscv_sret, riscv_uret): Likewise.
1399 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1401 * pretty-print.c (selftest::test_pp_format): Move save and restore
1402 of quotes to class auto_fix_quotes, and add an instance.
1403 * selftest.c: Include "intl.h".
1404 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
1405 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
1406 * selftest.h (selftest::auto_fix_quotes): New class.
1408 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
1410 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
1411 (aarch64_sve_prepare_conditional_op): Remove.
1412 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
1413 Allow aarch64_simd_reg_or_zero as select operand; remove
1414 the aarch64_sve_prepare_conditional_op call.
1415 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
1416 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1417 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
1418 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
1419 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
1420 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
1421 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
1422 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
1423 and a splitters to match all of the *_any patterns.
1424 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
1426 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
1427 (SVE_COND_FP_BINARY_REV): Remove.
1428 (sve_int_op_rev, sve_fp_op_rev): New.
1429 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
1430 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
1431 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
1432 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
1433 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
1434 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
1435 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
1436 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
1437 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
1439 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
1440 Remove match_dup 1 from the inner unspec.
1441 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
1443 * config/aarch64/aarch64.md (movprfx): New attr.
1444 (length): Default movprfx to 8.
1445 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
1446 (*madd<SVE_I>, *msub<SVE_I): Likewise.
1447 (*<su>mul<SVE_I>3_highpart): Likewise.
1448 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
1449 (*v<ASHIFT><SVE_I>3): Likewise.
1450 (*<su><MAXMIN><SVE_I>3): Likewise.
1451 (*<su><MAXMIN><SVE_F>3): Likewise.
1452 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
1453 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
1454 (*div<SVE_F>4): Likewise.
1456 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
1458 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
1461 2018-07-02 Richard Biener <rguenther@suse.de>
1463 PR tree-optimization/86363
1464 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
1465 memset argument refers to a non-variable address.
1467 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1469 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
1470 {PLUS,MINUS}_EXPR code to...
1471 (adjust_symbolic_bound): ...here,
1472 (combine_bound): ...here,
1473 (set_value_range_with_overflow): ...and here.
1475 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
1477 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
1479 (extract_range_from_abs_expr): ...here.
1481 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
1483 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
1484 -fno-omit-frame-pointer when not optimizing.
1486 2018-07-02 Martin Liska <mliska@suse.cz>
1489 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
1490 (propagate_nothrow): Likewise.
1492 2018-07-02 Martin Liska <mliska@suse.cz>
1495 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
1497 2018-07-02 David Malcolm <dmalcolm@redhat.com>
1499 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
1500 function in r262149, changing "loc" param from source_location to
1501 const dump_location_t &.
1502 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
1503 declaration, as above.
1505 2018-07-01 Paul Koning <ni1d@arrl.net>
1507 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1508 -munit-asm, -mgnu-asm, -mdec-asm.
1509 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
1510 (pdp11_output_labelref): New.
1511 (pdp11_output_def): New.
1512 (pdp11_output_addr_vec_elt): New.
1513 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
1514 %# and %@ format codes.
1515 (pdp11_option_override): New.
1516 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
1517 (pdp11_output_ident): New.
1518 (pdp11_asm_named_section): New.
1519 (pdp11_asm_init_sections): New.
1520 (pdp11_file_start): New.
1521 (pdp11_file_end): New.
1522 (output_ascii): Use .ascii/.asciz for -mdec-asm.
1523 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
1524 %o, like %c but octal.
1525 (pdp11_option_override): New.
1526 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
1528 (DATA_SECTION_ASM_OP): Ditto.
1529 (READONLY_DATA_SECTION_ASM_OP): New.
1530 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
1531 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
1532 (ASM_OUTPUT_LABELREF): Ditto.
1533 (ASM_OUTPUT_DEF): Ditto.
1534 (ASM_OUTPUT_EXTERNAL): New.
1535 (ASM_OUTPUT_SOURCE_FILENAME): New.
1536 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
1537 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
1538 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
1539 %# and %@ format codes.
1540 * config/pdp11/pdp11.opt (mgnu-asm): New.
1541 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
1542 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
1543 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
1545 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
1547 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
1548 dereferencing path[] beyond its length.
1550 (debug_all_paths): New.
1551 (rewire_first_differing_edge): New.
1552 (adjust_paths_after_duplication): New.
1553 (duplicate_thread_path): Call adjust_paths_after_duplication.
1555 (thread_through_all_blocks): Add new argument to
1556 duplicate_thread_path.
1558 2018-06-30 Jim Wilson <jimw@sifive.com>
1560 * config/riscv/predicates.md (p2m1_shift_operand): New.
1561 (high_mask_shift_operand): New.
1562 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
1563 pattern using p2m1_shift_operand.
1564 (lshsi3_zero_extend_3+2): New combiner pattern using
1565 high_mask_shift_operand.
1567 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1569 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
1571 (vect_recog_rotate_pattern): ...here.
1572 (vect_convert_input): Try to insert casts of invariants in the
1574 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
1575 preheader to be empty.
1577 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1579 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
1580 vector type. If given, install it in the new statement's
1582 (vect_element_precision): New function.
1583 (vect_unpromoted_value): New struct.
1584 (vect_unpromoted_value::vect_unpromoted_value): New function.
1585 (vect_unpromoted_value::set_op): Likewise.
1586 (vect_look_through_possible_promotion): Likewise.
1587 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
1588 (vect_widened_op_tree, vect_convert_input): Likewise.
1589 (vect_convert_inputs, vect_convert_output): Likewise.
1590 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
1591 to handle the optional cast of the multiplication result and
1592 vect_widened_op_tree to detect the widened multiplication itself.
1593 Do not require the input and output of promotion casts to have
1594 the same sign, but base the signedness of the operation on the
1595 input rather than the result. If the pattern includes two
1596 promotions, check that those promotions have the same sign.
1597 Do not restrict the MULT_EXPR handling to a double-width result;
1598 handle quadruple-width results and wider. Use vect_convert_inputs
1599 to convert the inputs to the common type.
1600 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
1601 to handle the optional cast of the ABS result. Also allow a sign
1602 change or a sign extension between the ABS and MINUS.
1603 Use vect_widened_op_tree to detect the widened subtraction and use
1604 vect_convert_inputs to convert the inputs to the common type.
1605 (vect_handle_widen_op_by_const): Delete.
1606 (vect_recog_widen_op_pattern): New function.
1607 (vect_recog_widen_mult_pattern): Use it.
1608 (vect_recog_widen_shift_pattern): Likewise.
1609 (vect_recog_widen_sum_pattern): Use
1610 vect_look_through_possible_promotion to handle the promoted
1613 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1615 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
1616 the containing gimple_seq *.
1617 * gimple-iterator.h (gsi_for_stmt): Declare it.
1618 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
1619 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
1620 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
1621 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
1622 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
1624 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
1626 (vect_mark_pattern_stmts): ...here. Handle cases in which the
1627 statement being replaced is part of an existing pattern
1628 definition sequence, inserting the new pattern statements before
1630 (vect_pattern_recog_1): Don't return a bool. If the statement
1631 is already part of a pattern, instead apply pattern matching
1632 to the pattern definition statements. Don't clear the
1633 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
1634 (vect_pattern_recog): Don't break after the first match;
1635 continue processing the pattern definition statements instead.
1636 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
1638 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1640 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
1641 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
1642 (vect_recog_widen_sum_pattern): Use it.
1644 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1646 * tree-vect-loop.c (vectorizable_reduction): Assert that the
1647 phi is not a pattern statement and has not been replaced by
1648 a pattern statement.
1649 * tree-vect-patterns.c (type_conversion_p): Don't check
1650 STMT_VINFO_IN_PATTERN_P.
1651 (vect_recog_vector_vector_shift_pattern): Likewise.
1652 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
1653 the pattern statement rather than the original statement; check
1654 directly for a WIDEN_MULT_EXPR here.
1655 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
1656 vect_is_simple_use to return the pattern statement rather
1657 than the original statement; use is_pattern_stmt_p to check
1658 for such a pattern statement.
1659 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
1660 to return the pattern statement rather than the original statement;
1661 don't do the same transformation here.
1662 (vect_is_simple_use): If the defining statement has been replaced
1663 by a pattern statement, return the pattern statement instead.
1664 Remove the corresponding (local) transformation from the vectype
1667 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
1669 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
1670 end and default to null.
1671 * tree-vect-loop.c (vect_create_epilog_for_reduction)
1672 (vectorizable_reduction): Update calls accordingly, dropping the
1673 gimple ** argument if the passed-back statement isn't needed.
1674 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
1675 (vect_recog_rotate_pattern): Likewise.
1676 (vect_recog_mask_conversion_pattern): Likewise.
1677 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1678 (vect_mask_constant_operand_p): Likewise.
1679 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
1680 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
1681 (get_group_load_store_type, get_load_store_type): Likewise.
1682 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
1683 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
1684 (vectorizable_conversion, vectorizable_assignment): Likewise.
1685 (vectorizable_shift, vectorizable_operation): Likewise.
1686 (vectorizable_store, vect_is_simple_cond): Likewise.
1687 (vectorizable_condition, vectorizable_comparison): Likewise.
1688 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
1689 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
1690 and move it to the end. Cope with null def_stmt_outs.
1692 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1694 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
1696 2018-06-29 Jeff Law <law@redhat.com>
1698 * config/v850/v850.c (v850_legitimate_address_p): Handle large
1699 displacements for TARGET_V850E2V3 and newer.
1700 (TARGET_LRA_P): Remove. Defaults to LRA now.
1701 * config/v850/v850.md (sign23byte_load): Remove.
1702 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
1703 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
1705 2018-06-29 Martin Liska <mliska@suse.cz>
1708 * coverage.c (coverage_init): Mangle full path name.
1709 * doc/invoke.texi: Document the change.
1710 * gcov-io.c (mangle_path): New.
1711 * gcov-io.h (mangle_path): Likewise.
1712 * gcov.c (mangle_name): Use mangle_path for path mangling.
1714 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1716 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
1717 if starting source register is not even.
1719 2018-06-29 Martin Liska <mliska@suse.cz>
1721 PR tree-optimization/86263
1722 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
1723 Make edge redirection.
1725 2018-06-29 David Malcolm <dmalcolm@redhat.com>
1727 * dumpfile.c (dump_loc): Add indentation based on scope depth.
1728 (dump_scope_depth): New variable.
1729 (get_dump_scope_depth): New function.
1730 (dump_begin_scope): New function.
1731 (dump_end_scope): New function.
1732 * dumpfile.h (get_dump_scope_depth): New declaration.
1733 (dump_begin_scope): New declaration.
1734 (dump_end_scope): New declaration.
1735 (class auto_dump_scope): New class.
1736 (AUTO_DUMP_SCOPE): New macro.
1737 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
1740 2018-06-29 Richard Biener <rguenther@suse.de>
1742 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
1743 compute_all_dependences succeeds.
1744 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
1745 exceed --param loop-max-datarefs-for-datadeps.
1747 2018-06-29 Jakub Jelinek <jakub@redhat.com>
1749 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
1751 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
1754 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
1755 alternative 0 in preferred_for_speed attribute.
1757 2018-06-28 Paul Koning <ni1d@arrl.net>
1759 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
1760 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
1761 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
1762 * config/pdp11/pdp11.md: Correct "length" attribute calculation
1763 for shift insn patterns.
1765 2018-06-28 David Malcolm <dmalcolm@redhat.com>
1767 * cgraph.c (cgraph_node::get_body): Replace assignments to
1768 "dump_file" with calls to set_dump_file.
1769 * dumpfile.c (alt_dump_file): Make static, and group with...
1770 (alt_flags): ...this definition.
1771 (dumps_are_enabled): New variable.
1772 (refresh_dumps_are_enabled): New function.
1773 (set_dump_file): New function.
1774 (set_alt_dump_file): New function.
1775 (gcc::dump_manager::dump_start): Replace assignments to
1776 "dump_file" and "alt_dump_file" with calls to set_dump_file and
1778 (gcc::dump_manager::dump_finish): Likewise.
1779 * dumpfile.h (alt_dump_file): Delete decl.
1780 (dumps_are_enabled): New variable decl.
1781 (set_dump_file): New function decl.
1782 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
1784 * tree-nested.c (lower_nested_functions): Replace assignments to
1785 "dump_file" with calls to set_dump_file.
1787 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1789 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
1790 goto_locus of each outgoing edge of each basic block.
1792 2018-06-28 Richard Biener <rguenther@suse.de>
1794 * dwarf2out.c (decl_scope_table): Remove.
1795 (push_decl_scope): Likewise.
1796 (pop_decl_scope): Likewise.
1797 (gen_type_die_for_member): Do not call push/pop_decl_scope.
1798 (gen_struct_or_union_type_die): Likewise.
1799 (gen_tagged_type_die): Likewise.
1800 (dwarf2out_init): Do not initialize decl_scope_table.
1801 (dwarf2out_c_finalize): Do not free it.
1803 2018-06-28 Richard Biener <rguenther@suse.de>
1805 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
1806 deciding whether to not re-use a DIE.
1808 2018-06-28 Richard Biener <rguenther@suse.de>
1810 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
1811 DW_AT_abstract_origin attribute.
1813 2018-06-28 Martin Liska <mliska@suse.cz>
1815 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
1816 Use newly introduced constants.
1817 * tree-switch-conversion.h (struct jump_table_cluster):
1818 Define max_ratio_for_size and max_ratio_for_speed.
1820 2018-06-28 Martin Liska <mliska@suse.cz>
1822 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
1823 Add new checking assert to catch invalid state.
1824 (jump_table_cluster::can_be_handled): Handle single case
1826 (jump_table_cluster::is_beneficial): Bail out for such case.
1827 (bit_test_cluster::find_bit_tests):
1828 Add new checking assert to catch invalid state.
1829 (bit_test_cluster::can_be_handled): Handle single case
1831 (bit_test_cluster::is_beneficial): Bail out for such case.
1832 (switch_decision_tree::analyze_switch_statement):
1835 2018-06-28 Martin Liska <mliska@suse.cz>
1837 * common.opt: Introduce -completion option.
1838 * gcc.c (driver_handle_option): Handle it.
1839 (driver::main): Print completions if completion
1841 * opt-suggestions.c (option_proposer::get_completions):
1843 (option_proposer::suggest_completion): Likewise.
1844 (option_proposer::find_param_completions): Likewise.
1845 (verify_autocompletions): Likewise.
1846 (test_completion_valid_options): Likewise.
1847 (test_completion_valid_params): Likewise.
1848 (in_completion_p): Likewise.
1849 (empty_completion_p): Likewise.
1850 (test_completion_partial_match): Likewise.
1851 (test_completion_garbage): Likewise.
1852 (opt_proposer_c_tests): Likewise.
1853 * opt-suggestions.h: Declare new functions.
1854 * opts.c (common_handle_option): Handle OPT__completion_.
1855 * selftest-run-tests.c (selftest::run_tests): Add
1856 opt_proposer_c_tests.
1857 * selftest.c (assert_str_startswith): New.
1858 * selftest.h (assert_str_startswith): Likewise.
1859 (opt_proposer_c_tests): New.
1860 (ASSERT_STR_STARTSWITH): Likewise.
1862 2018-06-28 Martin Liska <mliska@suse.cz>
1864 * Makefile.in: Add opt-suggestions.o.
1865 * gcc-main.c: Include opt-suggestions.h.
1866 * gcc.c (driver::driver): Likewise.
1867 (driver::~driver): Remove m_option_suggestions.
1868 (driver::build_option_suggestions): Moved to option_proposer.
1869 (driver::suggest_option): Likewise.
1870 (driver::handle_unrecognized_options): Use option_proposer.
1871 * gcc.h (class driver): Add new memver m_option_proposer.
1872 * opt-suggestions.c: New file.
1873 * opt-suggestions.h: New file.
1875 2018-06-28 Martin Liska <mliska@suse.cz>
1877 * vec.h (class auto_string_vec): New (moved from auto_argvec).
1878 (auto_string_vec::~auto_string_vec): Likewise.
1880 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
1882 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
1883 prevent_decl_creation_for_types fields up and add reset_location field.
1884 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
1885 statement if id->reset_location is true.
1886 (copy_edges_for_bb): Do not set goto_locus on the new edges if
1887 id->reset_location is true.
1888 (copy_phis_for_bb): Force input_location on the arguments if
1889 id->reset_location is true.
1890 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
1891 is set on the function to be inlined.
1893 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
1895 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
1897 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
1899 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
1900 registers for Pmode.
1901 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
1902 hard registers for the clobbered pseudo.
1904 2018-06-27 Paul Koning <ni1d@arrl.net>
1906 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
1907 mutually exclusive options.
1908 * config/pdp11/constraints.md (h): New constraint.
1909 (O): Update definition to match shift code generation.
1910 (D): New constraint.
1911 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
1913 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
1915 (output_jump): Change arguments.
1916 (pdp11_fixed_cc_regs): New function.
1917 (pdp11_cc_mode): Ditto.
1918 (pdp11_expand_shift): Ditto.
1919 (pdp11_assemble_shift): Ditto.
1920 (pdp11_small_shift): Ditto.
1921 (pdp11_branch_cost): Remove.
1922 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
1924 (pdp11_register_move_cost): Update for CC registers.
1925 (pdp11_rtx_costs): Add case for LSHIFTRT.
1926 (pdp11_output_jump): Add CCNZ mode conditional branches.
1927 (notice_update_cc_on_set): Remove.
1928 (pdp11_cc_mode): New function.
1929 (simple_memory_operand): Correct pre/post decrement case.
1930 (no_side_effect_operand): New function.
1931 (pdp11_regno_reg_class): Add CC_REGS class.
1932 (pdp11_fixed_cc_regs): New function.
1933 (pdp11_small_shift): New function.
1934 (pdp11_expand_shift): New function to expand shift insns.
1935 (pdp11_assemble_shift): New function to output shifts.
1936 (pdp11_branch_cost): Remove.
1937 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
1938 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
1939 (WCHAR_TYPE): Ditto.
1940 (PTRDIFF_TYPE): Ditto.
1941 (ADJUST_INSN_LENGTH): New macro.
1942 (FIXED_REGISTERS): Add CC registers.
1943 (CALL_USED_REGISTERS): Ditto.
1945 (REG_CLASS_NAMES): Ditto.
1946 (REG_CLASS_CONTENTS): Ditto.
1947 (SELECT_CC_MODE): Use new function.
1948 (TARGET_FLAGS_REGNUM): New macro.
1949 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
1950 (cc0_reg_rtx): Remove.
1951 (CC_STATUS_MDEP): Remove.
1952 (CC_STATUS_MDEFP_INIT): Remove.
1953 (CC_IN_FPU): Remove.
1954 (NOTICE_UPDATE_CC): Remove.
1955 (REGISTER_NAMES): Add CC registers.
1956 (BRANCH_COST): Change to constant 1.
1957 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
1959 * config/pdp11/pdp11.opt (mbcopy): Remove.
1960 (mbcopy-builtin): Remove.
1961 (mbranch-cheap): Remove.
1962 (mbranch-expensive): Remove.
1963 * config/pdp11/predicates.md (expand_shift_operand): Update to
1964 match shift code generation.
1965 (ccnz_operator): New predicate.
1966 * doc/invoke.texi (PDP-11 Options): Remove deleted options
1967 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
1968 Remove non-existent option -mabshi, -mno-abshi. Document mutually
1970 * doc/md.texi (PDP-11): Document new D and h constraints. Update
1971 description of O constraint.
1973 2018-06-27 Jeff Law <law@redhat.com>
1974 Austin Law <austinklaw@gmail.com>
1976 * config/v850/v850.md (addsi3_set_flags): New pattern.
1977 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
1978 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
1979 (zero_extendhisi2_v850_set_flags): Likewise.
1980 (zero_extendqisi2_v850_set_flags): Likewise.
1981 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
1982 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
1983 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
1985 * config/v850/v850-protos.h (notice_update_cc): Remove.
1986 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
1987 (v850_print_operand): Handle 'D' and "d".
1988 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
1989 Add handling of arithmetic/logical operations compared against zero.
1990 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
1991 Do not look at v850_compare_op, instead get mode from last argument.
1992 (v850_gen_compare): Remove
1993 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
1994 after reload for prologue insns.
1995 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
1997 (construct_save_jarl): Likewise.
1998 (TARGET_FLAGS_REGNUM): Define.
1999 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
2000 (NOTICE_UPDATE_CC): Remove.
2001 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
2002 than cc0. Conditionalize on reload_completed.
2003 (cmpsi_insn, setfcc_insn): Likewise.
2004 (tst1 splitter): Turn into define_and_split which sets the flags
2006 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
2007 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
2008 (cstoresf4, cstoredf4): Clobber the flags.
2009 (cmpsi, cmpsf, cmpdf): Remove expanders.
2010 (setf_insn): Remove pattern.
2011 (addsi3): Turn into define_and_split which clobbers the flags after
2012 reload and a suitable pattern (addsi3_clobber_flags) for use after
2014 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
2015 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
2016 (ashrsi3, ashrsi3_v850e2): Likewise.
2017 (bins): Clobber the flags.
2018 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
2019 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
2020 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
2021 (call_value_internal_short, call_value_internal_long): Likewise.
2022 (callt_save_interrupt, callt_return_interrupt): Likewise.
2023 (save_interrupt, return_interrupt): Likewise.
2024 (callt_save_all_interrupt, save_all_interrupt): Likewise.
2025 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
2026 (restore_all_interrupt, _restore_all_interrupt): Likewise.
2027 (All FP comparisons): Only allow after reload has completed.
2029 (divh, divhu): Tweak output template.
2030 (branch_z_normal, branch_z_invert): Remove
2031 (branch_nz_normal, branch_nz_invert): Likewise.
2032 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
2034 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
2035 * config/v850/v850.c (notice_update_cc): Remove.
2036 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
2037 (CC_NO_CARRY): Likewise.
2038 (NOTICE_UPDATE_CC): Define to nothing.
2039 * config/v850/v850.md: Remove block comment on cc0 handling
2040 Remove "cc" attribute from all patterns. Remove cc_status handling
2041 from all patterns. Minor formatting fixes.
2043 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2045 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
2046 (cortex-a76.cortex-a55): Likewise.
2047 * config/aarch64/aarch64-tune.md: Regenerate.
2048 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
2049 cortex-a76.cortex-a55.
2051 2018-06-27 Jeff Law <law@redhat.com>
2053 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
2054 (MULTILIB_DIRNAMES): Similarly.
2056 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
2058 * gimple.h (gimple_return_retbnd): Delete.
2059 (gimple_return_set_retbnd): Likewise.
2060 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
2061 gimple_return_set_retbnd.
2062 * gimple-pretty-print.c (dump_gimple_return): Remove call to
2063 gimple_return_retbnd and adjust.
2064 * tree-inline.h (struct copy_body_data): Remove retbnd field.
2065 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
2066 Explicitly return NULL in a couple more cases. Move assertion
2067 on debug statements and remove unreachable code.
2068 (reset_debug_binding): Do not test id->retbnd.
2069 (expand_call_inline): Do not set it.
2071 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
2073 * configure.ac: Add --disable-gcov option.
2074 * configure: Regenerate.
2075 * Makefile.in: Honour @enable_gcov@.
2076 * doc/install.texi: Document --disable-gcov.
2078 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2080 * config/arm/arm-cpus.in (cortex-a76): New entry.
2081 (cortex-a76.cortex-a55): Likewise.
2082 * config/arm/arm-tables.opt: Regenerate.
2083 * config/arm/arm-tune.md: Likewise.
2084 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
2085 * doc/invoke.texi (ARM Options): Document cortex-a76 and
2086 cortex-a76.cortex-a55.
2088 2018-06-27 Tamar Christina <tamar.christina@arm.com>
2091 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
2093 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
2095 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
2097 (EPILOGUE_USES): Likewise.
2099 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
2101 * tree-inline.c (remap_location): New function extracted from...
2102 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
2103 (copy_phis_for_bb): ...here. Call remap_location.
2104 (copy_cfg_body): Adjust call to copy_edges_for_bb.
2106 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
2108 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
2109 unaligned vsx for 16B memset.
2111 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
2114 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
2115 ieee128_float_type_node to long_double_type_node unless
2116 TARGET_LONG_DOUBLE_128 is set.
2118 2018-06-26 David Malcolm <dmalcolm@redhat.com>
2120 * cfgloop.c (get_loop_location): Convert return type from
2121 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
2122 by implicit construction from rtx_insn *, and using
2123 dump_user_location_t::from_function_decl for the fallback case.
2124 * cfgloop.h (get_loop_location): Convert return type from
2125 location_t to dump_user_location_t.
2126 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
2127 dump_printf_loc to pass in a dump_location_t rather than a
2128 location_t, via the gimple stmt.
2129 * coverage.c (get_coverage_counts): Update calls to
2130 dump_printf_loc to pass in dump_location_t rather than a
2132 * doc/optinfo.texi (Dump types): Convert example of
2133 dump_printf_loc from taking "locus" to taking "insn". Update
2134 description of the "_loc" calls to cover dump_location_t.
2135 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
2137 (dump_user_location_t::dump_user_location_t): New constructors,
2138 from gimple *stmt and rtx_insn *.
2139 (dump_user_location_t::from_function_decl): New function.
2140 (dump_loc): Make static.
2141 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
2142 const dump_location_t &.
2143 (dump_generic_expr_loc): Delete.
2144 (dump_printf_loc): Convert param "loc" from location_t to
2145 const dump_location_t &.
2146 (selftest::test_impl_location): New function.
2147 (selftest::dumpfile_c_tests): New function.
2148 * dumpfile.h: Include "profile-count.h".
2149 (class dump_user_location_t): New class.
2150 (struct dump_impl_location_t): New struct.
2151 (class dump_location_t): New class.
2152 (dump_printf_loc): Convert 2nd param from source_location to
2153 const dump_location_t &.
2154 (dump_generic_expr_loc): Delete.
2155 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
2156 const dump_location_t &.
2157 * gimple-fold.c (fold_gimple_assign): Update call to
2158 dump_printf_loc to pass in a dump_location_t rather than a
2159 location_t, via the gimple stmt.
2160 (gimple_fold_call): Likewise.
2161 * gimple-loop-interchange.cc
2162 (loop_cand::analyze_iloop_reduction_var): Update for change to
2163 check_reduction_path.
2164 (tree_loop_interchange::interchange): Update for change to
2166 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
2167 change in return-type of find_loop_location.
2168 (graphite_regenerate_ast_isl): Likewise.
2169 * graphite-optimize-isl.c (optimize_isl): Likewise.
2170 * graphite.c (graphite_transform_loops): Likewise.
2171 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
2172 pass in a dump_location_t rather than a location_t, via the
2174 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
2175 * ipa.c (walk_polymorphic_call_targets): Likewise.
2176 * loop-unroll.c (report_unroll): Convert "locus" param from
2177 location_t to dump_location_t.
2178 (decide_unrolling): Update for change to get_loop_location's
2180 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
2181 location_t to dump_user_location_t.
2182 (grid_find_single_omp_among_assignments_1): Updates calls to
2183 dump_printf_loc to pass in a dump_location_t rather than a
2184 location_t, via the gimple stmt.
2185 (grid_parallel_clauses_gridifiable): Convert "tloc" from
2186 location_t to dump_location_t. Updates calls to dump_printf_loc
2187 to pass in a dump_location_t rather than a location_t, via the
2189 (grid_inner_loop_gridifiable_p): Likewise.
2190 (grid_dist_follows_simple_pattern): Likewise.
2191 (grid_gfor_follows_tiling_pattern): Likewise.
2192 (grid_target_follows_gridifiable_pattern): Likewise.
2193 (grid_attempt_target_gridification): Convert initialization
2194 of local "grid" from memset to zero-initialization; FIXME: does
2195 this require C++11? Update call to dump_printf_loc to pass in a
2196 optinfo_location rather than a location_t, via the gimple stmt.
2197 * profile.c (read_profile_edge_counts): Updates call to
2198 dump_printf_loc to pass in a dump_location_t rather than a
2200 (compute_branch_probabilities): Likewise.
2201 * selftest-run-tests.c (selftest::run_tests): Call
2203 * selftest.h (dumpfile_c_tests): New decl.
2204 * tree-loop-distribution.c (pass_loop_distribution::execute):
2205 Update for change in return type of find_loop_location.
2206 * tree-parloops.c (parallelize_loops): Likewise.
2207 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
2208 "locus" from location_t to dump_user_location_t.
2209 (canonicalize_loop_induction_variables): Likewise.
2210 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
2211 for change in return type of find_loop_location.
2212 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
2213 to dump_printf_loc to pass in a dump_location_t rather than a
2214 location_t, via the stmt.
2215 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
2217 * tree-vect-loop-manip.c (find_loop_location): Convert return
2218 type from source_location to dump_user_location_t.
2219 (vect_do_peeling): Update for above change.
2220 (vect_loop_versioning): Update for change in type of
2222 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
2223 from location_t to dump_user_location_t.
2224 (vect_estimate_min_profitable_iters): Update for change in type
2226 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
2227 location_t to dump_location_t.
2228 (vect_slp_bb): Update for change in type of vect_location.
2229 * tree-vectorizer.c (vect_location): Convert from source_location
2230 to dump_user_location_t.
2231 (try_vectorize_loop_1): Update for change in vect_location's type.
2232 (vectorize_loops): Likewise.
2233 (increase_alignment): Likewise.
2234 * tree-vectorizer.h (vect_location): Convert from source_location
2235 to dump_user_location_t.
2236 (find_loop_location): Convert return type from source_location to
2237 dump_user_location_t.
2238 (check_reduction_path): Convert 1st param from location_t to
2239 dump_user_location_t.
2240 * value-prof.c (check_counter): Update call to dump_printf_loc to
2241 pass in a dump_user_location_t rather than a location_t; update
2242 call to error_at for change in type of "locus".
2243 (check_ic_target): Update call to dump_printf_loc to
2244 pass in a dump_user_location_t rather than a location_t, via the
2247 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
2249 * config/s390/s390.h (enum processor_flags): Do not use
2250 default tune parameter when -march was specified.
2252 2018-06-26 Jakub Jelinek <jakub@redhat.com>
2255 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
2256 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
2259 2018-06-26 Richard Biener <rguenther@suse.de>
2261 PR tree-optimization/86287
2263 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
2264 (vect_analyze_loop): Initialize n_stmts.
2266 2018-06-26 Richard Biener <rguenther@suse.de>
2269 * fold-const.c (fold_convertible_p): Pointer extension
2272 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
2275 * dwarf2out.c (loc_list_has_views): Adjust comments.
2276 (dw_loc_list): Split single cross-partition range with
2279 2018-06-25 Jeff Law <law@redhat.com>
2281 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
2282 on -mbig-switch by default.
2284 * config/v850/predicates.md (const_float_1_operand): Fix match_code
2286 (const_float_0_operand): Remove unused predicate.
2287 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
2288 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
2289 (recipsf2): New expander. Original pattern now called
2291 (recipdf2, recipdf2_insn): Similarly.
2292 (rsqrtsf2, rsqrtsf2_insn): Similarly
2293 (rsqrtdf2, rsqrtdf2_insn): Similarly
2295 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
2297 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
2298 Simplify logic for FreeBSD (twice).
2300 2018-06-25 Martin Sebor <msebor@redhat.com>
2302 PR tree-optimization/86204
2303 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
2304 a strnlen result if it's less than the length of the string.
2306 2018-06-25 Martin Sebor <msebor@redhat.com>
2308 PR tree-optimization/85700
2309 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
2310 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
2311 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
2313 2018-06-25 Martin Sebor <msebor@redhat.com>
2315 * doc/extend.texi (Zero-length arrays): Update and clarify.
2317 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
2319 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
2320 added IEEE/IBM long double multilib support on PowerPC little
2321 endian Linux systems.
2322 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
2323 (MULTILIB_DEFAULTS): Likewise.
2324 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2326 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
2327 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
2328 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
2330 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
2333 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
2334 (REORDER_45): Likewise.
2336 2018-06-25 Jeff Law <law@redhat.com>
2338 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
2339 dividend to 32 bits. Adjust length.
2340 (udivmodhi4): Cleanup output template. Fix length.
2342 2018-06-25 Carl Love <cel@us.ibm.com>
2344 * config/rs6000/vsx.md: Change word selector to prefered location.
2346 2018-06-25 Richard Biener <rguenther@suse.de>
2348 PR tree-optimization/86304
2349 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
2350 epilogue-if-converted loops as well.
2352 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
2354 * lto-section-out.c (lto_begin_section): Do not print section
2355 name for noaddr and unnumbered dumps.
2357 2018-06-25 Richard Biener <rguenther@suse.de>
2359 * tree-vectorizer.h (struct vec_info_shared): New structure
2360 with parts split out from struct vec_info and loop_nest from
2361 struct _loop_vec_info.
2362 (struct vec_info): Adjust accordingly.
2363 (struct _loop_vec_info): Likewise.
2364 (LOOP_VINFO_LOOP_NEST): Adjust.
2365 (LOOP_VINFO_DATAREFS): Likewise.
2366 (LOOP_VINFO_DDRS): Likewise.
2367 (struct _bb_vec_info): Likewise.
2368 (BB_VINFO_DATAREFS): Likewise.
2369 (BB_VINFO_DDRS): Likewise.
2370 (struct _stmt_vec_info): Add dr_aux member.
2371 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
2372 (DR_MISALIGNMENT_UNINITIALIZED): New.
2373 (set_dr_misalignment): Adjust.
2374 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
2375 (vect_analyze_loop): Adjust prototype.
2376 (vect_analyze_loop_form): Likewise.
2377 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
2378 Compute dependences lazily.
2379 (vect_record_base_alignments): Use shared datarefs/ddrs.
2380 (vect_verify_datarefs_alignment): Likewise.
2381 (vect_analyze_data_refs_alignment): Likewise.
2382 (vect_analyze_data_ref_accesses): Likewise.
2383 (vect_analyze_data_refs): Likewise.
2384 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
2385 constructor parameter for shared part.
2386 (vect_analyze_loop_form): Pass in shared part and adjust.
2387 (vect_analyze_loop_2): Pass in storage for the number of
2388 stmts. Move loop nest finding to the caller. Compute
2390 (vect_analyze_loop): Pass in shared part.
2391 (vect_transform_loop): Verify shared datarefs are unchanged.
2392 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
2393 constructor parameter for shared part.
2394 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
2395 (vect_slp_bb): Verify shared datarefs are unchanged before
2397 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
2399 (new_stmt_vec_info): Initialize DR_AUX misalignment to
2400 DR_MISALIGNMENT_UNINITIALIZED.
2401 * tree-vectorizer.c (vec_info::vec_info): Add constructor
2402 parameter for shared part.
2403 (vec_info::~vec_info): Adjust.
2404 (vec_info_shared::vec_info_shared): New.
2405 (vec_info_shared::~vec_info_shared): Likewise.
2406 (vec_info_shared::save_datarefs): Likewise.
2407 (vec_info_shared::check_datarefs): Likewise.
2408 (try_vectorize_loop_1): Construct shared part live for analyses
2409 of a single loop for multiple vector sizes.
2410 * tree-parloops.c (gather_scalar_reductions): Adjust.
2412 2018-06-25 Richard Biener <rguenther@suse.de>
2414 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
2415 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
2416 (vect_analyze_data_refs): Remove similar code from here and
2417 simplify accordingly.
2419 2018-06-25 Richard Biener <rguenther@suse.de>
2421 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
2422 for reverse storage order accesses rather than asserting
2423 they cannot happen here.
2425 2018-06-25 Tom de Vries <tdevries@suse.de>
2428 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
2429 Use data16 instead of .byte for insn prefix.
2431 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
2434 * parser.c (make_char_string_pack): Pass this literal chars
2435 through cpp_interpret_string.
2436 (cp_parser_userdef_numeric_literal): Check the result of
2437 make_char_string_pack.
2439 2018-06-24 Maya Rashish <coypu@sdf.org>
2441 * ginclude/stddef.h: Simplify conditions around avoiding
2442 re-definition of __size_t.
2444 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2446 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
2447 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
2449 2018-06-22 Maya Rashish <coypu@sdf.org>
2451 * doc/invoke.texi (mno-fancy-math-387): Update for changes
2452 made to OpenBSD and NetBSD through the years.
2454 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
2456 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2457 behavior of vec_pack (vector double, vector double) to match
2458 behavior of vec_float2 (vector double, vector double).
2460 2018-06-22 Olivier Hainque <hainque@adacore.com>
2462 * gimplify.c (gimplify_function_tree): Prevent creation
2463 of a trampoline for the address of the current function
2464 passed to entry/exit instrumentation hooks.
2466 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
2469 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
2472 2018-06-22 Martin Liska <mliska@suse.cz>
2474 PR tree-optimization/86263
2475 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2476 Bail out if is_enabled is false.
2477 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
2479 (jump_table_cluster::is_enabled): New function.
2481 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2483 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
2484 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
2485 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
2486 (lto_input_ts_binfo_tree_pointers): Likewise.
2487 * tree-streamer-out.c (streamer_write_tree_bitfields,
2488 write_ts_binfo_tree_pointers): Likewise.
2489 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
2491 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
2493 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
2495 2018-06-22 Martin Liska <mliska@suse.cz>
2497 * symbol-summary.h (get): Make it pure and inline move
2498 functionality from ::get function.
2499 (get): Remove and inline into ::get and ::get_create.
2500 (get_create): Move code from ::get function.
2502 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2505 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
2506 -x assembler-with-cpp.
2508 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2510 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
2511 _FILE_OFFSET_BITS=64 for C++.
2513 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
2515 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
2516 conversion insn that shows up when pr85657-3.c is compiled using
2517 IEEE 128-bit long double.
2518 (neg<mode>2_internal): Use the correct mode to check whether the
2519 mode is IBM extended.
2520 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
2521 multiply and divide external functions from being created more
2524 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2526 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
2528 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
2529 the edge can be forwarded.
2530 (cfg_layout_merge_blocks): Likewise.
2532 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2534 * except.c (finish_eh_generation): Commit edge insertions only after
2535 the EH edges have been redirected from post-landing to landing pads.
2537 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2539 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
2540 create_tmp_var_for to create the FRAME decl.
2541 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
2543 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
2545 * tree-inline.c (copy_edges_for_bb): Minor tweak.
2546 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
2547 debug statement when resetting its value.
2548 (expand_call_inline): Copy the locus of the call onto the assignment
2549 of the return value, if any. Use local variable in more cases.
2551 2018-06-21 Martin Liska <mliska@suse.cz>
2553 * ipa-pure-const.c (propagate_nothrow): Use
2554 funct_state_summaries->get.
2555 (dump_malloc_lattice): Likewise.
2556 (propagate_malloc): Likewise.
2558 2018-06-21 Richard Biener <rguenther@suse.de>
2560 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
2561 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
2562 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
2564 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
2565 BLOCK_ABSTRACT_ORIGIN unconditionally.
2567 2018-06-21 David Malcolm <dmalcolm@redhat.com>
2569 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
2571 * ipa-reference.c (ipa_reference_c_finalize): Delete
2572 ipa_ref_opt_sum_summaries and set it to NULL.
2574 2018-06-21 Tom de Vries <tdevries@suse.de>
2576 PR tree-optimization/85859
2577 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
2578 test with comment from bb_no_side_effects_p.
2580 2018-06-21 Richard Biener <rguenther@suse.de>
2582 PR tree-optimization/86232
2583 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
2584 max for constant niter.
2586 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2588 * config/aarch64/aarch64-simd.md
2589 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
2591 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
2593 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
2594 Make opernads of the unspec commutative.
2596 2018-06-21 Richard Biener <rguenther@suse.de>
2598 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
2599 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
2600 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
2601 (vect_analyze_data_ref_dependence): Re-order checks to deal with
2603 (vect_record_base_alignments): Do not record base alignment
2604 for gathers or scatters.
2605 (vect_compute_data_ref_alignment): Drop return value that is always
2606 true. Bail out early for gathers or scatters.
2607 (vect_enhance_data_refs_alignment): Bail out early for gathers
2609 (vect_find_same_alignment_drs): Likewise.
2610 (vect_analyze_data_refs_alignment): Remove dead code.
2611 (vect_slp_analyze_and_verify_node_alignment): Likewise.
2612 (vect_analyze_data_refs): For possible gathers or scatters do
2613 not create an alternate DR, just check their possible validity
2614 and mark them. Adjust DECL_NONALIASED handling to not rely
2616 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
2617 update inits of gathers or scatters.
2618 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
2619 Also copy gather/scatter flag to pattern vinfo.
2621 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
2623 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
2624 behavior of vec_packsu (vector unsigned long long, vector unsigned
2625 long long) to match behavior of vec_packs with same signature.
2627 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
2628 Thomas Schwinge <thomas@codesourcery.com>
2629 Cesar Philippidis <cesar@codesourcery.com>
2631 * gimplify.c (gimplify_scan_omp_clauses): Add support for
2632 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2633 (gimplify_adjust_omp_clauses): Likewise.
2634 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
2635 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
2636 (gimplify_omp_target_update): Update handling of acc update and
2638 * omp-low.c (install_var_field): Remove unused parameter
2639 base_pointers_restrict.
2640 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
2641 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
2643 (omp_target_base_pointers_restrict_p): Delete.
2644 (scan_omp_target): Update call to scan_sharing_clauses.
2645 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
2647 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
2648 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
2649 (convert_local_omp_clauses): Likewise.
2650 * tree-pretty-print.c (dump_omp_clause): Likewise.
2651 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
2653 (omp_clause_code_name): Likewise.
2655 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2658 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
2661 PR tree-optimization/86231
2662 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
2663 anti-range don't overwrite *vr0min before using it to compute *vr0max.
2665 2018-06-20 Tom de Vries <tdevries@suse.de>
2667 PR tree-optimization/86097
2668 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
2669 iv type if signedness of iv type is not the same as that of *nit.
2671 2018-06-20 Jakub Jelinek <jakub@redhat.com>
2673 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
2674 EDGE_EH edges, verify they are all EDGE_EH.
2676 2018-06-20 Maya Rashish <coypu@sdf.org>
2678 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
2680 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2682 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
2683 * config/aarch64/aarch64.c (xgene1_tunings): Add
2684 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
2685 (aarch64_mode_valid_for_sched_fusion_p):
2686 Allow 16-byte modes.
2687 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
2688 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
2690 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
2692 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
2693 * config/aarch64/iterators.md (VQ2): New mode iterator.
2695 2018-06-20 Martin Liska <mliska@suse.cz>
2697 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
2698 Change default ratio from 10 to 8.
2700 2018-06-20 Martin Liska <mliska@suse.cz>
2702 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
2704 (bit_test_cluster::find_bit_tests): Likewise.
2705 (switch_decision_tree::analyze_switch_statement): Find clusters.
2706 * tree-switch-conversion.h (struct jump_table_cluster): Document
2709 2018-06-20 Martin Liska <mliska@suse.cz>
2711 * tree-switch-conversion.c (switch_conversion::collect):
2712 Record m_uniq property.
2713 (switch_conversion::expand): Bail out for special conditions.
2714 (group_cluster::~group_cluster): New.
2715 (group_cluster::group_cluster): Likewise.
2716 (group_cluster::dump): Likewise.
2717 (jump_table_cluster::emit): New.
2718 (switch_decision_tree::fix_phi_operands_for_edges): New.
2719 (struct case_node): Remove struct.
2720 (jump_table_cluster::can_be_handled): New.
2721 (case_values_threshold): Moved to header.
2722 (reset_out_edges_aux): Likewise.
2723 (jump_table_cluster::is_beneficial): New.
2724 (bit_test_cluster::can_be_handled): Likewise.
2725 (add_case_node): Remove.
2726 (bit_test_cluster::is_beneficial): New.
2727 (case_bit_test::cmp): New.
2728 (bit_test_cluster::emit): New.
2729 (expand_switch_as_decision_tree_p): Remove.
2730 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
2731 (fix_phi_operands_for_edge): Likewise.
2732 (switch_decision_tree::analyze_switch_statement): New.
2733 (compute_cases_per_edge): Move ...
2734 (switch_decision_tree::compute_cases_per_edge): ... here.
2735 (try_switch_expansion): Likewise.
2736 (switch_decision_tree::try_switch_expansion): Likewise.
2737 (record_phi_operand_mapping): Likewise.
2738 (switch_decision_tree::record_phi_operand_mapping): Likewise.
2739 (emit_case_decision_tree): Likewise.
2740 (switch_decision_tree::emit): Likewise.
2741 (balance_case_nodes): Likewise.
2742 (switch_decision_tree::balance_case_nodes): Likewise.
2743 (dump_case_nodes): Likewise.
2744 (switch_decision_tree::dump_case_nodes): Likewise.
2745 (emit_jump): Likewise.
2746 (switch_decision_tree::emit_jump): Likewise.
2747 (emit_cmp_and_jump_insns): Likewise.
2748 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
2749 (emit_case_nodes): Likewise.
2750 (switch_decision_tree::emit_case_nodes): Likewise.
2751 (conditional_probability): Remove.
2752 * tree-switch-conversion.h (enum cluster_type): New.
2754 (struct cluster): Likewise.
2755 (cluster::cluster): Likewise.
2756 (struct simple_cluster): Likewise.
2757 (simple_cluster::simple_cluster): Likewise.
2758 (struct group_cluster): Likewise.
2759 (struct jump_table_cluster): Likewise.
2760 (struct bit_test_cluster): Likewise.
2761 (struct min_cluster_item): Likewise.
2762 (struct case_tree_node): Likewise.
2763 (case_tree_node::case_tree_node): Likewise.
2764 (jump_table_cluster::case_values_threshold): Likewise.
2765 (struct case_bit_test): Likewise.
2766 (struct switch_decision_tree): Likewise.
2767 (struct switch_conversion): Likewise.
2768 (switch_decision_tree::reset_out_edges_aux): Likewise.
2770 2018-06-20 Martin Liska <mliska@suse.cz>
2772 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
2773 (hoist_edge_and_branch_if_true): Likewise.
2774 (expand_switch_using_bit_tests_p): Likewise.
2775 (struct case_bit_test): Likewise.
2776 (case_bit_test_cmp): Likewise.
2777 (emit_case_bit_tests): Likewise.
2778 (switch_conversion::switch_conversion): New class.
2779 (struct switch_conv_info): Remove old struct.
2780 (collect_switch_conv_info): More to ...
2781 (switch_conversion::collect): ... this.
2782 (check_range): Likewise.
2783 (switch_conversion::check_range): Likewise.
2784 (check_all_empty_except_final): Likewise.
2785 (switch_conversion::check_all_empty_except_final): Likewise.
2786 (check_final_bb): Likewise.
2787 (switch_conversion::check_final_bb): Likewise.
2788 (create_temp_arrays): Likewise.
2789 (switch_conversion::create_temp_arrays): Likewise.
2790 (free_temp_arrays): Likewise.
2791 (gather_default_values): Likewise.
2792 (switch_conversion::gather_default_values): Likewise.
2793 (build_constructors): Likewise.
2794 (switch_conversion::build_constructors): Likewise.
2795 (constructor_contains_same_values_p): Likewise.
2796 (switch_conversion::contains_same_values_p): Likewise.
2797 (array_value_type): Likewise.
2798 (switch_conversion::array_value_type): Likewise.
2799 (build_one_array): Likewise.
2800 (switch_conversion::build_one_array): Likewise.
2801 (build_arrays): Likewise.
2802 (switch_conversion::build_arrays): Likewise.
2803 (gen_def_assigns): Likewise.
2804 (switch_conversion::gen_def_assigns): Likewise.
2805 (prune_bbs): Likewise.
2806 (switch_conversion::prune_bbs): Likewise.
2807 (fix_phi_nodes): Likewise.
2808 (switch_conversion::fix_phi_nodes): Likewise.
2809 (gen_inbound_check): Likewise.
2810 (switch_conversion::gen_inbound_check): Likewise.
2811 (process_switch): Use the newly created class.
2812 (switch_conversion::expand): New.
2813 (switch_conversion::~switch_conversion): New.
2814 * tree-switch-conversion.h: New file.
2816 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2818 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
2819 tree-vect-patterns.c.
2820 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
2821 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
2822 (vect_recog_sad_pattern): Likewise.
2823 (vect_recog_widen_sum_pattern): Likewise.
2824 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
2825 (vect_recog_widen_shift_pattern): Remove the type_in argument.
2826 (vect_recog_rotate_pattern): Likewise.
2827 (vect_recog_mult_pattern): Likewise.
2828 (vect_recog_vector_vector_shift_pattern): Likewise.
2829 (vect_recog_divmod_pattern): Likewise.
2830 (vect_recog_mixed_size_cond_pattern): Likewise.
2831 (vect_recog_bool_pattern): Likewise.
2832 (vect_recog_mask_conversion_pattern): Likewise.
2833 (vect_try_gather_scatter_pattern): Likewise.
2834 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
2835 (vect_recog_over_widening_pattern): Likewise.
2836 (vect_recog_gather_scatter_pattern): Likewise.
2837 (vect_recog_func_ptr): Move from tree-vectorizer.h
2838 (vect_vect_recog_func_ptrs): Move further down the file.
2839 (vect_recog_func): Likewise. Remove the third argument.
2840 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
2841 (vect_pattern_recog_1): Expect the pattern function to do any
2842 necessary target tests. Also expect it to provide a vector type.
2843 Remove the type_in handling.
2845 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2847 * tree-vect-patterns.c (vect_pattern_detected): New function.
2848 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
2849 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
2850 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
2851 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
2852 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
2853 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
2854 (vect_recog_mask_conversion_pattern)
2855 (vect_try_gather_scatter_pattern): Likewise.
2857 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2859 * tree-vect-patterns.c (vect_get_internal_def): New function.
2860 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
2861 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
2862 (search_type_for_mask_1): Use it.
2864 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2866 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2867 redundant WIDEN_SUM_EXPR handling.
2868 (vect_recog_sad_pattern): Likewise.
2870 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2872 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
2873 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
2874 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
2875 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
2876 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
2878 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2880 * tree-vect-stmts.c (vectorizable_call): Make sure that we
2881 use the stmt_vec_info of the original bb statement for the
2882 new zero assignment, even if the call is part of a pattern.
2884 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2886 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
2887 that the sequence is attached to the original statement rather
2888 than the pattern statement.
2889 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
2890 PATTERN_DEF_SEQ from the original statement rather than
2891 the main pattern statement.
2892 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
2893 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
2894 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
2896 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
2898 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
2899 definition statements before the early exit for statements that aren't
2901 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
2903 (vect_transform_loop): ...here. Process pattern definition
2904 statements without first checking whether the main pattern
2905 statement is live or relevant.
2907 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
2909 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
2910 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
2912 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
2914 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
2915 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
2916 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
2917 (expand_block_compare): Change select_block_compare_mode call.
2918 (expand_strncmp_align_check): Use new functions, fix comment.
2919 (emit_final_str_compare_gpr): New function.
2920 (expand_strn_compare): Refactor and clean up code.
2921 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
2923 2018-06-19 Tony Reix <tony.reix@atos.com>
2924 Damien Bergamini <damien.bergamini@atos.com>
2925 David Edelsohn <dje.gcc@gmail.com>
2927 * collect2.c (static_obj): New variable.
2928 (static_libs): New variable.
2929 (is_in_list): Uncomment declaration.
2930 (main): Track AIX libraries linked statically.
2931 (is_in_list): Uncomment definition.
2932 (scan_prog_file): Don't add AIX shared libraries initializer
2933 to constructor list if linking statically.
2935 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
2937 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
2939 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
2941 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
2943 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
2946 2018-06-19 Martin Liska <mliska@suse.cz>
2948 * config/i386/i386.c (ix86_can_inline_p): Do not use
2949 ipa_fn_summaries::get_create.
2950 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
2952 (devirtualization_time_bonus): Likewise.
2953 (ipcp_propagate_stage): Likewise.
2954 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
2955 (edge_set_predicate): Likewise.
2956 (evaluate_conditions_for_known_args): Likewise.
2957 (evaluate_properties_for_edge): Likewise.
2958 (ipa_call_summary::reset): Tranform to ...
2959 (ipa_call_summary::~ipa_call_summary): ... this.
2960 (ipa_fn_summary::reset): Transform to ...
2961 (ipa_fn_summary::~ipa_fn_summary): ... this.
2962 (ipa_fn_summary_t::remove): Rename to ...
2963 (ipa_fn_summary_t::remove_callees): ... this.
2964 (ipa_fn_summary_t::duplicate): Use placement new
2965 instead of memory copy.
2966 (ipa_call_summary_t::duplicate): Likewise.
2967 (ipa_call_summary_t::remove): Remove.
2968 (dump_ipa_call_summary): Change get_create to get.
2969 (ipa_dump_fn_summary): Dump only when summary exists.
2970 (analyze_function_body): Use symbol_summary::get instead
2972 (compute_fn_summary): Likewise.
2973 (estimate_edge_devirt_benefit): Likewise.
2974 (estimate_edge_size_and_time): Likewise.
2975 (inline_update_callee_summaries): Likewise.
2976 (remap_edge_change_prob): Likewise.
2977 (remap_edge_summaries): Likewise.
2978 (ipa_merge_fn_summary_after_inlining): Likewise.
2979 (write_ipa_call_summary): Likewise.
2980 (ipa_fn_summary_write): Likewise.
2981 (ipa_free_fn_summary): Likewise.
2982 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
2983 (struct ipa_call_summary): Likewise.
2984 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
2986 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
2987 (estimate_size_after_inlining): Likewise.
2988 (estimate_growth): Likewise.
2989 (growth_likely_positive): Likewise.
2990 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
2991 (inline_call): Likewise.
2992 * ipa-inline.c (caller_growth_limits): Likewise.
2993 (can_inline_edge_p): Likewise.
2994 (can_inline_edge_by_limits_p): Likewise.
2995 (compute_uninlined_call_time): Likewise.
2996 (compute_inlined_call_time): Likewise.
2997 (want_inline_small_function_p): Likewise.
2998 (edge_badness): Likewise.
2999 (update_caller_keys): Likewise.
3000 (update_callee_keys): Likewise.
3001 (inline_small_functions): Likewise.
3002 (inline_to_all_callers_1): Likewise.
3003 (dump_overall_stats): Likewise.
3004 (early_inline_small_functions): Likewise.
3005 (early_inliner): Likewise.
3006 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3007 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3008 * ipa-pure-const.c (malloc_candidate_p): Likewise.
3009 * ipa-split.c (execute_split_functions): Likewise.
3010 * symbol-summary.h: Likewise.
3011 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
3013 2018-06-19 Richard Biener <rguenther@suse.de>
3015 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
3016 (vectorize_loops): ... here. Fix dbgcnt handling.
3017 (try_vectorize_loop): Wrap try_vectorize_loop_1.
3019 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
3022 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
3023 ieee128 argument takes up only one (vector) register, not two (floating
3026 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
3028 * gimplify.c (gimplify_init_constructor): Really never clear for an
3029 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
3031 2018-06-19 Richard Biener <rguenther@suse.de>
3033 PR tree-optimization/86179
3034 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
3035 after failed recognition.
3037 2018-06-18 Martin Sebor <msebor@redhat.com>
3040 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
3041 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
3042 Handle integer subtraction.
3043 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
3044 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
3046 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3048 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
3049 param from rtx to rtx_insn *.
3050 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
3052 (frv_ifcvt_modify_insn): Likwise.
3053 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
3054 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
3055 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
3056 as_a <rtx_insn *> cast to local "unprotected_region" once
3057 it's been established that it's not NULL or pc_rtx.
3058 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
3059 param "sethi" from rtx to rtx_insn *.
3060 (nds32_group_float_insns): Likewise for param "insn".
3061 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
3063 (vax_output_int_subtract): Likewise.
3064 * config/vax/vax.c (vax_output_int_add): Likewise for param
3066 (vax_output_int_subtract): Likewise.
3067 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
3068 (emit_pattern_after): Likewise for param "after".
3069 (emit_insn_after): Likewise.
3070 (emit_jump_insn_after): Likewise.
3071 (emit_call_insn_after): Likewise.
3072 (emit_debug_insn_after): Likewise.
3073 (emit_pattern_before): Likewise for param "before".
3074 (emit_insn_before): Likewise.
3075 (emit_jump_insn_before): Likewise.
3076 * final.c (get_insn_template): Likewise for param "insn", removing
3078 * output.h (get_insn_template): Likewise for 2nd param.
3079 * rtl.h (emit_insn_before): Likewise.
3080 (emit_jump_insn_before): Likewise.
3081 (emit_debug_insn_before_noloc): Likewise.
3082 (emit_insn_after): Likewise.
3083 (emit_jump_insn_after): Likewise.
3084 (emit_call_insn_after): Likewise.
3085 (emit_debug_insn_after): Likewise.
3086 (set_insn_deleted): Likewise for param.
3088 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
3091 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
3092 floating point modes, so that IFmode is numerically greater than
3093 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
3094 to declare the ordering. This prevents IFmode from being
3095 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
3096 machine. Include rs6000-modes.h to share the fractional values
3097 between genmodes* and the rest of the compiler.
3101 * config/rs6000/rs6000-modes.h: New file.
3102 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
3103 meaning of rs6000_long_double_size so that 126..128 selects an
3104 appropriate 128-bit floating point type.
3105 (rs6000_option_override_internal): Likewise.
3106 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
3107 (TARGET_LONG_DOUBLE_128): Change the meaning of
3108 rs6000_long_double_size so that 126..128 selects an appropriate
3109 128-bit floating point type.
3110 (LONG_DOUBLE_TYPE_SIZE): Update comment.
3111 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
3112 source and destination to match the standard usage.
3113 (truncifkf2): Likewise.
3114 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
3115 ISA 2.07 to use an explicit clobber, instead of passing in a
3117 (copysign<mode>3_soft): Likewise.
3119 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3121 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
3122 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
3123 (vect_slp_analyze_instance_dependence): Likewise.
3124 (vect_enhance_data_refs_alignment): Likewise.
3125 (vect_analyze_data_refs_alignment): Likewise.
3126 (vect_slp_analyze_and_verify_instance_alignment
3127 (vect_analyze_data_ref_accesses): Likewise.
3128 (vect_prune_runtime_alias_test_list): Likewise.
3129 (vect_analyze_data_refs): Likewise.
3130 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
3131 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
3132 (vect_analyze_scalar_cycles_1): Likewise.
3133 (vect_get_loop_niters): Likewise.
3134 (vect_analyze_loop_form_1): Likewise.
3135 (vect_update_vf_for_slp): Likewise.
3136 (vect_analyze_loop_operations): Likewise.
3137 (vect_analyze_loop): Likewise.
3138 (vectorizable_induction): Likewise.
3139 (vect_transform_loop): Likewise.
3140 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
3141 * tree-vect-slp.c (vect_analyze_slp): Likewise.
3142 (vect_make_slp_decision): Likewise.
3143 (vect_detect_hybrid_slp): Likewise.
3144 (vect_slp_analyze_operations): Likewise.
3145 (vect_slp_bb): Likewise.
3146 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
3147 (vectorizable_bswap): Likewise.
3148 (vectorizable_call): Likewise.
3149 (vectorizable_simd_clone_call): Likewise.
3150 (vectorizable_conversion): Likewise.
3151 (vectorizable_assignment): Likewise.
3152 (vectorizable_shift): Likewise.
3153 (vectorizable_operation): Likewise.
3154 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
3156 2018-06-18 Martin Sebor <msebor@redhat.com>
3158 PR tree-optimization/81384
3159 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
3160 * builtins.c (expand_builtin_strnlen): New function.
3161 (expand_builtin): Call it.
3162 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
3163 * builtins.def (BUILT_IN_STRNLEN): New.
3164 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
3165 Warn for bounds in excess of maximum object size.
3166 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
3167 single-value ranges. Handle strnlen.
3168 (handle_builtin_strlen): Handle strnlen.
3169 (strlen_check_and_optimize_stmt): Same.
3170 * doc/extend.texi (Other Builtins): Document strnlen.
3172 2018-06-18 Maya Rashish <coypu@sdf.org>
3174 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
3175 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
3176 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
3178 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
3180 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
3182 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3184 * tree.c (escaped_string::escape): Replace cast to char * by
3185 const_cast<char *> (unescaped).
3187 2018-06-18 Nick Clifton <nickc@redhat.com>
3190 * tree.c (escaped_string): New class. Converts an unescaped
3191 string into its escaped equivalent.
3192 (warn_deprecated_use): Use the new class to convert the
3193 deprecation message, if present.
3194 (test_escaped_strings): New self test.
3195 (test_c_tests): Add test_escaped_strings.
3196 * doc/extend.texi (deprecated): Add a note that the
3197 deprecation message is affected by the -fmessage-length
3198 option, and that control characters will be escaped.
3199 (#pragma GCC error): Document this pragma.
3200 (#pragma GCC warning): Likewise.
3201 * doc/invoke.texi (-fmessage-length): Document this option's
3202 effect on the #warning and #error preprocessor directives and
3203 the deprecated attribute.
3205 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
3207 * tree.c (decl_value_expr_lookup): Revert latest change.
3208 (decl_value_expr_insert): Likewise.
3210 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
3212 * gimplify.c (nonlocal_vlas): Delete.
3213 (nonlocal_vla_vars): Likewise.
3214 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
3216 (gimplify_body): Do not create and destroy nonlocal_vlas.
3217 * tree-nested.c: Include diagnostic.h.
3218 (use_pointer_in_frame): Tweak.
3219 (lookup_field_for_decl): Add assertion and declare the transformation.
3220 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
3221 internal error when the reference is in a wrong context. Do not
3222 create a debug decl by default.
3223 (note_nonlocal_block_vlas): Delete.
3224 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
3225 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
3226 create a debug decl by default.
3227 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
3228 call is in a wrong context.
3229 (fixup_vla_decls): New function.
3230 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
3231 debug variables were created.
3232 * tree.c (decl_value_expr_lookup): Add checking assertion.
3233 (decl_value_expr_insert): Likewise.
3235 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3238 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
3239 * tree-scalar-evolution.c (interpret_expr): Likewise.
3240 (expression_expensive_p): Likewise.
3241 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
3242 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
3243 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
3244 (ssa_defined_by_minus_one_stmt_p): New.
3246 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
3249 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
3250 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
3251 * dojump.c (do_jump): Likewise.
3252 * expr.c (expand_expr_real_2): Check operand type's sign.
3253 * fold-const.c (const_unop): Handle ABSU_EXPR.
3254 (fold_abs_const): Likewise.
3255 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
3256 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
3257 (strip_sign_op_1): Likesise.
3258 * match.pd: Add new pattern to generate ABSU_EXPR.
3259 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
3260 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3261 * tree-eh.c (operation_could_trap_helper_p): Likewise.
3262 * tree-inline.c (estimate_operator_cost): Likewise.
3263 * tree-pretty-print.c (dump_generic_node): Likewise.
3264 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
3265 * tree.def (ABSU_EXPR): New.
3267 2018-06-16 Jakub Jelinek <jakub@redhat.com>
3270 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
3271 documented as preserved for backward compatibility only.
3272 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
3274 PR rtl-optimization/86108
3275 * bb-reorder.c (create_forwarder_block): Renamed to ...
3276 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
3277 jump from new landing pad to the second part.
3278 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
3281 2018-06-15 Jakub Jelinek <jakub@redhat.com>
3284 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
3285 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
3286 Only call store_expr for halves if the mode is the same.
3289 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
3290 Fix up comment formatting.
3292 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
3294 * typed-splay-tree.h (typed_splay_tree::remove): New function.
3295 (typed_splay_tree::closure,
3296 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
3297 (typed_splay_tree::typed_splay_tree,
3298 typed_splay_tree::operator =): Declared private.
3299 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
3300 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
3301 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
3302 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
3303 typed_splay_tree::splay_tree_splay,
3304 typed_splay_tree::splay_tree_foreach_helper,
3305 typed_splay_tree::splay_tree_insert,
3306 typed_splay_tree::splay_tree_remove,
3307 typed_splay_tree::splay_tree_lookup,
3308 typed_splay_tree::splay_tree_predecessor,
3309 typed_splay_tree::splay_tree_successor,
3310 typed_splay_tree::splay_tree_min,
3311 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
3312 (typed_splay_tree::root, typed_splay_tree::comp,
3313 typed_splay_tree::delete_key,
3314 typed_splay_tree::delete_value): New data members.
3315 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
3316 typed_splay_tree::remove.
3318 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
3320 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
3321 -mginv and -mno-ginv to the assembler.
3322 * config/mips/mips.opt (-mcrc): New option.
3324 * doc/invoke.text (-mcrc): Document.
3327 2018-06-15 Nick Clifton <nickc@redhat.com>
3330 * tree.c (escaped_string): New class. Converts an unescaped
3331 string into its escaped equivalent.
3332 (warn_deprecated_use): Use the new class to convert the
3333 deprecation message, if present.
3334 (test_escaped_strings): New self test.
3335 (test_c_tests): Add test_escaped_strings.
3336 * doc/extend.texi (deprecated): Add a note that the
3337 deprecation message is affected by the -fmessage-length
3338 option, and that control characters will be escaped.
3339 (#pragma GCC error): Document this pragma.
3340 (#pragma GCC warning): Likewise.
3341 * doc/invoke.texi (-fmessage-length): Document this option's
3342 effect on the #warning and #error preprocessor directives and
3343 the deprecated attribute.
3345 2018-06-15 Richard Biener <rguenther@suse.de>
3347 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
3348 here, also noting vector size used.
3349 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
3350 size used in MSG_OPTIMIZED_LOCATIONS dump.
3351 (pass_slp_vectorize::execute): Adjust.
3353 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
3356 * config/arc/arc.c (arc_return_address_register): Fix
3359 2018-06-15 Richard Biener <rguenther@suse.de>
3362 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
3363 leave useless conversion stripping to force_gimple_operand_gsi.
3364 (gimplify_build2): Likewise.
3365 (gimplify_build1): Likewise.
3367 2018-06-15 Richard Biener <rguenther@suse.de>
3370 * tree-cfg.c (move_stmt_op): unshare invariant addresses
3371 before adjusting their block.
3373 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
3375 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
3376 multilibs for *-*-rtems*.
3377 * config/riscv/t-rtems: New file.
3379 2018-06-14 Jakub Jelinek <jakub@redhat.com>
3382 * match.pd ((A +- CST1) +- CST2): Punt if last resort
3383 unsigned_type_for returns NULL.
3386 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
3387 subregs of multi-word pseudos unless the float mode has word size.
3389 2018-06-14 Richard Biener <rguenther@suse.de>
3392 * tree-vect-generic.c (build_word_mode_vector_type): Remove
3393 duplicate and harmful type_hash_canon.
3394 * tree.c (type_hash_canon): Assert we didn't find ourselves.
3396 2018-06-14 Richard Biener <rguenther@suse.de>
3399 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
3402 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
3404 * config/rtems.h (STDINT_LONG32): Define.
3406 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
3407 Prachi Godbole <prachi.godbole@imgtec.com>
3409 * config/mips/mips-cpus.def: Define P6600.
3410 * config/mips/mips-tables.opt: Regenerate.
3411 * config/mips/mips.c (mips_ucbranch_type): New enum.
3412 (mips_rtx_cost_data): Add support for P6600.
3413 (mips_issue_rate): Likewise.
3414 (mips_multipass_dfa_lookahead): Likewise.
3415 (mips_avoid_hazard): Likewise.
3416 (mips_reorg_process_insns): Likewise.
3417 (mips_classify_branch_p6600): New function.
3418 * config/mips/mips.h (TUNE_P6600): New define.
3419 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
3420 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
3421 * config/mips/mips.md: Include p6600.md.
3422 (processor): Add p6600.
3423 * config/mips/p6600.md: New file.
3424 * doc/invoke.texi: Add p6600 to supported architectures.
3426 2018-06-13 Martin Sebor <msebor@redhat.com>
3428 PR tree-optimization/86114
3429 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
3431 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
3433 2018-06-13 Richard Biener <rguenther@suse.de>
3435 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
3436 Properly set vector type of the intermediate stmt.
3437 * tree-vect-stmts.c (vectorizable_operation): The destination
3438 var always has vectype_out type.
3440 2018-06-13 Jeff Law <law@redhat.com>
3442 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
3443 integer 0 for argument to print_rtl_with_bb.
3444 (rl78_reorg): Likewise.
3446 2018-06-13 David Malcolm <dmalcolm@redhat.com>
3448 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
3449 from rtx to rtx_insn *.
3450 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
3452 (add_sched_insns_for_speculation): Likewise for local "target",
3453 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3454 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
3455 from rtx_insn ** to rtx_code_label **.
3456 (reorg_emit_nops): Likewise.
3457 (c6x_reorg): Likewise for local "call_labels".
3458 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
3460 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
3461 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
3462 the loops over LABEL_REFS.
3463 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
3465 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
3466 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
3467 (split_branches): Strengthen local "olabel" from rtx to
3468 rtx_insn *, adding a safe_as_a cast.
3469 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
3471 (add_insn_after): Likewise for first two params.
3472 (add_insn_before): Likewise.
3473 (remove_insn): Likewise for param.
3474 (emit_pattern_before_noloc): Likewise for second and third params.
3475 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
3476 (emit_call_insn_before_noloc): Likewise.
3477 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
3479 (emit_barrier_before): Likewise.
3480 (emit_label_before): Strengthen "label" param from "rtx" to
3481 "rtx_code_label *". Strengthen "before" param from "rtx" to
3483 (emit_insn_after_1): Strengthen "after" param from "rtx" to
3485 (emit_pattern_after_noloc): Likewise.
3486 (emit_insn_after_noloc): Likewise.
3487 (emit_jump_insn_after_noloc): Likewise.
3488 (emit_call_insn_after_noloc): Likewise.
3489 (emit_debug_insn_after_noloc): Likewise.
3490 (emit_barrier_after): Likewise.
3491 (emit_label_after): Likewise for both params.
3492 (emit_pattern_after_setloc): Likewise for "after" param. Convert
3493 "loc" param from "int" to "location_t".
3494 (emit_insn_after_setloc): Likewise.
3495 (emit_jump_insn_after_setloc): Likewise.
3496 (emit_call_insn_after_setloc): Likewise.
3497 (emit_debug_insn_after_setloc): Likewise.
3498 (emit_pattern_before_setloc): Likewise for "before" param. Convert
3499 "loc" param from "int" to "location_t".
3500 (emit_pattern_before): Convert NULL_RTX to NULL.
3501 (emit_insn_before_setloc): Convert "loc" param from "int" to
3503 (emit_jump_insn_before_setloc): Likewise.
3504 (emit_call_insn_before_setloc): Likewise.
3505 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
3506 rtx_insn *. Convert "loc" param from "int" to "location_t".
3507 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
3508 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
3509 Convert 3rd param from "int" to "location_t".
3510 (emit_barrier_before, emit_barrier_after, next_real_insn):
3511 Strengthen param from rtx to rtx_insn *.
3512 (emit_label_before): Strengthen 1st param from "rtx" to
3513 "rtx_code_label *". Strengthen 2nd param from "rtx" to
3515 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
3516 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
3517 Strengthen 2nd param from "rtx" to "rtx_insn *".
3518 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
3519 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
3520 Likewise. Convert 3rd param from "int" to "location_t".
3521 (emit_label_after): Strengthen 1st param from "rtx" to
3523 (next_real_insn, remove_insn): Strengthen param from "rtx" to
3525 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
3526 from "rtx" to "rtx_insn *".
3528 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
3530 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
3531 bodies streamed in with -Q.
3532 * dumpfile.c (dump_files): Add lto-stream-out dump file.
3533 * dumpfile.h (tree_dump_index): Add lto_stream_out.
3534 * gimple-streamer-out.c: Include gimple-pretty-print.h
3535 (output_bb): Dump stmts streamed.
3536 * lto-section-out.c: Include print-tree.h
3537 (lto_begin_section): Dump sections created.
3538 (lto_output_decl_index): Dump decl encoded.
3539 * lto-streamer-out.c: Include print-tree.h
3540 (create_output_block): Dump output block created.
3541 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
3542 (output_function): Dump function output.
3543 (output_constructor): Dump constructor streamed.
3544 (write_global_stream): Output indexes encoded.
3545 (produce_asm_for_decls): Dump streams encoded.
3546 * lto-streamer.c (streamer_dump_file): New global var.
3547 * lto-streamer.h (streamer_dump_file): Declare.
3548 * passes.c (ipa_write_summaries): Initialize streamer dump.
3549 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
3552 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
3555 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
3556 offsets for register save directives. Emit a second batch of save
3557 directives, if need be, when the function accesses prior frames.
3559 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3561 * config/arc/fpu.md (fmasf4): Force operand to register.
3562 (fnmasf4): Likewise.
3564 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3566 * config/arc/arc-protos.h (arc_pad_return): Remove.
3567 * config/arc/arc.c (machine_function): Remove force_short_suffix
3569 (arc_print_operand): Adjust printing of '&'.
3570 (arc_verify_short): Remove conditional printing of short suffix.
3571 (arc_final_prescan_insn): Remove reference to size_reason.
3572 (pad_return): New function.
3573 (arc_reorg): Call pad_return.
3574 (arc_pad_return): Remove.
3575 (arc_init_machine_status): Remove reference to force_short_suffix.
3576 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
3577 (attr length): When attribute iscompact is true force to 2
3578 regardless; in the case of maybe check if we want to force the
3579 instruction to have 4 bytes length.
3580 (nopv): Change it to generate 4 byte long nop as well.
3581 (blockage): New pattern.
3582 (simple_return): Remove call to arc_pad_return.
3583 (p_return_i): Likewise.
3585 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3587 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
3589 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3591 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
3594 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3596 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
3597 for ARC700 and ARCv2.
3599 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
3602 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
3603 operands[2] instead of operands[1].
3606 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3608 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
3609 case, check whether the outer register overlaps an unallocatable
3610 register, not just whether it fits the required class.
3612 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
3614 * poly-int.h (can_div_trunc_p): Add new overload in which all values
3616 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
3617 (memrefs_conflict_p): Likewise.
3618 (init_alias_analysis): Likewise.
3619 * cfgexpand.c (expand_debug_expr): Likewise.
3620 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
3621 * cse.c (fold_rtx): Likewise.
3622 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
3623 * expr.c (emit_block_move_hints): Likewise.
3624 (clear_storage_hints, push_block, emit_push_insn): Likewise.
3625 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
3626 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
3627 (emit_group_store): Likewise.
3628 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
3629 to read the PRE/POST_MODIFY increment.
3630 * calls.c (store_one_arg): Use strip_offset.
3631 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
3633 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
3635 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
3636 (simplify_binary_operation_1): Extend CONST_INT handling to
3638 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
3639 than a HOST_WIDE_INT.
3640 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
3642 (adjust_mems, add_stores): Update accodingly.
3643 (vt_canonicalize_addr): Track polynomial offsets.
3644 (emit_note_insn_var_location): Likewise.
3645 (vt_add_function_parameter): Likewise.
3646 (vt_initialize): Likewise.
3648 2018-06-12 Jeff Law <law@redhat.com>
3650 * config.gcc (alpha*-*-freebsd*): Remove.
3651 * config/alpha/freebsd.h: Remove.
3653 2018-06-12 David Malcolm <dmalcolm@redhat.com>
3656 * spellcheck-tree.c (levenshtein_distance): Rename to...
3657 (get_edit_distance): ...this, and update for underlying renaming.
3658 * spellcheck-tree.h (levenshtein_distance): Rename to...
3659 (get_edit_distance): ...this.
3660 * spellcheck.c (levenshtein_distance): Rename to...
3661 (get_edit_distance): ...this. Convert from Levenshtein distance
3662 to Damerau-Levenshtein distance by supporting transpositions of
3663 adjacent characters. Rename "v1" to "v_next" and "v0" to
3665 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
3666 (selftest::test_edit_distance_unit_test_oneway): ...this, and
3667 update for underlying renaming.
3668 (selftest::levenshtein_distance_unit_test): Rename to...
3669 (selftest::test_get_edit_distance_unit): ...this, and update for
3670 underlying renaming.
3671 (selftest::test_find_closest_string): Add example from PR 69968
3672 where transposition helps
3673 (selftest::test_metric_conditions): Update for renaming.
3674 (selftest::test_metric_conditions): Likewise.
3675 (selftest::spellcheck_c_tests): Likewise.
3676 * spellcheck.h (levenshtein_distance): Rename both overloads to...
3677 (get_edit_distance): ...this.
3678 (best_match::consider): Update for renaming.
3680 2018-06-12 Martin Sebor <msebor@redhat.com>
3682 PR tree-optimization/85259
3683 * builtins.c (compute_objsize): Handle constant offsets.
3684 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
3685 true iff a warning has been issued.
3686 * gimple.h (gimple_nonartificial_location): New function.
3687 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
3688 gimple_nonartificial_location and handle -Wno-system-headers.
3689 (handle_builtin_stxncpy): Same.
3691 2018-06-12 Martin Sebor <msebor@redhat.com>
3694 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
3696 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
3698 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
3699 BUILTIN_VEC_XST entries for pointer to double and long long.
3701 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
3704 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3706 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
3709 2018-06-12 Martin Liska <mliska@suse.cz>
3711 * doc/options.texi: Document IntegerRange.
3713 2018-06-12 Martin Liska <mliska@suse.cz>
3715 * config/i386/i386.opt: Make MPX-related options as Deprecated.
3716 * opt-functions.awk: Handle Deprecated flag.
3717 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
3719 (read_cmdline_option): Report warning for a deprecated option.
3720 * opts.h (struct cl_option): Add new field cl_deprecated.
3721 (CL_ERR_DEPRECATED): New.
3723 2018-06-12 Martin Liska <mliska@suse.cz>
3725 * doc/options.texi: Document Deprecated option flag.
3727 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3729 * config/arc/arc-arch.h (arc_extras): New enum.
3730 (arc_cpu_t):Add field extra.
3731 (arc_cpu_types): Consider the extras.
3732 * config/arc/arc-cpus.def: Add extras info.
3733 * config/arc/arc-opts.h (processor_type): Consider extra field.
3734 * config/arc/arc.c (arc_override_options): Handle extra field.
3736 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
3738 * config/arc/arc-arch.h: Update ARC_OPTX macro.
3739 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
3741 * config/arc/arc.c (arc_init): Update pic warning.
3742 (irq_range): Update irq range parsing warnings.
3743 (arc_override_options): Update various warning messages.
3744 (arc_handle_aux_attribute): Likewise.
3746 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
3748 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
3750 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
3752 * doc/sourcebuild.texi: Document usage of line number 0 in verify
3753 compiler messages directives.
3755 2018-06-12 Matthew Fortune <mfortune@gmail.com>
3757 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
3758 * config/mips/mips-tables.opt: Regenerate.
3759 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
3761 * doc/invoke.texi: Document -march=i6500.
3763 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
3765 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
3766 (i6400_gpmul): Add cpu_unit.
3767 (i6400_gpdiv): Likewise.
3768 (i6400_msa_add_d): Update reservations.
3769 (i6400_msa_int_add) Likewise.
3770 (i6400_msa_short_logic3) Likewise.
3771 (i6400_msa_short_logic2) Likewise.
3772 (i6400_msa_short_logic) Likewise.
3773 (i6400_msa_move) Likewise.
3774 (i6400_msa_cmp) Likewise.
3775 (i6400_msa_short_float2) Likewise.
3776 (i6400_msa_div_d) Likewise.
3777 (i6400_msa_long_logic1) Likewise.
3778 (i6400_msa_long_logic2) Likewise.
3779 (i6400_msa_mult) Likewise.
3780 (i6400_msa_long_float2) Likewise.
3781 (i6400_msa_long_float4) Likewise.
3782 (i6400_msa_long_float5) Likewise.
3783 (i6400_msa_long_float8) Likewise.
3784 (i6400_fpu_fadd): Include frint type.
3785 (i6400_fpu_store): New define_insn_reservation.
3786 (i6400_fpu_load): Likewise.
3787 (i6400_fpu_move): Likewise.
3788 (i6400_fpu_fcmp): Likewise.
3789 (i6400_fpu_fmadd): Likewise.
3790 (i6400_int_mult): Include imul3nc type and update reservation.
3791 (i6400_int_div): Include idiv3 type and update reservation.
3792 (i6400_int_load): Update to check type not move_type.
3793 (i6400_int_store): Likewise.
3794 (i6400_int_prefetch): Set zero latency.
3796 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
3798 * gcc.c: Document new %@{...} sequence.
3799 (LINK_COMMAND_SPEC): Use it for the -L switches.
3800 (cpp_unique_options): Use it for the -I switches.
3801 (at_file_argbuf): New global variable.
3802 (in_at_file): Likewise.
3803 (alloc_args): Create at_file_argbuf.
3804 (clear_args): Truncate at_file_argbuf.
3805 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
3806 (open_at_file): New function.
3807 (close_at_file): Likewise.
3808 (create_at_file): Delete.
3809 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
3812 (validate_switches_from_spec): Deal with %@{...} sequence.
3813 (validate_switches): Likewise.
3814 (driver::finalize): Call clear_args.
3816 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
3818 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
3820 2018-06-11 Martin Sebor <msebor@redhat.com>
3822 * doc/invoke.texi (-Wall): List -Wc++17-compat.
3823 (Wno-class-memaccess): Add @opindex.
3824 (Wno-templates, Wno-multiple-inheritance): Same.
3825 (Wno-virtual-inheritance, Wno-namespaces): Same.
3826 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
3827 (Wno-format-overflow, Wno-format-truncation): Same.
3828 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
3829 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
3830 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
3831 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
3832 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
3833 (Wno-misspelled-isr): Same.
3835 2018-06-11 Martin Sebor <msebor@redhat.com>
3837 * PR tree-optimization/86083
3838 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
3840 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
3842 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
3844 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
3847 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
3848 on the correct operand.
3849 (*movdi_internal64): Ditto.
3851 2018-06-11 Martin Liska <mliska@suse.cz>
3853 PR tree-optimization/86089
3854 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
3856 2018-06-11 Julia Koval <julia.koval@intel.com>
3858 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
3859 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
3860 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
3862 2018-06-11 Olivier Hainque <hainque@adacore.com>
3864 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
3865 for Ada with strict dwarf2.
3867 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
3870 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
3873 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
3875 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
3877 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
3879 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
3882 2018-06-08 Martin Liska <mliska@suse.cz>
3884 * tree-cfg.h (debug_function): Fix argument type to match
3887 2018-06-08 Martin Liska <mliska@suse.cz>
3889 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
3890 Remove usage of MPX-related (and removed) fields.
3891 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
3893 2018-06-08 David Malcolm <dmalcolm@redhat.com>
3895 * cfg.c (debug): Use TDF_NONE rather than 0.
3896 * cfghooks.c (debug): Likewise.
3897 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
3898 (struct dump_option_value_info): Convert to...
3899 (struct kv_pair): ...this template type.
3900 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
3902 (optinfo_verbosity_options): Likewise.
3903 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
3905 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
3906 than int for "optgroup_flags" param.
3907 (dump_generic_expr_loc): Use dump_flags_t rather than int for
3909 (dump_dec): Likewise.
3910 (dump_finish): Use TDF_NONE rather than 0.
3911 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
3912 rather than int for "optgroup_flags" param. Use TDF_NONE rather
3913 than 0. Update for change to option_ptr.
3914 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
3915 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
3916 0. Update for changes to optinfo_verbosity_options and
3918 (opt_info_switch_p): Convert optgroup_flags from int to
3920 (dump_basic_block): Use dump_flags_t rather than int
3921 for "dump_kind" param.
3922 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
3923 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
3924 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
3925 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
3926 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
3927 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
3928 TDF_NONE): Convert from macros to...
3929 (enum dump_flag): ...this new enum.
3930 (dump_flags_t): Update to use enum.
3931 (operator|, operator&, operator~, operator|=, operator&=):
3932 Implement for dump_flags_t.
3933 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
3934 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
3935 Convert from macros to...
3936 (enum optgroup_flag): ...this new enum.
3937 (optgroup_flags_t): New typedef.
3938 (operator|, operator|=): Implement for optgroup_flags_t.
3939 (struct dump_file_info): Convert field "alt_flags" to
3940 dump_flags_t. Convert field "optgroup_flags" to
3942 (dump_basic_block): Use dump_flags_t rather than int for param.
3943 (dump_generic_expr_loc): Likewise.
3944 (dump_dec): Likewise.
3945 (dump_register): Convert param "optgroup_flags" to
3947 (opt_info_enable_passes): Likewise.
3948 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
3950 * gimple-pretty-print.c (debug): Likewise.
3951 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
3952 (merged_store_group::apply_stores): Likewise.
3953 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
3954 * gimple.c (verify_gimple_pp): Likewise.
3955 * graphite-poly.c (print_pbb_body): Likewise.
3956 * passes.c (pass_manager::register_one_dump_file): Convert
3957 local "optgroup_flags" to optgroup_flags_t.
3958 * print-tree.c (print_node): Use TDF_NONE rather than 0.
3960 (debug_body): Likewise.
3961 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
3962 to optgroup_flags_t.
3963 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
3965 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
3966 (convert_mult_to_fma): Likewise.
3967 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
3968 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
3969 * tree-vect-data-refs.c (dump_lower_bound): Convert param
3970 "dump_kind" to dump_flags_t.
3972 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3974 * config/rs6000/rs6000.c (min, max): Delete.
3976 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
3978 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
3979 -mabi=spe and -mabi=no-spe.
3981 2018-06-08 Martin Liska <mliska@suse.cz>
3983 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
3984 where we expect an existing summary.
3986 2018-06-08 Martin Liska <mliska@suse.cz>
3988 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
3989 * ipa-inline.h (estimate_edge_growth): Likewise.
3991 2018-06-08 Martin Liska <mliska@suse.cz>
3993 * cgraph.c (function_version_hasher::hash): Use
3994 cgraph_node::get_uid ().
3995 (function_version_hasher::equal):
3996 * cgraph.h (cgraph_node::get_uid): New method.
3997 * ipa-inline.c (update_caller_keys): Use
3998 cgraph_node::get_uid ().
3999 (update_callee_keys): Likewise.
4000 * ipa-utils.c (searchc): Likewise.
4001 (ipa_reduced_postorder): Likewise.
4002 * lto-cgraph.c (input_node): Likewise.
4003 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
4004 * symbol-summary.h (symtab_insertion): Likewise.
4005 (symtab_removal): Likewise.
4006 (symtab_duplication): Likewise.
4007 * tree-pretty-print.c (dump_function_header): Likewise.
4008 * tree-sra.c (convert_callers_for_node): Likewise.
4010 2018-06-08 Martin Liska <mliska@suse.cz>
4012 * cgraph.c (symbol_table::create_edge): Always assign a new
4014 (symbol_table::free_edge): Do not recycle numbers.
4015 * cgraph.h (cgraph_edge::get): New method.
4016 * symbol-summary.h (symtab_removal): Use it.
4017 (symtab_duplication): Likewise.
4018 (call_summary::hashable_uid): Remove.
4020 2018-06-08 Martin Liska <mliska@suse.cz>
4022 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
4023 (initialize_growth_caches): Remove.
4024 (free_growth_caches): Likewise.
4025 (do_estimate_edge_time): Use edge_growth_cache.
4026 (do_estimate_edge_size): Likewise.
4027 (do_estimate_edge_hints): Likewise.
4028 * ipa-inline.c (reset_edge_caches): Likewise.
4029 (recursive_inlining): Likewise.
4030 (inline_small_functions): Likewise.
4031 * ipa-inline.h (initialize_growth_caches): Remove.
4032 (estimate_edge_size): Likewise.
4033 (estimate_edge_time): Likewise.
4034 (estimate_edge_hints): Likewise.
4035 (reset_edge_growth_cache): Likewise.
4036 * symbol-summary.h (call_summary::remove): New method.
4038 2018-06-08 Martin Liska <mliska@suse.cz>
4040 * ipa-cp.c (class edge_clone_summary): New summary.
4041 (grow_edge_clone_vectors): Remove.
4042 (ipcp_edge_duplication_hook): Remove.
4043 (class edge_clone_summary_t): New call_summary class.
4044 (ipcp_edge_removal_hook): Remove.
4045 (edge_clone_summary_t::duplicate): New function.
4046 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
4047 (create_specialized_node): Likewise.
4048 (ipcp_driver): Initialize edge_clone_summaries and do not
4051 2018-06-08 Martin Liska <mliska@suse.cz>
4053 * symbol-summary.h (get): New function.
4054 (call_summary::m_initialize_when_cloning): New class member.
4056 2018-06-08 Martin Liska <mliska@suse.cz>
4058 * cgraph.c (cgraph_node::remove): Do not recycle uid.
4059 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
4060 (symbol_table::allocate_cgraph_symbol): Do not set uid.
4061 * passes.c (uid_hash_t): Record removed_nodes by their uids.
4062 (remove_cgraph_node_from_order): Use the removed_nodes set.
4063 (do_per_function_toporder): Likwise.
4064 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
4065 instead of summary_uid.
4066 (symtab_removal): Likewise.
4067 (symtab_duplication): Likewise.
4069 2018-06-08 Martin Liska <mliska@suse.cz>
4071 * ipa-cp.c (ipcp_store_bits_results): Use
4072 ipcp_transformation_sum.
4073 (ipcp_store_vr_results): Likewise.
4074 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
4076 (ipcp_transformation_initialize): ... this.
4077 (ipa_set_node_agg_value_chain):
4078 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
4079 (write_ipcp_transformation_info): Likewise.
4080 (read_ipcp_transformation_info): Likewise.
4081 (ipcp_update_bits): Likewise.
4082 (ipcp_update_vr): Likewise.
4083 (ipcp_transform_function): Likewise.
4084 * ipa-prop.h: Rename ipcp_transformation_summary to
4085 ipcp_transformation.
4086 (class ipcp_transformation_t): New function summary.
4087 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
4088 (ipa_get_agg_replacements_for_node): Likewise.
4090 2018-06-08 Martin Liska <mliska@suse.cz>
4092 * ipa-pure-const.c (struct funct_state_d): Do it class instead
4094 (class funct_state_summary_t): New function_summary class.
4095 (has_function_state): Remove.
4096 (get_function_state): Likewise.
4097 (set_function_state): Likewise.
4098 (add_new_function): Likewise.
4099 (funct_state_summary_t::insert): New function.
4100 (duplicate_node_data): Remove.
4101 (remove_node_data): Remove.
4102 (funct_state_summary_t::duplicate): New function.
4103 (register_hooks): Create new funct_state_summaries.
4104 (pure_const_generate_summary): Use it.
4105 (pure_const_write_summary): Likewise.
4106 (pure_const_read_summary): Likewise.
4107 (propagate_pure_const): Likewise.
4108 (propagate_nothrow): Likewise.
4109 (dump_malloc_lattice): Likewise.
4110 (propagate_malloc): Likewise.
4111 (execute): Do not register hooks, just remove summary
4113 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
4116 2018-06-08 Martin Liska <mliska@suse.cz>
4118 * ipa-reference.c (remove_node_data): Remove.
4119 (duplicate_node_data): Likewise.
4120 (class ipa_ref_var_info_summary_t): New class.
4121 (class ipa_ref_opt_summary_t): Likewise.
4122 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
4123 (get_reference_optimization_summary): Use
4124 ipa_ref_opt_sum_summaries.
4125 (set_reference_vars_info): Remove.
4126 (set_reference_optimization_summary): Likewise.
4127 (ipa_init): Create summaries.
4128 (init_function_info): Use function summary.
4129 (ipa_ref_opt_summary_t::duplicate): New function.
4130 (ipa_ref_opt_summary_t::remove): New function.
4131 (get_read_write_all_from_node): Fix GNU coding style.
4132 (propagate): Use function summary.
4133 (write_node_summary_p): Fix GNU coding style.
4134 (stream_out_bitmap): Likewise.
4135 (ipa_reference_read_optimization_summary): Use function summary.
4136 (ipa_reference_c_finalize): Do not release hooks.
4138 2018-06-08 Martin Liska <mliska@suse.cz>
4140 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
4141 (analyze_function_body): Extract multiple calls of get_create.
4142 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4143 * ipa-inline.c (recursive_inlining): Use ::get method.
4144 * ipa-inline.h (estimate_edge_growth): Likewise.
4146 2018-06-08 Martin Liska <mliska@suse.cz>
4148 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
4150 (hsa_function_summary::hsa_function_summary): Use the new enum
4152 (hsa_gpu_implementation_p): Use hsa_summaries::get.
4153 * hsa-gen.c (hsa_get_host_function): Likewise.
4154 (get_brig_function_name): Likewise.
4155 * ipa-hsa.c (process_hsa_functions): Likewise.
4156 (ipa_hsa_write_summary): Likewise.
4157 * symbol-summary.h (symtab_duplication): Use ::get function/
4158 (get): New function.
4160 2018-06-08 Martin Liska <mliska@suse.cz>
4162 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
4164 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
4165 (hsa_register_kernel): Likewise.
4166 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
4167 * hsa-gen.c (hsa_get_host_function): Likewise.
4168 (get_brig_function_name): Likewise.
4169 (generate_hsa): Likewise.
4170 (pass_gen_hsail::execute): Likewise.
4171 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4172 (devirtualization_time_bonus): Likewise.
4173 (ipcp_propagate_stage): Likewise.
4174 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
4175 (edge_set_predicate): Likewise.
4176 (evaluate_conditions_for_known_args): Likewise.
4177 (evaluate_properties_for_edge): Likewise.
4178 (ipa_fn_summary::reset): Likewise.
4179 (ipa_fn_summary_t::duplicate): Likewise.
4180 (dump_ipa_call_summary): Likewise.
4181 (ipa_dump_fn_summary): Likewise.
4182 (analyze_function_body): Likewise.
4183 (compute_fn_summary): Likewise.
4184 (estimate_edge_devirt_benefit): Likewise.
4185 (estimate_edge_size_and_time): Likewise.
4186 (estimate_calls_size_and_time): Likewise.
4187 (estimate_node_size_and_time): Likewise.
4188 (inline_update_callee_summaries): Likewise.
4189 (remap_edge_change_prob): Likewise.
4190 (remap_edge_summaries): Likewise.
4191 (ipa_merge_fn_summary_after_inlining): Likewise.
4192 (ipa_update_overall_fn_summary): Likewise.
4193 (read_ipa_call_summary): Likewise.
4194 (inline_read_section): Likewise.
4195 (write_ipa_call_summary): Likewise.
4196 (ipa_fn_summary_write): Likewise.
4197 (ipa_free_fn_summary): Likewise.
4198 * ipa-hsa.c (process_hsa_functions): Likewise.
4199 (ipa_hsa_write_summary): Likewise.
4200 (ipa_hsa_read_section): Likewise.
4201 * ipa-icf.c (sem_function::merge): Likewise.
4202 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
4203 (do_estimate_edge_time): Likewise.
4204 (estimate_size_after_inlining): Likewise.
4205 (estimate_growth): Likewise.
4206 (growth_likely_positive): Likewise.
4207 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
4208 (inline_call): Likewise.
4209 * ipa-inline.c (caller_growth_limits): Likewise.
4210 (can_inline_edge_p): Likewise.
4211 (can_inline_edge_by_limits_p): Likewise.
4212 (compute_uninlined_call_time): Likewise.
4213 (compute_inlined_call_time): Likewise.
4214 (want_inline_small_function_p): Likewise.
4215 (edge_badness): Likewise.
4216 (update_caller_keys): Likewise.
4217 (update_callee_keys): Likewise.
4218 (recursive_inlining): Likewise.
4219 (inline_small_functions): Likewise.
4220 (inline_to_all_callers_1): Likewise.
4221 (dump_overall_stats): Likewise.
4222 (early_inline_small_functions): Likewise.
4223 (early_inliner): Likewise.
4224 * ipa-inline.h (estimate_edge_growth): Likewise.
4225 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4226 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4227 * ipa-prop.h (IPA_NODE_REF): Likewise.
4228 (IPA_EDGE_REF): Likewise.
4229 * ipa-pure-const.c (malloc_candidate_p): Likewise.
4230 (propagate_malloc): Likewise.
4231 * ipa-split.c (execute_split_functions): Likewise.
4232 * symbol-summary.h: Rename get to get_create.
4234 (get_create): Likewise.
4235 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
4237 2018-06-08 Martin Liska <mliska@suse.cz>
4239 * symbol-summary.h (release): Move definition out of class
4241 (symtab_removal): Likewise.
4242 (symtab_duplication): Likewise.
4244 2018-06-08 Martin Liska <mliska@suse.cz>
4246 * symbol-summary.h (function_summary): Move constructor
4247 implementation out of class declaration.
4248 (release): Likewise.
4249 (symtab_insertion): Likewise.
4250 (symtab_removal): Likewise.
4251 (symtab_duplication): Likewise.
4254 2018-06-08 Martin Liska <mliska@suse.cz>
4256 * Makefile.in: Remove support for MPX (macros, related functions,
4257 fields in cgraph_node, ...).
4258 * builtin-types.def (BT_BND): Likewise.
4259 (BT_FN_BND_CONST_PTR): Likewise.
4260 (BT_FN_CONST_PTR_BND): Likewise.
4261 (BT_FN_VOID_PTR_BND): Likewise.
4262 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
4263 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
4264 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
4265 (expand_builtin_mempcpy_with_bounds): Likewise.
4266 (expand_builtin_memset_with_bounds): Likewise.
4267 (expand_builtin_memset_args): Likewise.
4268 (std_expand_builtin_va_start): Likewise.
4269 (expand_builtin): Likewise.
4270 (expand_builtin_with_bounds): Likewise.
4271 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
4272 (DEF_LIB_BUILTIN_CHKP): Likewise.
4273 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
4274 (DEF_CHKP_BUILTIN): Likewise.
4275 (BUILT_IN_MEMCPY): Likewise.
4276 (BUILT_IN_MEMMOVE): Likewise.
4277 (BUILT_IN_MEMPCPY): Likewise.
4278 (BUILT_IN_MEMSET): Likewise.
4279 (BUILT_IN_STPCPY): Likewise.
4280 (BUILT_IN_STRCAT): Likewise.
4281 (BUILT_IN_STRCHR): Likewise.
4282 (BUILT_IN_STRCPY): Likewise.
4283 (BUILT_IN_STRLEN): Likewise.
4284 (BUILT_IN_MEMCPY_CHK): Likewise.
4285 (BUILT_IN_MEMMOVE_CHK): Likewise.
4286 (BUILT_IN_MEMPCPY_CHK): Likewise.
4287 (BUILT_IN_MEMSET_CHK): Likewise.
4288 (BUILT_IN_STPCPY_CHK): Likewise.
4289 (BUILT_IN_STRCAT_CHK): Likewise.
4290 (BUILT_IN_STRCPY_CHK): Likewise.
4291 * calls.c (store_bounds): Likewise.
4292 (emit_call_1): Likewise.
4293 (special_function_p): Likewise.
4294 (maybe_warn_nonstring_arg): Likewise.
4295 (initialize_argument_information): Likewise.
4296 (finalize_must_preallocate): Likewise.
4297 (compute_argument_addresses): Likewise.
4298 (expand_call): Likewise.
4299 * cfgexpand.c (expand_call_stmt): Likewise.
4300 (expand_return): Likewise.
4301 (expand_gimple_stmt_1): Likewise.
4302 (pass_expand::execute): Likewise.
4303 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
4304 (cgraph_node::remove): Likewise.
4305 (cgraph_node::dump): Likewise.
4306 (cgraph_node::verify_node): Likewise.
4307 * cgraph.h (chkp_function_instrumented_p): Likewise.
4308 (symtab_node::get_alias_target): Likewise.
4309 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
4310 (cgraph_local_p): Likewise.
4311 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
4312 (cgraph_edge::rebuild_references): Likewise.
4313 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
4314 (walk_polymorphic_call_targets): Likewise.
4315 (cgraph_node::expand_thunk): Likewise.
4316 (symbol_table::output_weakrefs): Likewise.
4317 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
4318 (ix86_handle_option): Likewise.
4319 * config/i386/constraints.md: Likewise.
4320 * config/i386/i386-builtin-types.def (BND): Likewise.
4324 * config/i386/i386-builtin.def (BDESC_END): Likewise.
4325 (BDESC_FIRST): Likewise.
4327 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
4328 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
4329 * config/i386/i386.c (enum reg_class): Likewise.
4330 (ix86_target_string): Likewise.
4331 (ix86_option_override_internal): Likewise.
4332 (ix86_conditional_register_usage): Likewise.
4333 (ix86_valid_target_attribute_inner_p): Likewise.
4334 (ix86_set_indirect_branch_type): Likewise.
4335 (ix86_set_current_function): Likewise.
4336 (ix86_function_arg_regno_p): Likewise.
4337 (init_cumulative_args): Likewise.
4338 (ix86_function_arg_advance): Likewise.
4339 (ix86_function_arg): Likewise.
4340 (ix86_pass_by_reference): Likewise.
4341 (ix86_function_value_regno_p): Likewise.
4342 (ix86_function_value_1): Likewise.
4343 (ix86_function_value_bounds): Likewise.
4344 (ix86_return_in_memory): Likewise.
4345 (ix86_setup_incoming_vararg_bounds): Likewise.
4346 (ix86_va_start): Likewise.
4347 (indirect_thunk_need_prefix): Likewise.
4348 (print_reg): Likewise.
4349 (ix86_print_operand): Likewise.
4350 (ix86_expand_call): Likewise.
4351 (ix86_output_function_return): Likewise.
4352 (reg_encoded_number): Likewise.
4353 (BDESC_VERIFYS): Likewise.
4354 (ix86_init_mpx_builtins): Likewise.
4355 (ix86_init_builtins): Likewise.
4356 (ix86_emit_cmove): Likewise.
4357 (ix86_emit_move_max): Likewise.
4358 (ix86_expand_builtin): Likewise.
4359 (ix86_builtin_mpx_function): Likewise.
4360 (ix86_get_arg_address_for_bt): Likewise.
4361 (ix86_load_bounds): Likewise.
4362 (ix86_store_bounds): Likewise.
4363 (ix86_load_returned_bounds): Likewise.
4364 (ix86_store_returned_bounds): Likewise.
4365 (ix86_class_likely_spilled_p): Likewise.
4366 (ix86_hard_regno_mode_ok): Likewise.
4367 (x86_order_regs_for_local_alloc): Likewise.
4368 (ix86_mitigate_rop): Likewise.
4369 (ix86_bnd_prefixed_insn_p): Likewise.
4370 (ix86_mpx_bound_mode): Likewise.
4371 (ix86_make_bounds_constant): Likewise.
4372 (ix86_initialize_bounds): Likewise.
4373 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
4374 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
4375 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
4376 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
4377 (TARGET_CHKP_BOUND_MODE): Likewise.
4378 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
4379 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
4380 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
4381 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
4382 * config/i386/i386.h (TARGET_MPX): Likewise.
4383 (TARGET_MPX_P): Likewise.
4384 (VALID_BND_REG_MODE): Likewise.
4385 (FIRST_BND_REG): Likewise.
4386 (LAST_BND_REG): Likewise.
4387 (enum reg_class): Likewise.
4388 (BND_REG_P): Likewise.
4389 (BND_REGNO_P): Likewise.
4390 (BNDmode): Likewise.
4391 (ADJUST_INSN_LENGTH): Likewise.
4392 * config/i386/i386.md: Likewise.
4393 * config/i386/i386.opt: Likewise.
4394 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
4395 (defined): Likewise.
4396 (LINK_MPX): Likewise.
4397 (MPX_SPEC): Likewise.
4398 (LIBMPX_SPEC): Likewise.
4399 (LIBMPXWRAPPERS_SPEC): Likewise.
4400 (CHKP_SPEC): Likewise.
4401 * config/i386/predicates.md: Likewise.
4402 * dbxout.c (dbxout_type): Likewise.
4403 * doc/extend.texi: Likewise.
4404 * doc/invoke.texi: Likewise.
4405 * doc/md.texi: Likewise.
4406 * doc/tm.texi: Likewise.
4407 * doc/tm.texi.in: Likewise.
4408 * dwarf2out.c (is_base_type): Likewise.
4409 (gen_formal_types_die): Likewise.
4410 (gen_subprogram_die): Likewise.
4411 (gen_type_die_with_usage): Likewise.
4412 (gen_decl_die): Likewise.
4413 (dwarf2out_late_global_decl): Likewise.
4414 * expr.c (expand_assignment): Likewise.
4415 (emit_storent_insn): Likewise.
4416 (store_expr_with_bounds): Likewise.
4417 (store_expr): Likewise.
4418 (expand_expr_real_1): Likewise.
4419 * expr.h (store_expr_with_bounds): Likewise.
4420 * function.c (use_register_for_decl): Likewise.
4421 (struct bounds_parm_data): Likewise.
4422 (assign_parms_augmented_arg_list): Likewise.
4423 (assign_parm_find_entry_rtl): Likewise.
4424 (assign_parm_is_stack_parm): Likewise.
4425 (assign_parm_load_bounds): Likewise.
4426 (assign_bounds): Likewise.
4427 (assign_parms): Likewise.
4428 (expand_function_start): Likewise.
4429 * gcc.c (CHKP_SPEC): Likewise.
4430 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
4431 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
4432 (wrestrict_dom_walker::check_call): Likewise.
4433 * gimple.c (gimple_build_call_from_tree): Likewise.
4434 * gimple.h (enum gf_mask): Likewise.
4435 (gimple_call_with_bounds_p): Likewise.
4436 (gimple_call_set_with_bounds): Likewise.
4437 * gimplify.c (gimplify_init_constructor): Likewise.
4438 * ipa-cp.c (initialize_node_lattices): Likewise.
4439 (propagate_constants_across_call): Likewise.
4440 (find_more_scalar_values_for_callers_subset): Likewise.
4441 * ipa-hsa.c (process_hsa_functions): Likewise.
4442 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
4443 * ipa-icf.c (sem_function::merge): Likewise.
4444 * ipa-inline.c (early_inliner): Likewise.
4445 * ipa-pure-const.c (warn_function_noreturn): Likewise.
4446 (warn_function_cold): Likewise.
4447 (propagate_pure_const): Likewise.
4448 * ipa-ref.h (enum GTY): Likewise.
4449 * ipa-split.c (find_retbnd): Likewise.
4450 (consider_split): Likewise.
4451 (split_function): Likewise.
4452 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
4453 * ipa.c (walk_polymorphic_call_targets): Likewise.
4454 (symbol_table::remove_unreachable_nodes): Likewise.
4455 (process_references): Likewise.
4456 (cgraph_build_static_cdtor_1): Likewise.
4457 * lto-cgraph.c (lto_output_node): Likewise.
4458 (output_refs): Likewise.
4459 (compute_ltrans_boundary): Likewise.
4460 (input_overwrite_node): Likewise.
4461 (input_node): Likewise.
4462 (input_cgraph_1): Likewise.
4463 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
4464 * passes.c (pass_manager::execute_early_local_passes): Likewise.
4465 (class pass_chkp_instrumentation_passes): Likewise.
4466 (make_pass_chkp_instrumentation_passes): Likewise.
4467 * passes.def: Likewise.
4468 * rtl.h (struct GTY): Likewise.
4469 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
4470 * stor-layout.c (layout_type): Likewise.
4471 * symtab.c: Likewise.
4472 * target.def: Likewise.
4473 * targhooks.c (default_chkp_bound_type): Likewise.
4474 (default_chkp_bound_mode): Likewise.
4475 (default_builtin_chkp_function): Likewise.
4476 (default_chkp_function_value_bounds): Likewise.
4477 (default_chkp_make_bounds_constant): Likewise.
4478 (default_chkp_initialize_bounds): Likewise.
4479 * targhooks.h (default_chkp_bound_type): Likewise.
4480 (default_chkp_bound_mode): Likewise.
4481 (default_builtin_chkp_function): Likewise.
4482 (default_chkp_function_value_bounds): Likewise.
4483 (default_chkp_make_bounds_constant): Likewise.
4484 (default_chkp_initialize_bounds): Likewise.
4485 * toplev.c (compile_file): Likewise.
4486 (process_options): Likewise.
4487 * tree-core.h (DEF_BUILTIN): Likewise.
4488 (DEF_BUILTIN_CHKP): Likewise.
4489 * tree-inline.c (declare_return_variable): Likewise.
4490 (remap_gimple_stmt): Likewise.
4491 (copy_bb): Likewise.
4492 (initialize_inlined_parameters): Likewise.
4493 (expand_call_inline): Likewise.
4494 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
4495 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
4496 (make_pass_ipa_chkp_produce_thunks): Likewise.
4497 (make_pass_chkp): Likewise.
4498 (make_pass_chkp_opt): Likewise.
4499 (make_pass_chkp_instrumentation_passes): Likewise.
4500 * tree-pretty-print.c (dump_generic_node): Likewise.
4501 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
4502 * tree-ssa-dce.c (propagate_necessity): Likewise.
4503 (eliminate_unnecessary_stmts): Likewise.
4504 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
4505 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
4506 * tree-ssa-sccvn.h: Likewise.
4507 * tree-ssa-strlen.c (get_string_length): Likewise.
4508 (valid_builtin_call): Likewise.
4509 (adjust_last_stmt): Likewise.
4510 (handle_builtin_strchr): Likewise.
4511 (handle_builtin_strcpy): Likewise.
4512 (handle_builtin_stxncpy): Likewise.
4513 (handle_builtin_memcpy): Likewise.
4514 (handle_builtin_strcat): Likewise.
4515 (strlen_check_and_optimize_stmt): Likewise.
4516 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
4517 * tree-streamer-in.c: Likewise.
4518 * tree-streamer.c (record_common_node): Likewise.
4519 * tree.c (tree_code_size): Likewise.
4520 (wide_int_to_tree_1): Likewise.
4521 (type_contains_placeholder_1): Likewise.
4522 (build_common_tree_nodes): Likewise.
4523 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
4524 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
4525 (POINTER_BOUNDS_P): Likewise.
4526 (BOUNDED_TYPE_P): Likewise.
4527 (BOUNDED_P): Likewise.
4528 (CALL_WITH_BOUNDS_P): Likewise.
4529 (pointer_bounds_type_node): Likewise.
4530 * value-prof.c (gimple_ic): Likewise.
4531 * var-tracking.c (vt_add_function_parameters): Likewise.
4532 * varasm.c (make_decl_rtl): Likewise.
4533 (assemble_start_function): Likewise.
4534 (output_constant): Likewise.
4535 (maybe_assemble_visibility): Likewise.
4536 * varpool.c (ctor_for_folding): Likewise.
4537 * chkp-builtins.def: Remove.
4538 * ipa-chkp.c: Remove.
4539 * ipa-chkp.h: Remove.
4540 * rtl-chkp.c: Remove.
4541 * rtl-chkp.h: Remove.
4542 * tree-chkp-opt.c: Remove.
4543 * tree-chkp.c: Remove.
4544 * tree-chkp.h: Remove.
4546 2018-06-07 Carl Love <cel@us.ibm.com>
4548 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
4549 vextract_fp_from_shortl): Add BE support.
4551 2018-06-07 Paul Koning <ni1d@arrl.net>
4553 * compare-elim.c (try_merge_compare): Don't merge compare if
4554 address contains a side effect.
4555 (try_eliminate_compare): Likewise.
4557 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
4559 * config.gcc: Support "tremont".
4560 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
4561 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4563 * config/i386/i386.c (m_TREMONT): Define.
4564 (processor_target_table): Add "tremont".
4565 (PTA_TREMONT): Define.
4566 (ix86_lea_outperforms): Add TARGET_TREMONT.
4567 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
4568 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
4569 and M_INTEL_GOLDMONT_PLUS.
4570 (fold_builtin_cpu): Add "tremont".
4571 (ix86_add_stmt_cost): Add TARGET_TREMONT.
4572 (ix86_option_override_internal): Add "tremont".
4573 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
4574 (processor_type): Add PROCESSOR_TREMONT.
4575 * config/i386/x86-tune.def: Add m_TREMONT.
4576 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
4578 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4580 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
4581 symbol defined for msp430i* devices to be lower case.
4583 gcc/testsuite/gcc.target/msp430/
4584 * msp430i-device-symbol.c: New test.
4585 * msp430f-device-symbol.c: New test.
4586 * msp430.h: New test header file.
4588 2018-06-07 Richard Biener <rguenther@suse.de>
4590 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
4591 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
4592 Properly wrap signed arithmetic if overflow wraps.
4594 2018-06-07 Jakub Jelinek <jakub@redhat.com>
4596 PR tree-optimization/69615
4597 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
4598 of a cast from a same precision integral SSA_NAME in a bb dominated
4599 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
4600 cast to utype if rhs2 has already a compatible type.
4602 2018-06-07 Richard Biener <rguenther@suse.de>
4604 PR tree-optimization/85935
4605 * graphite-scop-detection.c (find_params_in_bb): Analyze
4606 condition operands with respect to the correct loop. Assert
4607 the analysis doesn't fail.
4609 2018-06-04 Carl Love <cel@us.ibm.com>
4611 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
4612 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
4613 as it is slightly cheaper.
4614 (first_match_or_eos_index_<mode>):
4615 Calculate index using natural element order.
4616 (first_match_index_<mode>):
4617 Calculate index using natural element order.
4618 (first_match_or_eos_index_<mode>):
4619 Calculate index using natural order.
4620 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
4621 for BE and LE modes.
4622 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
4623 P9V_BUILTIN_VCLZLSBB_V16QI.
4624 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
4627 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4629 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
4630 indentation and line wrap for many prototypes. Add missing
4631 @smallexample directives around block of prototypes for vec_xl and
4634 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
4636 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
4637 track if we pass or return IEEE 128-bit floating point.
4638 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
4639 C++ mangling that is compatible with GCC 8.1.
4640 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
4641 (init_cumulative_args): Note if we pass or return IEEE 128-bit
4642 floating point types.
4643 (rs6000_function_arg_advance_1): Likewise.
4644 (rs6000_mangle_type): Optionally generate mangled names that match
4645 what GCC 8.1 generated for IEEE 128-bit floating point types.
4646 (rs6000_globalize_decl_name): If we have an external function that
4647 passes or returns IEEE 128-bit types, generate a weak reference
4648 from the mangled name used in GCC 8.1 to the current mangled
4650 (rs6000_init_builtins): Make __ibm128 use the long double type if
4651 long double is IBM extended double. Make __float128 use the long
4652 double type if long double is IEEE 128-bit.
4655 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
4656 macro for __ibm128 built-in functions.
4657 (PACK_IF): Add __ibm128 pack/unpack functions.
4658 (UNPACK_IF): Likewise.
4659 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
4660 enable long double built-in functions if long double is IEEE
4661 128-bit floating point.
4662 (rs6000_invalid_builtin): Update long double built-in function
4664 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
4665 functions, adjust the built-in function to use the long double
4666 built-in function if __ibm128 and long double are the same type.
4667 * doc/extend.texi (PowerPC builtins): Update documention for
4668 __builtin_{,un}pack_longdouble. Add documentation for
4669 __builtin_{,un}pack_ibm128.
4671 2018-06-06 Jim Wilson <jimw@sifive.com>
4673 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
4674 (struct machine_function): New field interrupt_mode.
4675 (riscv_handle_type_attribute): New function. Add forward declaration.
4676 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
4677 (riscv_expand_epilogue): Check interrupt_mode field.
4678 (riscv_set_current_function): Check interrupt attribute args and
4679 set interrupt_mode field.
4680 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
4681 (riscv_sret, riscv_uret): New.
4682 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
4683 new arguments to interrupt attribute.
4685 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
4688 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
4689 Don't handle -mcpu=power8 if -mpower9-vector is also used.
4691 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
4693 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
4694 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
4695 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
4696 several redundant entries.
4698 2018-06-06 David Malcolm <dmalcolm@redhat.com>
4700 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
4701 type from "rtx" to "rtx_insn *".
4702 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
4703 for local "call_insn", removing cast.
4704 (ix86_expand_call): Likewise, introducing a "call_insn" local.
4706 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
4708 PR tree-optimization/86066
4709 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
4710 for BIT_INSERT_EXPR stores.
4712 2018-06-06 Richard Biener <rguenther@suse.de>
4714 PR tree-optimization/86062
4715 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
4716 component refs ontop
4717 of to be offsetted base.
4719 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4721 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
4722 to be static and remove check on interrupt attribute name.
4724 gcc/testsuite/gcc.target/msp430/
4725 * function-attributes-4.c: New test.
4726 * static-interrupts.c: New test.
4728 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
4730 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
4731 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
4733 2018-06-05 Steve Ellcey <sellcey@cavium.com>
4736 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
4738 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
4739 Remove second argument, change how error is called.
4740 (aarch64_layout_arg): Remove second argument from
4741 aarch64_err_no_fpadvsimd call.
4742 (aarch64_init_cumulative_args): Ditto.
4743 (aarch64_gimplify_va_arg_expr): Ditto.
4744 * config/aarch64/aarch64.md (mov<mode>): Ditto.
4746 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
4748 * config/i386/i386.md (simple_return_indirect_internal): New expander.
4749 (*simple_return_indirect_internal<mode>): Rename from
4750 simple_return_indirect_internal. Use W mode iterator.
4751 (rstorssp): New expander.
4752 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
4753 (clrssbsy): New expander.
4754 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
4756 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
4758 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
4760 (cmse_check_pointed_object): Likewise.
4762 2018-06-05 Martin Liska <mliska@suse.cz>
4764 PR gcov-profile/47618
4765 * doc/invoke.texi: Document how -fprofile-dir format
4768 2018-06-05 Richard Biener <rguenther@suse.de>
4770 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
4771 removal pretend DOM info isn't available so we do not update
4772 it and only remove edges, not dominated blocks. Actually free
4773 DOM info in case we removed something. Remove unreachable blocks.
4774 (mfb_keep_latches): Work with either DOM info or marked backedges.
4775 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
4776 first. Mark backedges if DOM info isn't available.
4777 (Re-)compute DOM info after cleanup_control_flow_pre.
4779 2018-06-05 Richard Biener <rguenther@suse.de>
4781 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
4782 (locus_discrim_hasher::hash): Adjust.
4783 (locus_discrim_hasher::equal): Likewise.
4784 (next_discriminator_for_locus): Work on line directly.
4785 (same_line_p): Pass in expanded locus1 as well.
4786 (assign_discriminators): Avoid redundant location expansions.
4788 2018-06-05 Richard Biener <rguenther@suse.de>
4790 PR tree-optimization/86046
4791 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
4792 if required after clearing TREE_ADDRESSABLE.
4794 2018-06-05 Richard Biener <rguenther@suse.de>
4796 PR tree-optimization/86047
4797 * tree-ssa-loop.c (for_each_index): Glob handling of all
4798 decls and constants and really handle all of them.
4800 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4803 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
4804 qualifier_void_pointer and qualifier_const_void_pointer.
4805 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
4806 (arm_init_builtins): Handle the above.
4807 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
4808 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
4809 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
4812 2018-06-05 Martin Liska <mliska@suse.cz>
4814 * auto-profile.c (read_autofdo_file): Do not use
4815 gcov_ctr_summary struct.
4816 (afdo_callsite_hot_enough_for_early_inline): Likewise.
4817 * coverage.c (struct counts_entry): Likewise.
4818 (read_counts_file): Read just single summary entry.
4819 (get_coverage_counts): Use gcov_summary struct.
4820 * coverage.h (get_coverage_counts): Likewise.
4821 * gcov-dump.c (dump_working_sets): Likewise.
4822 (tag_summary): Dump just single summary.
4823 * gcov-io.c (gcov_write_summary): Write just histogram
4825 (gcov_read_summary): Read just single summary.
4826 (compute_working_sets): Use gcov_summary struct.
4827 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
4828 of GCOV_COUNTERS_SUMMABLE.
4829 (GCOV_COUNTERS_SUMMABLE): Remove.
4830 (GCOV_FIRST_VALUE_COUNTER): Replace with
4831 GCOV_COUNTER_V_INTERVAL.
4832 (struct gcov_ctr_summary): Remove.
4833 (struct gcov_summary): Directly use fields of former
4835 (compute_working_sets): Use gcov_summary struct.
4836 * gcov.c (read_count_file): Do not use ctrs fields.
4837 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
4839 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
4841 * profile.c: Likewise.
4842 * profile.h: Likewise.
4844 2018-06-05 Martin Liska <mliska@suse.cz>
4846 PR gcov-profile/84846
4847 * gcov.c (output_lines): Print working directory only
4848 in intermediate format.
4850 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
4852 * config/s390/s390-builtin-types.def: Add void function type.
4853 * config/s390/s390-builtins.def: Use the function type for the
4856 2018-06-04 Jim Wilson <jimw@sifive.com>
4858 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
4860 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
4861 and maybe_eh_return. Change regno to unsigned int. Use new args to
4862 handle EH_RETURN_DATA_REGNO registers properly.
4863 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
4864 (riscv_expand_epilogue): Update comment. Change argument name and
4865 type. Update code to use new name and type. Pass new args to
4866 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
4868 * config/riscv/riscv.md (NORMAL_RETURN): New.
4869 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
4870 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
4871 (eh_return): Call gen_eh_return_internal and emit barrier.
4872 (eh_return_internal): Call riscv_expand_epilogue.
4874 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
4876 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
4877 bit_insertion field and declare can_be_merged_into method.
4878 (merged_store_group::can_be_merged_into): New method.
4879 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
4880 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
4881 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
4883 2018-06-04 Richard Biener <rguenther@suse.de>
4885 PR tree-optimization/85955
4886 * builtins.c (fold_builtin_sincos): Convert pointers to
4887 destination to appropriate type before dereferencing.
4889 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
4891 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
4893 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
4895 * expr.c (expand_expr_real_1): Force the operand into memory if
4896 its TYPE_MODE is BLKmode and if there is no integer mode for
4897 the number of bits being extracted.
4899 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4903 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
4904 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
4906 2018-06-04 Richard Biener <rguenther@suse.de>
4908 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
4909 (cleanup_tree_cfg_noloop): ... single caller. Do
4910 start_recording_case_labels later.
4912 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
4914 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
4915 to _IMMINTRIN_H_INCLUDED.
4916 * config/i386/pconfigintrin.h: Ditto.
4917 * config/i386/waitpkgintrin.h: Ditto.
4918 * config/i386/immintrin.h: Add includes for sgxintrin.h,
4919 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
4920 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
4921 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
4922 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
4923 waitpkgintrin.h and cldemoteintrin.h.
4925 2018-06-04 Richard Biener <rguenther@suse.de>
4927 PR tree-optimization/86038
4928 * tracer.c (find_best_successor): Check probability for
4929 being initialized, bail out if not.
4931 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
4934 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
4935 of bits to ignore when comparing architectures.
4937 2018-06-04 Jakub Jelinek <jakub@redhat.com>
4939 PR tree-optimization/69615
4940 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
4941 maximum or minimum of the type, try to merge it also as if
4942 range1 is + [-, x - 1] or + [x + 1, -].
4945 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
4947 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
4949 PR tree-optimization/86034
4950 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
4951 the unsigned bitfield type in a bit insertion sequence if it does not
4952 have a larger precision than the bitfield size.
4953 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
4955 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
4957 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
4959 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4961 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
4962 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
4963 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
4964 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
4966 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
4968 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
4969 Disable -fdelete-null-pointer-checks for ELF toolchain.
4971 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
4972 Kito Cheng <kito.cheng@gmail.com>
4974 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
4975 (nds32le-*-*, nds32be-*-*): Integrate checking process.
4976 (nds32*-*-*): Add glibc and uclibc conditions.
4977 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
4978 (TARGET_EXCEPT_UNWIND_INFO): Define.
4979 * config/nds32/elf.h: New file.
4980 * config/nds32/linux.h: New file.
4981 * config/nds32/nds32-elf.opt: New file.
4982 * config/nds32/nds32-linux.opt: New file.
4983 * config/nds32/nds32-fp-as-gp.c
4984 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
4985 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
4987 (nds32_asm_file_end): Ditto.
4988 (nds32_print_operand): Ditto.
4989 (nds32_insert_attributes): Ditto.
4990 (nds32_init_libfuncs): New function.
4991 (TARGET_HAVE_TLS): Define.
4992 (TARGET_INIT_LIBFUNCS): Define.
4993 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
4995 (TARGET_ELF): Apply different mcmodel setting.
4996 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
4997 been migrated into elf.h and linux.h files.
4998 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
4999 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
5000 (mcmodel): The content has been migrated into nds32-elf.opt and
5001 nds32-linux.opt files.
5002 * config/nds32/t-elf: New file.
5003 * config/nds32/t-linux: New file.
5005 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
5006 Shiva Chen <shiva0217@gmail.com>
5008 * config/nds32/constants.md (unspec_volatile_element): Add
5009 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
5010 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
5012 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
5013 (make_pass_nds32_fp_as_gp): Declare.
5014 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
5016 (nds32_asm_function_end_prologue): Remove unused asm output.
5017 (nds32_asm_function_begin_epilogue): Remove unused asm output.
5018 (nds32_asm_file_start): Output necessary fp_as_gp information.
5019 (nds32_option_override): Adjust register usage.
5020 (nds32_expand_prologue): Consider fp_as_gp situation.
5021 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
5022 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
5025 (simple_return): Ditto.
5026 (omit_fp_begin): Output special directive for fp_as_gp.
5027 (omit_fp_end): Output special directive for fp_as_gp.
5028 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
5029 mforbid-fp-as-gp): New options.
5031 2018-06-01 Mark Wielaard <mark@klomp.org>
5033 * dwarf2out.c (dwarf2out_finish): Remove generation of
5034 DW_AT_loclists_base.
5036 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
5038 * gimple-ssa-store-merging.c: Include gimple-fold.h.
5039 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
5040 (struct merged_store_group): Add bit_insertion field.
5041 (dump_char_array): Use standard hexadecimal format.
5042 (merged_store_group::merged_store_group): Set bit_insertion to false.
5043 (merged_store_group::apply_stores): Use optimal buffer size. Deal
5044 with BIT_INSERT_EXPR stores. Move up code updating the mask and
5045 also print the mask in the dump file.
5046 (pass_store_merging::gate): Minor tweak.
5047 (imm_store_chain_info::coalesce_immediate): Fix wrong association
5048 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
5049 stores with INTEGER_CST stores.
5050 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
5051 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
5052 and use it throughout. Generate bit insertion sequences if need be.
5053 (pass_store_merging::process_store): Remove redundant condition.
5054 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
5056 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
5058 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
5059 the 128-bit floating point types. Fix function comment.
5061 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5063 * config/aarch64/aarch64-simd.md
5064 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
5066 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
5069 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5071 PR tree-optimization/85989
5072 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
5074 (backprop::intersect_uses): Check it when deciding whether this
5075 is a backedge reference.
5076 (backprop::process_block): Add each phi to m_visited_phis
5077 after visiting it, then clear it at the end.
5079 2018-06-01 Richard Biener <rguenther@suse.de>
5081 * tree-vectorizer.h (vect_dr_stmt): New function.
5082 (vect_get_load_cost): Adjust.
5083 (vect_get_store_cost): Likewise.
5084 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
5085 Use vect_dr_stmt instead of DR_SMTT.
5086 (vect_record_base_alignments): Likewise.
5087 (vect_calculate_target_alignment): Likewise.
5088 (vect_compute_data_ref_alignment): Likewise and make static.
5089 (vect_update_misalignment_for_peel): Likewise.
5090 (vect_verify_datarefs_alignment): Likewise.
5091 (vector_alignment_reachable_p): Likewise.
5092 (vect_get_data_access_cost): Likewise. Pass down
5093 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
5094 (vect_get_peeling_costs_all_drs): Likewise.
5095 (vect_peeling_hash_get_lowest_cost): Likewise.
5096 (vect_enhance_data_refs_alignment): Likewise.
5097 (vect_find_same_alignment_drs): Likewise.
5098 (vect_analyze_data_refs_alignment): Likewise.
5099 (vect_analyze_group_access_1): Likewise.
5100 (vect_analyze_group_access): Likewise.
5101 (vect_analyze_data_ref_access): Likewise.
5102 (vect_analyze_data_ref_accesses): Likewise.
5103 (vect_vfa_segment_size): Likewise.
5104 (vect_small_gap_p): Likewise.
5105 (vectorizable_with_step_bound_p): Likewise.
5106 (vect_prune_runtime_alias_test_list): Likewise.
5107 (vect_analyze_data_refs): Likewise.
5108 (vect_supportable_dr_alignment): Likewise.
5109 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5110 (vect_gen_prolog_loop_niters): Likewise.
5111 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5112 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
5114 (vect_recog_mask_conversion_pattern): Likewise.
5115 (vect_try_gather_scatter_pattern): Likewise.
5116 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
5117 to vect_get_store_cost.
5118 (vect_get_store_cost): Get stmt_info instead of DR.
5119 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
5120 (vect_get_load_cost): Get stmt_info instead of DR.
5122 2018-06-01 Richard Biener <rguenther@suse.de>
5125 * gimple-fold.c (var_decl_component_p): Also allow offsetted
5126 vars wrapped in MEM_REFs.
5128 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
5130 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5131 Fix subreg tests so that we only return a choice between
5132 GENERAL_REGS and FP_REGS if the original classes included both.
5134 2018-06-01 Richard Biener <rguenther@suse.de>
5137 * tree-ssa-structalias.c (get_function_part_constraint):
5138 Handle NULL fi->decl.
5139 (find_func_aliases_for_call): Properly handle indirect
5140 fi from direct call.
5141 (find_func_clobbers): Likewise.
5142 (ipa_pta_execute): Likewise.
5143 (create_variable_info_for): For functions that are ifunc_resolver
5144 resolve to a varinfo that contains the result of the resolver call.
5145 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
5148 2018-05-31 Michael Collison <michael.collison@arm.com>
5150 * config/aarch64/aarch64.md:
5151 (*fix_to_zero_extenddfdi2): New pattern.
5152 * gcc.target/aarch64/fix_extend1.c: New testcase.
5154 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
5158 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
5159 and BUILT_IN_STRNCMP_EQ.
5160 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
5161 BUILT_IN_STRNCMP_EQ.
5162 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
5163 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5164 (gimple_fold_builtin): Likewise.
5165 * tree-ssa-strlen.c (compute_string_length): New function.
5166 (determine_min_obsize): New function.
5167 (handle_builtin_string_cmp): New function to handle calls to
5168 string compare functions.
5169 (strlen_optimize_stmt): Add handling to builtin string compare
5171 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
5172 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
5173 * tree.c (build_common_builtin_nodes): Add new defines of
5174 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
5176 2018-05-31 Jakub Jelinek <jakub@redhat.com>
5179 * bb-reorder.c (pass_partition_blocks::gate): Return false for
5180 functions with naked attribute.
5182 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
5184 * config/i386/sse.md (avx_vec_concat<mode>):
5185 Substitute concat_tg_mode mode attribute with xtg_mode.
5186 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
5187 (concat_tg_mode): Remove mode attribute.
5189 2018-05-31 Martin Sebor <msebor@redhat.com>
5192 * calls.c (alloc_max_size): Correct a logic error/typo.
5193 Treat excessive arguments as infinite. Warn for invalid arguments.
5194 * doc/invoke.texi (-Walloc-size-larger-than): Update.
5196 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
5199 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
5200 and movx for Haswell.
5202 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
5203 Cesar Philippidis <cesar@codesourcery.com>
5206 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
5207 when emitting error on private/firstprivate reductions.
5208 * omp-low.c (lower_omp_target): Avoid reference-type processing
5209 on pointers for firstprivate clause.
5211 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5213 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
5216 * config/aarch64/aarch64-simd.md
5217 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
5218 (aarch64_ld1_x3_<mode>): Likewise
5219 (aarch64_st1x2<VALLDIF:mode>): Likewise
5220 (aarch64_st1_x2_<mode>): Likewise
5221 (aarch64_st1x3<VALLDIF:mode>): Likewise
5222 (aarch64_st1_x3_<mode>): Likewise
5223 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
5224 (vld1_s8_x3): Likewise.
5225 (vld1_u16_x3): Likewise.
5226 (vld1_s16_x3): Likewise.
5227 (vld1_u32_x3): Likewise.
5228 (vld1_s32_x3): Likewise.
5229 (vld1_u64_x3): Likewise.
5230 (vld1_s64_x3): Likewise.
5231 (vld1_f16_x3): Likewise.
5232 (vld1_f32_x3): Likewise.
5233 (vld1_f64_x3): Likewise.
5234 (vld1_p8_x3): Likewise.
5235 (vld1_p16_x3): Likewise.
5236 (vld1_p64_x3): Likewise.
5237 (vld1q_u8_x3): Likewise.
5238 (vld1q_s8_x3): Likewise.
5239 (vld1q_u16_x3): Likewise.
5240 (vld1q_s16_x3): Likewise.
5241 (vld1q_u32_x3): Likewise.
5242 (vld1q_s32_x3): Likewise.
5243 (vld1q_u64_x3): Likewise.
5244 (vld1q_s64_x3): Likewise.
5245 (vld1q_f16_x3): Likewise.
5246 (vld1q_f32_x3): Likewise.
5247 (vld1q_f64_x3): Likewise.
5248 (vld1q_p8_x3): Likewise.
5249 (vld1q_p16_x3): Likewise.
5250 (vld1q_p64_x3): Likewise.
5251 (vst1_s64_x2): Likewise.
5252 (vst1_u64_x2): Likewise.
5253 (vst1_f64_x2): Likewise.
5254 (vst1_s8_x2): Likewise.
5255 (vst1_p8_x2): Likewise.
5256 (vst1_s16_x2): Likewise.
5257 (vst1_p16_x2): Likewise.
5258 (vst1_s32_x2): Likewise.
5259 (vst1_u8_x2): Likewise.
5260 (vst1_u16_x2): Likewise.
5261 (vst1_u32_x2): Likewise.
5262 (vst1_f16_x2): Likewise.
5263 (vst1_f32_x2): Likewise.
5264 (vst1_p64_x2): Likewise.
5265 (vst1q_s8_x2): Likewise.
5266 (vst1q_p8_x2): Likewise.
5267 (vst1q_s16_x2): Likewise.
5268 (vst1q_p16_x2): Likewise.
5269 (vst1q_s32_x2): Likewise.
5270 (vst1q_s64_x2): Likewise.
5271 (vst1q_u8_x2): Likewise.
5272 (vst1q_u16_x2): Likewise.
5273 (vst1q_u32_x2): Likewise.
5274 (vst1q_u64_x2): Likewise.
5275 (vst1q_f16_x2): Likewise.
5276 (vst1q_f32_x2): Likewise.
5277 (vst1q_f64_x2): Likewise.
5278 (vst1q_p64_x2): Likewise.
5279 (vst1_s64_x3): Likewise.
5280 (vst1_u64_x3): Likewise.
5281 (vst1_f64_x3): Likewise.
5282 (vst1_s8_x3): Likewise.
5283 (vst1_p8_x3): Likewise.
5284 (vst1_s16_x3): Likewise.
5285 (vst1_p16_x3): Likewise.
5286 (vst1_s32_x3): Likewise.
5287 (vst1_u8_x3): Likewise.
5288 (vst1_u16_x3): Likewise.
5289 (vst1_u32_x3): Likewise.
5290 (vst1_f16_x3): Likewise.
5291 (vst1_f32_x3): Likewise.
5292 (vst1_p64_x3): Likewise.
5293 (vst1q_s8_x3): Likewise.
5294 (vst1q_p8_x3): Likewise.
5295 (vst1q_s16_x3): Likewise.
5296 (vst1q_p16_x3): Likewise.
5297 (vst1q_s32_x3): Likewise.
5298 (vst1q_s64_x3): Likewise.
5299 (vst1q_u8_x3): Likewise.
5300 (vst1q_u16_x3): Likewise.
5301 (vst1q_u32_x3): Likewise.
5302 (vst1q_u64_x3): Likewise.
5303 (vst1q_f16_x3): Likewise.
5304 (vst1q_f32_x3): Likewise.
5305 (vst1q_f64_x3): Likewise.
5306 (vst1q_p64_x3): Likewise.
5308 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5310 * config/msp430/msp430.c (msp430_output_labelref): Prepend
5311 user_label_prefix to name.
5313 * tree-core.h: Update comment about the format of NAME string
5314 passed to handler in attribute_spec.
5316 * config/msp430/msp430.md: Remove erroneous subreg expression from
5317 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
5318 zero_extend{q,h}isi2.
5320 2018-05-30 Borislav Petkov <bp@suse.de>
5322 * doc/extend.texi: Document some architecture specific
5323 constraints and sort entries.
5325 2018-05-30 Martin Sebor <msebor@redhat.com>
5328 * builtins.c (expand_builtin_stpcpy_1): New function.
5329 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
5330 only if the former succeeds.
5332 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
5334 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
5337 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5339 * doc/invoke.texi (-flinker-output): Document
5341 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5343 * passes.c (ipa_write_summaries): Only modify statements if body
5345 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
5346 incrementally linking.
5347 (ipa_passes): Likewise.
5348 * lto-cgraph.c (lto_output_node): When incrementally linking do not
5349 pass down resolution info.
5350 * common.opt (flag_incremental_link): Update info.
5351 * gcc.c (plugin specs): Turn flinker-output=* to
5352 -plugin-opt=-linker-output-known
5353 * toplev.c (compile_file): Also cut compilation when doing incremental
5355 * flag-types. (enum lto_partition_model): Add
5356 LTO_LINKER_OUTPUT_NOLTOREL.
5357 (invoke.texi): Add -flinker-output docs.
5358 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
5359 link same way as WPA; do not stream in dead initializers.
5361 * dwarf2out.c (dwarf2out_die_ref_for_decl,
5362 darf2out_register_external_decl): Support incremental link.
5364 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5366 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
5368 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
5370 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
5371 it down to simple_object_copy_lto_debug_sections.
5372 (run_gcc): Determine incremental LTO link time and configure
5373 lto1 into non-wpa mode, disable renaming of debug sections.
5375 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
5377 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
5378 descriptions of various incorrectly documented functions.
5380 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
5383 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
5384 address check not strict.
5386 2018-05-30 Richard Biener <rguenther@suse.de>
5388 PR tree-optimization/85964
5389 * tracer.c (better_p): Drop initialized count check, we only
5390 call the function with initialized counts now.
5391 (find_best_successor): Do find a best edge if one
5392 has uninitialized count.
5393 (find_best_predecessor): Likewise. Do BB frequency check only
5394 if count is initialized.
5396 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
5398 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
5399 (aarch64_ldrstr_offset_compare): New.
5400 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
5401 load/store orderings.
5402 (aarch64_gen_adjusted_ldpstp): Likewise.
5404 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
5406 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
5407 Check for subset of GENERAL_REGS and FP_REGS.
5408 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
5411 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
5413 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
5414 and wi::to_poly_offset. Add the current offset and then check
5415 whether the sum fits, rather than using an unchecked addition of
5416 a checked term. Check for a shwi rather than a uhwi.
5417 * expr.c (get_bit_range): Use tree_to_poly_uint64.
5418 (store_constructor): Use poly_int_tree_p.
5419 (expand_expr_real_1): Likewise.
5420 * function.c (assign_temp): Likewise.
5421 * fold-const.c (const_binop): Use poly_int_tree_p and
5423 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
5425 * ipa-icf-gimple.c (func_checker::compare_operand): Use
5426 to_poly_offset for MEM offsets.
5427 * ipa-icf.c (sem_variable::equals): Likewise.
5428 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
5429 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
5430 wi::to_poly_offset for BIT_FIELD_REF offsets.
5431 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
5433 * var-tracking.c (emit_note_insn_var_location): Use
5434 tree_to_poly_uint64.
5436 2018-05-29 Jim Wilson <jimw@sifive.com>
5438 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
5440 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
5443 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
5444 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
5446 (sse4_1_round<mode>2): Use nonimmediate_operand
5447 for operand 1 predicate.
5449 2018-05-29 Martin Sebor <msebor@redhat.com>
5450 Richard Biener <rguenther@suse.de>
5453 * calls.c (get_size_range): Call determine_value_range instead
5454 of get_value_range..
5455 * tree-vrp.h (determine_value_range): Declared new function.
5456 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
5458 2018-05-29 Richard Biener <rguenther@suse.de>
5460 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
5461 sure to use non-pattern stmts for get_earlier_stmt arguments.
5462 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
5463 called on pattern stmts.
5464 (get_later_stmt): Likewise.
5466 2018-05-29 Martin Liska <mliska@suse.cz>
5468 PR gcov-profile/85759
5469 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
5472 2018-05-29 Jakub Jelinek <jakub@redhat.com>
5474 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
5476 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
5480 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5481 VEC_PACK_FLOAT_EXPR): New tree codes.
5482 * tree-pretty-print.c (op_code_prio): Handle
5483 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5484 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5485 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5486 * tree-inline.c (estimate_operator_cost): Likewise.
5487 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
5488 * fold-const.c (const_binop): Likewise.
5489 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
5490 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
5491 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
5492 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
5493 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
5494 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
5495 * expr.c (expand_expr_real_2): Likewise.
5496 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
5497 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
5498 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
5500 * optabs.c (expand_widen_pattern_expr): For
5501 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
5502 sign from result type rather than operand's type.
5503 (expand_binop_directly): For vec_packu_float_optab and
5504 vec_packs_float_optab allow result type to be different from operand's
5506 * optabs-tree.c (optab_for_tree_code): Handle
5507 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5508 VEC_PACK_FLOAT_EXPR. Formatting fixes.
5509 * tree-vect-generic.c (expand_vector_operations_1): Handle
5510 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
5511 VEC_PACK_FLOAT_EXPR.
5512 * tree-vect-stmts.c (supportable_widening_operation): Handle
5514 (supportable_narrowing_operation): Handle FLOAT_EXPR.
5515 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
5516 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
5517 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
5518 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
5520 (vec_pack<floatprefix>_float_<mode>): New expander.
5521 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
5523 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
5524 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
5525 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
5526 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
5527 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
5529 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
5530 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
5531 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
5532 VEC_PACK_FLOAT_EXPR): Document.
5534 2018-05-29 Richard Biener <rguenther@suse.de>
5536 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
5538 (stmt_vec_info_vec): Make pointer.
5539 (init_stmt_vec_info_vec): Remove.
5540 (free_stmt_vec_info_vec): Likewise.
5541 (set_stmt_vec_info_vec): New function.
5542 (free_stmt_vec_infos): Likewise.
5543 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
5544 (set_vinfo_for_stmt): Likewise.
5545 (get_earlier_stmt): Likewise.
5546 (get_later_stmt): Likewise.
5547 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
5548 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
5549 (vec_info::~vec_info): Free stmt_vec_infos.
5550 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
5551 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
5552 (pass_slp_vectorize::execute): Likewise.
5553 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
5554 (free_stmt_vec_info_vec): Likewise.
5555 (set_stmt_vec_info_vec): New function.
5556 (free_stmt_vec_infos): Likewise.
5557 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
5558 the global stmt_vec_info_vec.
5559 * tree-parloops.c (gather_scalar_reductions): Use
5560 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
5563 2018-05-29 Richard Biener <rguenther@suse.de>
5565 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
5567 2018-05-29 Martin Liska <mliska@suse.cz>
5568 David Malcolm <dmalcolm@redhat.com>
5570 * vec.c (test_reverse): New.
5571 (vec_c_tests): Add new test.
5572 * vec.h (vl_ptr>::reverse): New function.
5574 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
5576 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
5578 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
5581 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
5583 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
5585 2018-05-28 Richard Biener <rguenther@suse.de>
5587 PR tree-optimization/85933
5588 * tree-vect-data-refs.c (vect_record_base_alignments): Only
5589 look at stmts marked as vectorizable.
5591 2018-05-28 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/85934
5594 * tree-vect-generic.c (expand_vector_operations_1): Hoist
5595 vector boolean check before scalar optimization.
5597 2018-05-28 Jakub Jelinek <jakub@redhat.com>
5599 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
5602 2018-05-28 Mark Wielaard <mark@klomp.org>
5604 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
5605 if it is an expression containing a minus sign.
5607 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
5609 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
5611 2018-05-27 Paul Koning <ni1d@arrl.net>
5613 * config/pdp11/pdp11.md (truncsihi2): Remove.
5615 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
5616 Chung-Ju Wu <jasonwucj@gmail.com>
5618 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
5620 (unaligned_store_dw): Ditto.
5621 * config/nds32/nds32-memory-manipulation.c
5622 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
5623 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
5624 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
5625 (emit_setmem_word_loop): Rename to ...
5626 (emit_setmem_doubleword_loop): ... this.
5627 (nds32_gen_dup_4_byte_to_word_value): New function.
5628 (nds32_gen_dup_8_byte_to_double_word_value): New function.
5629 (nds32_expand_setmem_loop): Refine implementation.
5630 (nds32_expand_setmem_loop_v3m): Ditto.
5631 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
5634 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5636 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
5638 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
5640 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
5641 (nds32_init_machine_status): Initialize machine->attr_naked_p and
5642 machine->attr_no_prologue_p.
5643 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
5644 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
5645 (nds32_expand_epilogue): Consider attr_naked_p.
5646 (nds32_expand_epilogue_v3pop): Likewise.
5647 (nds32_can_use_return_insn): Likewise.
5648 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
5649 attr_no_prologue_p fields.
5650 * config/nds32/nds32.opt (mret-in-naked-func): New option.
5652 2018-05-27 Jakub Jelinek <jakub@redhat.com>
5655 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
5657 * config/i386/sse.md
5658 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
5660 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
5662 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
5664 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
5666 (*<floatsuffix>floatv2div2sf2): Rename to ...
5667 (*float<floatunssuffix>v2div2sf2): ... this.
5668 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
5669 (float<floatunssuffix>v2div2sf2_mask): ... this.
5670 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
5671 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
5672 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
5674 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
5676 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5678 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
5680 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5682 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
5684 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
5685 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
5686 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
5687 gen_ufix_truncv8dfv8si2.
5688 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
5689 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
5690 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
5691 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
5692 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
5693 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
5694 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
5695 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
5697 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5701 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
5703 2018-05-25 Jim Wilson <jimw@sifive.com>
5705 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
5706 * config/riscv/riscv.c (struct machine_function): Add
5707 interrupt_handler_p and attribute_checked_p fields.
5708 (riscv_attribute_table): Add interrupt.
5709 (riscv_interrupt_type_p): New.
5710 (riscv_save_reg_p): Save extra regs for interrupt handler.
5711 (riscv_use_save_libcall): Return false for interrupt handler.
5712 (riscv_first_stack_step): Add forward declaration.
5713 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
5714 for interrupt handler with large frame. Use it for saved reg list.
5715 (riscv_expand_prologue): Move flag_stack_usage_info support to
5716 eliminate duplication.
5717 (riscv_expand_epilogue): Generate mret for interrupt handler.
5718 (riscv_epilogue_uses): New.
5719 (riscv_can_use_return_insn): Return false for interrupt handler.
5720 (riscv_function_ok_for_sibcall): Likewise.
5721 (riscv_set_current_function): Add interrupt handler support.
5722 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
5723 * config/riscv/riscv.md (UNSPECV_MRET): New.
5725 (riscv_frflags, riscv_fsflags): Use tab after opcode.
5727 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
5729 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
5731 PR tree-optimization/85712
5732 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
5733 this candidate has already been replaced in-situ by a copy.
5735 2018-05-25 Jason Merrill <jason@redhat.com>
5737 PR c++/80485 - inline function non-zero address.
5738 * symtab.c (nonzero_address): Check DECL_COMDAT.
5740 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
5743 * config/alpha/alpha.md (ashlsi3): New insn pattern.
5744 (*ashlsi_se): Rename from *ashldi_se. Define as sign
5745 extension of SImode operation. Use const123_operand predicate.
5746 (*saddsi_1): Remove.
5747 (*saddl_se_1): Ditto.
5749 (*ssubl_se_1): Ditto.
5750 * config/alpha/predicates.md (const123_operand): New predicate.
5751 * config/alpha/constraints.md (P): Use IN_RANGE.
5753 2018-05-25 Richard Biener <rguenther@suse.de>
5755 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
5757 (ref_maybe_used_by_stmt_p): Likewise.
5758 (stmt_may_clobber_ref_p): Likewise.
5759 (stmt_may_clobber_ref_p_1): Likewise.
5760 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
5762 (ref_maybe_used_by_stmt_p): Likewise.
5763 (stmt_may_clobber_ref_p): Likewise.
5764 (stmt_may_clobber_ref_p_1): Likewise.
5765 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
5766 the alias oracle to disambiguate DRs with stmts DR analysis
5768 (vect_analyze_data_refs): Do not give up on not analyzable
5769 DRs for BB vectorization. Remove code truncating the dataref
5772 2018-05-25 Jakub Jelinek <jakub@redhat.com>
5775 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
5776 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
5777 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
5779 2018-05-25 Richard Biener <rguenther@suse.de>
5781 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
5782 function, combining stmt data ref gathering and fatal analysis
5784 (vect_analyze_data_refs): Remove now redudnant code and simplify.
5785 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
5786 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
5787 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
5788 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
5790 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5792 PR tree-optimization/85720
5793 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
5794 SCC if all partitions are builtins.
5795 (version_loop_by_alias_check): New parameter. Generate cancelable
5796 runtime alias check if all partitions are builtins.
5797 (distribute_loop): Update call to above function.
5799 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5801 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
5802 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
5803 (parm_default_def_partition_arg): Ditto.
5804 (set_parm_default_def_partition): Ditto.
5805 (get_parm_default_def_partitions): Ditto and make it static.
5806 (get_undefined_value_partitions): Ditto and make it static.
5807 (remove_ssa_form): Refactor call to init_var_map here.
5808 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
5809 computation for loop region.
5810 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
5811 (register_default_def): Delete.
5812 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
5813 (parm_default_def_partition_arg): Ditto.
5814 (set_parm_default_def_partition): Ditto.
5815 (get_parm_default_def_partitions): Ditto and make it static.
5816 (get_undefined_value_partitions): Ditto and make it static.
5817 (coalesce_with_default, coalesce_with_default): Update comment.
5818 (create_coalesce_list_for_region): New func factored out from
5819 create_outofssa_var_map.
5820 (populate_coalesce_list_for_outofssa): New func factored out from
5821 create_outofssa_var_map and coalesce_ssa_name.
5822 (create_outofssa_var_map): Delete.
5823 (coalesce_ssa_name): Refactor to support live range computation.
5824 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
5825 (get_parm_default_def_partitions): Delete.
5826 (get_undefined_value_partitions): Ditto.
5827 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
5828 computation for loop region.
5829 (new_tree_live_info, loe_visit_block): Ditto.
5830 (live_worklist, set_var_live_on_entry): Ditto.
5831 (calculate_live_on_exit, verify_live_on_entry): Ditto.
5832 * tree-ssa-live.h (struct _var_map): New fields.
5833 (init_var_map): Change decl.
5834 (region_contains_p): New.
5836 2018-05-25 Bin Cheng <bin.cheng@arm.com>
5838 * tree-ssa-live.h (live_merge_and_clear): Delete.
5840 2018-05-25 Richard Biener <rguenther@suse.de>
5843 * tree-dump.c (dequeue_and_dump): Remove access to removed
5844 operand 2 of a SWITCH_EXPR.
5846 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5848 * doc/sourcebuild.texi (vect_double_cond_arith): Include
5849 multiplication and division.
5850 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
5851 (cond_udiv@var{m}, cond_umod@var{m}): Document.
5852 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
5853 (cond_udiv_optab, cond_umod_optab): New optabs.
5854 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
5855 (IFN_COND_RDIV): New internal functions.
5856 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
5857 TRUNC_MOD_EXPR and RDIV_EXPR.
5858 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
5859 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
5861 (SVE_INT_BINARY): Include mult.
5862 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
5863 (optab, sve_int_op): Handle mult.
5864 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
5866 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
5867 for SVE_INT_BINARY_SD.
5869 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5871 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
5872 (optab, sve_int_op): Handle div and udiv.
5873 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
5874 for SVE_INT_BINARY_SD.
5875 (*<optab><mode>3): New insn for the same.
5877 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5879 * tree-vect-patterns.c: Include predict.h.
5880 (vect_recog_divmod_pattern): Restrict check for division support
5881 to when optimizing for size.
5883 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5885 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
5886 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
5887 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
5888 (gimple_match_op::set_op): Likewise.
5889 (gimple_resimplify4): Declare.
5890 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
5891 (expr::gen_transform): Likewise.
5892 (decision_tree::gen): Generate a simplification routine for 4 operands.
5893 * gimple-match-head.c (gimple_simplify): Add an overload for
5894 4 operands. In the top-level function, handle up to 4 call
5895 arguments and call gimple_resimplify4.
5896 (gimple_resimplify4): New function.
5897 (build_call_internal): Pass a fourth operand.
5898 (maybe_push_to_seq): Likewise.
5899 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
5900 Fold VEC_COND_EXPRs of an operation and a default value into
5901 an IFN_COND_* function if possible.
5902 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
5904 (SVE_COND_FP_BINARY): Include them.
5905 (optab, sve_fp_op): Handle them.
5906 (SVE_INT_BINARY_REV): New code iterator.
5907 (SVE_COND_FP_BINARY_REV): New int iterator.
5908 (commutative): New int attribute.
5909 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
5911 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
5913 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
5914 (*cond_<optab><mode>): New patterns for reversed operands.
5916 2018-05-25 Richard Biener <rguenther@suse.de>
5918 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
5919 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
5920 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
5921 (STMT_VINFO_GROUPED_ACCESS): Adjust.
5922 * tree-vect-data-refs.c (everywhere): Adjust users.
5923 * tree-vect-loop.c (everywhere): Likewise.
5924 * tree-vect-slp.c (everywhere): Likewise.
5925 * tree-vect-stmts.c (everywhere): Likewise.
5926 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5928 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5930 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
5932 (gcc_cv_as_section_exclude): ... this.
5933 Try Solaris as #exclude syntax.
5934 * configure: Regenerate.
5935 * config.in: Regenerate.
5936 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
5938 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
5939 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
5941 * varasm.c (default_elf_asm_named_section): Don't check if
5942 HAVE_GAS_SECTION_EXCLUDE is defined.
5944 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5946 * doc/md.texi: Update the documentation of the cond_* optabs
5947 to mention the new final operand. Fix GET_MODE_NUNITS call.
5948 Describe the scalar case too.
5949 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
5950 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
5952 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
5953 (get_conditional_internal_fn): Update comment.
5954 * tree-vect-loop.c (vectorizable_reduction): Pass the original
5955 accumulator value as a final argument to conditional functions.
5956 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
5957 a define_expand and add an "else" operand. Assert for now that
5958 the else operand is equal to operand 2. Use SVE_INT_BINARY and
5959 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
5960 (*cond_<optab><mode>): New patterns.
5961 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
5962 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
5963 (UNSPEC_COND_EOR): Delete.
5964 (optab): Remove associated mappings.
5965 (SVE_INT_BINARY): New code iterator.
5966 (sve_int_op): Remove int attribute and add "minus" to the code
5968 (SVE_COND_INT_OP): Delete.
5969 (SVE_COND_FP_OP): Rename to...
5970 (SVE_COND_FP_BINARY): ...this.
5972 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
5974 * optabs.c (can_reuse_operands_p): New function.
5975 (maybe_legitimize_operands): Try to reuse the results for
5978 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
5980 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
5981 Add {q} suffix to insn mnemonic.
5983 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5985 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
5986 (msp430_warn_func_return): New.
5988 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
5990 * fold-const.c (tree_nonzero_bits): New function.
5991 * fold-const.h (tree_nonzero_bits): Likewise.
5992 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
5993 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
5995 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
5999 * varasm.c (assemble_alias): Check ifunc_resolver only on
6002 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
6005 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
6006 when memory input operand is handled.
6008 2018-05-24 Luis Machado <luis.machado@linaro.org>
6010 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
6012 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
6014 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6016 * match.pd: Delay FMA folds until after vectorization.
6018 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
6021 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6022 address check not strict.
6024 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
6026 * gimple-match.h (gimple_match_op): New class.
6027 (mprts_hook): Replace parameters with a gimple_match_op *.
6028 (maybe_build_generic_op): Likewise.
6029 (gimple_simplified_result_is_gimple_val): Replace parameters with
6030 a const gimple_match_op *.
6031 (gimple_simplify): Replace code_helper * and tree * parameters with
6032 a gimple_match_op * parameter.
6033 (gimple_resimplify1): Replace code_helper *, tree and tree *
6034 parameters with a gimple_match_op * parameter.
6035 (gimple_resimplify2): Likewise.
6036 (gimple_resimplify3): Likewise.
6037 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
6038 parameters with a gimple_match_op * parameter.
6039 * gimple-match-head.c (gimple_simplify): Change prototypes of
6040 auto-generated functions to take a gimple_match_op * instead of
6041 separate code_helper * and tree * parameters. Make the same
6042 change in the top-level overload and update calls to the
6043 gimple_resimplify routines. Update calls to the auto-generated
6044 functions and to maybe_push_res_to_seq in the publicly-facing
6045 operation-specific gimple_simplify overloads.
6046 (gimple_match_op::MAX_NUM_OPS): Define.
6047 (gimple_resimplify1): Replace rcode and ops with a single res_op
6048 parameter. Update call to gimple_simplify.
6049 (gimple_resimplify2): Likewise.
6050 (gimple_resimplify3): Likewise.
6051 (mprts_hook): Replace parameters with a gimple_match_op *.
6052 (maybe_build_generic_op): Likewise.
6053 (build_call_internal): Replace type, nargs and ops with
6054 a gimple_match_op *.
6055 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
6056 with a single gimple_match_op *. Update calls to mprts_hook,
6057 build_call_internal and gimple_simplified_result_is_gimple_val.
6058 Factor out code that is common to the tree_code and combined_fn cases.
6059 * genmatch.c (expr::gen_transform): Replace tem_code and
6060 tem_ops with a gimple_match_op called tem_op. Update calls
6061 to the gimple_resimplify functions and maybe_push_res_to_seq.
6062 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
6063 res_ops. Update call to the gimple_resimplify functions.
6064 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
6065 (decision_tree::gen): Make the functions take a gimple_match_op *
6066 called res_op instead of separate res_code and res_ops parameters.
6067 Update call accordingly.
6068 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
6069 and ops with a single res_op parameter. Update calls to
6070 maybe_build_generic_op and maybe_push_res_to_seq.
6071 (fold_stmt_1): Update calls to gimple_simplify and
6072 replace_stmt_with_simplification.
6073 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
6074 and gimple_simplified_result_is_gimple_val.
6075 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
6077 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
6078 with a gimple_match_op *.
6079 (vn_nary_build_or_lookup): Likewise. Update call to
6080 vn_nary_build_or_lookup_1.
6081 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
6082 gimple_match_op *. Update calls to the gimple_resimplify routines
6083 and to gimple_simplified_result_is_gimple_val.
6084 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
6085 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
6086 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
6087 (visit_nary_op): Likewise.
6088 (visit_reference_op_load): Likewise.
6090 2018-05-23 Luis Machado <luis.machado@linaro.org>
6092 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
6093 modifier for printing the step amount.
6095 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
6098 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
6101 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
6103 * doc/sourcebuild.texi (Endianness): New subsubsection.
6105 2018-05-23 Luis Machado <luis.machado@linaro.org>
6107 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6108 <prefetch_dynamic_strides>: New const bool field.
6109 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6110 prefetch_dynamic_strides.
6111 (exynosm1_prefetch_tune): Likewise.
6112 (thunderxt88_prefetch_tune): Likewise.
6113 (thunderx_prefetch_tune): Likewise.
6114 (thunderx2t99_prefetch_tune): Likewise.
6115 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
6117 (aarch64_override_options_internal): Update to set
6118 PARAM_PREFETCH_DYNAMIC_STRIDES.
6119 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
6120 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
6121 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
6122 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
6123 prefetch-dynamic-strides setting.
6125 2018-05-23 Luis Machado <luis.machado@linaro.org>
6127 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
6128 <minimum_stride>: New const int field.
6129 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
6130 minimum_stride field defaulting to -1.
6131 (exynosm1_prefetch_tune): Likewise.
6132 (thunderxt88_prefetch_tune): Likewise.
6133 (thunderx_prefetch_tune): Likewise.
6134 (thunderx2t99_prefetch_tune): Likewise.
6135 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
6136 <default_opt_level>: Set to 3.
6137 (aarch64_override_options_internal): Update to set
6138 PARAM_PREFETCH_MINIMUM_STRIDE.
6139 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
6140 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
6141 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
6142 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
6143 stride is constant and is below the minimum stride threshold.
6145 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6147 * config/arm/arm-cpus.in (mode26): Delete.
6148 (armv4): Delete mode26 reference.
6149 * config/arm/arm.c (arm_configure_build_target): Delete use of
6152 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6154 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
6156 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
6157 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
6159 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
6160 (floatunsdidf2): Ditto.
6162 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
6164 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
6165 (fixuns_trunc<mode>si2_avx512f): Ditto.
6166 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
6167 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
6168 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
6170 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6172 PR rtl-optimization/79985
6173 * df-scan.c (df_insn_refs_collect): Remove special case for
6174 global registers and asm statements.
6176 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
6178 * extend.texi (Global Register Variables): Rewrite the bullet list.
6179 Note that the register is available for allocation. Note that access
6180 via inline asm must use constraints. Add note about async-signal
6181 handlers. Remove paragraph about automagic register selection.
6183 2018-05-23 Richard Biener <rguenther@suse.de>
6185 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
6186 of fixed offset from memset VN.
6188 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
6190 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
6192 (alloc_cand_and_find_basis): Initialize first_interp field.
6193 (slsr_process_mul): Modify first_interp field.
6194 (slsr_process_add): Likewise.
6195 (slsr_process_cast): Modify first_interp field for each new
6197 (slsr_process_copy): Likewise.
6198 (dump_candidate): Dump first_interp field.
6199 (replace_mult_candidate): Process all interpretations, not just
6201 (replace_rhs_if_not_dup): Likewise.
6202 (replace_one_candidate): Likewise.
6204 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
6206 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
6208 (aarch64_needs_frame_chain): New function.
6209 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
6211 2018-05-23 Sudakshina Das <sudi.das@arm.com>
6214 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
6215 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
6216 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
6217 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
6218 as true for strict-align.
6219 (aarch64_can_inline_p): Perform checks even when callee has no
6220 attributes to check for strict alignment.
6221 * doc/extend.texi (AArch64 Function Attributes): Document
6223 * doc/invoke.texi: (AArch64 Options): Likewise.
6225 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
6227 PR tree-optimization/85853
6228 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
6229 the handling of the root of the node to...
6230 (vect_slp_analyze_node_operations_1): ...this new function,
6231 and run the whole thing with the child nodes' def types
6232 set according to their SLP node's def type.
6234 2018-05-23 Richard Biener <rguenther@suse.de>
6237 * tree-data-ref.c (create_runtime_alias_checks): Defer
6238 and ignore overflow warnings.
6240 2018-05-23 Yury Gribov <tetra2005@gmail.com>
6242 PR tree-optimization/85822
6243 * tree-vrp.c (is_masked_range_test): Fix handling of negative
6246 2018-05-23 Richard Biener <rguenther@suse.de>
6248 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
6249 memset constants via native_interpret_expr.
6251 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
6254 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
6256 (cgraph_node::create_alias): Likewise.
6257 (cgraph_node::get_availability): Check ifunc_resolver instead
6258 of looking up ifunc attribute.
6259 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
6260 * varasm.c (do_assemble_alias): Likewise.
6261 (assemble_alias): Likewise.
6262 (default_binds_local_p_3): Likewise.
6263 * cgraph.h (cgraph_node): Add ifunc_resolver.
6264 (cgraph_node::only_called_directly_or_aliased_p): Return false
6266 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
6268 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
6269 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
6270 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
6271 instead of looking up ifunc attribute.
6273 2018-05-22 Luis Machado <luis.machado@linaro.org>
6275 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
6277 2018-05-22 Martin Sebor <msebor@redhat.com>
6280 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
6281 only when expasion succeeds.
6282 (expand_builtin_strcmp): Same.
6283 (expand_builtin_strncmp): Same.
6285 2018-05-22 Martin Sebor <msebor@redhat.com>
6287 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
6289 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6290 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6292 * config/aarch64/aarch64-ldpstp.md: Replace uses of
6293 aarch64_mem_pair_operand with memory_operand and delete operand swapping
6295 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6296 Add check for legitimate_address.
6297 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
6298 (aarch64_swap_ldrstr_operands): New.
6299 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
6302 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
6303 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6305 * config/aarch64/aarch64.md: New patterns to generate stp
6307 (store_pair_sw, store_pair_dw): New patterns to generate stp for
6308 single words and double words.
6309 (load_pair_sw, load_pair_dw): Likewise.
6310 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
6312 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
6314 * config/aarch64/aarch64-ldpstp.md: Modify peephole
6315 for different mode ldpstp and add peephole for merged zero stores.
6317 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
6319 (aarch64_gen_store_pair): Rename calls to match new patterns.
6320 (aarch64_gen_load_pair): Rename calls to match new patterns.
6321 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
6322 (load_pair<DREG:mode><DREG2:mode>): ... This.
6323 (store_pair<mode>): Rename to...
6324 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
6325 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
6327 (V_INT_EQUIV): Handle SImode.
6328 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
6331 2018-05-22 Martin Sebor <msebor@redhat.com>
6334 * calls.c (maybe_warn_nonstring_arg): Use string length to set
6335 or ajust the presumed bound on an operation to avoid unnecessary
6338 2018-05-22 Martin Sebor <msebor@redhat.com>
6340 PR tree-optimization/85826
6341 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
6342 assuming that a DECL necesarily has a constant size.
6344 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
6347 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
6349 2018-05-22 Richard Biener <rguenther@suse.de>
6351 PR tree-optimization/85834
6352 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
6353 non-constant and non-zero memset arguments.
6355 2018-05-22 Martin Liska <mliska@suse.cz>
6358 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
6360 2018-05-22 Richard Biener <rguenther@suse.de>
6362 PR tree-optimization/85863
6363 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
6364 comparisons when vectype is specified.
6365 (vectorizable_condition): Do not specify vectype for
6366 vect_is_simple_cond when SLP vectorizing.
6368 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
6371 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
6372 define __ibm128 as long double.
6373 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
6374 as a distinct type when IEEE 128-bit support is enabled.
6375 (init_float128_ieee): Fix up conversions between IFmode and IEEE
6376 128-bit types to use the correct functions.
6377 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
6378 convert between 128-bit floating point types that have different
6379 modes but the same representation, instead of using gen_lowpart to
6381 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
6383 (IFKF_reg): New attributes to give the register constraints for
6385 (extend<mode>tf2_internal): New insns to mark an explicit
6386 conversion between 128-bit floating point types that have a
6387 different mode but share the same representation.
6389 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
6391 PR tree-optimization/85814
6392 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
6393 a null return from get_strinfo when unsharing the next
6394 strinfo in the chain.
6396 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
6399 * varasm.c (weak_finish): Clean up weak_decls.
6401 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6403 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
6404 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
6405 UNSPEC_UADALP values.
6406 * config/aarch64/iterators.md (ABAL): New int iterator.
6409 (sur): Add mappings for the above.
6410 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
6412 (aarch64_<sur>abal<mode>_4): Likewise.
6413 (aarch64_<sur>adalp<mode>_3): Likewise.
6414 (<sur>sadv16qi): New define_expand.
6416 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
6418 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
6419 (*movdf_internal): Ditto.
6420 (*rcpsf2_sse): Ditto.
6421 (*rsqrtsf2_sse): Ditto.
6422 (*sqrt<mode>2_sse): Ditto.
6424 2018-05-21 Tamar Christina <tamar.christina@arm.com>
6426 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
6428 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
6429 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
6430 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
6431 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
6433 * config/aarch64/arm_neon.h: Likewise.
6434 * config/aarch64/iterators.md (VQ_I): New.
6436 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
6438 * config.gcc: Add arc/t-multilib-linux to tmake_file for
6440 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
6443 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6445 * config/nds32/constraints.md (S): New constraint.
6446 * config/nds32/nds32.md (call_internal): Use constraint S.
6447 (call_value_internal): Likewise.
6448 (sibcall_internal): Likewise.
6449 (sibcall_value_internal): Likewise.
6451 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6452 Chung-Ju Wu <jasonwucj@gmail.com>
6454 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
6457 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
6458 Chung-Ju Wu <jasonwucj@gmail.com>
6460 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
6461 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
6462 (nds32_rtx_costs_impl): Simplify.
6463 (nds32_address_cost_impl): Simplify.
6464 (nds32_init_rtx_costs): New function.
6465 (nds32_rtx_costs_speed_prefer): Likewise.
6466 (nds32_rtx_costs_size_prefer): Likewise.
6467 (nds32_address_cost_speed_prefer): Likewise.
6468 (nds32_address_cost_speed_fwprop): Likewise.
6469 (nds32_address_cost_size_prefer): Likewise.
6470 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
6471 * config/nds32/nds32.c (nds32_option_override): Use
6472 nds32_init_rtx_costs function.
6474 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
6476 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
6477 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
6478 (TARGET_PIPELINE_N8): Likewise.
6479 (TARGET_PIPELINE_N10): Likewise.
6480 (TARGET_PIPELINE_N13): Likewise.
6481 (TARGET_PIPELINE_GRAYWOLF): Likewise.
6483 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
6485 * config/nds32/nds32-fpu.md: Update copyright year.
6487 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6489 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
6491 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6493 * config/nds32/nds32.c
6494 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
6495 * config/nds32/nds32.opt (minline-asm-r15): New option.
6497 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6499 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
6501 * config/nds32/nds32.md (abssi2): New pattern.
6503 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
6505 * config/i386/i386.md (rex64namesuffix): New mode attribute.
6506 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
6507 Merge insn pattern from sse_cvtsi2ss<round_name> and
6508 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
6509 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
6510 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
6511 using SWI48 mode iterator.
6512 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
6513 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
6514 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
6515 pattern from sse_cvttss2si<round_saeonly_name>
6516 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
6517 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
6518 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
6519 using SWI48 mode iterator.
6520 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
6521 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
6522 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
6523 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
6524 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
6525 using SWI48 mode iterator.
6526 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
6527 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
6528 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
6529 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
6530 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
6531 SWI48 mode iterator.
6532 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
6533 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
6534 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
6535 pattern from sse_cvttsd2si<round_saeonly_name>
6536 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
6538 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6540 * config/nds32/nds32-md-auxiliary.c
6541 (nds32_valid_smw_lwm_base_p): Refine.
6542 (nds32_output_smw_single_word): Refine.
6543 (nds32_output_smw_double_word): New.
6544 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
6546 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6548 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
6549 (nds32_output_stack_pop): Refine.
6550 (nds32_expand_unaligned_load): Refine.
6551 (nds32_expand_unaligned_store): Refine.
6553 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6554 Chung-Ju Wu <jasonwucj@gmail.com>
6556 * config/nds32/constants.md: Add TP_REGNUM constant.
6557 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
6558 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
6560 * config/nds32/nds32-doubleword.md: Consider flag_pic.
6561 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
6562 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
6563 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
6564 and PIC code generation.
6565 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
6567 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
6569 * config/nds32/nds32.md: Support TLS and PIC.
6570 * config/nds32/nds32.c: Support TLS and PIC.
6571 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
6572 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
6575 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
6577 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
6578 mode with E_ prefix.
6580 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6581 Chung-Ju Wu <jasonwucj@gmail.com>
6583 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
6584 * config/nds32/nds32-md-auxiliary.c
6585 (symbolic_reference_mentioned_p): New.
6586 (nds32_legitimize_ict_address): New.
6587 (nds32_expand_ict_move): New.
6588 (nds32_indirect_call_referenced_p): New.
6589 (nds32_symbol_binds_local_p): Delete.
6590 (nds32_long_call_p): Modify.
6591 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
6592 * config/nds32/nds32-protos.h
6593 (symbolic_reference_mentioned_p): Declare.
6594 (nds32_legitimize_ict_address): Declare.
6595 (nds32_expand_ict_move): Declare.
6596 (nds32_indirect_call_referenced_p): Declare.
6597 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
6598 (nds32_relax_group): Use nds32_ict_const_p as condition.
6599 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
6600 (nds32_asm_file_start): Output ict_model directive in asm code.
6601 (nds32_legitimate_address_p): Consider indirect call.
6602 (nds32_print_operand): Consider indirect call.
6603 (nds32_print_operand_address): Consider indirect call.
6604 (nds32_insert_attributes): Handle "indirect_call" attribute.
6605 (TARGET_LEGITIMATE_ADDRESS_P): Define.
6606 (TARGET_LEGITIMATE_CONSTANT_P): Define.
6607 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
6608 (TARGET_DELEGITIMIZE_ADDRESS): Define.
6609 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
6610 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
6611 (TARGET_ICT_MODEL_SMALL): Define.
6612 (TARGET_ICT_MODEL_LARGE): Define.
6613 * config/nds32/nds32.md (movsi): Consider ict model.
6614 (call, call_value): Consider ict model.
6615 (sibcall, sibcall_value): Consider ict model.
6616 * config/nds32/nds32.opt (mict-model): New option.
6617 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
6620 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
6621 Monk Chiang <sh.chiang04@gmail.com>
6622 Jim Wilson <jimw@sifive.com>
6624 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
6625 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
6626 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
6627 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
6628 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
6629 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
6630 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
6631 compute save_libcall_adjustment properly.
6632 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
6633 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
6634 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
6635 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
6636 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
6637 (ABI_SPEC): Handle mabi=ilp32e.
6638 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
6639 (RVE): Add RVE mask.
6640 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
6641 <-march>: Add rv32e as an example.
6643 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
6646 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
6647 (intra_create_variable_infos): Handle C++ constructors.
6649 2018-05-18 Martin Liska <mliska@suse.cz>
6651 * passes.def: Remove a redundant pass.
6653 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
6656 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
6658 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6660 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
6662 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
6664 (armv2, armv2a, armv3, armv3m): Delete architectures.
6665 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
6666 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
6667 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
6669 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
6670 (*mulsidi3adddi): Likewise.
6671 (mulsidi3): Likewise.
6672 (*mulsidi3_nov6): Likewise.
6673 (umulsidi3): Likewise.
6674 (umulsidi3_nov6): Likewise.
6675 (umaddsidi4): Likewise.
6676 (*umulsidi3adddi): Likewise.
6677 (smulsi3_highpart): Likewise.
6678 (*smulsi3_highpart_nov6): Likewise.
6679 (umulsi3_highpart): Likewise.
6680 (*umulsi3_highpart_nov6): Likewise.
6681 * config/arm/arm.h (arm_arch3m): Delete.
6682 * config/arm/arm.c (arm_arch3m): Delete.
6683 (arm_option_override_internal): Update armv3-related comment.
6684 (arm_configure_build_target): Delete use of isa_bit_mode32.
6685 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
6686 (arm_rtx_costs_internal): Delete check of arm_arch3m.
6687 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
6689 (mulusa3): Likewise.
6690 * config/arm/arm-protos.h (arm_arch3m): Delete.
6691 * config/arm/arm-tables.opt: Regenerate.
6692 * config/arm/arm-tune.md: Likewise.
6693 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
6694 deleted architectures.
6696 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6698 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
6699 (armv5t, armv5te): New features.
6700 (ARMv5, ARMv5e): Delete fgroups.
6701 (ARMv5t, ARMv5te): Adjust for above changes.
6703 (armv5, armv5e): Delete arches.
6704 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
6706 (*call_reg_arm): Likewise.
6707 (*call_value_reg_armv5): Likewise.
6708 (*call_value_reg_arm): Likewise.
6709 (*call_symbol): Likewise.
6710 (*call_value_symbol): Likewise.
6711 (*sibcall_insn): Likewise.
6712 (*sibcall_value_insn): Likewise.
6714 (prefetch): Likewise.
6715 (define_split and define_peephole2 dependent on arm_arch5):
6717 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
6719 (TARGET_ARM_QBIT): Likewise.
6720 (TARGET_DSP_MULTIPLY): Likewise.
6721 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
6722 (arm_arch5, arm_arch5e): Delete.
6723 (arm_arch5t, arm_arch5te): Declare.
6724 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
6725 (arm_arch5t): Declare.
6726 (arm_option_reconfigure_globals): Update for the above.
6727 (arm_options_perform_arch_sanity_checks): Update comment, replace
6728 use of arm_arch5 with arm_arch5t.
6729 (use_return_insn): Likewise.
6730 (arm_emit_call_insn): Likewise.
6731 (output_return_instruction): Likewise.
6732 (arm_final_prescan_insn): Likewise.
6733 (arm_coproc_builtin_available): Likewise.
6734 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
6735 arm_arch5e with arm_arch5t and arm_arch5te.
6736 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
6737 (arm_arch5t, arm_arch5te): Declare.
6738 * config/arm/arm-tables.opt: Regenerate.
6739 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
6740 * config/arm/t-multilib: Likewise.
6741 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
6742 instead of arm_arch5.
6743 (*call_reg_thumb1): Likewise.
6744 (*call_value_reg_thumb1_v5): Likewise.
6745 (*call_value_reg_thumb1): Likewise.
6746 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
6748 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
6750 2018-05-18 Martin Liska <mliska@suse.cz>
6752 PR gcov-profile/84846
6753 * doc/gcov.texi: Document -t option of gcov tool.
6755 2018-05-18 Martin Liska <mliska@suse.cz>
6757 PR gcov-profile/84846
6758 * gcov.c (print_usage): Add new -t option.
6759 (process_args): Handle the option.
6760 (generate_results): Use stdout as output when requested by
6763 2018-05-18 Martin Liska <mliska@suse.cz>
6765 PR gcov-profile/84846
6766 * coverage.c (coverage_init): Write PWD to .gcno file.
6767 * doc/gcov.texi: Document how working directory is printed.
6768 * gcov-dump.c (dump_gcov_file): Print PWD.
6769 * gcov.c (output_intermediate_file): Likewise.
6770 (read_graph_file): Read PWD string.
6771 (output_lines): Print PWD.
6773 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6776 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
6777 for retval and return false if all args to phi are zero.
6779 2018-05-18 Richard Biener <rguenther@suse.de>
6781 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
6783 (evrp_dom_walker::before_dom_children): Call it.
6785 2018-05-18 Richard Biener <rguenther@suse.de>
6787 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
6788 results when processing array refs with variable index.
6790 2018-05-18 Toon Moene <toon@moene.org>
6792 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
6793 directly after that of -floop-interchange. Indicate that both
6794 options are enabled by default when specifying -O3.
6796 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6798 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
6799 iterator. Delete separate integer-mode vec_set<mode> expander.
6800 (aarch64_simd_vec_setv2di): Delete.
6801 (vec_setv2di): Delete.
6802 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
6803 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
6804 the "w, r" alternative.
6806 2018-05-18 Martin Liska <mliska@suse.cz>
6808 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
6809 * tree-pass.h (make_pass_lower_switch_O0): New function.
6810 * tree-switch-conversion.c (node_has_low_bound): Remove.
6811 (node_has_high_bound): Likewise.
6812 (node_is_bounded): Likewise.
6813 (class pass_lower_switch): Make it a template type and create
6815 (pass_lower_switch::execute): Add template argument.
6816 (make_pass_lower_switch): New function.
6817 (make_pass_lower_switch_O0): New function.
6818 (do_jump_if_equal): Remove.
6819 (emit_case_nodes): Simplify to just handle all 3 cases and leave
6820 all the hard work to tree optimization passes.
6822 2018-05-18 Martin Liska <mliska@suse.cz>
6824 * dbgcnt.c (limit_low): Renamed from limit.
6825 (limit_high): New variable.
6826 (dbg_cnt_is_enabled): Check for upper limit.
6827 (dbg_cnt): Adjust dumping.
6828 (dbg_cnt_set_limit_by_index): Add new argument for high
6830 (dbg_cnt_set_limit_by_name): Likewise.
6831 (dbg_cnt_process_single_pair): Parse new format.
6832 (dbg_cnt_process_opt): Use strtok.
6833 (dbg_cnt_list_all_counters): Remove 'value' and add
6835 * doc/invoke.texi: Document changes.
6837 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
6839 * doc/sourcebuild.texi (scalar_all_fma): Document.
6840 * tree.def (FMA_EXPR): Delete.
6841 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
6842 * internal-fn.c (ternary_direct): New macro.
6843 (expand_ternary_optab_fn): Likewise.
6844 (direct_ternary_optab_supported_p): Likewise.
6845 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
6846 * builtins.c (fold_builtin_fma): Delete.
6847 (fold_builtin_3): Don't call it.
6848 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
6849 * expr.c (expand_expr_real_2): Likewise.
6850 * fold-const.c (operand_equal_p): Likewise.
6851 (fold_ternary_loc): Likewise.
6852 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
6853 * gimple.c (DEFTREECODE): Likewise.
6854 * gimplify.c (gimplify_expr): Likewise.
6855 * optabs-tree.c (optab_for_tree_code): Likewise.
6856 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
6857 * tree-eh.c (operation_could_trap_p): Likewise.
6858 (stmt_could_throw_1_p): Likewise.
6859 * tree-inline.c (estimate_operator_cost): Likewise.
6860 * tree-pretty-print.c (dump_generic_node): Likewise.
6861 (op_code_prio): Likewise.
6862 * tree-ssa-loop-im.c (stmt_cost): Likewise.
6863 * tree-ssa-operands.c (get_expr_operands): Likewise.
6864 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
6865 * fold-const-call.h (fold_fma): Delete.
6866 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
6867 CFN_FNMA and CFN_FNMS.
6869 * genmatch.c (combined_fn): New enum.
6870 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
6871 (commutative_op): New function.
6872 (commutate): Use it. Handle more than 2 operands.
6873 (dt_operand::gen_gimple_expr): Use commutative_op.
6874 (parser::parse_expr): Allow :c to be used with non-binary
6875 operators if the commutative operand is known.
6876 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
6877 CFN_FMS, CFN_FNMA and CFN_FNMS.
6878 (backprop::process_assign_use): Remove FMA_EXPR handling.
6879 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
6880 (gen_hsa_fma): New function.
6881 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
6882 IFN_FNMA and IFN_FNMS.
6883 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
6884 * gimple-fold.h (follow_all_ssa_edges): Declare.
6885 * gimple-fold.c (follow_all_ssa_edges): New function.
6886 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
6887 gimple_build interface and use follow_all_ssa_edges to fold the result.
6888 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
6889 instead of checking for optabs directly.
6890 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
6891 rather than FMA_EXPRs.
6892 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
6893 call to IFN_FMA instead of an FMA_EXPR.
6895 2018-05-17 Jim Wilson <jimw@sifive.com>
6897 * expr.c (do_tablejump): When converting index to Pmode, if we have a
6898 sign extended promoted subreg, and the range does not have the sign bit
6899 set, then do a sign extend.
6901 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
6902 test, check for sign extended subreg and/or constant operands, and
6903 do a sign extend in that case.
6905 2018-05-17 Steve Ellcey <sellcey@cavium.com>
6907 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
6908 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
6910 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
6911 Change logics_shift_reg to logics_shift_imm.
6912 (thunderx2t99_fp_loadpair_basic): Delete.
6913 (thunderx2t99_fp_storepair_basic): Delete.
6914 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
6915 (thunderx2t99_asimd_polynomial): Delete.
6916 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
6917 and neon_fp_mul_d_scalar_q.
6918 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
6919 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
6920 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
6921 (thunderx2t99_asimd_lut): Add missing tbl types.
6922 (thunderx2t99_asimd_ext): Delete.
6923 (thunderx2t99_asimd_load1_1_mult): Delete.
6924 (thunderx2t99_asimd_load1_2_mult): Delete.
6925 (thunderx2t99_asimd_load1_ldp): New.
6926 (thunderx2t99_asimd_load1): New.
6927 (thunderx2t99_asimd_load2): Add missing *load2* types.
6928 (thunderx2t99_asimd_load3): New.
6929 (thunderx2t99_asimd_load4): New.
6930 (thunderx2t99_asimd_store1_1_mult): Delete.
6931 (thunderx2t99_asimd_store1_2_mult): Delete.
6932 (thunderx2t99_asimd_store2_mult): Delete.
6933 (thunderx2t99_asimd_store2_onelane): Delete.
6934 (thunderx2t99_asimd_store_stp): New.
6935 (thunderx2t99_asimd_store1): New.
6936 (thunderx2t99_asimd_store2): New.
6937 (thunderx2t99_asimd_store3): New.
6938 (thunderx2t99_asimd_store4): New.
6940 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
6942 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
6943 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
6945 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
6946 Segher Boessenkool <segher@kernel.crashing.org>
6949 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
6952 2018-05-17 Richard Biener <rguenther@suse.de>
6954 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
6955 for pruning loop and prune defs feeding only already visited PHIs.
6957 2018-05-17 Richard Biener <rguenther@suse.de>
6959 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
6961 2018-05-17 Bin Cheng <bin.cheng@arm.com>
6962 Richard Biener <rguenther@suse.de>
6964 PR tree-optimization/85793
6965 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
6966 for VMAT_ELEMENTWISE.
6968 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6970 * internal-fn.h (lookup_internal_fn): Declare
6971 * internal-fn.c (lookup_internal_fn): New function.
6972 * gimple.c (gimple_build_call_from_tree): Handle calls to
6974 * gimple-pretty-print.c (dump_gimple_call): Print "." before
6975 internal function names.
6976 * tree-pretty-print.c (dump_generic_node): Likewise.
6977 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
6979 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
6981 * gimple-fold.h (gimple_build): Make the function forms take
6982 combined_fn rather than built_in_function.
6983 (gimple_simplify): Likewise.
6984 * gimple-match-head.c (gimple_simplify): Likewise.
6985 * gimple-fold.c (gimple_build): Likewise.
6986 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
6987 rather than gimple_build_call_internal.
6988 (get_initial_defs_for_reduction): Likewise.
6989 (vect_create_epilog_for_reduction): Likewise.
6990 (vectorizable_live_operation): Likewise.
6992 2018-05-17 Martin Liska <mliska@suse.cz>
6994 * gimple-ssa-sprintf.c (format_directive): Do not use
6995 space in between 'G_' and '('.
6997 2018-05-17 Jakub Jelinek <jakub@redhat.com>
7000 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
7001 even if the mask is not all ones.
7004 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
7006 (ix86_gimple_fold_builtin): Likewise.
7009 * config/i386/i386.c: Include tree-vector-builder.h.
7010 (ix86_vector_shift_count): New function.
7011 (ix86_fold_builtin): Fold shift builtins by scalar count.
7012 (ix86_gimple_fold_builtin): Likewise.
7014 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
7015 _mm512_setzero): New intrinsics.
7017 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
7018 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7020 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
7021 code generation for cases where splatting a value is not useful.
7022 * simplify-rtx.c (simplify_ternary_operation): Simplify
7023 vec_merge across a vec_duplicate and a paradoxical subreg forming
7024 a vector mode to a vec_concat.
7026 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
7028 * config.gcc: Support "goldmont-plus".
7029 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7031 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7032 PROCESSOR_GOLDMONT_PLUS.
7033 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
7034 (processor_target_table): Add "goldmont-plus".
7035 (PTA_GOLDMONT_PLUS): Define.
7036 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
7037 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
7038 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
7039 (fold_builtin_cpu): Add "goldmont-plus".
7040 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
7041 (ix86_option_override_internal): Add "goldmont-plus".
7042 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
7043 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
7044 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
7045 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
7047 2018-05-17 Richard Biener <rguenther@suse.de>
7049 PR tree-optimization/85757
7050 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
7051 remove defs that only feed that PHI from further processing.
7053 2018-05-16 Jim Wilson <jimw@sifive.com>
7055 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
7057 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
7059 2018-05-16 Mark Wielaard <mark@klomp.org>
7061 * dwarf2out.c (count_index_strings): New function.
7062 (output_indirect_strings): Call count_index_strings and generate
7063 header for dwarf_version >= 5.
7065 2018-05-16 Mark Wielaard <mark@klomp.org>
7067 * dwarf2out.c (dwarf_FORM): New function.
7068 (set_indirect_string): Use dwarf_FORM.
7069 (reset_indirect_string): Likewise.
7070 (size_of_die): Likewise.
7071 (value_format): Likewise.
7072 (output_die): Likewise.
7073 (add_skeleton_AT_string): Likewise.
7074 (output_macinfo_op): Likewise.
7075 (index_string): Likewise.
7076 (output_index_string_offset): Likewise.
7077 (output_index_string): Likewise.
7078 (count_index_strings): Likewise.
7080 2018-05-16 Carl Love <cel@us.ibm.com>
7082 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7083 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
7085 2018-05-16 Martin Jambor <mjambor@suse.cz>
7087 * ipa-prop.c (ipa_free_all_edge_args): Remove.
7088 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
7090 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
7092 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
7093 (fnma<mode>4): Likewise.
7094 (fms<mode>4): Likewise.
7095 (fnms<mode>4): Likewise.
7096 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
7097 (aarch64_fnma<mode>4): Likewise.
7098 (aarch64_fms<mode>4): Likewise.
7099 (aarch64_fnms<mode>4): Likewise.
7100 (aarch64_fnmadd<mode>4): Likewise.
7102 2018-05-16 Jason Merrill <jason@redhat.com>
7104 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
7106 2018-05-16 Richard Biener <rguenther@suse.de>
7108 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
7109 (dump_stmt_cost): Declare.
7110 (add_stmt_cost): Dump cost we add.
7111 (add_stmt_costs): New function.
7112 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
7114 (vect_analyze_stmt): Adjust prototype.
7115 (vectorizable_condition): Likewise.
7116 (vectorizable_live_operation): Likewise.
7117 (vectorizable_reduction): Likewise.
7118 (vectorizable_induction): Likewise.
7119 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
7120 cost vector to pass to vectorizable_ and record afterwards.
7121 (vect_model_reduction_cost): Take cost vector argument and adjust.
7122 (vect_model_induction_cost): Likewise.
7123 (vectorizable_reduction): Likewise.
7124 (vectorizable_induction): Likewise.
7125 (vectorizable_live_operation): Likewise.
7126 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
7127 SLP_TREE_NUMBER_OF_VEC_STMTS.
7128 (vect_analyze_slp_cost_1): Remove.
7129 (vect_analyze_slp_cost): Likewise.
7130 (vect_slp_analyze_node_operations): Take visited args and
7131 a target cost vector. Avoid processing already visited stmt sets.
7132 (vect_slp_analyze_operations): Use a local cost vector to gather
7133 costs and register those of non-discarded instances.
7134 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
7135 (vect_schedule_slp_instance): Remove copying of
7136 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
7138 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
7139 adding cost. Record cost entry location.
7140 (vect_prologue_cost_for_slp_op): Function to compute cost of
7141 a constant or invariant generated for SLP vect in the prologue,
7142 split out from vect_analyze_slp_cost_1.
7143 (vect_model_simple_cost): Make static. Adjust for SLP costing.
7144 (vect_model_promotion_demotion_cost): Likewise.
7145 (vect_model_store_cost): Likewise, make static.
7146 (vect_model_load_cost): Likewise.
7147 (vectorizable_bswap): Add cost vector arg and adjust.
7148 (vectorizable_call): Likewise.
7149 (vectorizable_simd_clone_call): Likewise.
7150 (vectorizable_conversion): Likewise.
7151 (vectorizable_assignment): Likewise.
7152 (vectorizable_shift): Likewise.
7153 (vectorizable_operation): Likewise.
7154 (vectorizable_store): Likewise.
7155 (vectorizable_load): Likewise.
7156 (vectorizable_condition): Likewise.
7157 (vectorizable_comparison): Likewise.
7158 (can_vectorize_live_stmts): Likewise.
7159 (vect_analyze_stmt): Likewise.
7160 (vect_transform_stmt): Adjust calls to vectorizable_*.
7161 * tree-vectorizer.c: Include gimple-pretty-print.h.
7162 (dump_stmt_cost): New function.
7164 2018-05-16 Richard Biener <rguenther@suse.de>
7166 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
7167 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
7168 * tree-ssa-dse.c: Include tree-ssa-loop.h.
7169 (check_name): New callback.
7170 (dse_classify_store): Track cycles via a visited bitmap of PHI
7171 defs and simplify handling of in-loop and across loop dead stores
7172 and properly fail for loop-variant refs. Handle byte-tracking with
7173 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
7176 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
7178 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
7179 (vect_get_mask_type_for_stmt): Likewise.
7180 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
7182 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
7183 to determine the statement's vector type and the vector type that
7184 should be used for calculating nunits. Deal with cases in which
7185 the type has to be deferred.
7186 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
7187 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
7188 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
7189 (vect_determine_vf_for_stmt): New functions, split out from...
7190 (vect_determine_vectorization_factor): ...here.
7191 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
7192 (vect_get_mask_type_for_stmt): New functions, split out from
7193 vect_determine_vectorization_factor.
7195 2018-05-16 Richard Biener <rguenther@suse.de>
7197 * tree-cfg.c (verify_gimple_assign_ternary): Properly
7198 verify the [VEC_]COND_EXPR embedded comparison.
7200 2018-05-15 Martin Sebor <msebor@redhat.com>
7202 PR tree-optimization/85753
7203 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
7204 RECORD_TYPE in addition to ARRAY_TYPE.
7206 2018-05-15 Martin Sebor <msebor@redhat.com>
7209 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
7211 2018-05-15 Richard Biener <rguenther@suse.de>
7213 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
7214 add by_clobber_p one. Change algorithm to collect all defs
7215 representing uses we need to walk and try reducing them to
7216 a single one before failing.
7217 (dse_dom_walker::dse_optimize_stmt): Adjust.
7219 2018-05-13 Mark Wielaard <mark@klomp.org>
7221 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
7222 (size_of_loc_descr): Likewise.
7223 (output_loc_operands): Likewise.
7224 (output_loc_operands_raw): Likewise.
7225 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
7226 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
7227 (hash_loc_operands): Likewise.
7228 (compare_loc_operands): Likewise.
7230 2018-05-14 Mark Wielaard <mark@klomp.org>
7232 * dwarf2out.c (count_index_addrs): New function.
7233 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
7235 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7237 PR tree-optimization/83648
7238 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
7239 return value as malloc candidate.
7241 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7244 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
7245 param as true in call to suggest_attribute.
7247 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
7249 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
7250 -mreadonly-in-sdata.
7252 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7254 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
7256 (aarch64_crypto_aesd_fused): Likewise.
7258 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
7260 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
7261 (movsi_aarch64): Likewise.
7262 (load_pairsi): Likewise.
7263 (load_pairdi): Likewise.
7264 (store_pairsi): Likewise.
7265 (store_pairdi): Likewise.
7266 (load_pairsf): Likewise.
7267 (load_pairdf): Likewise.
7268 (store_pairsf): Likewise.
7269 (store_pairdf): Likewise.
7270 (zero_extend): Likewise.
7271 (trunc): Swap alternatives.
7272 (fcvt_target): Add '?' to prefer w over r.
7274 2018-05-14 Jakub Jelinek <jakub@redhat.com>
7277 * config/i386/i386.md: Disallow non-commutative arithmetics in
7278 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
7279 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
7280 in the peephole2 before it.
7282 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
7284 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
7285 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
7286 (ix86_handle_option): Handle -mcldemote.
7287 * config.gcc: New header.
7288 * config/i386/cldemoteintrin.h: New file.
7289 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
7290 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
7292 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7293 OPTION_MASK_ISA_CLDEMOTE.
7294 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
7295 (ix86_valid_target_attribute_inner_p): Ditto.
7296 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
7297 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
7298 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
7299 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
7300 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
7302 * config/i386/i386.opt: Add -mcldemote.
7303 * config/i386/x86intrin.h: New header.
7304 * doc/invoke.texi: Add -mcldemote.
7306 2018-05-14 Richard Biener <rguenther@suse.de>
7308 * doc/match-and-simplify.texi: Adjust :s documentation.
7310 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
7312 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
7313 intended memcpy size.
7314 (REORDER_45): Likewise.
7316 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
7318 * sort.cc: New file.
7319 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
7320 * vec.c (qsort_chk): Use gcc_qsort.
7321 * Makefile.in (OBJS-libcommon): Add sort.o.
7322 (build/sort.o): New target. Use it...
7323 (BUILD_RTL): ... here, and...
7324 (build/gencfn-macros): ... here, and...
7325 (build/genmatch): ... here.
7327 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7328 Chung-Ju Wu <jasonwucj@gmail.com>
7330 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
7331 * config/nds32/nds32-graywolf.md: New file.
7332 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
7333 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
7335 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
7336 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
7337 * config/nds32/nds32.md (pipeline_model): Add graywolf.
7338 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
7339 * config/nds32/pipelines.md: Include n15 settings.
7341 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7342 Chung-Ju Wu <jasonwucj@gmail.com>
7344 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
7345 * config/nds32/nds32-n13.md: New file.
7346 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
7347 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
7349 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
7350 * config/nds32/nds32.md (pipeline_model): Add n13.
7351 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
7352 * config/nds32/pipelines.md: Include n13 settings.
7354 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
7355 Chung-Ju Wu <jasonwucj@gmail.com>
7357 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
7358 * config/nds32/nds32-n10.md: New file.
7359 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
7360 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
7362 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
7363 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
7364 * config/nds32/nds32.md (pipeline_model): Add n10.
7365 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
7366 * config/nds32/pipelines.md: Include n10 settings.
7368 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
7369 Kito Cheng <kito.cheng@gmail.com>
7370 Chung-Ju Wu <jasonwucj@gmail.com>
7372 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
7373 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
7374 Add enum values for DSP extension instructions.
7375 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
7377 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
7378 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
7380 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
7381 * config/nds32/nds32-dspext.md: New file for DSP implementation.
7382 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
7383 * config/nds32/nds32-intrinsic.md: Likewise.
7384 * config/nds32/nds32_intrinsic.h: Likewise.
7385 * config/nds32/nds32-md-auxiliary.c: Likewise.
7386 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
7387 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
7388 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
7389 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
7390 * config/nds32/nds32-protos.h: New declarations for DSP extension.
7391 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
7392 TYPE_DMAC in switch statement.
7393 * config/nds32/nds32.c: New checking and implementation for DSP
7394 extension instructions.
7395 * config/nds32/nds32.h: Likewise.
7396 * config/nds32/nds32.md: Likewise.
7397 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
7398 * config/nds32/predicates.md: Implement new predicates for DSP
7401 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7403 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
7404 Reformat alternatives and attributes so it is easier to identify
7405 which constraints/attributes go with which instruction.
7406 (mov<mode>_hardfloat32, FMOVE64): Likewise.
7407 (mov<mode>_softfloat32, FMOVE64): Likewise.
7408 (mov<mode>_hardfloat64, FMOVE64): Likewise.
7409 (mov<mode>_softfloat64, FMOVE64): Likewise.
7411 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
7413 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7415 (Basic PowerPC Built-in Functions): The new name of the
7416 subsection previously known as "PowerPC Built-in Functions".
7417 (Basic PowerPC Built-in Functions Available on all Configurations):
7419 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
7420 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
7421 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
7422 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
7424 2018-05-11 Martin Jambor <mjambor@suse.cz>
7427 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
7430 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7433 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
7435 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
7437 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
7438 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
7439 (ix86_handle_option): Handle -mwaitpkg.
7440 * config.gcc: New header.
7441 * config/i386/cpuid.h (bit_WAITPKG): New bit.
7442 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
7443 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
7445 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7446 OPTION_MASK_ISA_WAITPKG.
7447 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
7448 (ix86_option_override_internal): Add PTA_WAITPKG.
7449 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
7450 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
7451 IX86_BUILTIN_TPAUSE.
7452 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
7453 __builtin_ia32_umwait and __builtin_ia32_tpause.
7454 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
7455 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
7456 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
7457 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
7458 UNSPECV_TPAUSE): New.
7459 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
7460 * config/i386/i386.opt: Add -mwaitpkg.
7461 * config/i386/waitpkgintrin.h: New file.
7462 * config/i386/x86intrin.h: New header.
7463 * doc/invoke.texi: Add -mwaitpkg.
7465 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
7468 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
7470 (cortex-m0): Use armv6s-m isa.
7471 (cortex-m0plus): Likewise.
7472 (cortex-m1): Likewise.
7473 (cortex-m0.small-multiply): Likewise.
7474 (cortex-m0plus.small-multiply): Likewise.
7475 (cortex-m1.small-multiply): Likewise.
7477 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
7478 Jakub Jelinek <jakub@redhat.com>
7480 PR tree-optimization/85692
7481 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
7482 source permute as well.
7484 2018-05-11 Martin Liska <mliska@suse.cz>
7487 * doc/extend.texi: Document LLVM style format for no_sanitize
7490 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
7492 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
7493 mode_supports_vsx_dform_quad to mode_supports_dq_form.
7494 (mode_supports_vsx_dform_quad): Likewise.
7495 (mode_supports_vmx_dform): Move these functions to be next to the
7496 other mode_supports functions.
7497 (mode_supports_dq_form): Likewise.
7498 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
7499 mode_supports_dq_form.
7500 (reg_offset_addressing_ok_p): Likewise.
7501 (offsettable_ok_by_alignment): Likewise.
7502 (rs6000_legitimate_offset_address_p): Likewise.
7503 (legitimate_lo_sum_address_p): Likewise.
7504 (rs6000_legitimize_address): Likewise.
7505 (rs6000_legitimize_reload_address): Likewise.
7506 (rs6000_secondary_reload_inner): Likewise.
7507 (rs6000_preferred_reload_class): Likewise.
7508 (rs6000_output_move_128bit): Likewise.
7510 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7512 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
7513 Generate SImode target register for null target.
7514 <case IX86_BUILTIN_XGETBV>: Ditto.
7515 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
7516 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
7518 2018-05-10 Carl Love <cel@us.ibm.com>
7520 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
7521 dcbtt and dcbtstt if operands[2] is 0.
7523 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
7526 * config/i386/sse.md (usadv64qi): New expander.
7528 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
7530 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
7531 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
7532 -maltivec=be support.
7533 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
7534 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
7535 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
7536 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
7537 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
7538 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
7539 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
7540 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
7541 altivec_vsumsws): Adjust.
7542 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
7543 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
7544 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
7546 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
7547 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
7548 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
7549 (altivec_lve<VI_char>x): Delete expand.
7550 (*altivec_lve<VI_char>x_internal): Rename to...
7551 (altivec_lve<VI_char>x): ... this.
7552 (altivec_lvxl_<mode>): Delete expand.
7553 (*altivec_lvxl_<mode>_internal): Rename to ...
7554 (altivec_lvxl_<mode>): ... this.
7555 (altivec_stvxl_<mode>): Delete expand.
7556 (*altivec_stvxl_<mode>_internal): Rename to ...
7557 (altivec_stvxl_<mode>): ... this.
7558 (altivec_stve<VI_char>x): Delete expand.
7559 (*altivec_stve<VI_char>x_internal): Rename to ...
7560 (altivec_stve<VI_char>x): ... this.
7561 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
7562 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
7563 reduc_plus_scal_<mode>): Adjust.
7564 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
7566 (rs6000_cpu_cpp_builtins): Adjust.
7567 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
7568 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
7569 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
7570 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
7571 -maltivec=be support.
7572 (rs6000_split_vec_extract_var): Adjust.
7573 (rs6000_split_v4si_init): Adjust.
7574 (swap_selector_for_mode): Delete.
7575 (altivec_expand_lvx_be, altivec_expand_stvx_be,
7576 altivec_expand_stvex_be): Delete.
7577 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
7578 -maltivec=be support.
7579 (rs6000_gimple_fold_builtin): Ditto.
7580 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
7582 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
7583 (TARGET_DIRECT_MOVE_64BIT): Adjust.
7584 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
7585 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
7586 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
7587 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
7588 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
7589 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
7590 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
7591 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
7592 anonymous split): Adjust.
7593 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
7594 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
7596 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
7598 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
7599 when --with-gxx-include-dir is also specified.
7600 * configure: Regenerate.
7602 2018-05-09 Jim Wilson <jimw@sifive.com>
7605 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
7606 * config/riscv/t-withmultilib: New.
7607 * config/riscv/withmultilib.h: New.
7608 * doc/install.texi: Document RISC-V --with-multilib-list support.
7610 2018-05-09 Richard Biener <rguenther@suse.de>
7612 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
7614 (vect_bb_vectorization_profitable_p): Adjust. Compute
7615 actual scalar cost using the cost vector and the add_stmt_cost
7618 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7620 PR rtl-optimization/85645
7621 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
7622 in the REG_CFA_REGISTER note for LR, don't leave it empty.
7624 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7626 PR rtl-optimization/85645
7627 * shrink-wrap.c (spread_components): Return a boolean saying if
7628 anything was changed.
7629 (try_shrink_wrapping_separate): Iterate spread_components until
7630 nothing changes anymore.
7632 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7634 PR rtl-optimization/85645
7635 * regrename.c (build_def_use): Also kill the chains that include the
7636 destination of a REG_CFA_REGISTER note.
7638 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
7640 PR rtl-optimization/85645
7641 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
7642 insn that has a REG_CFA_REGISTER note.
7644 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
7646 * cfgexpand.c (expand_clobber): New function.
7647 (expand_gimple_stmt_1): Use it.
7648 * tree-vect-stmts.c (vect_clobber_variable): New function,
7650 (vectorizable_simd_clone_call): ...here.
7651 (vectorizable_store): Emit a clobber either side of an
7652 IFN_STORE_LANES sequence.
7653 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
7655 2018-05-09 Tom de Vries <tom@codesourcery.com>
7658 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
7659 (define_insn "trap_if_false"): Add exit after trap.
7661 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
7663 PR rtl-optimization/85638
7664 * bb-reorder.c: Include common/common-target.h.
7665 (create_forwarder_block): New function extracted from...
7666 (fix_up_crossing_landing_pad): ...here. Rename into...
7667 (dw2_fix_up_crossing_landing_pad): ...this.
7668 (sjlj_fix_up_crossing_landing_pad): New function.
7669 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
7670 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
7671 from both partitions and exit the loop after one iteration.
7673 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7676 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7678 (Basic PowerPC Built-in Functions): The new name of the
7679 subsection previously known as "PowerPC Built-in Functions".
7680 (Basic PowerPC Built-in Functions Available on all Configurations):
7682 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7684 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7686 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7688 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7691 2018-05-08 Jim Wilson <jimw@sifive.com>
7693 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
7694 (LD_EMUL_SUFFIX): New.
7695 (LINK_SPEC): Use it.
7697 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
7699 * doc/extend.texi (PowerPC Built-in Functions): Rename this
7701 (Basic PowerPC Built-in Functions): The new name of the
7702 subsection previously known as "PowerPC Built-in Functions".
7703 (Basic PowerPC Built-in Functions Available on all Configurations):
7705 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
7707 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
7709 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
7711 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
7714 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7717 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
7718 after cmpelim optimization.
7720 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
7722 * config.gcc: Support "goldmont".
7723 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
7724 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
7726 * config/i386/i386.c (m_GOLDMONT): Define.
7727 (processor_target_table): Add "goldmont".
7728 (PTA_GOLDMONT): Define.
7729 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
7730 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
7731 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
7732 (fold_builtin_cpu): Add "goldmont".
7733 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
7734 (ix86_option_override_internal): Add "goldmont".
7735 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
7736 (processor_type): Add PROCESSOR_GOLDMONT.
7737 * config/i386/i386.md: Add CPU "glm".
7738 * config/i386/glm.md: New file.
7739 * config/i386/x86-tune.def: Add m_GOLDMONT.
7740 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
7742 2018-05-08 Jakub Jelinek <jakub@redhat.com>
7745 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
7747 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
7748 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
7749 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
7752 * config/i386/i386.c (ix86_fold_builtin): Handle
7753 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
7756 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
7757 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
7759 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
7762 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
7763 (check_arch): Likewise.
7764 (check_fpu): Return the result rather than printing it.
7765 (end arch): Fix operator precedence.
7766 (end cpu): Likewise.
7767 (END): Print the result from check_fpu.
7769 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7770 Alan Hayward <alan.hayward@arm.com>
7771 David Sherwood <david.sherwood@arm.com>
7773 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
7774 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
7775 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
7776 (*fcmuo<mode>_and): New patterns.
7778 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
7780 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
7781 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
7782 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
7783 (cmp_op, sve_imm_con): New code attributes.
7784 (SVE_COND_INT_CMP, imm_con): Delete.
7785 (cmp_op): Remove above unspecs from int attribute.
7786 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
7788 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
7789 comparison-specific unspecs.
7790 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
7791 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
7792 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
7793 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
7794 (*vec_fcm<cmp_op><mode>): Rename to...
7795 (*fcm<cmp_op><mode>): ...this and adjust likewise.
7796 (*vec_fcmuo<mode>): Rename to...
7797 (*fcmuo<mode>): ...this and adjust likewise.
7798 (*pred_fcm<cmp_op><mode>): New pattern.
7799 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
7800 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
7802 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
7804 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
7805 (aarch64_emit_sve_predicated_cond): New function.
7806 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
7807 (aarch64_emit_unspec_cond_or): Replace with...
7808 (aarch64_emit_sve_or_conds): ...this new function. Use
7809 aarch64_emit_sve_ptrue_op for the individual comparisons and
7810 aarch64_emit_binop to OR them together.
7811 (aarch64_emit_inverted_unspec_cond): Replace with...
7812 (aarch64_emit_sve_inverted_cond): ...this new function. Use
7813 aarch64_emit_sve_ptrue_op for the comparison and
7814 aarch64_emit_unop to invert the result.
7815 (aarch64_expand_sve_vec_cmp_float): Update after the above
7816 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
7818 2018-05-07 Nathan Sidwell <nathan@acm.org>
7820 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
7821 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
7822 (Backwards Compatibility): Likewise.
7824 2018-05-07 Luis Machado <luis.machado@linaro.org>
7828 2018-05-07 Luis Machado <luis.machado@linaro.org>
7830 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7831 <prefetch_dynamic_strides>: New const bool field.
7832 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7833 prefetch_dynamic_strides.
7834 (exynosm1_prefetch_tune): Likewise.
7835 (thunderxt88_prefetch_tune): Likewise.
7836 (thunderx_prefetch_tune): Likewise.
7837 (thunderx2t99_prefetch_tune): Likewise.
7838 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7840 (aarch64_override_options_internal): Update to set
7841 PARAM_PREFETCH_DYNAMIC_STRIDES.
7842 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7843 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7844 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7845 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7846 prefetch-dynamic-strides setting.
7848 2018-05-07 Luis Machado <luis.machado@linaro.org>
7850 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7851 <minimum_stride>: New const int field.
7852 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7853 minimum_stride field.
7854 (exynosm1_prefetch_tune): Likewise.
7855 (thunderxt88_prefetch_tune): Likewise.
7856 (thunderx_prefetch_tune): Likewise.
7857 (thunderx2t99_prefetch_tune): Likewise.
7858 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7859 (aarch64_override_options_internal): Update to set
7860 PARAM_PREFETCH_MINIMUM_STRIDE.
7861 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7862 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7863 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7864 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7865 stride is constant and is below the minimum stride threshold.
7867 2018-05-07 Luis Machado <luis.machado@linaro.org>
7869 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
7872 2018-05-07 Luis Machado <luis.machado@linaro.org>
7874 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7875 <prefetch_dynamic_strides>: New const bool field.
7876 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7877 prefetch_dynamic_strides.
7878 (exynosm1_prefetch_tune): Likewise.
7879 (thunderxt88_prefetch_tune): Likewise.
7880 (thunderx_prefetch_tune): Likewise.
7881 (thunderx2t99_prefetch_tune): Likewise.
7882 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
7884 (aarch64_override_options_internal): Update to set
7885 PARAM_PREFETCH_DYNAMIC_STRIDES.
7886 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
7887 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
7888 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
7889 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
7890 prefetch-dynamic-strides setting.
7892 2018-05-07 Luis Machado <luis.machado@linaro.org>
7894 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
7895 <minimum_stride>: New const int field.
7896 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
7897 minimum_stride field.
7898 (exynosm1_prefetch_tune): Likewise.
7899 (thunderxt88_prefetch_tune): Likewise.
7900 (thunderx_prefetch_tune): Likewise.
7901 (thunderx2t99_prefetch_tune): Likewise.
7902 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
7903 (aarch64_override_options_internal): Update to set
7904 PARAM_PREFETCH_MINIMUM_STRIDE.
7905 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
7906 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
7907 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
7908 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
7909 stride is constant and is below the minimum stride threshold.
7911 2018-05-06 Jakub Jelinek <jakub@redhat.com>
7914 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
7915 the type is addressable. Don't force op into register if it has
7918 2018-05-05 Roland McGrath <mcgrathr@google.com>
7921 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
7922 any section for which we don't know a specific type it should have,
7923 regardless of name. Previously this was done only for the exact
7924 names ".init_array", ".fini_array", and ".preinit_array".
7925 (default_elf_asm_named_section): Add comment about
7926 relationship with default_section_type_flags and SECTION_NOTYPE.
7927 (get_section): Don't consider it a type conflict if one side has
7928 SECTION_NOTYPE and the other doesn't, as long as neither has the
7929 SECTION_BSS et al used in the default_section_type_flags logic.
7931 2018-05-05 Tom de Vries <tom@codesourcery.com>
7934 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
7935 (workaround_barsyncs): New function.
7936 (nvptx_reorg): Use workaround_barsyncs.
7937 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
7938 (define_expand "nvptx_membar_cta"): New define_expand.
7939 (define_insn "*nvptx_membar_cta"): New insn.
7941 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
7943 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
7944 To improve optimization opportunities.
7945 * builtin-types.def: The new needed builtin types for the above.
7947 2018-05-04 Richard Biener <rguenther@suse.de>
7949 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
7950 * gimple-ssa-store-merging.c
7951 (imm_store_chain_info::output_merged_store): Remove redundant create,
7952 release split_store vector contents on failure.
7953 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
7954 scalar stmt vector on cache hit.
7956 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
7958 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
7960 * config.gcc (powerpc-xilinx-eabi*): Remove.
7961 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
7963 (fusion_addis_mem_combo_load): Ditto.
7964 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
7966 (rs6000_cpu_cpp_builtins): Ditto.
7967 * config/rs6000/rs6000-linux.c
7968 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
7969 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
7970 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
7972 (rs6000_setup_reg_addr_masks): Ditto.
7973 (rs6000_init_hard_regno_mode_ok): Ditto.
7974 (rs6000_option_override_internal): Ditto.
7975 (legitimate_lo_sum_address_p): Ditto.
7976 (rs6000_legitimize_address): Ditto.
7977 (rs6000_legitimize_reload_address): Ditto.
7978 (rs6000_legitimate_address_p): Ditto.
7979 (abi_v4_pass_in_fpr): Ditto.
7980 (setup_incoming_varargs): Ditto.
7981 (rs6000_gimplify_va_arg): Ditto.
7982 (rs6000_split_multireg_move): Ditto.
7983 (rs6000_savres_strategy): Ditto.
7984 (rs6000_emit_prologue_components): Ditto.
7985 (rs6000_emit_epilogue_components): Ditto.
7986 (rs6000_emit_prologue): Ditto.
7987 (rs6000_emit_epilogue): Ditto.
7988 (rs6000_elf_file_end): Ditto.
7989 (rs6000_function_value): Ditto.
7990 (rs6000_libcall_value): Ditto.
7991 * config/rs6000/rs6000.h: Ditto.
7992 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
7993 (TARGET_MINMAX): ... this. New.
7994 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
7995 * config/rs6000/rs6000.md: Remove Xilinx FP support.
7996 (*movsi_internal1_single): Delete.
7997 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
7998 mfpu=, mxilinx-fpu): Delete.
7999 * config/rs6000/singlefp.h: Delete.
8000 * config/rs6000/sysv4.h: Remove Xilinx FP support.
8001 * config/rs6000/t-rs6000: Ditto.
8002 * config/rs6000/t-xilinx: Delete.
8003 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
8004 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
8005 (VStype_simple): Delete.
8006 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
8007 * config/rs6000/xfpu.h: Delete.
8008 * config/rs6000/xfpu.md: Delete.
8009 * config/rs6000/xilinx.h: Delete.
8010 * config/rs6000/xilinx.opt: Delete.
8011 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
8012 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
8014 2018-05-04 Tom de Vries <tom@codesourcery.com>
8017 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
8020 2018-05-04 Richard Biener <rguenther@suse.de>
8023 * tree-complex.c (update_complex_assignment): We are always in SSA form.
8024 (expand_complex_div_wide): Likewise.
8025 (expand_complex_operations_1): Likewise.
8026 (expand_complex_libcall): Preserve EH info of the original stmt.
8027 (tree_lower_complex): Handle removed blocks.
8028 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
8029 on complex multiplication and division libcall builtins.
8031 2018-05-04 Richard Biener <rguenther@suse.de>
8034 * fold-const.c (negate_expr_p): Restrict negation of operand
8035 zero of a division to when we know that can happen without
8037 (fold_negate_expr_1): Likewise.
8039 2018-05-04 Jakub Jelinek <jakub@redhat.com>
8042 * real.h (real_nextafter): Declare.
8043 * real.c (real_nextafter): New function.
8044 * fold-const-call.c (fold_const_nextafter): New function.
8045 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
8046 CASE_CFN_NEXTTOWARD.
8047 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
8048 even when arg1_mode is different from arg0_mode.
8050 2018-05-03 Nathan Sidwell <nathan@acm.org>
8052 * doc/extend.texi (Deprecated Features): Remove
8054 (Backwards Compatibility): Likewise.
8055 * doc/invoke.texi (C++ Language Options): Likewise.
8056 (C++ Dialect Options): Likewise.
8058 2018-05-03 Jakub Jelinek <jakub@redhat.com>
8061 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
8062 _mm512_mask_mullox_epi64): New intrinsics.
8064 2018-05-03 Tom de Vries <tom@codesourcery.com>
8067 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8068 dump files): Add offload-tree.
8070 2018-05-03 Richard Biener <rguenther@suse.de>
8072 PR tree-optimization/85615
8073 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
8074 to loops not nested in BBs loop father to avoid creating multi-entry
8077 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8079 PR tree-optimization/70291
8080 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
8081 arguments. Change return type to tree. Emit libcall as a new
8082 statement rather than replacing existing one when inplace_p is true.
8083 (expand_complex_multiplication_components): New function.
8084 (expand_complex_multiplication): Expand floating-point complex
8085 multiplication using the above.
8086 (expand_complex_division): Rename inner_type parameter to type.
8087 Update expand_complex_libcall call-site.
8088 (expand_complex_operations_1): Update expand_complex_multiplication
8089 and expand_complex_division call-sites.
8091 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8094 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8095 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8096 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
8097 the highest significant bit of the shift count mask is clear. In
8098 check whether and[sq]i3 is needed verify that all significant bits
8099 of the shift count other than the highest are set.
8101 2018-05-02 Tom de Vries <tom@codesourcery.com>
8104 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
8105 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
8106 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
8107 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
8108 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
8109 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
8110 __builtin_goacc_parlevel_size.
8112 2018-05-02 Richard Biener <rguenther@suse.de>
8114 PR tree-optimization/85597
8115 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
8116 do not use split vect_get_vec_defs call but call vect_get_slp_defs
8119 2018-05-02 Tom de Vries <tom@codesourcery.com>
8122 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8123 dump files): Add ltrans-tree.
8125 2018-05-02 Tom de Vries <tom@codesourcery.com>
8128 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
8129 dump files): Add wpa-ipa.
8131 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
8133 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
8134 powerpc*-*-linux*paired* target.
8135 * config/rs6000/750cl.h: Delete.
8136 * config/rs6000/paired.h: Delete.
8137 * config/rs6000/paired.md: Delete.
8138 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
8140 * config/rs6000/rs6000-builtin.def: Remove paired float support.
8141 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
8142 comment. Remove paired float support.
8143 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
8144 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
8146 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
8147 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
8149 * config/rs6000/rs6000.c: Remove paired float support.
8150 (paired_expand_vector_init, paired_expand_vector_move,
8151 paired_emit_vector_compare, paired_emit_vector_cond_expr,
8152 (paired_expand_lv_builtin, paired_expand_stv_builtin,
8153 paired_expand_builtin, paired_expand_predicate_builtin,
8154 paired_init_builtins): Delete.
8155 * config/rs6000/rs6000.h: Remove paired float support.
8156 * config/rs6000/rs6000.md: Remove paired float support.
8157 (move_from_CR_ov_bit): Delete.
8158 * config/rs6000/rs6000.opt (mpaired): Delete.
8159 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
8160 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
8162 2018-05-02 Richard Biener <rguenther@suse.de>
8165 * gimplify.c (gimplify_save_expr): When in SSA form allow
8166 SAVE_EXPRs to compute to SSA vars.
8168 2018-05-02 Jakub Jelinek <jakub@redhat.com>
8171 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
8172 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
8173 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
8174 clobber operands[2], instead use a new pseudo. Formatting fixes.
8176 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
8178 PR tree-optimization/85586
8179 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
8180 exit early for statements in the same group if the accesses are
8183 2018-05-02 Tom de Vries <tom@codesourcery.com>
8186 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
8189 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
8191 PR tree-optimization/85143
8192 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
8194 2018-05-01 Tom de Vries <tom@codesourcery.com>
8197 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
8198 not found" error message.
8200 2018-05-01 Tom de Vries <tom@codesourcery.com>
8203 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
8204 * vec.c (test_ordered_remove_if): New function.
8205 (vec_c_tests): Call test_ordered_remove_if.
8206 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
8207 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
8208 * tree-vect-patterns.c (vect_pattern_recog_1): Use
8209 VEC_ORDERED_REMOVE_IF.
8211 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8213 PR tree-optimization/82665
8214 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
8215 pointer subtraction where arguments come from a memchr call.
8217 2018-05-01 Jakub Jelinek <jakub@redhat.com>
8219 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
8220 --push-state --as-needed and --pop-state instead of --as-needed and
8221 --no-as-needed if ld supports it.
8222 * configure: Regenerated.
8225 * doc/install.texi2html: Replace _002d with - and _002a with * in
8226 generated html files using sed.
8228 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8231 * gcc-rich-location.c (blank_line_before_p): New function.
8232 (use_new_line): New function.
8233 (gcc_rich_location::add_fixit_insert_formatted): New function.
8234 * gcc-rich-location.h
8235 (gcc_rich_location::add_fixit_insert_formatted): New function.
8237 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8239 * selftest.c (assert_streq): Rename "expected" and "actual" to
8240 "val1" and "val2". Extend NULL-handling to cover both inputs
8241 symmetrically, while still requiring both to be non-NULL for a pass.
8242 * selftest.h (assert_streq): Rename "expected" and "actual" to
8244 (ASSERT_EQ): Likewise.
8245 (ASSERT_EQ_AT): Likewise.
8246 (ASSERT_KNOWN_EQ): Likewise.
8247 (ASSERT_KNOWN_EQ_AT): Likewise.
8248 (ASSERT_NE): Likewise.
8249 (ASSERT_MAYBE_NE): Likewise.
8250 (ASSERT_MAYBE_NE_AT): Likewise.
8251 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
8252 the assertion to pass.
8253 (ASSERT_STREQ_AT): Likewise.
8255 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8257 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
8258 interaction with -pie.
8260 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8262 * selftest.h: Fix alphabetization of per-source-file selftest
8265 2018-04-30 Jason Merrill <jason@redhat.com>
8267 PR c++/61982 - dead stores to destroyed objects.
8268 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
8271 2018-04-30 Jason Merrill <jason@redhat.com>
8273 * tree.c (build_clobber): New.
8274 * tree.h: Declare it.
8275 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
8277 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8279 * diagnostic-show-locus.c (layout::layout): Update for
8280 location_get_source_line returning a char_span.
8281 (struct char_span): Move to input.h.
8282 (struct correction): Update for fields in char_span becoming
8284 (struct source_line): Update for location_get_source_line
8285 returning a char_span.
8286 (layout::print_line): Likewise.
8287 * edit-context.c (edited_file::print_content): Likewise.
8288 (edited_file::print_diff_hunk): Likewise.
8289 (edited_file::print_run_of_changed_lines): Likewise.
8290 (edited_file::get_num_lines): Likewise.
8291 (edited_line::edited_line): Likewise.
8292 * final.c (asm_show_source): Likewise.
8293 * input.c (location_get_source_line): Convert return type
8294 from const char * to char_span, losing the final "line_len"
8296 (dump_location_info): Update for the above.
8297 (get_substring_ranges_for_loc): Likewise. Use a char_span
8298 when handling the literal within the line.
8299 (test_reading_source_line): Update for location_get_source_line
8300 returning a char_span.
8301 * input.h (class char_span): Move here from
8302 diagnostic-show-locus.c, converting from a struct to a class.
8303 Make data members private.
8304 (char_span::operator bool): New.
8305 (char_span::length): New.
8306 (char_span::get_buffer): New.
8307 (char_span::operator[]): New.
8308 (char_span::subspan): Make const.
8309 (char_span::xstrdup): New.
8310 (location_get_source_line): Convert return type from const char *
8311 to char_span, losing the final "line_size" param.
8313 2018-04-30 Jan Hubicka <jh@suse.cz>
8315 * lto-wrapper.c (ltrans_priorities): New static var.
8316 (cmp_priority): New.
8317 (run_gcc): Read priorities and if doing parallel build order
8318 the Makefile by them.
8320 2018-04-30 David Malcolm <dmalcolm@redhat.com>
8322 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
8324 2018-04-30 Richard Biener <rguenther@suse.de>
8326 * tree-cfg.c (verify_address): Remove base argument, add
8327 flag whether to check TREE_ADDRESSABLE and do that.
8328 (verify_expr): Remove.
8329 (verify_types_in_gimple_reference): Add pieces from verify_expr.
8330 (verify_gimple_assign_single): Likewise.
8331 (verify_gimple_switch): Likewise.
8332 (verify_expr_location_1): Dereference tp once. Add (disabled)
8333 piece from verify_expr.
8334 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
8336 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8338 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
8340 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8342 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
8343 (small_data_pattern): Likewise.
8344 (arc_rewrite_small_data): Likewise.
8345 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
8346 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
8347 (get_symbol_alignment): New function.
8348 (legitimate_small_data_address_p): Likewise.
8349 (legitimate_scaled_address): Update, call
8350 legitimate_small_data_address_p.
8351 (output_sdata): New static variable.
8352 (arc_print_operand): Update how we handle small data operands.
8353 (arc_print_operand_address): Likewise.
8354 (arc_legitimate_address_p): Update, use
8355 legitimate_small_data_address_p.
8356 (arc_rewrite_small_data_p): Remove.
8357 (arc_rewrite_small_data_1): Likewise.
8358 (arc_rewrite_small_data): Likewise.
8359 (small_data_pattern): Likewise.
8360 (compact_sda_memory_operand): Update to use
8361 legitimate_small_data_address_p and get_symbol_alignment.
8362 (prepare_move_operands): Don't rewite sdata pattern.
8363 (prepare_extend_operands): Remove.
8364 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
8366 (zero_extendqisi2): Likewise.
8367 (zero_extendhisi2): Likewise.
8368 (extendqihi2): Likewise.
8369 (extendqisi2): Likewise.
8370 (extendhisi2): Likewise.
8374 * config/arc/constraints.md (Usd): Change it to memory constraint.
8376 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
8378 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
8379 as source of std instructions.
8380 * config/arc/arc.md (movsi_insn): Update pattern predicate to
8381 allow 6-bit constants as source for store instructions.
8382 (movdi_insn): Update instruction pattern to allow 6-bit constants
8383 as source for store instructions.
8385 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
8387 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
8389 2018-04-30 Nathan Sidwell <nathan@acm.org>
8390 Sandra Loosemore <sandra@codesourcery.com>
8392 * dumpfile.c (dump_open): Allow '-' for stdout.
8393 * doc/invoke.texi (Developer Options): Document dump filename
8394 determination early. Document stdin/stdout selection.
8396 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
8398 Microblaze Target: PIC data text relative
8400 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
8401 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
8403 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
8404 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
8405 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
8406 New addressing mode for data-text relative position indepenedent code.
8407 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
8408 'ADDRESS_SYMBOLIC_TXT_REL'.
8409 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
8410 (microblaze_legitimate_pic_operand): Exclude function calls from
8411 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
8412 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
8414 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8415 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
8416 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
8417 for 'address + offset'.
8418 (microblaze_expand_prologue): Add new function prologue call for
8420 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
8421 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
8422 table in case of TARGET_PIC_DATA_TEXT_REL.
8423 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
8424 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
8425 Add new macros 'UNSPEC_TEXT',
8426 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
8427 + exclude function calls from 'UNSPEC_PLT' in case of data text
8429 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8430 new target hook for generating address diff vector tables in case of
8432 * doc/tm.texi : Regenerate.
8433 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
8434 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
8435 of addr diff vector generation.
8436 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
8437 target hook definition.
8438 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
8439 Add default function for generate_pic_addr_diff_vec -> flag_pic.
8440 * doc/invoke.texi (Add new pic option): Add new microblaze pic
8441 option for data text relative.
8443 2018-04-30 Richard Biener <rguenther@suse.de>
8445 * tree-chrec.h (evolution_function_is_constant_p): Remove
8447 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
8449 2018-04-30 Richard Biener <rguenther@suse.de>
8452 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
8454 2018-04-30 Richard Biener <rguenther@suse.de>
8456 PR tree-optimization/28364
8457 PR tree-optimization/85275
8458 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
8459 copying first exit test.
8461 2018-04-28 Mark Wielaard <mark@klomp.org>
8463 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
8465 (dwarf_AT): Handle DW_AT_addr_base.
8466 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
8468 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
8471 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
8472 (*ashl<dwi>3_doubleword_mask_1): Ditto.
8473 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
8474 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
8476 2018-04-28 Richard Biener <rguenther@suse.de>
8478 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
8479 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
8480 to reflect use. Only add interesting stmts.
8482 2018-04-27 Martin Jambor <mjambor@suse.cz>
8485 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
8486 the jump function allows for passing through aggregate values.
8488 2018-04-27 David Malcolm <dmalcolm@redhat.com>
8490 * input.h (in_system_header_at): Convert from macro to inline
8492 (from_macro_expansion_at): Likewise.
8493 (from_macro_definition_at): Likewise.
8495 2018-04-27 Jeff Law <law@redhat.com>
8497 * config.gcc: Mark tile* targets as deprecated/obsolete.
8499 2018-04-27 Richard Biener <rguenther@suse.de>
8501 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
8504 2018-04-27 Richard Biener <rguenther@suse.de>
8506 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
8508 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
8510 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
8511 with Yd constraint. Set "preferred_for_speed" attribute from
8512 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
8514 (*movdi_internal): Ditto.
8515 (movti_interunit splitters): Remove
8516 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
8517 (movdi_interunit splitters): Ditto.
8518 * config/i386/constraints.md (Ye): Remove.
8519 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
8521 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8524 * config/aarch64/constraints.md (Usg): Limit to 31.
8527 2018-04-27 Jakub Jelinek <jakub@redhat.com>
8529 PR tree-optimization/85529
8530 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
8531 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
8532 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
8533 zero extension or masking of the MSB bit.
8534 (optimize_range_tests): Add FIRST_BB argument, pass it through
8535 to optimize_range_tests_var_bound.
8536 (maybe_optimize_range_tests, reassociate_bb): Adjust
8537 optimize_range_tests callers.
8539 2018-04-26 Richard Biener <rguenther@suse.de>
8540 Jakub Jelinek <jakub@redhat.com>
8542 * cgraph.h (symbol_table): Just declare debug method here.
8543 * symtab.c (symbol_table::debug): Define.
8545 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
8547 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
8549 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
8551 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
8552 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
8553 (*movdi_internal): Substitute Yi and Yj constraint with x
8554 and Ym and Yn constraint with y constraint. Update "isa"
8555 attribute and set "preferred_for_speed" attribute from
8556 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
8557 (*movsi_internal): Ditto.
8558 (*movdf_internal): Ditto.
8559 (*movsf_internal): Ditto.
8560 (*zero_extendsidi2): Ditto.
8561 * config/i386/sse.md (vec_set<mode>_0): Ditto.
8562 (sse2_loadld): Ditto.
8563 (*vec_extract<ssevecmodelower>_0): Ditto.
8564 (*vec_extractv4si_0_zext_sse4): Ditto.
8565 (vec_concatv2di): Ditto.
8566 (*vec_dup<mode>): Ditto.
8567 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
8568 * config/i386/constraints.md (Yi): Remove.
8573 2018-04-26 Nathan Sidwell <nathan@acm.org>
8575 * dumpfile.c (dump_open): New.
8576 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
8577 (dump_finish): Detect stdio/stderr by value not name.
8579 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
8581 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
8583 2018-04-26 Tom de Vries <tom@codesourcery.com>
8586 * config/nvptx/nvptx.c (verify_neutering_jumps)
8587 (verify_neutering_labels): New function
8588 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
8590 2018-04-26 Tom de Vries <tom@codesourcery.com>
8593 * config/nvptx/nvptx.c (needs_neutering_p): New function.
8594 (nvptx_single): Use needs_neutering_p to skip over insns that do not
8597 2018-04-26 Richard Biener <rguenther@suse.de>
8598 Tom de Vries <tom@codesourcery.com>
8601 * lto-streamer-out.c (output_function): Fixup loops if required to match
8602 discovery done in the reader.
8604 2018-04-26 Richard Biener <rguenther@suse.de>
8606 PR tree-optimization/85116
8607 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
8608 have a loop exit from the single latch predecessor. Remove
8609 case of header with just condition.
8610 (ch_base::copy_headers): Exclude infinite loops from any
8612 (pass_ch::execute): Record exits.
8614 2018-04-26 Richard Biener <rguenther@suse.de>
8616 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
8617 prologue cost vector and pass it to vect_get_load_cost.
8618 (vect_get_peeling_costs_all_drs): Likewise.
8619 (vect_peeling_hash_get_lowest_cost): Likewise.
8620 (vect_enhance_data_refs_alignment): Likewise.
8622 2018-04-26 Richard Biener <rguenther@suse.de>
8625 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
8626 checking of integer<->pointer conversions.
8627 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
8628 sign-/zero-extending pointer types.
8629 (expand_omp_for_static_chunk): Likewise.
8631 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
8632 Jean Lee <xiaoyur347@gmail.com>
8634 * config/mips/mips.c (mips_asan_shadow_offset): New function.
8635 (TARGET_ASAN_SHADOW_OFFSET): Define.
8636 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
8637 true for -fsanitize=address.
8639 2018-04-25 Mark Wielaard <mark@klomp.org>
8641 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
8644 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8646 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
8647 than "alu", remove explicit "memory" and "imm_disp" attributes.
8648 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
8651 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
8652 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
8653 gen_lowpart_no_emit.
8655 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
8658 * config/i386/i386.c (ix86_expand_builtin): Change memory
8659 operand to XI, extend p0 to Pmode.
8660 * config/i386/i386.md: Change unspec volatile and operand
8661 1 mode to XI, change operand 0 mode to P.
8663 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8665 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
8666 GET_MODE_MASK before any checking.
8667 (nds32_can_use_bset_p): Likewise.
8668 (nds32_can_use_btgl_p): Likewise.
8670 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8672 * config/nds32/nds32-doubleword.md: New define_split pattern for
8673 illegal register number.
8675 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8677 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
8679 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
8681 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
8683 2018-04-25 Richard Biener <rguenther@suse.de>
8685 * lto-streamer.h (LTO_major_version): Bump to 8.
8687 2018-04-25 Jakub Jelinek <jakub@redhat.com>
8689 * BASE-VER: Set to 9.0.0.
8691 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
8693 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
8694 in __abskf2 and __powikf2.
8696 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8699 * config/aarch64/constraints.md (Usg, Usj): New constraints.
8700 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
8701 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
8702 Use the above on operand 2. Reindent.
8703 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
8705 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8708 * common/config/i386/i386-common.c (ix86_handle_option): Don't
8710 * config/i386/i386.opt (mcet): Removed.
8711 * doc/install.texi: Remove -mcet documentation.
8712 * doc/invoke.texi: Likewise.
8714 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
8717 * doc/install.texi: Remove -mcet from bootstrap-cet.
8719 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8722 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
8723 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
8727 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
8728 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
8729 containing a CONST_VECTOR.
8731 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
8733 * doc/install.texi: Update newlib dependency for nvptx.
8735 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8738 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
8739 instead of INTVAL when shifting x left.
8741 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
8743 PR tree-optimization/85478
8744 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
8745 vect_grouped_store_supported for single element vectors.
8747 2018-04-24 Richard Biener <rguenther@suse.de>
8750 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
8751 load cost increase to the case of non-constant step.
8753 2018-04-24 Jakub Jelinek <jakub@redhat.com>
8756 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
8757 destination if any_malformed_asm.
8759 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
8762 * expr.c (store_field): In the bitfield case, if the value comes from
8763 a function call and is returned in registers by means of a PARALLEL,
8764 do not change the mode of the temporary unless BLKmode and VOIDmode.
8766 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
8768 PR rtl-optimization/85423
8769 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
8770 dependencies to debug insns when the previous insn is non-debug.
8772 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8774 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
8775 enums into a single definition.
8776 (fls): Fix predicates and printing.
8779 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
8781 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
8782 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
8783 and short u6 immediate.
8784 (check_if_valid_sleep_operand): Remove.
8785 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
8787 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8789 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
8790 flag_always_save_lp condition.
8791 * config/nds32/nds32.opt (malways-save-lp): New option.
8793 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8795 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
8796 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
8797 * config/nds32/nds32.h
8798 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
8799 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
8801 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8803 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
8804 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
8806 2018-04-22 Shiva Chen <shiva0217@gmail.com>
8807 Chung-Ju Wu <jasonwucj@gmail.com>
8809 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
8811 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
8812 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
8814 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8816 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
8818 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8820 * config/nds32/nds32-protos.h (nds32_data_alignment,
8821 nds32_local_alignment): Declare.
8822 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
8823 nds32_local_alignment): New functions.
8824 (TARGET_CONSTANT_ALIGNMENT): Define.
8825 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
8827 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8829 * config/nds32/nds32.c
8830 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
8831 (TARGET_MODES_TIEABLE_P): Likewise.
8833 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
8835 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
8838 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
8839 Chung-Ju Wu <jasonwucj@gmail.com>
8841 * config/nds32/constants.md (unspec_volatile_element): Add enum values
8842 for unaligned access.
8843 * config/nds32/nds32-intrinsic.c: Implementation of expanding
8845 * config/nds32/nds32-intrinsic.md: Likewise.
8846 * config/nds32/nds32_intrinsic.h: Likewise.
8847 * config/nds32/nds32.h (nds32_builtins): Likewise.
8848 * config/nds32/nds32.opt (munaligned-access): New option.
8849 * config/nds32/nds32.c (nds32_asm_file_start): Display
8850 flag_unaligned_access status.
8852 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8854 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
8855 -mno-relax is present.
8856 * config/riscv/linux.h (LINK_SPEC): Ditto.
8858 2018-04-20 Martin Sebor <msebor@redhat.com>
8861 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
8863 (gimple_fold_builtin_stxcpy_chk): Same.
8864 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
8866 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
8869 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
8870 __powikf2 when long double is IEEE 128-bit.
8872 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
8874 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
8875 step to make sure stack always aligned.
8877 2018-04-20 Carl Love <cel@us.ibm.com>
8880 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
8881 size check for arg0.
8883 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
8884 Tom de Vries <tom@codesourcery.com>
8887 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
8888 Emit insns for calls too.
8889 (nvptx_find_par): Always look for worker-level predecessor insn.
8890 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
8892 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
8893 (nvptx_process_pars): Propagate frames for calls.
8895 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
8898 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
8900 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
8901 (ix86_handle_option): Don't handle OPT_mibt.
8902 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
8904 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
8906 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
8907 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
8908 (ix86_target_macros): Define __CET__ with flag_cf_protection
8909 for -fcf-protection.
8910 * config/i386/i386.c (isa2_opts): Remove -mibt.
8911 * config/i386/i386.h (TARGET_IBT): Removed.
8912 (TARGET_IBT_P): Likewise.
8913 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
8914 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
8915 * config/i386/i386.opt (mcet): Update help message.
8918 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
8919 -mcet as an alias for -mshstk.
8921 2018-04-20 Richard Biener <rguenther@suse.de>
8924 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
8925 complexity by forcing a single use of the multiply operand.
8927 2018-04-20 Martin Jambor <mjambor@suse.cz>
8930 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
8931 recursion dependency to only apply to non-clones.
8933 2018-04-20 Martin Jambor <mjambor@suse.cz>
8936 * ipa-cp.c (create_specialized_node): Check that clones of
8937 self-recursive edges exist during IPA-CP.
8939 2018-04-19 Toon Moene <toon@moene.org>
8941 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
8944 2018-04-19 Jakub Jelinek <jakub@redhat.com>
8946 PR tree-optimization/85467
8947 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
8948 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
8949 VECTOR_CST element to type.
8951 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8954 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
8955 * config/i386/i386.md (builtin_setjmp_setup): Removed.
8956 (builtin_longjmp): Likewise.
8957 (save_stack_nonlocal): New pattern.
8958 (restore_stack_nonlocal): Likewise.
8960 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8963 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8964 Replace ASM_OUTPUT_LABEL with fprintf.
8966 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
8969 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
8970 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
8971 * config/i386/i386-c.c (ix86_target_macros_internal): Also
8972 define __IBT__ and __SHSTK__ for -fcf-protection.
8973 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
8975 (ix86_trampoline_init): Likewise.
8976 (x86_output_mi_thunk): Likewise.
8977 (ix86_notrack_prefixed_insn_p): Likewise.
8978 (ix86_option_override_internal): Don't disallow -fcf-protection.
8979 * config/i386/i386.md (rdssp<mode>): Also enable for
8981 (incssp<mode>): Likewise.
8982 (nop_endbr): Likewise.
8983 * config/i386/i386.opt (mcet): Change help message to built-in
8987 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
8988 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
8989 enable CET built-in functions.
8991 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
8993 * common/config/i386/i386-common.c
8994 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
8995 OPTION_MASK_ISA_MOVDIRI_UNSET,
8996 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
8997 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
8998 * config.gcc (movdirintrin.h): New header.
8999 * config/i386/cpuid.h (bit_MOVDIRI,
9000 bit_MOVDIR64B): New bits.
9001 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
9003 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
9004 (VOID, PVOID, PCVOID)): New function types.
9005 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
9006 __builtin_ia32_directstoreu_u64,
9007 __builtin_ia32_movdir64b): New builtins.
9008 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
9009 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
9011 (ix86_valid_target_attribute_inner_p): Ditto.
9012 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
9013 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
9014 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
9015 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
9016 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
9017 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
9018 (movdiri<mode>, movdir64b_<mode>): New.
9019 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
9020 * config/i386/immintrin.h: Include movdirintrin.h.
9021 * config/i386/movdirintrin.h: New file.
9022 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
9024 2018-04-19 Richard Biener <rguenther@suse.de>
9027 * cfg.c (clear_bb_flags): When loop state says we have
9028 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
9030 2018-04-19 Richard Biener <rguenther@suse.de>
9032 PR tree-optimization/84737
9033 * tree-vect-data-refs.c (vect_copy_ref_info): New function
9034 copying restrict info.
9035 (vect_setup_realignment): Use it.
9036 * tree-vectorizer.h (vect_copy_ref_info): Declare.
9037 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
9038 the first DR to all generated stores.
9039 (vectorizable_load): Likewise for loads.
9041 2018-04-19 Jakub Jelinek <jakub@redhat.com>
9043 PR tree-optimization/85446
9044 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
9045 the integral and pointer types to have the same precision.
9047 * doc/install.texi: Document --disable-cet being the default and
9050 2018-04-18 Martin Liska <mliska@suse.cz>
9052 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
9055 2018-04-18 Martin Liska <mliska@suse.cz>
9058 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9061 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9062 arguments if they are comparable.
9064 2018-04-18 Martin Liska <mliska@suse.cz>
9067 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
9070 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
9073 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
9076 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
9077 ENDBR after calling __morestack.
9079 2018-04-18 David Malcolm <dmalcolm@redhat.com>
9082 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
9083 by using gcc_base_ver to generate a gcc_driver_version, and use
9084 it when generating GCC_DRIVER_NAME.
9085 * configure: Regenerate.
9087 2018-04-18 Jakub Jelinek <jakub@redhat.com>
9090 * config.gcc: Obsolete powerpc*-*-*spe*.
9092 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9095 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
9096 (stabstr_D): Change type of unum from unsigned int to
9097 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
9100 2018-04-17 Jim Wilson <jimw@sifive.com>
9103 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
9104 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
9105 Set arg_pointer_offset after using pretend_args_size.
9107 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9109 PR rtl-optimization/85431
9110 * dse.c (record_store): Ignore zero width stores.
9113 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
9114 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
9115 __builtin_stack_restore rather than after it.
9116 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
9117 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
9118 argument instead of virtual_dynamic_stack_rtx.
9120 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
9122 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
9124 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
9125 Add note to error message to explain internal mapping of overloaded
9126 built-in function name to non-overloaded built-in function name.
9127 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
9130 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
9133 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
9134 where the inputs overlap with the output.
9136 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9138 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
9139 (=v, v) alternative and explicit "memory" attribute.
9140 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
9141 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9143 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
9144 "sselog1" type instead of "sselog".
9145 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
9146 "sselog". Remove explicit "memory" attribute.
9147 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
9148 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9150 (vec_extract_hi_v32hi): Merge all alternatives into one, use
9151 "sselog1" type instead of "sselog". Remove explicit "memory"
9153 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
9154 use "sselog1" type instead of "sselog". Remove explicit "memory"
9156 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
9157 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
9159 (vec_extract_hi_v64qi): Merge all alternatives into one, use
9160 "sselog1" type instead of "sselog". Remove explicit "memory"
9162 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
9163 use "sselog1" type instead of "sselog". Remove explicit "memory"
9167 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
9170 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
9173 2018-04-17 Martin Jambor <mjambor@suse.cz>
9176 * ipa-cp.c (create_specialized_node): Call
9177 expand_all_artificial_thunks if necessary.
9179 2018-04-17 Martin Liska <mliska@suse.cz>
9182 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
9183 in message, remote space in between '_G' and '('.
9185 2018-04-17 Jakub Jelinek <jakub@redhat.com>
9188 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
9189 avx512f_vmcmp<mode>3<round_saeonly_name>,
9190 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
9191 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
9192 avx512f_rndscale<mode><round_saeonly_name>,
9193 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
9194 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
9195 Use %<iptr>2 instead of %2 for -masm=intel.
9196 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
9197 avx512f_vcvttss2usi<round_saeonly_name>,
9198 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
9200 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
9201 avx512f_vcvttsd2usi<round_saeonly_name>,
9202 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
9203 Use %q1 instead of %1 for -masm=intel.
9204 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
9205 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
9206 of %3 for -masm=intel.
9207 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
9209 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
9211 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
9213 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
9214 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
9216 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
9218 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
9219 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
9220 %g1 and one with %0 and %1.
9221 (avx512er_vmrcp28<mode><round_saeonly_name>,
9222 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
9224 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
9225 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
9226 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
9227 of %0 and %{%4%} for -masm=intel.
9228 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
9229 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
9230 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
9231 order of %0 and %{%5%}%{z%} for -masm=intel.
9233 2018-04-17 Jan Hubicka <jh@suse.cz>
9236 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
9238 2018-04-17 Martin Liska <mliska@suse.cz>
9241 * multiple_target.c (create_dispatcher_calls): Set apostrophes
9242 for target_clone error message. Make default implementation
9243 clone to be a local declaration.
9244 (separate_attrs): Add new argument and check for an empty
9246 (expand_target_clones): Handle it.
9247 (ipa_target_clone): Make redirection just for target_clones
9250 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
9251 Tom de Vries <tom@codesourcery.com>
9254 * omp-expand.c (expand_oacc_for): Add dummy false branch for
9255 tiled basic blocks without omp continue statements.
9257 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
9260 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
9261 vec_extract expression as having side effects to make sure it gets
9264 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
9267 * config/i386/i386.c (get_builtin_code_for_version): Check
9270 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
9273 * gcc/config.gcc: Support "skylake".
9274 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
9276 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
9277 (processor_target_table): Add "skylake".
9278 (ix86_option_override_internal): Add "skylake".
9279 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
9280 PROCESSOR_CANNONLAKE.
9281 (get_builtin_code_for_version): Fix priority for
9282 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
9283 PROCESSOR_SKYLAKE-AVX512.
9284 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
9285 (processor_type): Add PROCESSOR_SKYLAKE.
9287 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
9288 Jason Merrill <jason@redhat.com>
9291 * convert.c (convert_to_integer_1): Use direct recursion for
9292 enumeral types and types with a precision less than the number
9293 of bits in their mode.
9295 2018-04-16 Julia Koval <julia.koval@intel.com>
9298 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
9299 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
9301 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
9304 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
9305 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
9306 and -mno-direct-move.
9308 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
9311 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
9312 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
9313 Ensure negative shifts result in {0}.
9315 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
9317 PR rtl-optimization/79916
9318 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
9319 regs (if any) to define how to gnerate SD moves when LRA is in
9322 2018-04-13 Jakub Jelinek <jakub@redhat.com>
9324 PR rtl-optimization/85393
9325 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
9326 * except.c (expand_dw2_landing_pad_for_region): Make static.
9327 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
9328 a label and unconditional jump to old_bb, rather than
9329 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
9332 PR rtl-optimization/85376
9333 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
9334 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
9335 instead of a specific value.
9337 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9338 Bin Cheng <bin.cheng@arm.com>
9340 PR tree-optimization/82965
9341 PR tree-optimization/83991
9342 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
9343 by_profile_only parameter.
9344 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
9345 information if the loop was predicted to iterate too many times.
9346 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
9348 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
9351 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
9354 2018-04-13 Martin Liska <mliska@suse.cz>
9355 Jakub Jelinek <jakub@redhat.com>
9358 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
9359 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
9360 * builtins.c (expand_builtin_memory_copy_args): Use
9361 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
9362 handle dest_addr == pc_rtx.
9364 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
9367 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
9368 asked to not generate direct moves.
9369 (fix_trunc<mode>si2_stfiwx): Similar.
9370 (fix_trunc<mode>si2_internal): Similar.
9372 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9375 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
9376 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
9377 lookup if dest in some wider mode is known to be const0_rtx and
9378 if so, record permanent equivalence for it to be ZERO_EXTEND of
9379 the narrower mode destination.
9381 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9383 * lto-streamer-out.c (output_function): Revert 259346.
9384 * omp-expand.c (expand_oacc_for): Likewise.
9386 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
9388 PR rtl-optimization/85354
9389 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
9390 * sel-sched.c (sel_global_init): ... here.
9392 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
9395 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
9396 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
9397 mode for PE-COFF targets.
9398 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
9399 (i386_pe_asm_lto_end): Likewise.
9400 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
9401 (TARGET_ASM_LTO_END): Likewise.
9402 * config/i386/winnt.c (saved_debug_info_level): New static variable.
9403 (i386_pe_asm_lto_start): New function.
9404 (i386_pe_asm_lto_end): Likewise.
9406 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
9407 Richard Biener <rguenther@suse.de>
9410 * lto-streamer-out.c (output_function): Fix CFG loop state before
9412 * omp-expand.c (expand_oacc_for): Handle calls to internal
9413 functions like regular functions.
9415 2018-04-12 Richard Biener <rguenther@suse.de>
9418 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
9419 for the early LTO debug to properly generate references to it
9420 during DIE emission. Do not re-use that for the skeleton for
9422 (dwarf2out_early_finish): Likewise.
9424 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9427 * config/i386/sse.md
9428 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
9429 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
9430 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
9431 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9432 and output is a reg, avoid creating invalid lowpart subreg, but
9433 instead split into a 512-bit move. Don't split if not AVX512VL,
9434 input is xmm16+ reg and output is a mem.
9435 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
9436 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
9437 xmm16+ reg and output is a mem.
9439 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
9441 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
9442 also for flag_dwarf2_cfi_asm.
9444 2018-04-12 Jakub Jelinek <jakub@redhat.com>
9446 PR rtl-optimization/85342
9447 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
9448 a bool scalar var inside of the loop instead. Don't try to update
9449 recog_data.operand after failed apply_change_group.
9451 2018-04-12 Tom de Vries <tom@codesourcery.com>
9454 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
9455 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
9456 array with flexible array member as array without given dimension.
9457 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
9458 argument for undefined param to true.
9460 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
9463 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9464 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
9465 from PowerPC section.
9466 * config/rs6000/sysv4.opt (mcall-): Improve help text.
9467 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
9468 help text that is too long.
9469 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
9470 help text that is too long.
9471 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
9472 help text that is too long.
9474 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
9476 * config/alpha/alpha.md (stack_probe_internal): Rename
9477 from "probe_stack". Update all callers.
9479 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9481 PR rtl-optimization/84566
9482 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
9483 sched_macro_fuse_insns.
9485 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9488 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
9489 (compute_block_dependences): ... from here.
9491 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9493 PR tree-optimization/85331
9494 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
9495 from int to HOST_WIDE_INT.
9497 2018-04-11 Martin Jambor <mjambor@suse.cz>
9500 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
9501 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
9502 not the same as the source val.
9503 (cgraph_edge_brings_value_p): New parameter.
9504 (gather_edges_for_value): Pass destination value to
9505 cgraph_edge_brings_value_p.
9506 (perhaps_add_new_callers): Likewise.
9507 (get_info_about_necessary_edges): Likewise and exclude values brought
9508 only by self-recursive edges.
9509 (create_specialized_node): Redirect only clones of self-calling edges.
9510 (+self_recursive_pass_through_p): New function.
9511 (find_more_scalar_values_for_callers_subset): Use it.
9512 (find_aggregate_values_for_callers_subset): Likewise.
9513 (known_aggs_to_agg_replacement_list): Removed.
9514 (decide_whether_version_node): Re-calculate known constants for all
9515 remaining context clones.
9517 2018-04-11 Richard Biener <rguenther@suse.de>
9520 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
9521 from early DWARF output.
9522 (dwarf2out_early_finish): Output line info unconditionally into
9523 early DWARF and add reference to it.
9525 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9528 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
9529 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
9530 other than V2DFmode using iptr mode attribute.
9531 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
9533 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
9535 PR rtl-optimization/84659
9536 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
9538 2018-04-11 Jakub Jelinek <jakub@redhat.com>
9541 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
9543 (output_loc_list): Pass address of a dummy size variable even in the
9544 locview handling loop.
9545 (index_location_lists): Add comment on why skip_loc_list_entry can't
9548 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
9551 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
9554 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
9557 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
9558 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
9559 and -mstring-compare-inline-limit.
9561 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9564 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
9565 for stack clash protection in a register whenever we need it to be in
9568 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9570 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
9571 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
9573 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
9576 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
9578 (mlong-double-): Ditto.
9579 * config/rs6000/sysv4.opt (msdata=): Ditto.
9580 (mtls-size=): Ditto.
9582 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
9584 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
9585 erroneous entries for
9586 "vector int vec_ldl (int, long int *)", and
9587 "vector unsigned int vec_ldl (int, unsigned long int *)".
9588 Add comments and entries for
9589 "vector bool char vec_ldl (int, bool char *)",
9590 "vector bool short vec_ldl (int, bool short *)",
9591 "vector bool int vec_ldl (int, bool int *)",
9592 "vector bool long long vec_ldl (int, bool long long *)",
9593 "vector pixel vec_ldl (int, pixel *)",
9594 "vector long long vec_ldl (int, long long *)",
9595 "vector unsigned long long vec_ldl (int, unsigned long long *)".
9596 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
9597 type tree bool_long_long_type_node and correct definition of
9598 bool_V2DI_type_node to make reference to this new type tree.
9599 (rs6000_mangle_type): Replace erroneous reference to
9600 bool_long_type_node with bool_long_long_type_node.
9601 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
9602 comments to emphasize sign distinctions for char and int types and
9603 replace RS6000_BTI_bool_long constant with
9604 RS6000_BTI_bool_long_long constant. Also add comment to restrict
9605 use of RS6000_BTI_pixel.
9606 (bool_long_type_node): Remove this macro definition.
9607 (bool_long_long_type_node): New macro definition
9609 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9611 PR rtl-optimization/85300
9612 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
9613 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
9614 simplify_unary_operation fails.
9616 2018-04-10 Martin Liska <mliska@suse.cz>
9618 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
9619 cgraph_edge and ipa_ref.
9621 2018-04-10 Jakub Jelinek <jakub@redhat.com>
9625 * config/i386/sse.md
9626 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
9627 computation of the VEC_MERGE selector from mask.
9628 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
9629 Fix decoding of the VEC_MERGE selector into mask.
9631 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9633 PR tree-optimization/85286
9634 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
9636 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
9638 * final.c (final_1): Set insn_last_address as well as
9639 insn_current_address.
9641 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9644 * explow.c (emit_stack_probe): Call validize_mem on memory location
9645 before passing it to gen_probe_stack. Create address operand and
9646 legitimize it for the probe_stack_address case.
9648 2018-04-09 Jan Hubicka <jh@suse.cz>
9651 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
9652 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
9653 * tree.c (free_lang_data_in_type): Fix handling of binfos;
9655 (free_lang_data): Rebuild type inheritance graph.
9657 2018-04-09 Martin Sebor <msebor@redhat.com>
9659 * invoke.texi (-finline-small-functions): Mention other optimization
9661 (-findirect-inlining, -fpartial-inlining): Same.
9662 (-finline-functions-called-once): Same.
9663 (-freorder-blocks-and-partition): Same.
9665 2018-04-09 Jan Hubicka <jh@suse.cz>
9668 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
9669 jumps; choose last target that matches the criteria (i.e.
9670 no partition changes for non-crossing jumps).
9671 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
9672 support for redirecting crossing jumps to non-crossing.
9674 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
9676 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
9677 also for naked functions.
9679 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
9681 * config/arc/arc.md (add_shift): New pattern.
9682 (add_shift2): Likewise.
9683 (sub_shift): Likewise.
9684 (sub_shift_cmp0_noout): Likewise.
9685 (compare_si_ashiftsi): Likewise.
9686 (xbfu_cmp0_noout): New combine pattern.
9687 (xbfu_cmp0"): Likewise.
9688 (movsi_set_cc_insn): Place the predicable variant first.
9689 (commutative_binary_cmp0_noout): Remove clobber.
9690 (commutative_binary_cmp0): New pattern.
9691 (noncommutative_binary_cmp0): Likewise.
9692 (noncommutative_binary_cmp0_noout): Likewise.
9693 (noncommutative_binary_comparison_result_used): Removed.
9694 (rsub_cmp0): New pattern.
9695 (rsub_cmp0_noout): Likewise.
9696 (extzvsi): Changed, keep only meaningful variants.
9697 (SQH, SEZ): New iterators.
9698 (SQH_postfix): New mode attribute.
9699 (SEZ_prefix): New code attribute.
9700 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
9701 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
9702 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
9704 (noncommutative_operator): Check the availability of barrel
9707 2018-04-09 Richard Biener <rguenther@suse.de>
9709 PR tree-optimization/85284
9710 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
9711 Only use the niter constraining form of simple_iv when the exit
9714 2018-04-09 Tom de Vries <tom@codesourcery.com>
9717 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
9718 (define_expand "*memory_barrier"): New define_expand.
9719 (define_insn "memory_barrier"): New insn.
9721 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9723 PR rtl-optimization/80463
9724 PR rtl-optimization/83972
9725 PR rtl-optimization/83480
9727 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
9728 correct producer for the insn.
9729 (tidy_control_flow): Fixup seqnos in case of debug insns.
9731 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9733 PR rtl-optimization/83913
9735 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
9736 different sched-times when merging exprs.
9738 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9740 PR rtl-optimization/83962
9742 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
9743 tidy_fallthru_edge and tidy_control_flow.
9745 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
9747 PR rtl-optimization/83530
9749 * sel-sched.c (force_next_insn): New global variable.
9750 (remove_insn_for_debug): When force_next_insn is true, also leave only
9751 next insn in the ready list.
9752 (sel_sched_region): When the region wasn't scheduled, make another pass
9753 over it with force_next_insn set to 1.
9755 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9757 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
9759 * config/nds32/constants.md (unspec_volatile_element): Add enum values
9760 for interrupt control.
9761 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9762 functions for interrupt control.
9763 * config/nds32/nds32-intrinsic.md: Likewise.
9764 * config/nds32/nds32_intrinsic.h: Likewise.
9765 * config/nds32/nds32.h (nds32_builtins): Likewise.
9767 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9769 * config/nds32/nds32.c (nds32_init_machine_status,
9770 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
9771 strict_aligned_p field.
9772 (nds32_expand_to_rtl_hook): New function.
9773 (TARGET_EXPAND_TO_RTL_HOOK): Define.
9774 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
9776 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9777 Chung-Ju Wu <jasonwucj@gmail.com>
9779 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
9780 * config/nds32/nds32-n7.md: New file.
9781 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
9782 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
9784 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
9785 * config/nds32/nds32.md (pipeline_model): Add n7.
9786 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
9787 * config/nds32/pipelines.md: Include n7 settings.
9789 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9790 Chung-Ju Wu <jasonwucj@gmail.com>
9792 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
9793 * config/nds32/nds32-e8.md: New file.
9794 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
9795 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
9797 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
9798 * config/nds32/nds32.md (pipeline_model): Add e8.
9799 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
9800 * config/nds32/pipelines.md: Include e8 settings.
9802 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9803 Chung-Ju Wu <jasonwucj@gmail.com>
9805 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
9806 * config/nds32/nds32-n8.md: New file.
9807 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
9808 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
9810 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
9811 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
9812 * config/nds32/nds32.md (pipeline_model): Add n8.
9813 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
9814 * config/nds32/pipelines.md: Include n8 settings.
9816 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
9817 Chung-Ju Wu <jasonwucj@gmail.com>
9819 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
9820 * config/nds32/nds32-n9-2r1w.md: New file.
9821 * config/nds32/nds32-n9-3r2w.md: New file.
9822 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
9823 nds32_register_ports): New or modify for cpu n9.
9824 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
9826 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
9827 * config/nds32/nds32-utils.c: New file.
9828 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
9829 TARGET_MUL_SLOW): Define.
9830 * config/nds32/nds32.md (pipeline_model): New attribute.
9831 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
9832 New options that support cpu n9.
9833 * config/nds32/pipelines.md: Include n9 settings.
9834 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
9836 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
9838 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
9839 information if necessary.
9840 (output_cond_branch_compare_zero): Likewise.
9841 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
9842 (nds32_target_alignment): Refine for alignment.
9843 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
9844 (FUNCTION_BOUNDARY): Modify.
9845 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
9847 * config/nds32/nds32.opt (malways-align, malign-functions): New.
9849 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
9851 * config/nds32/constants.md (unspec_volatile_element): Add values for
9852 TLB operation and data prefetch.
9853 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
9854 functions for TLB operation and data prefetch.
9855 * config/nds32/nds32-intrinsic.md: Likewise.
9856 * config/nds32/nds32_intrinsic.h: Likewise.
9857 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
9858 (nds32_print_operand): Likewise.
9859 * config/nds32/nds32.h (nds32_builtins): Likewise.
9861 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
9862 Andrew Pinski <pinsika@gcc.gnu.org>
9865 * match.pd: Use constant_boolean_node of correct type instead of
9866 boolean_true_node or boolean_false_node for simplifying
9867 pointer comparisons to zero.
9869 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9871 PR tree-optimization/80021
9872 * tree.c (verify_type_variant): Make error call in verify_variant_match
9873 translatable and remove final full stop.
9875 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9877 * config/nds32/constants.md (unspec_volatile_element): Add
9878 UNSPEC_VOLATILE_EH_RETURN.
9879 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
9880 nds32_output_stack_pop): Support dwarf exception handling process.
9881 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
9882 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
9883 exception handling process.
9884 (nds32_compute_stack_frame): Likewise.
9885 (nds32_return_addr_rtx): Likewise.
9886 (nds32_initial_elimination_offset): Likewise.
9887 (nds32_expand_prologue): Likewise.
9888 (nds32_expand_epilogue): Likewise.
9889 (nds32_dynamic_chain_address): New function.
9890 * config/nds32/nds32.h (machine_function): Add fields for dwarf
9892 (DYNAMIC_CHAIN_ADDRESS): Define.
9893 (EH_RETURN_DATA_REGNO): Define.
9894 (EH_RETURN_STACKADJ_RTX): Define.
9895 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
9896 patterns for dwarf exception handling.
9898 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9900 * config/nds32/nds32.h: Clean up obsolete macros.
9902 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9904 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9905 Add enum values for particular instructions.
9906 * config/nds32/nds32-intrinsic.c: Implementation of expanding
9907 particular intrinsic functions.
9908 * config/nds32/nds32-intrinsic.md: Likewise.
9909 * config/nds32/nds32_intrinsic.h: Likewise.
9910 * config/nds32/nds32.h (nds32_builtins): Likewise.
9911 * config/nds32/nds32.md (type): Add pbsad and pbsada.
9912 (btst, ave): New patterns for particular instructions.
9914 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9916 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
9917 Add enum values for atomic load/store and memory sync.
9918 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
9920 * config/nds32/nds32-intrinsic.md: Likewise.
9921 * config/nds32/nds32_intrinsic.h: Likewise.
9922 * config/nds32/nds32.h (nds32_builtins): Likewise.
9924 2018-04-07 Jakub Jelinek <jakub@redhat.com>
9926 PR tree-optimization/85257
9927 * fold-const.c (native_encode_vector): If not all elts could fit
9928 and off is -1, return 0 rather than offset.
9929 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
9930 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
9931 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
9932 adjust buffer in native_interpret_expr call.
9934 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9936 * config/nds32/constants.md (unspec_volatile_element): Add cache
9937 control enum values.
9938 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
9939 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
9940 * config/nds32/nds32.c (nds32_cctl_names): New.
9941 (nds32_print_operand): Handle cache control register names.
9942 * config/nds32/nds32.h (nds32_builtins): New enum values.
9943 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
9945 * config/nds32/nds32.md (type): Add mmu.
9946 * config/nds32/pipelines.md (simple_insn): Add mmu.
9948 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9950 * config/nds32/nds32.md (type): Remove call.
9951 * config/nds32/pipelines.md (simple_insn): Likewise.
9953 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9955 * config/nds32/constants.md (unspec_volatile_element): Add
9956 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
9957 UNSPEC_VOLATILE_FMFCFG.
9958 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
9959 description for fmfcfg and fmfcsr.
9960 (bdesc_1arg): Add fmtcsr.
9961 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
9962 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
9963 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
9964 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
9965 unspec_fmfcfg): New patterns.
9966 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
9967 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
9968 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
9969 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
9970 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
9971 __nds32__fmfcfg): Define.
9973 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
9975 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
9976 intrinsic register names.
9977 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
9978 intrinsic register enum values and macros.
9980 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
9982 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
9983 for load/store addressing form.
9984 (nds32_print_operand_address): Likewise.
9986 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
9989 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
9990 based on LABEL_REF. Remove useless assertion.
9991 (pic_address_needs_scratch): Fix formatting.
9992 (sparc_legitimize_pic_address): Minor tweaks.
9993 (sparc_delegitimize_address): Adjust assertion accordingly.
9994 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
9995 into symbolic_operand.
9996 (movsi_high_pic_label_ref): Likewise.
9997 (movsi_lo_sum_pic_label_ref): Likewise.
9998 (movdi_pic_label_ref): Likewise.
9999 (movdi_high_pic_label_ref): Likewise.
10000 (movdi_lo_sum_pic_label_ref): Likewise.
10002 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
10004 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
10005 custom LIB_SPEC setup.
10007 2018-04-06 Ruslan Bukin <br@bsdpad.com>
10008 Kito Cheng <kito.cheng@gmail.com>
10010 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
10011 * config/riscv/freebsd.h: New.
10013 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10015 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
10016 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
10019 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10020 Kito Cheng <kito.cheng@gmail.com>
10022 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
10023 nds32_output_call, nds32_symbol_binds_local_p): New functions.
10024 * config/nds32/nds32-protos.h (nds32_output_call,
10025 nds32_output_return): Declare.
10026 * config/nds32/nds32.md: Refine all the call and return patterns.
10028 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10031 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
10032 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
10034 PR rtl-optimization/84872
10035 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
10036 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
10037 EDGE_CROSSING edge.
10039 2018-04-06 Tamar Christina <tamar.christina@arm.com>
10041 * expr.c (copy_blkmode_to_reg): Revert 254862.
10042 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
10044 2018-04-06 Richard Biener <rguenther@suse.de>
10046 PR middle-end/85244
10047 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
10048 after seeing a component reference with an adjacent field. Treat
10049 refs to arrays at struct end of external decls similar to
10050 refs to unconstrained commons.
10052 2018-04-06 Jakub Jelinek <jakub@redhat.com>
10055 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
10056 look through SAVE_EXPRs with non-side-effects argument. Adjust
10058 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
10061 2018-04-06 Richard Biener <rguenther@suse.de>
10063 PR middle-end/85180
10064 * alias.c (find_base_term): New wrapper around find_base_term
10065 unwinding CSELIB_VAL_PTR changes.
10066 (find_base_term): Do not restore CSELIB_VAL_PTR during the
10069 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10071 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
10073 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
10074 constant definitions.
10075 ("nop"): lr 0,0 -> nopr r0
10076 ("nop_lr0", "nop_lr1"): New insn definitions.
10078 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
10080 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
10081 NDS32_V3PUSH_AVAILABLE_P macro.
10083 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
10084 Chung-Ju Wu <jasonwucj@gmail.com>
10086 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
10087 (nds32*-*-*): Add float and fpu_config into supported_defaults.
10088 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
10089 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
10090 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
10091 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
10092 * config/nds32/constraints.md: New constraints and checking for hard
10093 float configuration.
10094 * config/nds32/iterators.md: New mode iterator and attribute for hard
10095 float configuration.
10096 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
10098 * config/nds32/nds32-fpu.md: New file.
10099 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
10100 deal with hard float code generation.
10101 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
10103 (abi_type, float_reg_number): New enum type.
10104 * config/nds32/nds32-predicates.c: New predicates for hard float.
10105 * config/nds32/nds32-protos.h: Declare functions for hard float.
10106 * config/nds32/nds32.c: Implementation for hard float configuration.
10107 * config/nds32/nds32.h: Definitions for hard float configuration.
10108 * config/nds32/nds32.md: Include hard float machine description and
10109 modify patterns for hard float configuration.
10110 * config/nds32/nds32.opt: New options for hard float configuration.
10111 * config/nds32/predicates.md: New predicates for hard float
10114 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10116 * common/config/nds32/nds32-common.c
10117 (nds32_option_optimization_table): Enable -mreleax-hint by default.
10119 2018-04-05 Jakub Jelinek <jakub@redhat.com>
10121 PR middle-end/85195
10122 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
10123 CONSTRUCTOR_ELT (ctor, ...)->value.
10125 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
10128 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
10130 2018-04-05 Tom de Vries <tom@codesourcery.com>
10133 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
10136 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10137 Kito Cheng <kito.cheng@gmail.com>
10139 * config/nds32/constraints.md (U33): Fine-tune checking condition.
10140 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
10141 * config/nds32/nds32.h (nds32_16bit_address_type): Add
10142 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
10144 2018-04-05 Shiva Chen <shiva0217@gmail.com>
10145 Kito Cheng <kito.cheng@gmail.com>
10147 * config/nds32/constraints.md (Ufe): New memory constraint.
10148 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
10149 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
10150 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
10152 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
10153 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
10155 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10157 * config/nds32/nds32.md: Use optimize_size in the condition for
10158 alu-shift instructions.
10160 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10162 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
10164 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10166 * config/nds32/nds32.md (negsi2): Refine pattern.
10168 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10169 Chung-Ju Wu <jasonwucj@gmail.com>
10171 * config/nds32/iterators.md (shift_rotate): New code iterator.
10172 (shift): New code attribute.
10173 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
10174 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
10175 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
10176 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
10177 bit-wise operations.
10178 (andsi3, *andsi3): Ditto.
10179 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
10180 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
10181 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
10182 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
10183 nds32_ior_operand, nds32_xor_operand): New predicates.
10185 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10187 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
10188 (addsi3, subsi3): ... this.
10190 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10192 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
10194 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10196 * config/nds32/nds32.md: Adjust indention.
10198 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
10200 * config/nds32/nds32.md (feature): New attribute.
10202 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
10204 * config/nds32/nds32.md (subtype): New attribute.
10206 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
10209 * config/arm/arm-builtins.c (arm_expand_builtin): Change
10210 expansion to perform a bitwise AND of the argument followed by a
10211 boolean negation of the result.
10213 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
10215 PR rtl-optimization/84878
10216 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
10217 the basic block. Assert the use reference is not artificial and that
10218 it has an associated insn.
10220 2018-04-04 Michael Matz <matz@suse.de>
10222 * builtins.c (compute_objsize): Pass correct operand
10223 to array_at_struct_end_p.
10225 2018-04-04 Richard Biener <rguenther@suse.de>
10228 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
10229 from contexts for DINFO_LEVEL_TERSE and below.
10231 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10233 * config/nds32/nds32-doubleword.md (move_<mode>): Require
10234 resiter_operand condition.
10235 * config/nds32/nds32.md (*move<mode>): Ditto.
10237 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10238 Monk Chiang <sh.chiang04@gmail.com>
10240 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
10242 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10244 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
10246 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10247 Kito Cheng <kito.cheng@gmail.com>
10249 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
10250 nds32_cond_code_str, output_cond_branch,
10251 output_cond_branch_compare_zero, nds32_expand_cbranch,
10252 nds32_expand_cstore, nds32_expand_movcc,
10253 nds32_output_cbranchsi4_equality_zero,
10254 nds32_output_cbranchsi4_equality_reg,
10255 nds32_output_cbranchsi4_equality_reg_or_const_int,
10256 nds32_output_cbranchsi4_greater_less_zero: New functions.
10257 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
10258 nds32_expand_cstore, nds32_expand_movcc,
10259 nds32_output_cbranchsi4_equality_zero,
10260 nds32_output_cbranchsi4_equality_reg,
10261 nds32_output_cbranchsi4_equality_reg_or_const_int,
10262 nds32_output_cbranchsi4_greater_less_zero): Declare.
10263 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
10264 nds32_rimm11s_operand): New predicates.
10265 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
10266 * config/nds32/nds32.md: Rewrite all the branch and conditional move
10269 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
10271 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
10272 * config/nds32/nds32.md: Ditto.
10273 * config/nds32/pipelines.md: Ditto.
10275 2018-04-04 Richard Biener <rguenther@suse.de>
10277 PR tree-optimization/85168
10278 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
10279 propagating abnormals.
10281 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10283 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
10285 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
10286 Kito Cheng <kito.cheng@gmail.com>
10288 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
10289 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
10290 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
10291 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
10292 * config/nds32/nds32.md (sibcall_internal): New.
10293 (sibcall_register): Remove.
10294 (sibcall_immediate): Remove.
10295 (sibcall_value_internal): New.
10296 (sibcall_value_register): Remove.
10297 (sibcall_value_immediate): Remove.
10298 * config/nds32/predicates.md (nds32_general_register_operand): New.
10299 (nds32_call_address_operand): New.
10301 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10303 PR rtl-optimization/85167
10304 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
10305 bb_defs if *split_p, instead preinitialize it to NULL.
10307 PR tree-optimization/85156
10308 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
10309 evaluating the argument multiple times.
10311 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
10313 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
10315 (_mm_cvtpd_ps): Likewise.
10316 (_mm_cvttpd_epi32): Likewise.
10317 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
10318 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
10319 vector, pixel, and bool following altivec.h include.
10321 2018-04-03 Martin Sebor <msebor@redhat.com>
10323 * doc/extend.texi (Common Function Attributes): Clarify.
10324 (const attribute): Likewise.
10325 (pure attribute): Likewise.
10327 2018-04-03 Jakub Jelinek <jakub@redhat.com>
10330 * config/i386/i386.c (ix86_expand_vector_set): Use
10331 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
10333 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
10335 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
10336 instructions when changing rounding bits to preserve precision bits
10337 in the x87 control word.
10339 2018-04-03 Martin Liska <mliska@suse.cz>
10341 PR tree-optimization/82491
10342 * rtl.h (strip_offset_and_add): Replace += suboffset with
10343 poly_uint64 () + suboffset.
10345 2018-03-29 Martin Liska <mliska@suse.cz>
10346 Martin Jambor <mjambor@suse.cz>
10349 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
10350 param_type is not an integral or pointer type.
10352 2018-04-03 Richard Biener <rguenther@suse.de>
10354 * sese.h (recompute_all_dominators): Remove.
10356 2018-04-02 Martin Sebor <msebor@redhat.com>
10358 * doc/invoke.texi (-Wrestrict): Fix typos.
10360 2018-04-02 Jim Wilson <jimw@sifive.com>
10362 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
10363 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
10364 (<optab>di3, <optab>si3_extend): Likewise.
10365 (<optab>si3_mask, <optab>si3_mask_1): New.
10366 (<optab>di3_mask, <optab>di3_mask_1): New.
10367 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
10368 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
10369 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
10371 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
10373 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
10376 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
10378 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
10379 (nds32_canonicalize_comparison): New function.
10381 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10382 Kito Cheng <kito.cheng@gmail.com>
10383 Kuan-Lin Chen <kuanlinchentw@gmail.com>
10385 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
10386 * config/nds32/constants.md (unspec_volatile_element): Add
10387 UNSPEC_VOLATILE_RELAX_GROUP.
10388 * config/nds32/nds32-relax-opt.c: New file.
10389 * config/nds32/nds32-predicates.c
10390 (nds32_symbol_load_store_p): New function.
10391 * config/nds32/nds32-protos.h
10392 (nds32_symbol_load_store_p): Declare function.
10393 (make_pass_nds32_relax_opt): Declare new rtl pass function.
10394 * config/nds32/nds32.c
10395 (nds32_register_pass): New function to register pass.
10396 (nds32_register_passes): New function to register passes.
10397 * config/nds32/nds32.md (relax_group): New pattern.
10398 * config/nds32/nds32.opt (mrelax-hint): New option.
10399 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
10401 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
10403 * config/nds32/t-nds32: Modify files dependency.
10405 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10407 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
10408 (PROFILE_HOOK): Define its implementation.
10410 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
10412 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
10413 type and 32-bit size.
10415 2018-04-01 Jakub Jelinek <jakub@redhat.com>
10417 PR middle-end/85090
10418 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
10419 (V_128_256): New mode iterator.
10420 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
10421 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
10422 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
10424 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
10425 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
10427 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
10430 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
10431 NaN inputs correctly.
10433 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
10436 * config/rs6000/vsx.md (??r): New mode attribute.
10437 (*vsx_mov<mode>_64bit): Use it.
10438 (*vsx_mov<mode>_32bit): Likewise.
10440 2018-03-30 Martin Sebor <msebor@redhat.com>
10442 PR tree-optimization/84818
10443 * builtins.c (check_access): Use warning_n.
10445 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10448 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
10450 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
10453 2018-03-30 Julia Koval <julia.koval@intel.com>
10456 * x86-tune.def (movx, partial_reg_dependency): Enable for
10459 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
10461 PR inline-asm/84985
10462 * lra-constraints.c (process_alt_operands): Move setting
10463 this_alternative_matches below.
10465 2018-03-29 Martin Liska <mliska@suse.cz>
10468 * doc/invoke.texi: Document how LTO works with debug info.
10469 Describe auto-load support of binutils. Mention 'x86-64'
10470 as valid option value of -march option.
10472 2018-03-29 Jakub Jelinek <jakub@redhat.com>
10474 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
10477 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
10478 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
10479 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
10482 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10485 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
10486 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
10487 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
10488 for RS6000_BTM_POWERPC64.
10489 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
10490 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
10491 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
10494 (DIVDEU): Likewise.
10496 2018-03-28 Carl Love <cel@us.ibm.com>
10499 2017-09-27 Carl Love <cel@us.ibm.com>
10501 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
10502 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
10503 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
10506 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10508 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
10509 instead of __vector bool.
10510 (_mm_max_pu8): Likewise.
10511 (_mm_min_pi16): Likewise.
10513 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
10516 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
10517 (DIVWEUO): Likewise.
10518 (DIVDEO): Likewise.
10519 (DIVDEUO): Likewise.
10520 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
10521 DIVWEUO and DIVDEUO.
10522 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
10523 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
10524 (div_extend): Likewise.
10525 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
10527 (__builtin_divweuo): Likewise.
10528 (__builtin_divdeo): Likewise.
10529 (__builtin_divdeuo): Likewise.
10531 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10534 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
10535 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
10537 PR tree-optimization/82004
10538 * gimple-match-head.c (optimize_pow_to_exp): New function.
10539 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
10540 Don't fold to exp if optimize_pow_to_exp is false.
10542 2018-03-28 Martin Liska <mliska@suse.cz>
10545 * calls.c (initialize_argument_information): Fix trailing space.
10546 * common.opt: Fix typo and provide better explanation for
10547 -fsanitize-coverage option.
10548 * config/i386/i386.opt: Fix typo.
10550 2018-03-28 Jakub Jelinek <jakub@redhat.com>
10551 Martin Liska <mliska@suse.cz>
10554 * gimplify.c (asan_poison_variable): Don't do the check for
10555 gimplify_omp_ctxp here.
10556 (gimplify_decl_expr): Do it here.
10557 (gimplify_target_expr): Likewise.
10559 2018-03-28 Martin Liska <mliska@suse.cz>
10562 * config/i386/i386.c (ix86_function_arg_advance): Do not call
10563 chkp_type_bounds_count if MPX is not enabled.
10565 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
10567 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
10569 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
10572 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
10573 function to create the function decl for complex long double
10574 multiply and divide for -mabi=ieeelongdouble.
10575 (init_float128_ieee): Call it.
10577 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
10580 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
10581 -fcf-protection=branch -mibt.
10582 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
10584 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10587 * config/arm/arm.c (arm_valid_symbolic_address): Handle
10588 arm_word_relocations.
10590 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
10593 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
10594 extern array declarations.
10596 2018-03-27 Richard Biener <rguenther@suse.de>
10598 PR middle-end/84067
10599 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
10600 explicit single_use checks.
10602 2018-03-27 Richard Biener <rguenther@suse.de>
10604 PR tree-optimization/85082
10605 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
10608 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
10610 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
10611 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
10612 Turn on fasynchronous-unwind-tables and funwind-tables.
10614 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
10617 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
10618 (*bmi_blsr_<mode>_ccz): Ditto.
10620 2018-03-26 Tom de Vries <tom@codesourcery.com>
10622 PR tree-optimization/85063
10623 * omp-general.c (offloading_function_p): New function. Factor out
10625 * omp-offload.c (pass_omp_target_link::gate): ... here.
10626 * omp-general.h (offloading_function_p): Declare.
10627 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
10628 with attribute omp declare target for offloading functions.
10630 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
10632 PR tree-optimization/84005
10633 * tree-data-ref.h (get_base_for_alignment): Declare.
10634 * tree-data-ref.c (get_base_for_alignment_1): New function.
10635 (get_base_for_alignment): Likewise.
10636 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
10637 get_base_for_alignment to find a suitable base object, instead
10638 of always using drb->base_address.
10640 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10642 PR inline-asm/85022
10643 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
10644 known size by default.
10646 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
10648 PR inline-asm/85030
10649 * lra-constraints.c (process_alt_operands): Don't match BLKmode
10650 and non BLKmode operands.
10652 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10655 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
10656 Clean up attributes.
10658 2018-03-23 Richard Biener <rguenther@suse.de>
10661 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
10662 we are going to emit early debug for LTO.
10664 2018-03-23 Jakub Jelinek <jakub@redhat.com>
10666 PR inline-asm/85034
10667 * function.c (match_asm_constraints_1): Don't optimize if input
10668 doesn't satisfy general_operand predicate for output's mode.
10670 PR inline-asm/85022
10671 * alias.c (write_dependence_p): Don't require for x_canonicalized
10672 non-VOIDmode if x has VOIDmode.
10675 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
10676 just don't try to optimize it rather than assert it never happens.
10678 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10680 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
10681 macro expansions for definition of ST_INTERNAL_<mode> and
10682 LD_INTERNAL_<mode> builtins.
10683 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
10685 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
10687 (altivec_expand_st_builtin): Likewise.
10688 (altivec_expand_builtin): Remove calls to deleted functions.
10689 (rs6000_address_for_altivec): Delete this function.
10690 * config/rs6000/vector.md: Remove expands for
10691 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
10693 2018-03-22 Sudakshina Das <sudi.das@arm.com>
10696 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
10697 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
10698 re-computing once computed.
10699 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
10700 (arm_init_machine_status): Initialize
10701 machine->static_chain_stack_bytes.
10703 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
10706 * doc/extend.texi: Add four new prototypes for vec_ld.
10707 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
10708 definitions for more logical presentation.
10709 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
10710 entries for V1TI variants of __builtin_altivec_ld builtin.
10711 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
10712 handling of V1TI variant of LVX icode pattern.
10713 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
10714 (rs6000_gimple_fold_builtin): Likewise.
10715 (altivec_init_builtins): Add code to define
10716 __builtin_altivec_lvx_v1ti function.
10718 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10720 PR inline-asm/84941
10721 * function.c (match_asm_constraints_1): Don't do the optimization
10722 if input isn't a REG, SUBREG, MEM or constant.
10724 2018-03-22 Tom de Vries <tom@codesourcery.com>
10726 PR tree-optimization/84956
10727 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
10728 bb_has_abnormal_pred.
10730 2018-03-22 Jakub Jelinek <jakub@redhat.com>
10733 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
10734 DECL_INITIAL (decl) to decl at the end.
10735 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
10736 adjust the comment.
10738 2018-03-21 Joseph Myers <joseph@codesourcery.com>
10740 * doc/extend.texi (__builtin_tgmath): Document when complex
10741 integer types are treated as _Complex _Float64.
10743 2018-03-21 Tom de Vries <tom@codesourcery.com>
10745 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
10747 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10749 PR tree-optimization/84960
10750 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
10751 if it is ENTRY block, move them into single succ of ENTRY in that case.
10753 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
10755 PR tree-optimization/84811
10756 * poly-int.h (poly_span_traits): Remove the T3 parameter and
10757 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
10758 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
10759 (known_subrange_p): Update accordingly. Cast each value involved
10760 in the size comparison, rather than casting the result of the
10763 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10765 PR tree-optimization/84982
10766 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
10767 by flipping the least significant bit rather than all bits from
10768 bitpos to bitpos + bitsize - 1.
10770 2018-03-21 Nathan Sidwell <nathan@acm.org>
10772 * doc/extend.texi (Deprecated Features): Remove mention of
10773 long-deleted deprecations.
10775 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10778 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
10779 * configure: Regenerate.
10781 2018-03-21 Tom de Vries <tom@codesourcery.com>
10783 PR tree-optimization/83126
10784 * tree-parloops.c (num_phis): New function.
10785 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
10787 2018-03-21 Nathan Sidwell <nathan@acm.org>
10789 * doc/extend.texi (Deprecated Features): Update deprecated flags,
10790 mention anon-struct/union members and trailing attributes.
10792 2018-03-21 Bin Cheng <bin.cheng@arm.com>
10794 PR tree-optimization/84969
10795 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
10796 builtin memset partitions if they set different rhs values.
10798 2018-03-21 Jakub Jelinek <jakub@redhat.com>
10800 PR rtl-optimization/84989
10801 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
10802 VEC_DUPLICATE with scalar result mode.
10804 2018-03-21 Martin Liska <mliska@suse.cz>
10807 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
10808 not intended return statement.
10810 2018-03-21 Martin Liska <mliska@suse.cz>
10813 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
10814 (chkp_find_bound_slots_1): Limit number of iterations.
10816 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
10819 * Minor grammar fixes for x86 options.
10821 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10824 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
10825 holding REG_CFA_RESTORE notes, instead turn them into a USE.
10827 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
10830 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
10831 (altivec_lvx_<mode>_1op): Likewise.
10832 (altivec_stvx_<mode>_2op): Likewise.
10833 (altivec_stvx_<mode>_1op): Likewise.
10834 (altivec_lvx_<VM2:mode>): New define_expand.
10835 (altivec_stvx_<VM2:mode>): Likewise.
10836 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
10837 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10838 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
10839 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
10840 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
10841 (rs6000_gen_lvx): Likewise.
10842 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
10843 (altivec_expand_stv_builtin): Likewise.
10844 (altivec_expand_builtin): Likewise.
10845 * config/rs6000/vector.md: Likewise.
10847 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10850 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
10853 2018-03-20 Richard Biener <rguenther@suse.de>
10856 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
10857 sign-conversions as zero, fall back to standard scalar_stmt
10860 2018-03-20 Martin Liska <mliska@suse.cz>
10863 * predict.c (rebuild_frequencies): Handle case when we have
10864 PROFILE_ABSENT, but flag_guess_branch_prob is false.
10866 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10869 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
10870 flag_section_anchors.
10871 * varasm.c (use_blocks_for_decl_p): Remove hack for
10872 dw2_force_const_mem.
10875 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
10877 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
10878 be created, use lowpart_subreg of operands[0] rather than operands[0]
10880 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
10881 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
10882 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
10883 and n constraint instead of aarch64_shift_imm_di and Usd.
10884 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
10885 (*aarch64_<optab>_reg_minus<mode>3): ... this.
10887 2018-03-20 Sudakshina Das <sudi.das@arm.com>
10890 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
10891 to favor GPR over NEON registers.
10892 (<shift>di3_neon): Likewise.
10894 2018-03-20 Tom de Vries <tom@codesourcery.com>
10897 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
10898 (nvptx_process_pars): Emit bar.sync asap and alap.
10900 2018-03-20 Tom de Vries <tom@codesourcery.com>
10903 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
10904 seen_label if seen_label is already set.
10906 2018-03-20 Jakub Jelinek <jakub@redhat.com>
10909 * config/i386/i386.c (fold_builtin_cpu): For features above 31
10910 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
10911 Use 1U instead of 1. Formatting fixes.
10914 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
10915 instead of TREE_TYPE (s1) for the return value.
10917 2018-03-19 Jakub Jelinek <jakub@redhat.com>
10919 PR tree-optimization/84946
10920 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
10921 bitsize + bitsize in poly_uint64 rather than poly_int64.
10924 * dwarf2asm.c: Include fold-const.c.
10925 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
10926 of decl rather than decl itself.
10928 PR rtl-optimization/84643
10929 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
10931 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
10934 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
10935 calling assemble_variable.
10937 2018-03-19 Sudakshina Das <sudi.das@arm.com>
10940 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
10941 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
10943 2018-03-19 Jim Wilson <jimw@sifive.com>
10946 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
10947 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
10948 (riscv_first_stack_step): Likewise.
10949 (riscv_option_override): Use STACK_BOUNDARY instead of
10950 MIN_STACK_BOUNDARY.
10951 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
10952 MIN_STACK_BOUNDARY.
10953 (BIGGEST_ALIGNMENT): Set to 128.
10954 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
10955 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
10958 2018-03-19 Richard Biener <rguenther@suse.de>
10960 PR tree-optimization/84933
10961 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
10962 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
10964 2018-03-19 Richard Biener <rguenther@suse.de>
10966 PR tree-optimization/84859
10967 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
10968 (cond_if_else_store_replacement): Perform sinking operation on
10969 single-store BBs regardless of MAX_STORES_TO_SINK setting.
10970 Generalize what a BB with a single eligible store is.
10972 2018-03-19 Richard Biener <rguenther@suse.de>
10974 PR tree-optimization/84929
10975 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
10976 chrec_is_positive against non-chrec arg.
10978 2018-03-19 Tamar Christina <tamar.christina@arm.com>
10981 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
10983 2018-03-18 Martin Liska <mliska@suse.cz>
10985 PR rtl-optimization/84635
10986 * regrename.c (build_def_use): Use matches_mode only when
10989 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
10991 PR tree-optimization/84913
10992 * tree-vect-loop.c (vectorizable_reduction): Don't try to
10993 vectorize chains of COND_EXPRs.
10995 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
10997 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
10999 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11001 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
11003 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
11005 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
11007 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11008 Kito Cheng <kito.cheng@gmail.com>
11010 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
11011 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
11012 (nds32_adjust_reg_alloc_order): New function.
11013 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
11015 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
11017 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
11018 nds32_print_operand, nds32_print_operand_address): Use
11019 HOST_WIDE_INT_PRINT_DEC instead.
11021 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
11023 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
11025 2018-03-17 Jakub Jelinek <jakub@redhat.com>
11028 * config/i386/i386.c (initial_ix86_tune_features,
11029 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
11030 unsigned long long.
11031 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
11032 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
11033 rather than 1u << ix86_tune. Formatting fix.
11034 (ix86_option_override_internal): Change ix86_arch_mask from
11035 unsigned int to unsigned HOST_WIDE_INT, initialize to
11036 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
11037 (ix86_function_specific_restore): Likewise.
11039 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11042 * postreload.c (reload_combine_recognize_pattern): Perform
11043 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
11044 truncate_int_for_mode the result for the destination's mode.
11047 * hsa-gen.c (mem_type_for_type): Fix comment typo.
11048 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
11050 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11053 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
11056 * lra-assigns.c (lra_split_hard_reg_for): Don't use
11057 regno_allocno_class_array and sorted_pseudos.
11058 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
11059 insns where regno is used.
11061 2018-03-16 Martin Liska <mliska@suse.cz>
11064 * multiple_target.c (create_dispatcher_calls): Redirect
11065 reference in the symbol table.
11067 2018-03-16 Martin Liska <mliska@suse.cz>
11070 * multiple_target.c (create_dispatcher_calls): Redirect also
11073 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11077 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
11078 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
11079 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
11081 2018-03-16 Julia Koval <julia.koval@intel.com>
11083 * doc/invoke.texi (Skylake Server): Add CLWB.
11084 Cannonlake): Remove CLWB.
11086 2018-03-16 Jakub Jelinek <jakub@redhat.com>
11088 PR tree-optimization/84841
11089 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
11091 (FLOAT_ONE_CONST_TYPE): Define.
11092 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
11093 (sort_by_operand_rank): Put entries with higher constant_type last
11094 rather than first to match comments.
11096 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
11098 * config/nios2/nios2.md (movsi_internal): Fix thinko in
11101 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11104 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
11105 check and use address of target always.
11107 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
11110 * config/i386/i386.c (indirect_thunk_needed): Update comments.
11111 (indirect_thunk_bnd_needed): Likewise.
11112 (indirect_thunks_used): Likewise.
11113 (indirect_thunks_bnd_used): Likewise.
11114 (indirect_return_needed): New.
11115 (indirect_return_bnd_needed): Likewise.
11116 (output_indirect_thunk_function): Add a bool argument for
11118 (output_indirect_thunk_function): Don't generate alias for
11119 function return thunk.
11120 (ix86_code_end): Call output_indirect_thunk_function to generate
11121 function return thunks.
11122 (ix86_output_function_return): Set indirect_return_bnd_needed
11123 and indirect_return_needed instead of indirect_thunk_bnd_needed
11124 and indirect_thunk_needed.
11126 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
11128 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
11129 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
11130 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
11132 2018-03-15 David Malcolm <dmalcolm@redhat.com>
11133 Paul Hua <paul.hua.gm@gmail.com>
11136 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
11138 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
11140 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
11141 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
11142 resp. SFmode cases.
11144 2018-03-15 Tamar Christina <tamar.christina@arm.com>
11147 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
11148 instead of GET_MODE_SIZE when comparing Units.
11150 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
11153 * varasm.c (hash_section): Return an unchangeble hash value
11154 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
11155 Return !aarch64_can_use_per_function_literal_pools_p ().
11157 2018-03-15 Jakub Jelinek <jakub@redhat.com>
11160 * optabs.c (emit_conditional_move): Pass address of cmode's copy
11161 rather than address of cmode as last argument to prepare_cmp_insn.
11163 2018-03-15 Julia Koval <julia.koval@intel.com>
11165 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
11166 F_AVX512VNNI, F_AVX512BITALG): New.
11168 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
11171 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
11172 insn for floating-point loads and stores.
11174 2018-03-14 Carl Love <cel@us.ibm.com>
11176 * config/rs6000/rs6000-c.c: Add macro definitions for
11177 ALTIVEC_BUILTIN_VEC_PERMXOR.
11178 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
11179 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
11180 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
11181 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
11183 * config/doc/extend.texi: Add prototypes for vec_permxor.
11185 2018-03-14 David Malcolm <dmalcolm@redhat.com>
11188 * diagnostic-show-locus.c (class layout_point): Convert m_line
11189 from int to linenum_type.
11190 (line_span::comparator): Use linenum "compare" function when
11191 comparing line numbers.
11192 (test_line_span): New function.
11193 (layout_range::contains_point): Convert param "row" from int to
11195 (layout_range::intersects_line_p): Likewise.
11196 (layout::will_show_line_p): Likewise.
11197 (layout::print_source_line): Likewise.
11198 (layout::should_print_annotation_line_p): Likewise.
11199 (layout::print_annotation_line): Likewise.
11200 (layout::print_leading_fixits): Likewise.
11201 (layout::annotation_line_showed_range_p): Likewise.
11202 (struct line_corrections): Likewise for field m_row.
11203 (line_corrections::line_corrections): Likewise for param "row".
11204 (layout::print_trailing_fixits): Likewise.
11205 (layout::get_state_at_point): Likewise.
11206 (layout::get_x_bound_for_row): Likewise.
11207 (layout::print_line): Likewise.
11208 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
11209 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
11210 * input.c (selftest::test_linenum_comparisons): New function.
11211 (selftest::input_c_tests): Call it.
11212 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
11213 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
11214 * selftest.h (ASSERT_GT): New macro.
11215 (ASSERT_GT_AT): New macro.
11216 (ASSERT_LT): New macro.
11217 (ASSERT_LT_AT): New macro.
11219 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
11221 PR rtl-optimization/84780
11222 * combine.c (distribute_links): Don't make a link based on pc_rtx.
11224 2018-03-14 Martin Liska <mliska@suse.cz>
11226 * tree.c (record_node_allocation_statistics): Use
11227 get_stats_node_kind.
11228 (get_stats_node_kind): New function extracted from
11229 record_node_allocation_statistics.
11230 (free_node): Use get_stats_node_kind.
11232 2018-03-14 Richard Biener <rguenther@suse.de>
11234 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
11235 that the value-set of ANTIC_IN doesn't grow.
11238 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11240 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11241 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11243 2018-03-14 Julia Koval <julia.koval@intel.com>
11245 * config.gcc (icelake-client, icelake-server): New.
11247 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
11248 (initial_ix86_arch_features): Ditto.
11249 (PTA_SKYLAKE): Add SGX.
11250 (PTA_ICELAKE): Remove.
11251 (PTA_ICELAKE_CLIENT): New.
11252 (PTA_ICELAKE_SERVER): New.
11253 (ix86_option_override_internal): Split up icelake on icelake client and
11255 (get_builtin_code_for_version): Ditto.
11256 (fold_builtin_cpu): Ditto.
11257 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
11258 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
11259 * config/i386/i386.h (processor_type): Ditto.
11260 * doc/invoke.texi: Ditto.
11262 2018-03-14 Jakub Jelinek <jakub@redhat.com>
11265 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
11266 INTEGER_CST offset, add it together with bitpos / 8 and
11267 sign extend based on POINTER_SIZE.
11271 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
11274 * config/i386/constraints.md (Yc): New register constraint.
11275 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
11276 Use Yc constraint for alternative 2 of operand 0. Remove
11277 preferred_for_speed attribute.
11279 2018-03-14 Richard Biener <rguenther@suse.de>
11281 PR tree-optimization/84830
11282 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
11283 with the old one to avoid oscillations.
11285 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
11288 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
11290 (assign_by_spills): Return a flag of reload assignment failure.
11291 Do not process the reload assignment failures. Do not spill other
11292 reload pseudos if they has the same reg class. Update n if
11294 (lra_assign): Add a return arg. Set up from the result of
11295 assign_by_spills call.
11296 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11297 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11298 usage_insns if it is not NULL.
11299 (spill_hard_reg_in_range): New function.
11300 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11301 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11302 function prototypes.
11303 (lra_assign): Change prototype.
11304 * lra.c (lra): Add code to deal with fails by splitting hard reg
11307 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
11309 * config/riscv/riscv.opt (mrelax): New option.
11310 * config/riscv/riscv.c (riscv_file_start): Emit ".option
11311 "norelax" when riscv_mrelax is disabled.
11312 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
11314 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11317 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
11318 reassociation for int modes.
11320 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11322 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
11323 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
11325 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
11326 * config/aarch64/aarch64-sve.md
11327 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
11328 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
11329 (*extend<mode><Vwide>2): Rename to...
11330 (aarch64_sve_extend<mode><Vwide>2): ...this.
11331 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
11332 renaming the old pattern to...
11333 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
11335 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
11336 define_expand, renaming the old pattern to...
11337 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
11338 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
11339 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
11340 account when deciding which SVE instruction the optab should use.
11341 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
11343 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11345 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
11346 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
11347 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
11348 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
11349 (tlsdesc_small_<mode>): Turn a define_expand and use
11350 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
11351 (tlsdesc_small_advsimd_<mode>): ...this.
11352 (tlsdesc_small_sve_<mode>): New pattern.
11354 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
11356 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
11357 (UNSPEC_UMUL_HIGHPART): New constants.
11358 (MUL_HIGHPART): New int iteraor.
11359 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
11360 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
11362 (*<su>mul<mode>3_highpart): New define_insn.
11364 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
11367 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
11370 2018-03-13 Martin Liska <mliska@suse.cz>
11373 * (sem_item_optimizer::sem_item_optimizer): Initialize new
11375 (sem_item_optimizer::~sem_item_optimizer): Release it.
11376 (sem_item_optimizer::merge_classes): Register variable aliases.
11377 (sem_item_optimizer::fixup_pt_set): New function.
11378 (sem_item_optimizer::fixup_points_to_sets): Likewise.
11379 * ipa-icf.h: Declare new variables and functions.
11381 2018-03-13 Jakub Jelinek <jakub@redhat.com>
11383 PR middle-end/84834
11384 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
11385 integer_pow2p@2 and test integer_pow2p in condition.
11386 (A < 0 ? C : 0): Similarly for @1.
11388 PR middle-end/84831
11389 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
11390 characters starting at p contain '\0' character, don't look beyond
11394 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
11395 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
11398 * reg-stack.c (change_stack): Change update_end var from int to
11399 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
11400 also call set_block_for_insn on the newly added insns and rescan.
11403 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
11404 on the last operand.
11407 * tree.c (stabilize_reference_1): Return save_expr (e) for
11408 STATEMENT_LIST even if it doesn't have side-effects.
11410 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
11412 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
11414 2018-03-12 Renlin Li <renlin.li@arm.com>
11416 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
11417 aarch64_output_scalar_simd_mov_immediate.
11419 2018-03-12 Martin Sebor <msebor@redhat.com>
11421 PR tree-optimization/83456
11422 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
11423 for perfectly overlapping calls to memcpy.
11424 (gimple_fold_builtin_memory_chk): Same.
11425 (gimple_fold_builtin_strcpy): Handle no-warning.
11426 (gimple_fold_builtin_stxcpy_chk): Same.
11427 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
11429 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11431 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
11432 parameter. Use it for SFmode.
11433 (rs6000_function_arg_advance_1): Adjust.
11434 (rs6000_function_arg): Adjust.
11435 (rs6000_gimplify_va_arg): Pass false for that new parameter.
11437 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
11439 PR rtl-optimization/84169
11440 PR rtl-optimization/84780
11441 * combine.c (can_combine_p): Check for a 2-insn combination whether
11442 the destination register is used between the two insns, too.
11444 2018-03-12 Richard Biener <rguenther@suse.de>
11446 PR tree-optimization/84803
11447 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
11448 for refs DR analysis didn't process.
11450 2018-03-12 Richard Biener <rguenther@suse.de>
11452 PR tree-optimization/84777
11453 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
11454 force-vectorize loops ignore whether we are optimizing for size.
11456 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
11458 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
11459 (TARGET_MD_ASM_ADJUST): Define.
11461 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
11462 Kito Cheng <kito.cheng@gmail.com>
11463 Chung-Ju Wu <jasonwucj@gmail.com>
11465 * config/nds32/nds32.c (nds32_compute_stack_frame,
11466 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
11467 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
11468 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
11469 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
11470 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
11471 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
11472 * config/nds32/nds32.md (prologue, epilogue): Use macro
11473 NDS32_V3PUSH_AVAILABLE_P to do checking.
11475 2018-03-11 Jakub Jelinek <jakub@redhat.com>
11478 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
11479 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
11480 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
11481 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
11482 addition of most attributes on !orig_type_die or the attribute not
11483 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
11485 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11486 Chung-Ju Wu <jasonwucj@gmail.com>
11488 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
11489 __NDS32_VH__ macro.
11490 * config/nds32/nds32.opt (mvh): New option.
11492 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11493 Chung-Ju Wu <jasonwucj@gmail.com>
11495 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
11497 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
11498 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
11501 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
11502 Chung-Ju Wu <jasonwucj@gmail.com>
11504 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
11506 * config/nds32/nds32-multiple.md (strlensi): New pattern.
11507 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
11509 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
11510 Kito Cheng <kito.cheng@gmail.com>
11511 Chung-Ju Wu <jasonwucj@gmail.com>
11513 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
11514 UNSPEC_FFMISM and UNSPEC_FLMISM.
11515 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
11516 for ffb, ffmism and flmism.
11517 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
11518 (unspec_ffmism): Ditto.
11519 (unspec_flmism): Ditto.
11520 (nds32_expand_builtin_impl): Check if string extension is available.
11521 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
11522 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
11524 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
11527 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11530 * lra-assigns.c (assign_by_spills): Return a flag of reload
11531 assignment failure. Do not process the reload assignment
11532 failures. Do not spill other reload pseudos if they has the same
11534 (lra_assign): Add a return arg. Set up from the result of
11535 assign_by_spills call.
11536 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11537 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11538 usage_insns if it is not NULL.
11539 (spill_hard_reg_in_range): New function.
11540 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11541 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11542 function prototypes.
11543 (lra_assign): Change prototype.
11544 * lra.c (lra): Add code to deal with fails by splitting hard reg
11547 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
11550 * config/i386/i386.opt: Replace Enforcment with Enforcement.
11552 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
11555 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
11556 (dw_val_node): Add val_symbolic_view.
11557 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
11558 (symview_upper_bound): New.
11559 (new_line_info_table): Initialize symviews_since_reset.
11560 (dwarf2out_source_line): Count symviews_since_reset and set
11561 symview_upper_bound.
11562 (dw_val_equal_p): Handle symview.
11563 (add_AT_symview): New.
11564 (print_dw_val): Handle symview.
11565 (attr_checksum, attr_checksum_ordered): Likewise.
11566 (same_dw_val_p, size_of_die): Likewise.
11567 (value_format, output_die): Likewise.
11568 (add_high_low_attributes): Use add_AT_symview for entry_view.
11569 (dwarf2out_finish): Reset symview_upper_bound, clear
11572 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
11575 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
11576 Add strict argument and use it.
11577 (rs6000_split_multireg_move): Update for new strict argument.
11578 (mem_operand_gpr): Disallow all non-offsettable addresses.
11579 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
11581 2018-03-09 Jakub Jelinek <jakub@redhat.com>
11584 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
11585 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
11586 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
11589 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
11590 decl, use remap_type if we want to use the type.
11592 2018-03-09 Martin Sebor <msebor@redhat.com>
11594 PR tree-optimization/84526
11595 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
11597 (builtin_access::generic_overlap): Be prepared to handle non-array
11600 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
11602 PR rtl-optimization/84682
11603 * lra-constraints.c (process_address_1): Check is_address flag
11604 for address constraints.
11605 (process_alt_operands): Likewise.
11606 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
11607 preprocess_constraints.
11608 * recog.h (preprocess_constraints): Add oploc parameter.
11610 * recog.c (preprocess_constraints): Test address_operand for
11611 CT_ADDRESS constraints.
11613 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
11616 * lra-assigns.c (assign_by_spills): Return a flag of reload
11617 assignment failure. Do not process the reload assignment
11618 failures. Do not spill other reload pseudos if they has the same
11620 (lra_assign): Add a return arg. Set up from the result of
11621 assign_by_spills call.
11622 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
11623 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
11624 usage_insns if it is not NULL.
11625 (spill_hard_reg_in_range): New function.
11626 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
11627 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
11628 function prototypes.
11629 (lra_assign): Change prototype.
11630 * lra.c (lra): Add code to deal with fails by splitting hard reg
11633 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11636 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
11637 Accept complain bool parameter. Only emit errors if it is true.
11638 (arm_parse_cpu_option_name): Likewise.
11639 (arm_target_thumb_only): Adjust callers of the above.
11640 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
11641 prototype to take a default true bool parameter.
11642 (arm_parse_arch_option_name): Likewise.
11644 2018-03-09 David Malcolm <dmalcolm@redhat.com>
11645 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
11649 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
11650 * configure: Regenerate.
11651 * configure.ac ("linker --version-script option"): New.
11652 ("linker soname option"): New.
11654 2018-03-09 Richard Biener <rguenther@suse.de>
11656 PR tree-optimization/84775
11657 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
11658 immediate uses of predicate stmts and mark them modified.
11661 PR tree-optimization/84178
11662 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11664 (version_loop_for_if_conversion): Delay update_ssa call.
11665 (tree_if_conversion): Delay update_ssa until after predicate
11668 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
11671 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
11672 when the function accesses prior frames.
11674 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11677 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11678 gen_llsym, otherwise call maybe_gen_llsym.
11680 PR inline-asm/84742
11681 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
11682 has ',' character inside of it.
11684 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11687 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11688 as clobbering CC_REGNUM.
11690 2018-03-08 Richard Biener <rguenther@suse.de>
11692 PR middle-end/84552
11693 * tree-scalar-evolution.c: Include tree-into-ssa.h.
11694 (follow_copies_to_constant): Do not follow SSA names registered
11697 2018-03-08 Richard Biener <rguenther@suse.de>
11699 PR tree-optimization/84178
11700 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
11702 (version_loop_for_if_conversion): Delay update_ssa call.
11703 (tree_if_conversion): Delay update_ssa until after predicate
11706 2018-03-08 David Malcolm <dmalcolm@redhat.com>
11708 PR tree-optimization/84178
11709 * tree-if-conv.c (release_bb_predicate): Remove the
11710 the assertion that the stmts have NULL use_ops.
11711 Discard the statements, asserting that they haven't
11712 yet been added to a BB.
11714 2018-03-08 Richard Biener <rguenther@suse.de>
11716 PR tree-optimization/84746
11717 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
11718 (phi_translate): Pass in destination ANTIC_OUT set.
11719 (phi_translate_1): Likewise. For a simplified result lookup
11720 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
11721 (phi_translate_set): Adjust.
11722 (do_pre_regular_insertion): Likewise.
11723 (do_pre_partial_partial_insertion): Likewise.
11725 2018-03-08 Martin Liska <mliska@suse.cz>
11727 PR gcov-profile/84735
11728 * doc/gcov.texi: Document usage of profile files.
11729 * gcov-io.h: Document changes in the format.
11731 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
11735 * dwarf2out.c (struct dw_line_info_table): Update comments for
11737 (FORCE_RESET_NEXT_VIEW): New.
11738 (FORCE_RESETTING_VIEW_P): New.
11739 (RESETTING_VIEW_P): Check for -1 too.
11740 (ZERO_VIEW_P): Likewise.
11741 (new_line_info_table): Force-reset next view.
11742 (dwarf2out_begin_function): Likewise.
11743 (dwarf2out_source_line): Simplify zero_view_p initialization.
11744 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
11745 view directly. Omit view when omitting .loc at line 0.
11747 2018-03-08 Jakub Jelinek <jakub@redhat.com>
11749 PR tree-optimization/84740
11750 * tree-switch-conversion.c (process_switch): Call build_constructors
11751 only if info.phi_count is non-zero.
11753 PR tree-optimization/84739
11754 * tree-tailcall.c (find_tail_calls): Check call arguments against
11755 DECL_ARGUMENTS (current_function_decl) rather than
11756 DECL_ARGUMENTS (func) when checking for tail recursion.
11758 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11760 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
11761 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
11762 Volker Reichelt's entry and add entries for people that perform
11763 GCC fuzzy testing and report numerous bugs.
11765 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
11768 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
11769 readonly data in sdata, if that is disabled.
11770 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
11771 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
11772 -mreadonly-in-sdata option.
11774 2018-03-07 Martin Sebor <msebor@redhat.com>
11776 PR tree-optimization/84468
11777 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
11778 basic block when looking for nul assignment.
11780 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
11783 * except.h (output_function_exception_table): Adjust prototype.
11784 * except.c (output_function_exception_table): Remove FNNAME parameter
11785 and add SECTION parameter. Ouput one part of the table at a time.
11786 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
11787 the first part of the exception table and emit unwind directives.
11788 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
11789 (i386_pe_seh_cold_init): Likewise.
11790 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
11791 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
11792 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
11793 (ix86_output_call_insn): Emit a nop in one more case for SEH.
11794 * config/i386/winnt.c: Include except.h.
11795 (struct seh_frame_state): Add reg_offset, after_prologue and
11796 in_cold_section fields.
11797 (i386_pe_seh_end_prologue): Set seh->after_prologue.
11798 (i386_pe_seh_cold_init): New function.
11799 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
11800 to seh->in_cold_section.
11801 (seh_emit_push): Record the offset of the push.
11802 (seh_emit_save): Record the offet of the save.
11803 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
11804 Test seh->after_prologue to disregard the epilogue.
11805 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
11806 (i386_pe_end_cold_function): New function.
11808 2018-03-07 Jakub Jelinek <jakub@redhat.com>
11811 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
11812 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
11815 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
11817 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
11818 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
11820 PR middle-end/84723
11821 * multiple_target.c: Include tree-inline.h and intl.h.
11822 (expand_target_clones): Diagnose and fail if node->definition and
11823 !tree_versionable_function_p (node->decl).
11825 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
11827 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
11829 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
11830 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
11831 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
11833 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11836 * combine.c (try_combine): Use reg_or_subregno instead of handling
11837 just paradoxical SUBREGs and REGs.
11839 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
11841 * config/arc/arc.c (arc_finalize_pic): Remove function.
11842 (arc_must_save_register): We use single base PIC register, remove
11843 checks to save/restore the PIC register.
11844 (arc_expand_prologue): Likewise.
11845 * config/arc/arc-protos.h (arc_set_default_type_attributes):
11847 (arc_verify_short): Likewise.
11848 (arc_attr_type): Likewise.
11849 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
11850 (walk_stores): Likewise.
11851 (arc_address_cost): Make it static.
11852 (arc_verify_short): Likewise.
11853 (branch_dest): Likewise.
11854 (arc_attr_type): Likewise.
11855 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
11856 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
11857 (arc_final_prescan_insn): Remove inserting the nops due to
11858 hardware hazards. It is done in reorg step.
11859 (insn_length_variant_t): Remove.
11860 (insn_length_parameters_t): Likewise.
11861 (arc_insn_length_parameters): Likewise.
11862 (arc_get_insn_variants): Likewise.
11863 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
11865 2018-03-06 Jakub Jelinek <jakub@redhat.com>
11867 PR inline-asm/84683
11868 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
11871 PR tree-optimization/84687
11872 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
11874 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
11876 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11878 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11879 Rename to ppc_speculation_barrier.
11880 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
11881 __builtin_ppc_speculation_barrier.
11883 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11886 * combine.c (combine_simplify_rtx): Don't try to simplify if
11887 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
11890 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
11892 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
11893 to 32 bytes when compiling for POWER9.
11895 2018-03-05 Jakub Jelinek <jakub@redhat.com>
11898 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
11899 regparm >= 3 with no arg reg available also for calls with
11900 flag_force_indirect_call. Pass decl to ix86_function_regparm.
11903 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
11905 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
11907 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
11910 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
11912 2018-03-05 Richard Biener <rguenther@suse.de>
11914 PR tree-optimization/84486
11915 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
11916 When inserting a __builtin_assume_aligned call set the LHS
11917 SSA name alignment info accordingly.
11919 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
11921 PR tree-optimization/84114
11922 * config/aarch64/aarch64.c (aarch64_reassociation_width)
11923 Avoid reassociation of FLOAT_MODE addition.
11925 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
11927 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
11928 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
11929 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
11930 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
11931 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
11932 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
11933 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
11935 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
11936 __builtin_ia32_wbinvd): New builtins.
11937 (SPECIAL_ARGS2): New.
11938 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
11939 (SPECIAL_ARGS2): New.
11940 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
11941 (ix86_valid_target_attribute_inner_p): Ditto.
11942 (ix86_init_mmx_sse_builtins): Add special_args2.
11943 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
11944 TARGET_WBNOINVD_P): New.
11945 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
11946 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
11947 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
11948 * config/i386/immintrin.h (_wbinvd): New intrinsic.
11949 * config/i386/pconfigintrin.h: New file.
11950 * config/i386/wbnoinvdintrin.h: Ditto.
11951 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
11953 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
11955 2018-03-05 Richard Biener <rguenther@suse.de>
11957 PR tree-optimization/84670
11958 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
11960 (BB_VISITED_WITH_VISITED_SUCCS): New define.
11961 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
11962 (compute_antic_aux): Only assert the number of values in ANTIC_IN
11963 doesn't grow if all successors (recursively) were visited at least
11966 2018-03-05 Richard Biener <rguenther@suse.de>
11968 PR tree-optimization/84650
11969 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
11970 if executed in the loop pipeline.
11972 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
11974 * doc/configfiles.texi (Configuration Files): Move info about
11975 conditionalizing $target-protos.h to...
11976 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
11977 differs from $target-protos.h.
11979 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
11980 Chung-Ju Wu <jasonwucj@gmail.com>
11982 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
11983 * config/nds32/nds32-multiple.md (setmemsi): Define.
11984 * config/nds32/nds32-memory-manipulation.c
11985 (nds32_gen_dup_4_byte_to_word_value): New.
11986 (emit_setmem_word_loop): New.
11987 (emit_setmem_byte_loop): New.
11988 (nds32_expand_setmem_loop): New.
11989 (nds32_expand_setmem_loop_v3m): New.
11990 (nds32_expand_setmem_unroll): New.
11991 (nds32_expand_setmem): New.
11993 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
11994 Chung-Ju Wu <jasonwucj@gmail.com>
11996 * config/nds32/nds32-memory-manipulation.c
11997 (nds32_emit_load_store): New.
11998 (nds32_emit_post_inc_load_store): New.
11999 (nds32_emit_mem_move): New.
12000 (nds32_emit_mem_move_block): New.
12001 (nds32_expand_movmemsi_loop_unknown_size): New.
12002 (nds32_expand_movmemsi_loop_known_size): New.
12003 (nds32_expand_movmemsi_loop): New.
12004 (nds32_expand_movmemsi_unroll): New.
12005 (nds32_expand_movmemqi): Rename ...
12006 (nds32_expand_movmemsi): ... to this.
12007 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
12008 (movmemsi): ... to this.
12009 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
12010 (nds32_expand_movmemsi): ... to this.
12012 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12013 Monk Chiang <sh.chiang04@gmail.com>
12014 Chung-Ju Wu <jasonwucj@gmail.com>
12016 * config/nds32/nds32-protos.h
12017 (nds32_expand_load_multiple): New arguments.
12018 (nds32_expand_store_multiple): Ditto.
12019 (nds32_valid_multiple_load_store): Rename ...
12020 (nds32_valid_multiple_load_store_p): ... to this.
12021 * config/nds32/nds32-memory-manipulation.c
12022 (nds32_expand_load_multiple): Refine implementation.
12023 (nds32_expand_store_multiple): Ditto.
12024 * config/nds32/nds32-multiple.md
12025 (load_multiple): Update nds32_expand_load_multiple interface.
12026 (store_multiple): Update nds32_expand_store_multiple interface.
12027 * config/nds32/nds32-predicates.c
12028 (nds32_valid_multiple_load_store): Rename ...
12029 (nds32_valid_multiple_load_store_p): ... to this and refine
12031 * config/nds32/predicates.md
12032 (nds32_load_multiple_and_update_address_operation): New predicate.
12033 (nds32_store_multiple_and_update_address_operation): New predicate.
12035 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
12036 Chung-Ju Wu <jasonwucj@gmail.com>
12038 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
12039 (combo): New attribute.
12040 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
12042 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
12044 * config/nds32/nds32.opt: Change -mcmodel= default value.
12046 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
12047 Monk Chiang <sh.chiang04@gmail.com>
12048 Chung-Ju Wu <jasonwucj@gmail.com>
12050 * config/nds32/constants.md (unspec_element): New enum.
12051 * config/nds32/constraints.md (Umw): New constraint.
12052 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
12053 * config/nds32/nds32-intrinsic.md: Likewise.
12054 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
12055 (nds32_valid_smw_lwm_base_p): New.
12056 (nds32_output_smw_single_word): New.
12057 (nds32_output_lmw_single_word): New.
12058 (nds32_expand_unaligned_load): New.
12059 (nds32_expand_unaligned_store): New.
12060 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
12061 (nds32_output_smw_single_word): Declare.
12062 (nds32_output_lmw_single_word): Declare.
12063 (nds32_expand_unaligned_load): Declare.
12064 (nds32_expand_unaligned_store): Declare.
12065 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
12066 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
12067 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
12068 NDS32_BUILTIN_UASTORE_DW.
12069 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
12072 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
12073 Kito Cheng <kito.cheng@gmail.com>
12074 Chung-Ju Wu <jasonwucj@gmail.com>
12076 * config/nds32/nds32-intrinsic.c
12077 (nds32_expand_builtin_null_ftype_reg): Delete.
12078 (nds32_expand_builtin_reg_ftype_imm): Ditto.
12079 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
12080 (nds32_read_argument): New.
12081 (nds32_legitimize_target): Ditto.
12082 (nds32_legitimize_argument): Ditto.
12083 (nds32_check_constant_argument): Ditto.
12084 (nds32_expand_unop_builtin): Ditto.
12085 (nds32_expand_unopimm_builtin): Ditto.
12086 (nds32_expand_binop_builtin): Ditto.
12087 (nds32_builtin_decl_impl): Ditto.
12088 (builtin_description): Ditto.
12089 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
12090 (nds32_init_builtins_impl): Ditto.
12091 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
12092 (nds32_builtin_decl): New.
12093 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
12094 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
12096 2018-03-02 Jeff Law <law@redhat.com>
12098 * reorg.c (stop_search_p): Handle DEBUG_INSN.
12099 (redundant_insn, fill_simple_delay_slots): Likewise.
12100 (fill_slots_from_thread): Likewise.
12101 * resource.c (mark_referenced_resources): Likewise.
12102 (mark_set_resources, find_dead_or_set_registers): Likewise.
12104 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12106 * substring-locations.h (format_warning_va): Formatting fix for
12107 ATTRIBUTE_GCC_DIAG.
12108 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
12110 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
12111 * substring-locations.c: Include intl.h.
12112 (format_warning_va): Turned into small wrapper around
12113 format_warning_n_va, renamed to ...
12114 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
12115 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
12117 (format_warning_at_substring_n): New function.
12118 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
12119 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
12120 format_warning_at_substring with just a shorter name instead of
12121 const function pointer.
12122 (fmtwarn_n): New function.
12123 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
12124 appropriate, get rid of all the fmtstr temporaries, move conditionals
12125 with G_() wrapped string literals directly into fmtwarn arguments,
12126 cast dir.len to (int), formatting fixes.
12128 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
12130 * doc/invoke.texi: Remove "Cilk Plus" references.
12132 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12133 Richard Biener <rguenther@suse.de>
12136 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
12137 for error or warning attributes if CALL_FROM_THUNK_P is set.
12140 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12143 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
12144 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
12147 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
12148 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
12150 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
12151 instead of -1U in last predictors element's probability member.
12153 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
12156 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
12157 arguments if they are comparable.
12159 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12161 PR tree-optimization/84634
12162 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
12163 masks and masked_loop_p with a single loop_masks, making sure it's
12164 null for bb vectorization.
12166 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
12168 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
12169 (vect_analyze_data_ref_access): Use loop->safe_len rather than
12170 loop->force_vectorize to check whether there is no alias.
12172 2018-03-02 Jakub Jelinek <jakub@redhat.com>
12175 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
12177 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
12179 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
12180 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
12181 instead of a loop around prev_real_insn.
12182 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
12185 PR inline-asm/84625
12186 * config/i386/i386.c (ix86_print_operand): Use conditional
12187 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
12190 2018-03-02 Richard Biener <rguenther@suse.de>
12192 PR tree-optimization/84427
12193 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
12194 (bitmap_set_subtract_values): Rewrite to handle multiple
12197 (prune_clobbered_mems): Likewise.
12198 (phi_translate): Take edge instead of pred/phiblock.
12199 (phi_translate_1): Likewise.
12200 (phi_translate_set): Likewise. Insert all translated
12201 exprs for a value into the set, keeping possibly multiple
12202 expressions per value.
12203 (compute_antic_aux): Adjust for phi_translate changes.
12204 When intersecting union the expressions and prune those
12205 not in the final value set, keeping possibly multiple
12206 expressions per value. Do not use value-insertion
12207 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
12208 all expressions. Add verification that the value-sets
12209 only shrink during iteration.
12210 (compute_partial_antic_aux): Adjust for the phi_translate changes.
12211 (do_pre_regular_insertion): Likewise.
12212 (do_pre_partial_partial_insertion): Likewise.
12214 2018-03-02 Richard Biener <rguenther@suse.de>
12217 * config/darwin.c (saved_debug_info_level): New static global.
12218 (darwin_asm_lto_start): Disable debug info generation for LTO out.
12219 (darwin_asm_lto_end): Restore debug info generation settings.
12221 2018-03-01 Martin Liska <mliska@suse.cz>
12224 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12225 volatile arguments.
12227 2018-03-01 Richard Biener <rguenther@suse.de>
12230 * dwarf2out.c (gen_variable_die): Properly handle late VLA
12231 type annotation with LTO when debug was disabled at compile-time.
12233 2018-03-01 Matthew Fortune <mfortune@gmail.com>
12235 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
12237 (mips_final_postscan_insn): Likewise.
12239 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
12241 PR rtl-optimization/84528
12242 * alias.c (init_alias_target): Add commentary.
12243 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
12244 a unique base value if the frame pointer is not eliminated
12245 to the stack pointer.
12247 2018-03-01 Tom de Vries <tom@codesourcery.com>
12249 PR rtl-optimization/83327
12250 * lra-int.h (hard_regs_spilled_into): Declare.
12251 * lra.c (hard_regs_spilled_into): Define.
12252 (init_reg_info): Init hard_regs_spilled_into.
12253 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
12254 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
12255 (process_bb_lives): Handle hard_regs_spilled_into.
12256 (lra_create_live_ranges_1): Before doing liveness propagation, clear
12257 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
12259 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
12261 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
12262 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
12263 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
12264 * config/rs6000/aix72.h: New file.
12266 2018-02-28 Jakub Jelinek <jakub@redhat.com>
12268 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
12269 instead of warning_at with conditional singular and plural messages
12273 * stor-layout.c (update_alignment_for_field): For
12274 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
12275 && !DECL_PACKED (field), do the alignment update, just use
12276 only desired_align instead of MAX (type_align, desired_align)
12278 (place_field): Don't do known_align < desired_align handling
12279 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
12280 is non-NULL, instead do it after rli->prev_field handling and
12281 only if not within a bitfield word. For DECL_PACKED (field)
12282 use type_align of BITS_PER_UNIT.
12284 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
12286 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
12287 superfluous parentheses and trailing spaces.
12289 2018-02-28 Richard Biener <rguenther@suse.de>
12291 PR tree-optimization/84584
12292 * graphite-scop-detection.c (scop_detection::add_scop): Discard
12293 SCoPs with fake exit edge.
12295 2018-02-28 Martin Liska <mliska@suse.cz>
12298 * timevar.c (timer::print): Fix format to properly print 100%
12301 2018-02-28 Richard Biener <rguenther@suse.de>
12303 PR middle-end/84607
12304 * genmatch.c (capture_info::walk_match): Do not mark
12305 captured expressions without operands as expr_p given
12306 they act more like predicates and should be subject to
12307 "lost tail" side-effect preserving.
12309 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
12311 PR rtl-optimization/81611
12312 * auto-inc-dec.c (attempt_change): Move dead note from
12313 mem_insn if it's the next use of regno
12314 (find_address): Take address use of reg holding
12315 non-incremented value. Add parm to limit search to the named
12317 (merge_in_block): Attempt to use a mem insn that is the next
12318 use of the original regno.
12320 2018-02-27 Martin Sebor <msebor@redhat.com>
12323 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
12324 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
12326 2018-02-27 Martin Sebor <msebor@redhat.com>
12328 PR translation/84207
12329 * diagnostic-core.h (warning_n, error_n, inform_n): Change
12330 n argument to unsigned HOST_WIDE_INT.
12331 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
12332 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
12333 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
12334 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
12336 2018-02-27 Richard Biener <rguenther@suse.de>
12338 PR tree-optimization/84512
12339 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
12340 Do not use the estimate returned from record_stmt_cost for
12341 the scalar iteration cost but sum properly using add_stmt_cost.
12343 2018-02-27 Richard Biener <rguenther@suse.de>
12345 PR tree-optimization/84466
12346 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12347 Adjust last change to less strictly validate use operands.
12349 2018-02-27 Martin Liska <mliska@suse.cz>
12351 PR gcov-profile/84548
12352 * gcov.c (process_file): Allow partial overlap and consider it
12353 also as group functions.
12354 (output_lines): Properly calculate range of lines for a group.
12356 2018-02-27 Martin Liska <mliska@suse.cz>
12358 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
12359 'ggc' suffixes. Change first column width.
12360 (timer::print): Fix formatting of the column.
12362 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
12364 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
12365 preserve inline entry blocks for the sake of debug inline
12366 entry point markers alone.
12367 (remove_unused_locals): Suggest in comments a better place to
12368 force the preservation of inline entry blocks that are
12369 otherwise unused, but do not preserve them.
12371 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12373 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
12375 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12378 * config/i386/constraints.md (Bs): Replace
12379 ix86_indirect_branch_register with
12380 TARGET_INDIRECT_BRANCH_REGISTER.
12382 * config/i386/i386.md (indirect_jump): Likewise.
12383 (tablejump): Likewise.
12384 (*sibcall_memory): Likewise.
12385 (*sibcall_value_memory): Likewise.
12386 Peepholes of indirect call and jump via memory: Likewise.
12387 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
12388 (*sibcall_value_GOT_32): Likewise.
12389 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12390 (GOT_memory_operand): Likewise.
12391 (call_insn_operand): Likewise.
12392 (sibcall_insn_operand): Likewise.
12393 (GOT32_symbol_operand): Likewise.
12394 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
12396 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12398 PR rtl-optimization/83496
12399 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
12400 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
12401 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
12402 redundant insn, if any.
12403 (relax_delay_slots): Likewise.
12404 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
12406 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
12408 PR tree-optimization/83965
12409 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
12410 that grouped statements are part of a reduction chain. Return
12411 true if the statement is not marked as a reduction itself but
12412 is part of a group.
12413 (vect_recog_dot_prod_pattern): Don't check whether the statement
12414 is part of a group here.
12415 (vect_recog_sad_pattern): Likewise.
12416 (vect_recog_widen_sum_pattern): Likewise.
12418 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
12421 * final.c (rest_of_clean_state): Also look for calls inside sequences.
12423 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
12426 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
12428 (ix86_output_indirect_function_return): New prototype.
12429 (ix86_split_simple_return_pop_internal): Likewise.
12430 * config/i386/i386.c (indirect_return_via_cx): New.
12431 (indirect_return_via_cx_bnd): Likewise.
12432 (indirect_thunk_name): Handle return va CX_REG.
12433 (output_indirect_thunk_function): Create alias for
12434 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
12435 (ix86_output_indirect_jmp): Remove the bool argument.
12436 (ix86_output_indirect_function_return): New function.
12437 (ix86_split_simple_return_pop_internal): Likewise.
12438 * config/i386/i386.md (*indirect_jump): Don't pass false
12439 to ix86_output_indirect_jmp.
12440 (*tablejump_1): Likewise.
12441 (simple_return_pop_internal): Change it to define_insn_and_split.
12442 Call ix86_split_simple_return_pop_internal to split it for
12443 -mfunction-return=.
12444 (simple_return_indirect_internal): Call
12445 ix86_output_indirect_function_return instead of
12446 ix86_output_indirect_jmp.
12448 2018-02-26 Jakub Jelinek <jakub@redhat.com>
12451 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
12452 memset and value initialization afterwards.
12454 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
12456 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
12458 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12461 * common/config/aarch64/aarch64-common.c
12462 (aarch_option_optimization_table[]): Switch
12463 off fomit-frame-pointer
12465 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12466 Chung-Ju Wu <jasonwucj@gmail.com>
12468 * config/nds32/nds32-multiple.md (load_multiple): Disallow
12470 (store_multiple): Ditto.
12472 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
12474 * config.gcc: Add --with-cpu support for nds32 target.
12475 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
12476 * config/nds32/nds32.opt: Add -mcpu= option.
12478 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
12480 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
12481 isel=yes): Warn for these deprecated options.
12483 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
12485 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
12486 ISA_2_5_MASKS_EMBEDDED.
12488 2018-02-23 Jakub Jelinek <jakub@redhat.com>
12490 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
12491 p->max as pointers rather than using iterative_hash_expr.
12493 2018-02-23 Carl Love <cel@us.ibm.com>
12495 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
12496 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
12498 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12499 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
12500 P8V_BUILTIN_VEC_VUNSIGNED2.
12502 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
12505 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
12506 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
12508 * lra-constraints.c (curr_insn_transform): Set up
12509 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
12511 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
12512 * lra-eliminations.c (spill_pseudos): Ditto.
12513 (process_insn_for_elimination): Ditto.
12514 * lra-lives.c (reg_early_clobber_p): Use the new macros.
12515 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
12516 LRA_NON_CLOBBERED_ALT.
12518 2018-02-22 Martin Sebor <msebor@redhat.com>
12520 PR tree-optimization/84480
12521 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
12522 to maybe_diag_stxncpy_trunc. Call it.
12523 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
12524 from gimple_fold_builtin_strcpy. Print inlining stack.
12525 (handle_builtin_stxncpy): Print inlining stack.
12526 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
12528 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
12531 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
12532 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
12533 and -fcheck-pointer-bounds are used together.
12534 (indirect_thunk_prefix): New enum.
12535 (indirect_thunk_need_prefix): New function.
12536 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
12537 "_nt" instead of "_bnd" for NOTRACK prefix.
12538 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
12539 (output_indirect_thunk_function): Likewise.
12541 (ix86_code_end): Update output_indirect_thunk_function calls.
12542 (ix86_output_indirect_branch_via_reg): Replace
12543 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
12544 (ix86_output_indirect_branch_via_push): Likewise.
12545 (ix86_output_function_return): Likewise.
12546 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
12547 incompatible with -fcf-protection=branch and
12548 -fcheck-pointer-bounds.
12550 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12553 * config/aarch64/aarch64.c (aarch64_print_address_internal):
12554 Change gcc_assert call to output_operand_lossage.
12556 2018-02-22 Steve Ellcey <sellcey@cavium.com>
12558 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
12560 2018-02-22 DJ Delorie <dj@redhat.com>
12561 Sebastian Perta <sebastian.perta@renesas.com>
12562 Oleg Endo <olegendo@gcc.gnu.org>
12564 * config/rx/rx.c (rx_rtx_costs): New function.
12565 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
12567 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
12569 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
12571 2018-02-22 Martin Liska <mliska@suse.cz>
12574 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
12575 Add "native" as a possible value.
12577 2018-02-22 Martin Liska <mliska@suse.cz>
12580 * config/i386/i386.c (ix86_option_override_internal):
12581 Add "native" as a possible value for -march and -mtune.
12583 2018-02-22 Jakub Jelinek <jakub@redhat.com>
12586 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
12587 to all type variants.
12589 PR tree-optimization/84503
12590 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
12591 width as info->bitpos + info->bitsize - start.
12592 (merged_store_group::merge_overlapping): Simplify width computation.
12593 (check_no_overlap): New function.
12594 (imm_store_chain_info::try_coalesce_bswap): Compute expected
12595 start + width and last_order of the group, fail if check_no_overlap
12597 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
12598 to group if check_no_overlap fails.
12600 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12602 * config/rs6000/altivec.md: Delete contraint arguments to
12603 define_expand, define_split, and define_peephole2, and in
12604 define_insn_and_split if always unused.
12605 * config/rs6000/darwin.md: Ditto.
12606 * config/rs6000/dfp.md: Ditto.
12607 * config/rs6000/rs6000.md: Ditto.
12608 * config/rs6000/sync.md: Ditto.
12609 * config/rs6000/vector.md: Ditto.
12610 * config/rs6000/vsx.md: Ditto.
12612 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
12614 * config/rs6000/altivec.md: Write output control strings as braced
12615 blocks instead of double-quoted strings.
12616 * config/rs6000/darwin.md: Ditto.
12617 * config/rs6000/rs6000.md: Ditto.
12618 * config/rs6000/vector.md: Ditto.
12619 * config/rs6000/vsx.md: Ditto.
12621 2018-02-21 Jason Merrill <jason@redhat.com>
12623 PR c++/84314 - ICE with templates and fastcall attribute.
12624 * attribs.c (build_type_attribute_qual_variant): Remove assert.
12626 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12628 * ipa-cp.c (determine_versionability): Fix comment typos.
12630 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
12633 * ipa-cp.c (determine_versionability): Do not version functions caling
12636 2018-02-21 Martin Liska <mliska@suse.cz>
12639 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
12640 Add "native" as a possible value.
12641 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
12642 the macro when native cpu detection is available.
12644 2018-02-21 Martin Liska <mliska@suse.cz>
12647 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
12648 Add "native" as a possible value.
12649 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
12650 when native cpu detection is available.
12652 2018-02-21 Jakub Jelinek <jakub@redhat.com>
12653 Martin Sebor <msebor@redhat.com>
12655 PR tree-optimization/84478
12656 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
12658 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
12659 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
12660 support which is conservatively correct, for 2 only stay conservative
12661 for maxlen. Formatting and comment capitalization fixes. Add STRICT
12662 argument to the 2 argument get_range_strlen, adjust 6 arg
12663 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
12665 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
12666 (gimple_fold_builtin_strlen): Pass true as last argument to
12669 2018-02-20 Martin Sebor <msebor@redhat.com>
12671 PR middle-end/84095
12672 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
12673 (builtin_memref::set_base_and_offset): Same. Handle inner references.
12674 (builtin_memref::builtin_memref): Factor out parts into
12675 set_base_and_offset and call it.
12677 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12679 PR middle-end/84406
12680 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
12681 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
12682 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
12683 search at the associated MODE_INT.
12685 2018-02-20 Jeff Law <law@redhat.com>
12687 PR middle-end/82123
12688 PR tree-optimization/81592
12689 PR middle-end/79257
12690 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
12691 for range data rather than using global data.
12692 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
12693 range data rather than using global data.
12694 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
12695 pass it to children as needed.
12696 (struct directive::fmtresult): Similarly.
12697 (struct directive::set_width): Similarly.
12698 (struct directive::set_precision): Similarly.
12699 (format_integer, format_directive, parse_directive): Similarly.
12700 (format_none): Accept unnamed vr_values parameter.
12701 (format_percent, format_floating, format_character): Similarly.
12702 (format_string, format_plain): Similarly.
12703 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
12704 the EVRP range analyzer for range data rather than using global data.
12705 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
12706 gimple-ssa-evrp-analyze.h
12707 (class sprintf_dom_walker): Add after_dom_children member function.
12708 Add evrp_range_analyzer member.
12709 (sprintf_dom_walker::before_dom_children): Call into the EVRP
12710 range analyzer as needed.
12711 (sprintf_dom_walker::after_dom_children): New member function.
12712 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
12714 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
12715 (evrp_range_analyzer::pop_to_marker): Likewise.
12717 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
12719 PR tree-optimization/84419
12720 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
12721 with the required type if its current type is compatible but
12724 2018-02-20 Jakub Jelinek <jakub@redhat.com>
12726 PR middle-end/82004
12727 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
12728 after vectorization.
12730 2018-02-20 Martin Liska <mliska@suse.cz>
12733 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
12734 possible values if we don't have a hint.
12736 2018-02-20 Martin Liska <mliska@suse.cz>
12740 * final.c (shorten_branches): Build align_tab array with one
12742 * opts.c (finish_options): Add alignment option limit check.
12743 (MAX_CODE_ALIGN): Likewise.
12744 (MAX_CODE_ALIGN_VALUE): Likewise.
12745 * doc/invoke.texi: Document maximum allowed option value for
12746 all -falign-* options.
12748 2018-02-19 Jakub Jelinek <jakub@redhat.com>
12751 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
12752 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
12753 * var-tracking.c (emit_note_insn_var_location): Remove all references
12754 to NOTE_INSN_CALL_ARG_LOCATION.
12755 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
12756 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
12757 Use copy_rtx_if_shared.
12758 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
12759 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
12760 (dwarf2out_var_location): Remove handling of
12761 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
12763 * final.c (final_scan_insn): Remove all references to
12764 NOTE_INSN_CALL_ARG_LOCATION.
12765 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
12766 before dumping final insns.
12767 * except.c (emit_note_eh_region_end): Remove all references to
12768 NOTE_INSN_CALL_ARG_LOCATION.
12769 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
12770 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
12771 * config/arc/arc.c (hwloop_optimize): Likewise.
12772 * config/arm/arm.c (create_fix_barrier): Likewise.
12773 * config/s390/s390.c (s390_chunkify_start): Likewise.
12774 * config/sh/sh.c (find_barrier): Likewise.
12775 * config/i386/i386.c (rest_of_insert_endbranch,
12776 ix86_seh_fixup_eh_fallthru): Likewise.
12777 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
12778 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
12779 * config/frv/frv.c (frv_function_prologue): Likewise.
12780 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
12782 (note_outside_basic_block_p): Remove all references to
12783 NOTE_INSN_CALL_ARG_LOCATION.
12784 * gengtype.c (adjust_field_rtx_def): Likewise.
12785 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
12787 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
12788 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
12791 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
12794 PR tree-optimization/84452
12795 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
12796 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
12799 2018-02-19 Martin Liska <mliska@suse.cz>
12802 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
12804 2018-02-19 Martin Liska <mliska@suse.cz>
12805 Richard Sandiford <richard.sandiford@linaro.org>
12807 PR tree-optimization/82491
12808 * gimple-fold.c (get_base_constructor): Make earlier bail out
12811 2018-02-19 Carl Love <cel@us.ibm.com>
12813 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
12814 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
12816 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
12817 P8V_BUILTIN_VEC_NEG.
12819 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
12821 * config/rl78/rl78.md (movdf): New define expand.
12823 2018-02-19 Martin Liska <mliska@suse.cz>
12826 * doc/invoke.texi: Fix typo.
12827 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
12829 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
12831 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
12832 handle rs6000_single_float and rs6000_double_float specially for
12835 2018-02-16 Jeff Law <law@redhat.com>
12837 * config/rx/rx.c (add_pop_cfi_notes): New function.;
12838 (pop_regs): Use it.
12840 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12843 * ipa-inline.c (inline_small_functions): Fix a typo.
12845 2018-02-16 Nathan Sidwell <nathan@acm.org>
12847 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
12849 2018-02-16 Carl Love <cel@us.ibm.com>
12851 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
12852 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
12853 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
12854 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
12855 expansion to P8V_BUILTIN_VEC_FLOAT2.
12857 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
12859 PR rtl-optimization/70023
12860 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
12861 src_regno into account.
12863 2018-02-16 Carl Love <cel@us.ibm.com>
12865 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
12866 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
12867 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
12868 * config/rs6000/rs6000.c: Remove case statements for
12869 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
12870 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
12871 and P9V_BUILTIN_VEC_VINSERT4B.
12872 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
12873 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
12874 * config/rs6000/vsx.md:
12875 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
12878 2018-02-16 Carl Love <cel@us.ibm.com>
12880 * config/rs6000/altivec.h: Add builtin names vec_extract4b
12882 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
12884 * config/rs6000/rs6000-c.c: Add the definitions for
12885 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
12886 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
12887 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
12888 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
12889 definition for insert4b and define insn *insert3b_internal.
12890 * doc/extend.texi: Add documentation for vec_extract4b.
12892 2018-02-16 Nathan Sidwell <nathan@acm.org>
12894 * doc/extend.texi (Backwards Compatibility): Mention friend
12895 injection. Note for-scope is deprecated.
12896 * doc/invoke.texi (-ffriend-injection): Deprecate.
12898 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
12900 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
12901 that moved to I2, also allow destinations that are a paradoxical
12902 subreg (instead of a normal reg).
12904 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
12907 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
12910 2018-02-16 Richard Biener <rguenther@suse.de>
12912 PR tree-optimization/84037
12913 PR tree-optimization/84016
12915 * config/i386/i386.c (ix86_builtin_vectorization_cost):
12916 Adjust vec_construct for the fact we need additional higher latency
12917 128bit inserts for AVX256 and AVX512 vector builds.
12918 (ix86_add_stmt_cost): Scale vector construction cost for
12921 2018-02-16 Richard Biener <rguenther@suse.de>
12923 PR tree-optimization/84417
12924 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
12925 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
12926 (non_rewritable_lvalue_p): Likewise, use poly-ints.
12928 2018-02-16 Martin Liska <mliska@suse.cz>
12931 * internal-fn.def (ASAN_CHECK): Set proper flags.
12932 (ASAN_MARK): Likewise.
12934 2018-02-16 Julia Koval <julia.koval@intel.com>
12936 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
12937 from PTA_CANNONLAKE.
12939 2018-02-16 Jakub Jelinek <jakub@redhat.com>
12942 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
12943 Use ++iter rather than iter++ for std::list iterators.
12944 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
12945 defer deleting them until all nodes in the forest are processed. Do
12946 free even leaf nodes. Change to_process into auto_vec.
12949 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
12950 * vec.h (vec_default_construct): Use memset instead of placement new
12951 if BROKEN_VALUE_INITIALIZATION is defined.
12952 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
12953 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
12956 PR rtl-optimization/83723
12957 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
12958 * lra.c (lra_substitute_pseudo): Likewise. If true, use
12959 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
12961 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
12963 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
12965 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
12967 PR rtl-optimization/81443
12968 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
12969 from inner REGs to paradoxical SUBREGs.
12971 2018-02-16 Richard Biener <rguenther@suse.de>
12973 PR tree-optimization/84399
12974 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12975 For operands we can analyze at their definition make sure we can
12976 analyze them at each use as well.
12978 2018-02-16 Richard Biener <rguenther@suse.de>
12980 PR tree-optimization/84190
12981 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
12982 volatile accesses if the decl isn't volatile.
12984 2018-02-15 Jason Merrill <jason@redhat.com>
12986 PR c++/84314 - ICE with templates and fastcall attribute.
12987 * attribs.c (build_type_attribute_qual_variant): Don't clobber
12988 TYPE_CANONICAL on an existing type.
12990 2018-02-15 Jakub Jelinek <jakub@redhat.com>
12992 PR tree-optimization/84383
12993 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
12994 dstoff nor call operand_equal_p if dstbase is NULL.
12996 PR tree-optimization/84334
12997 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
12998 also a CONSTANT_CLASS_P, punt.
13000 2018-02-14 Jim Wilson <jimw@sifive.com>
13002 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
13003 first SMALL_OPERAND check. New local min_second_step. Move assert
13004 to where locals are set. Add TARGET_RVC support.
13005 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
13007 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
13009 * doc/invoke.texi: Correct -Wformat-overflow code sample.
13011 2018-02-14 Martin Sebor <msebor@redhat.com>
13013 PR tree-optimization/83698
13014 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
13015 arrays constrain the offset range to their bounds.
13016 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
13017 (builtin_access::overlap): Avoid setting the size of overlap if it's
13019 (maybe_diag_overlap): Also consider arrays when deciding what values
13020 of offsets to include in diagnostics.
13022 2018-02-14 Martin Sebor <msebor@redhat.com>
13025 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
13026 that correspond to the kind of a declaration.
13028 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
13031 * config/pa/pa.md: Load address of PIC label using the linkage table
13032 if the label is nonlocal.
13034 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
13036 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
13037 warning message if user requests -maltivec=be.
13038 * doc/invoke.texi: Document deprecation of -maltivec=be.
13040 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
13043 * config/rs6000/rs6000-c.c: Update definitions for
13044 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
13045 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
13047 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13050 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
13051 add _get_ssp intrinsics. Remove argument from
13052 __builtin_ia32_rdssp[d|q].
13053 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
13054 * config/i386/i386-builtin.def: Remove argument from
13055 __builtin_ia32_rdssp[d|q].
13056 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
13057 ix86_expand_special_args_builtin for _rdssp[d|q].
13058 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
13059 Clear register before usage.
13060 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
13061 Add documentation for new _get_ssp and _inc_ssp intrinsics.
13063 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
13065 PR tree-optimization/84357
13066 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
13067 operand 1 of an ARRAY_REF too.
13069 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
13072 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
13073 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13075 (set_of_reg): New struct.
13076 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
13077 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
13078 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
13080 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
13081 Split into bitclr, bitset, bitinvert patterns if appropriate.
13082 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
13083 use rx_fuse_in_memory_bitop.
13084 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
13085 to named insn, correct maximum insn length.
13087 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
13090 * machmode.def: Define a complex mode for PARTIAL_INT.
13091 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
13093 * doc/rtl.texi: Document CSPImode.
13094 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
13096 (msp430_hard_regno_nregs_with_padding): Likewise.
13098 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
13101 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
13103 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
13105 PR rtl-optimization/84169
13106 * combine.c (try_combine): New variable split_i2i3. Set it to true if
13107 we generated a parallel as new i3 and we split that to new i2 and i3
13108 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
13109 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
13110 those to i2, not i1. Partially rewrite this scan code.
13112 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13115 * stor-layout.c (place_field): For variable length fields, adjust
13116 offset_align afterwards not just based on the field's alignment,
13117 but also on the size.
13119 PR middle-end/84309
13120 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
13121 of exps and logs in the use_exp2 case.
13123 2018-02-13 Jeff Law <law@redhat.com>
13125 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
13126 entry for "vector".
13128 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
13131 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
13135 * common.opt (gas-loc-support, gas-locview-support): New.
13136 (ginline-points, ginternal-reset-location-views): New.
13137 * doc/invoke.texi: Document them. Use @itemx where intended.
13138 (gvariable-location-views): Adjust.
13139 * target.def (reset_location_view): New.
13140 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
13141 (TARGET_RESET_LOCATION_VIEW): New.
13142 * doc/tm.texi: Rebuilt.
13143 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
13144 (dwarf2out_default_as_locview_support): New.
13145 (output_asm_line_debug_info): Use option variables.
13146 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
13147 (output_loc_list): Likewise.
13148 (add_high_low_attributes): Check option variables.
13149 Don't output entry view attribute in strict mode.
13150 (gen_inlined_subroutine_die): Check option variables.
13151 (dwarf2out_inline_entry): Likewise.
13152 (init_sections_and_labels): Likewise.
13153 (dwarf2out_early_finish): Likewise.
13154 (maybe_reset_location_view): New, from...
13155 (dwarf2out_var_location): ... here. Call it.
13156 * debug.h (dwarf2out_default_as_loc_support): Declare.
13157 (dwarf2out_default_as_locview_support): Declare.
13158 * hooks.c (hook_int_rtx_insn_0): New.
13159 * hooks.h (hook_int_rtx_insn_0): Declare.
13160 * toplev.c (process_options): Take -gas-loc-support and
13161 -gas-locview-support from dwarf2out. Enable
13162 -gvariable-location-views by default only with locview
13163 assembler support. Enable -ginternal-reset-location-views by
13164 default only if the target defines the corresponding hook.
13165 Enable -ginline-points by default if location views are
13166 enabled; force it disabled if statement frontiers are
13168 * tree-inline.c (expand_call_inline): Check option variables.
13169 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
13171 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
13173 PR tree-optimization/84321
13174 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
13175 handling. Also check whether the anti-range contains any values
13176 that satisfy the mask; switch to a VR_RANGE if not.
13178 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
13181 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
13183 2018-02-13 Martin Jambor <mjambor@suse.cz>
13186 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
13187 of call statements, also set location of a load to a temporary.
13189 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
13191 * config/rl78/rl78.c (add_vector_labels): New function.
13192 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
13193 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
13194 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
13195 which checks that no arguments are passed.
13196 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
13197 * doc/extend.texi: Documentation for the new attribute.
13199 2018-02-13 Andreas Schwab <schwab@suse.de>
13201 * config/riscv/linux.h (CPP_SPEC): Define.
13203 2018-02-13 Jakub Jelinek <jakub@redhat.com>
13206 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
13207 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
13208 OPTION_MASK_ISA_AES as first argument to def_builtin_const
13209 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
13210 instead of OPTION_MASK_ISA_PCLMUL as first argument to
13211 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
13212 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
13213 temporarily for AES and PCLMUL builtins.
13215 PR tree-optimization/84339
13216 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
13217 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
13220 PR middle-end/84309
13221 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
13222 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
13223 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
13225 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
13227 * omp-simd-clone.h: New file.
13228 * omp-simd-clone.c: Include omp-simd-clone.h.
13229 (expand_simd_clones): No longer static.
13230 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
13231 cgraph.h and omp-simd-clone.h.
13232 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
13233 (vect_recog_widen_shift_pattern): Formatting fix.
13234 (vect_pattern_recog_1): Don't check optab for calls.
13237 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
13238 operands[2] into a REG before using gen_lowpart on it.
13240 2018-02-12 Jeff Law <law@redhat.com>
13243 * config/sh/sh.c (find_barrier): Consider a sibling call
13246 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
13247 successfully back substituting a reg.
13249 2018-02-12 Richard Biener <rguenther@suse.de>
13251 PR tree-optimization/84037
13252 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
13253 parameter, move visited init to caller.
13254 (vect_slp_analyze_operations): Separate cost from validity
13255 check, initialize visited once for all instances.
13256 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
13258 * tree-vect-stmts.c (vect_model_simple_cost): Make early
13260 (vect_model_promotion_demotion_cost): Likewise.
13261 (vectorizable_bswap): Guard cost modeling with !slp_node
13262 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
13264 (vectorizable_call): Likewise.
13265 (vectorizable_conversion): Likewise.
13266 (vectorizable_assignment): Likewise.
13267 (vectorizable_shift): Likewise.
13268 (vectorizable_operation): Likewise.
13269 (vectorizable_store): Likewise.
13270 (vectorizable_load): Likewise.
13271 (vectorizable_condition): Likewise.
13272 (vectorizable_comparison): Likewise.
13274 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
13277 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
13278 (ASAN_MARK): Fix fnspec to account for return value, change pointer
13279 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
13281 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13283 PR middle-end/83665
13284 * params.def (inline-min-speedup): Increase from 8 to 15.
13285 (max-inline-insns-auto): Decrease from 40 to 30.
13286 * ipa-split.c (consider_split): Add some buffer for function to
13287 be considered inlining candidate.
13288 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
13291 2018-02-12 Richard Biener <rguenther@suse.de>
13293 PR tree-optimization/84037
13294 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
13295 matched stmts if we cannot swap the non-matched ones.
13297 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13299 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
13300 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
13301 _mm_maskz_scalef_round_ss): New intrinsics.
13302 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
13303 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
13304 __builtin_ia32_scalefss_round): Remove.
13305 (__builtin_ia32_scalefsd_mask_round,
13306 __builtin_ia32_scalefss_mask_round): New intrinsics.
13307 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
13308 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
13309 ((match_operand:VF_128 2 "<round_nimm_predicate>"
13310 "<round_constraint>")): Changed to ...
13311 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
13312 "<round_scalar_constraint>")): ... this.
13313 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
13314 %0, %1, %2<round_op3>}"): Changed to ...
13315 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
13316 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
13317 %2<round_scalar_mask_op3>}"): ... this.
13318 * config/i386/subst.md (round_scalar_nimm_predicate): New.
13320 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
13322 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
13323 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
13324 (_mm_maskz_sqrt_round_ss): New intrinsics.
13325 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
13326 (__builtin_ia32_sqrtsd_mask_round)
13327 (__builtin_ia32_sqrtss_mask_round): New builtins.
13328 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
13329 (__builtin_ia32_sqrtss_round): Remove.
13330 (__builtin_ia32_sqrtsd_mask_round)
13331 (__builtin_ia32_sqrtss_mask_round): New builtins.
13332 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
13333 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
13334 ((match_operand:VF_128 1 "vector_operand"
13335 "xBm,<round_constraint>")): Changed to ...
13336 ((match_operand:VF_128 1 "vector_operand"
13337 "xBm,<round_scalar_constraint>")): ... this.
13338 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
13339 %0, %2, %<iptr>1<round_op3>}): Changed to ...
13340 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
13341 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
13342 %<iptr>1<round_scalar_mask_op3>}): ... this.
13343 ((set_attr "prefix" "<round_prefix>")): Changed to ...
13344 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
13346 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
13349 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
13350 Cast vec_cmpeq result to correct type.
13351 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
13352 Cast vec_cmpgt result to correct type.
13354 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
13356 * final.c (final_scan_insn_1): Renamed from...
13357 (final_scan_insn): ... this. New wrapper, to recover
13358 seen from the outermost call in recursive ones.
13359 * config/sparc/sparc.c (output_return): Drop seen from call.
13360 (output_sibcall): Likewise.
13361 * config/visium/visium.c (output_branch): Likewise.
13363 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
13365 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
13368 2018-02-10 Alan Modra <amodra@gmail.com>
13371 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
13372 Specify LR as an input.
13374 2018-02-10 Jakub Jelinek <jakub@redhat.com>
13377 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
13378 remove_member_access_dummy_vars): New functions.
13379 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
13380 lower_omp_1, execute_lower_omp): Use them.
13382 PR rtl-optimization/84308
13383 * shrink-wrap.c (spread_components): Release todo vector.
13385 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
13387 PR rtl-optimization/57193
13388 * ira-color.c (struct allocno_color_data): Add member
13389 conflict_allocno_hard_prefs.
13390 (update_conflict_allocno_hard_prefs): New.
13391 (bucket_allocno_compare_func): Add a preference based on
13392 conflict_allocno_hard_prefs.
13393 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
13394 (color_allocnos): Remove a dead code. Initiate
13395 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
13397 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13400 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
13401 constraint from =wa to wa. Avoid a subreg on the output operand,
13402 instead use a pseudo and subreg it in a move.
13403 (p9_xxbrd_<mode>): Changed to ...
13404 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
13405 (p9_xxbrd_v2df): New expander.
13406 (p9_xxbrw_<mode>): Changed to ...
13407 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
13408 (p9_xxbrw_v4sf): New expander.
13410 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13412 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
13414 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
13417 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
13418 multiply in 32-bit mode.
13419 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
13420 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
13423 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
13425 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
13426 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
13427 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
13428 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
13430 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
13433 * dwarf2out.c (is_trivial_indirect_ref): New function.
13434 (dwarf2out_late_global_decl): Do not generate a location
13435 attribute for variables that have a non-trivial DECL_VALUE_EXPR
13436 and that are not defined in the current unit.
13438 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13440 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
13441 instead of a libcall for UNORDERED.
13443 2018-02-09 Tamar Christina <tamar.christina@arm.com>
13446 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
13447 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
13449 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13452 * config/s390/s390.c (s390_set_current_function): Invoke
13453 s390_indirect_branch_settings also if fndecl didn't change.
13455 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13457 * config/rs6000/rs6000.md (blockage): Set length to zero.
13459 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
13461 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
13463 2018-02-09 Jakub Jelinek <jakub@redhat.com>
13466 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
13467 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
13471 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
13472 PARALLEL incoming that failed vt_get_decl_and_offset check.
13474 PR middle-end/84237
13475 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
13476 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
13478 (get_variable_section): For decls in named .bss* sections pass true as
13479 second argument to bss_initializer_p.
13481 2018-02-09 Marek Polacek <polacek@redhat.com>
13482 Jakub Jelinek <jakub@redhat.com>
13485 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
13486 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
13487 Sync some changes from cxx_fold_indirect_ref.
13489 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
13491 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
13493 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
13494 (BLOCK_INLINE_ENTRY_LABEL): New.
13495 (dwarf2out_var_location): Disregard inline entry markers.
13496 (inline_entry_data): New struct.
13497 (inline_entry_data_hasher): New hashtable type.
13498 (inline_entry_data_hasher::hash): New.
13499 (inline_entry_data_hasher::equal): New.
13500 (inline_entry_data_table): New variable.
13501 (add_high_low_attributes): Add DW_AT_entry_pc and
13502 DW_AT_GNU_entry_view attributes if a pending entry is found
13503 in inline_entry_data_table. Add old entry_pc attribute only
13504 if debug nonbinding markers are disabled.
13505 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
13506 markers are enabled.
13507 (block_within_block_p, dwarf2out_inline_entry): New.
13508 (dwarf2out_finish): Check that no entries remained in
13509 inline_entry_data_table.
13510 * final.c (reemit_insn_block_notes): Handle inline entry notes.
13511 (final_scan_insn, notice_source_line): Likewise.
13512 (rest_of_clean_state): Skip inline entry markers.
13513 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
13515 * gimple.c (gimple_build_debug_inline_entry): New.
13516 * gimple.h (enum gimple_debug_subcode): Add
13517 GIMPLE_DEBUG_INLINE_ENTRY.
13518 (gimple_build_debug_inline_entry): Declare.
13519 (gimple_debug_inline_entry_p): New.
13520 (gimple_debug_nonbind_marker_p): Adjust.
13521 * insn-notes.def (INLINE_ENTRY): New.
13522 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
13523 inline entry marker notes.
13524 (print_insn): Likewise.
13525 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
13526 (INSN_DEBUG_MARKER_KIND): Likewise.
13527 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
13528 * tree-inline.c (expand_call_inline): Build and insert
13529 debug_inline_entry stmt.
13530 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
13531 inline entry blocks early, if nonbind markers are enabled.
13532 (dump_scope_block): Dump fragment info.
13533 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
13534 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
13535 (gimple_build_debug_inline_entry): New.
13536 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
13537 Enable/disable inline entry points too.
13538 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
13539 (DEBUG_INSN): Describe inline entry markers.
13541 * common.opt (gvariable-location-views): New.
13542 (gvariable-location-views=incompat5): New.
13543 * config.in: Rebuilt.
13544 * configure: Rebuilt.
13545 * configure.ac: Test assembler for view support.
13546 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
13547 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
13548 * dwarf2out.c (var_loc_view): New typedef.
13549 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
13550 (dwarf2out_locviews_in_attribute): New.
13551 (dwarf2out_locviews_in_loclist): New.
13552 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
13553 (enum dw_line_info_opcode): Add LI_adv_address.
13554 (struct dw_line_info_table): Add view.
13555 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
13556 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
13557 (zero_view_p): New variable.
13558 (ZERO_VIEW_P): New macro.
13559 (output_asm_line_debug_info): New.
13560 (struct var_loc_node): Add view.
13561 (add_AT_view_list, AT_loc_list): New.
13562 (add_var_loc_to_decl): Add view param. Test it against last.
13563 (new_loc_list): Add view params. Record them.
13564 (AT_loc_list_ptr): Handle loc and view lists.
13565 (view_list_to_loc_list_val_node): New.
13566 (print_dw_val): Handle dw_val_class_view_list.
13567 (size_of_die): Likewise.
13568 (value_format): Likewise.
13569 (loc_list_has_views): New.
13570 (gen_llsym): Set vl_symbol too.
13571 (maybe_gen_llsym, skip_loc_list_entry): New.
13572 (dwarf2out_maybe_output_loclist_view_pair): New.
13573 (output_loc_list): Output view list or entries too.
13574 (output_view_list_offset): New.
13575 (output_die): Handle dw_val_class_view_list.
13576 (output_dwarf_version): New.
13577 (output_compilation_unit_header): Use it.
13578 (output_skeleton_debug_sections): Likewise.
13579 (output_rnglists, output_line_info): Likewise.
13580 (output_pubnames, output_aranges): Update version comments.
13581 (output_one_line_info_table): Output view numbers in asm comments.
13582 (dw_loc_list): Determine current endview, pass it to new_loc_list.
13583 Call maybe_gen_llsym.
13584 (loc_list_from_tree_1): Adjust.
13585 (add_AT_location_description): Create view list attribute if
13586 needed, check it's absent otherwise.
13587 (convert_cfa_to_fb_loc_list): Adjust.
13588 (maybe_emit_file): Call output_asm_line_debug_info for test.
13589 (dwarf2out_var_location): Reset views as needed. Precompute
13590 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
13591 attribute. Set view.
13592 (new_line_info_table): Reset next view.
13593 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
13594 (dwarf2out_source_line): Likewise. Output view resets and labels to
13595 the assembler, or select appropriate line info opcodes.
13596 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
13597 (optimize_string_length): Catch it. Adjust.
13598 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
13599 dw_val_class_view_list, and remove it if no longer needed.
13600 (hash_loc_list): Hash view numbers.
13601 (loc_list_hasher::equal): Compare them.
13602 (optimize_location_lists): Check whether a view list symbol is
13603 needed, and whether the locview attribute is present, and
13604 whether they match. Remove the locview attribute if no longer
13606 (index_location_lists): Call skip_loc_list_entry for test.
13607 (dwarf2out_finish): Call output_asm_line_debug_info for test.
13608 Use output_dwarf_version.
13609 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
13610 (struct dw_val_node): Add val_view_list.
13611 * final.c (SEEN_NEXT_VIEW): New.
13612 (set_next_view_needed): New.
13613 (clear_next_view_needed): New.
13614 (maybe_output_next_view): New.
13615 (final_start_function): Rename to...
13616 (final_start_function_1): ... this. Take pointer to FIRST,
13617 add SEEN parameter. Emit param bindings in the initial view.
13618 (final_start_function): Reintroduce SEEN-less interface.
13619 (final): Rename to...
13620 (final_1): ... this. Take SEEN parameter. Output final pending
13621 next view at the end.
13622 (final): Reintroduce seen-less interface.
13623 (final_scan_insn): Output pending next view before switching
13624 sections or ending a block. Mark the next view as needed when
13625 outputting variable locations. Notify debug backend of section
13626 changes, and of location view changes.
13627 (rest_of_handle_final): Adjust.
13628 * toplev.c (process_options): Autodetect value for debug variable
13629 location views option. Warn on incompat5 without -gdwarf-5.
13630 * doc/invoke.texi (gvariable-location-views): New.
13631 (gvariable-location-views=incompat5): New.
13632 (gno-variable-location-views): New.
13634 2018-02-08 David Malcolm <dmalcolm@redhat.com>
13636 PR tree-optimization/84136
13637 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
13638 that the result of find_edge is non-NULL.
13640 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13643 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
13644 storing integer register in SImode. Fix cost of 256 and 512
13645 byte aligned SSE register store.
13647 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
13649 * config/i386/i386.c (ix86_multiplication_cost): Fix
13650 multiplication cost for TARGET_AVX512DQ.
13652 2018-02-08 Marek Polacek <polacek@redhat.com>
13654 PR tree-optimization/84238
13655 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
13658 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13660 PR tree-optimization/84265
13661 * tree-vect-stmts.c (vectorizable_store): Don't treat
13662 VMAT_CONTIGUOUS accesses as grouped.
13663 (vectorizable_load): Likewise.
13665 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13667 PR tree-optimization/81635
13668 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
13669 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
13670 (test_round_for_mask): New functions.
13671 (wide_int_cc_tests): Call test_round_for_mask.
13672 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
13673 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
13674 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
13675 range returned by get_range_info.
13677 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
13680 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
13681 * symtab.c: Include builtins.h
13682 (symtab_node::output_to_lto_symbol_table_p): Move here
13683 from lto-streamer-out.c:output_symbol_p.
13684 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
13685 (output_symbol_p): Move all logic to symtab.c
13686 (produce_symtab): Update.
13688 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
13690 * config/s390/s390-opts.h (enum indirect_branch): Define.
13691 * config/s390/s390-protos.h (s390_return_addr_from_memory)
13692 (s390_indirect_branch_via_thunk)
13693 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
13694 (enum s390_indirect_branch_type): Define.
13695 * config/s390/s390.c (struct s390_frame_layout, struct
13696 machine_function): Remove.
13697 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
13698 (indirect_branch_table_label_no, indirect_branch_table_name):
13700 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
13701 (enum s390_indirect_branch_option): Define.
13702 (s390_return_addr_from_memory): New function.
13703 (s390_handle_string_attribute): New function.
13704 (s390_attribute_table): Add new attribute handler.
13705 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
13706 (s390_indirect_branch_via_thunk): New function.
13707 (s390_indirect_branch_via_inline_thunk): New function.
13708 (s390_function_ok_for_sibcall): When jumping via thunk disallow
13709 sibling call optimization for non z10 compiles.
13710 (s390_emit_call): Force indirect branch target to be a single
13711 register. Add r1 clobber for non-z10 compiles.
13712 (s390_emit_epilogue): Emit return jump via return_use expander.
13713 (s390_reorg): Handle JUMP_INSNs as execute targets.
13714 (s390_option_override_internal): Perform validity checks for the
13715 new command line options.
13716 (s390_indirect_branch_attrvalue): New function.
13717 (s390_indirect_branch_settings): New function.
13718 (s390_set_current_function): Invoke s390_indirect_branch_settings.
13719 (s390_output_indirect_thunk_function): New function.
13720 (s390_code_end): Implement target hook.
13721 (s390_case_values_threshold): Implement target hook.
13722 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
13724 * config/s390/s390.h (struct s390_frame_layout)
13725 (struct machine_function): Move here from s390.c.
13726 (TARGET_INDIRECT_BRANCH_NOBP_RET)
13727 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
13728 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
13729 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
13730 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
13731 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
13732 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
13733 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
13734 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
13735 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
13736 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
13737 (mnemonic attribute): Add values which aren't recognized
13739 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
13740 pattern for branch conversion. Fix mnemonic attribute.
13741 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
13742 indirect branch via thunk if requested.
13743 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
13744 ("*indirect_jump"): Disable for branch conversion using out of
13746 ("indirect_jump_via_thunk<mode>_z10")
13747 ("indirect_jump_via_thunk<mode>")
13748 ("indirect_jump_via_inlinethunk<mode>_z10")
13749 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
13750 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
13751 ("casesi_jump_via_inlinethunk<mode>_z10")
13752 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
13753 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
13754 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
13755 ("*indirect2_jump"): Disable for branch conversion.
13756 ("casesi_jump"): Turn into expander and expand patterns for branch
13758 ("return_use"): New expander.
13759 ("*return"): Emit return via thunk and rename it to ...
13760 ("*return<mode>"): ... this one.
13761 * config/s390/s390.opt: Add new options and and enum for the
13764 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
13766 * lra-constraints.c (match_reload): Unconditionally use
13767 gen_lowpart_SUBREG, rather than selecting between that
13768 and equivalent gen_rtx_SUBREG code.
13770 2018-02-08 Richard Biener <rguenther@suse.de>
13772 PR tree-optimization/84233
13773 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
13774 changed flag instead of boguously re-using phi_inserted.
13776 2018-02-08 Martin Jambor <mjambor@suse.cz>
13778 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
13779 static local variables.
13781 2018-02-08 Richard Biener <rguenther@suse.de>
13783 PR tree-optimization/84278
13784 * tree-vect-stmts.c (vectorizable_store): When looking for
13785 smaller vector types to perform grouped strided loads/stores
13786 make sure the mode is supported by the target.
13787 (vectorizable_load): Likewise.
13789 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13791 * config/aarch64/aarch64.c (aarch64_components_for_bb):
13792 Increase LDP/STP opportunities by adding adjacent callee-saves.
13794 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
13796 PR rtl-optimization/84068
13797 PR rtl-optimization/83459
13798 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
13800 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
13802 PR tree-optimization/84224
13803 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
13804 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
13805 non-zero arguments.
13807 2018-02-07 Iain Sandoe <iain@codesourcery.com>
13810 * config/rs6000/altivec.md (*restore_world): Remove LR use.
13811 * config/rs6000/predicates.md (restore_world_operation): Adjust op
13812 count, remove one USE.
13814 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
13816 * doc/install.texi (Configuration): Document the
13817 --with-long-double-format={ibm,ieee} PowerPC configuration
13821 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
13822 Convert from define_expand to be define_insn_and_split. Rework
13823 float/double/_Float128 conversions to QI/HI/SImode to work with
13824 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
13825 conversions to QI/HImode types did a store and then a load to
13826 truncate the value. For conversions to VSX registers, don't split
13827 the insn, instead emit the code directly. Use the code iterator
13828 any_fix to combine signed and unsigned conversions.
13829 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
13830 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
13831 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13832 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
13833 (fix_<mode>di2_hw): Likewise.
13834 (fixuns_<mode>di2_hw): Likewise.
13835 (fix_<mode>si2_hw): Likewise.
13836 (fixuns_<mode>si2_hw): Likewise.
13837 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
13838 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
13839 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
13840 fix<uns>_trunc<SFDF:mode>si2_p8.
13841 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
13842 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
13843 (fix<uns>_<mode>_mem): Likewise.
13844 (fctiw<u>z_<mode>_mem): Likewise.
13845 (fix<uns>_<mode>_mem): Likewise.
13846 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
13847 the register allocator from doing a direct move to the GPRs to do
13848 a store, and instead use the ISA 3.0 store byte/half-word from
13849 vector register instruction. For IEEE 128-bit floating point,
13850 also optimize stores of 32-bit ints.
13851 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
13853 2018-02-07 Alan Hayward <alan.hayward@arm.com>
13855 * genextract.c (push_pathstr_operand): New function to support
13857 (walk_rtx): Call push_pathstr_operand.
13858 (print_path): Support [a-zA-Z].
13860 2018-02-07 Richard Biener <rguenther@suse.de>
13862 PR tree-optimization/84037
13863 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
13864 (cse_and_gimplify_to_preheader): Declare.
13865 (vect_get_place_in_interleaving_chain): Likewise.
13866 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13868 (_loop_vec_info::~_loop_vec_info): Delete it.
13869 (cse_and_gimplify_to_preheader): New function.
13870 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
13871 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
13872 (vectorizable_load): Likewise. For grouped stores always base
13873 the IV on the first element.
13874 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
13875 condition before gimplifying.
13877 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13879 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
13880 *DIV_EXPR and *MOD_EXPR.
13882 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
13885 * config/i386/i386.c (ix86_option_override_internal): Mask out
13886 the CF_SET bit when checking -fcf-protection.
13888 2018-02-07 Tom de Vries <tom@codesourcery.com>
13891 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
13894 2018-02-07 Richard Biener <rguenther@suse.de>
13896 PR tree-optimization/84204
13897 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
13900 PR tree-optimization/84205
13901 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
13902 special-case isl_ast_op_zdiv_r.
13904 PR tree-optimization/84223
13905 * graphite-scop-detection.c (gather_bbs::before_dom_children):
13906 Only add conditions from within the region.
13907 (gather_bbs::after_dom_children): Adjust.
13909 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
13912 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
13913 * config/avr/avr.md: Only post-reload split REG-REG moves if
13914 either register is GENERAL_REG_P.
13916 2018-02-07 Jakub Jelinek <jakub@redhat.com>
13918 PR tree-optimization/84235
13919 * tree-ssa-scopedtables.c
13920 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
13921 if the subtraction is performed in floating point type where NaNs are
13922 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
13923 build 1. Formatting fix.
13925 2018-02-06 Jakub Jelinek <jakub@redhat.com>
13928 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
13929 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
13930 and skip it regardless of bb boundaries. Use CALL_P macro,
13931 don't test INSN_P (insn) together with CALL_P or JUMP_P check
13932 unnecessarily, formatting fix.
13934 2018-02-06 Michael Collison <michael.collison@arm.com>
13936 * config/arm/thumb2.md:
13937 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
13938 (*thumb_mov_notscc): Ditto.
13940 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
13943 * config/rs6000/rs6000.md (su code attribute): Use "u" for
13944 unsigned_fix, not "s".
13946 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13948 * configure.ac (gcc_fn_eh_frame_ro): New function.
13949 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
13950 correct .eh_frame permissions.
13951 * configure: Regenerate.
13953 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
13955 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
13956 irrelevant options.
13958 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13960 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13961 Display warning message for -mno-speculate-indirect-jumps.
13963 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
13965 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
13967 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
13969 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
13971 PR tree-optimization/84225
13972 * tree-eh.c (find_trapping_overflow): Only call
13973 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
13975 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
13978 * config/i386/i386.c: Reimplement the check of possible options
13979 -mibt/-mshstk conbination. Change error messages.
13980 * doc/invoke.texi: Fix a typo: remove extra '='.
13982 2018-02-06 Marek Polacek <polacek@redhat.com>
13984 PR tree-optimization/84228
13985 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
13987 2018-02-06 Tamar Christina <tamar.christina@arm.com>
13990 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
13991 emitted arch directives.
13992 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
13993 __ARM_FEATURE_COPROC before changing architectures.
13995 2018-02-06 Richard Biener <rguenther@suse.de>
13997 * config/i386/i386.c (print_reg): Fix typo.
13998 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
14000 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
14002 * configure: Regenerate.
14004 2018-02-05 Martin Sebor <msebor@redhat.com>
14006 PR tree-optimization/83369
14007 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
14010 2018-02-05 Martin Liska <mliska@suse.cz>
14012 * doc/invoke.texi: Cherry-pick upstream r323995.
14014 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
14016 * ira.c (ira_init_register_move_cost): Adjust comment.
14018 2018-02-05 Martin Liska <mliska@suse.cz>
14020 PR gcov-profile/84137
14021 * doc/gcov.texi: Fix typo in documentation.
14023 2018-02-05 Martin Liska <mliska@suse.cz>
14025 PR gcov-profile/83879
14026 * doc/gcov.texi: Document necessity of --dynamic-list-data when
14027 using dlopen functionality.
14029 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
14031 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
14032 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
14033 _mm_maskz_range_ss, _mm_mask_range_round_ss,
14034 _mm_maskz_range_round_ss): New intrinsics.
14035 (__builtin_ia32_rangesd128_round)
14036 (__builtin_ia32_rangess128_round): Remove.
14037 (__builtin_ia32_rangesd128_mask_round,
14038 __builtin_ia32_rangess128_mask_round): New builtins.
14039 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
14040 __builtin_ia32_rangess128_round): Remove.
14041 (__builtin_ia32_rangesd128_mask_round,
14042 __builtin_ia32_rangess128_mask_round): New builtins.
14043 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
14044 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
14045 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
14046 "<round_saeonly_constraint>")): Changed to ...
14047 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
14048 "<round_saeonly_scalar_constraint>")): ... this.
14049 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
14050 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
14051 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
14052 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
14053 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
14055 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
14057 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
14059 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
14060 Remove all values except native, 8540 and 8548.
14062 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
14064 * config/i386/i386.c (ix86_output_function_return): Pass
14065 INVALID_REGNUM, instead of -1, as invalid register number to
14066 indirect_thunk_name and output_indirect_thunk.
14068 2018-02-02 Julia Koval <julia.koval@intel.com>
14070 * config.gcc: Add -march=icelake.
14071 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
14072 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
14073 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
14074 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
14075 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
14076 (processor_target_table): Add icelake.
14077 (ix86_option_override_internal): Handle new PTAs.
14078 (get_builtin_code_for_version): Handle icelake.
14079 (M_INTEL_COREI7_ICELAKE): New.
14080 (fold_builtin_cpu): Handle icelake.
14081 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
14082 * doc/invoke.texi: Add -march=icelake.
14084 2018-02-02 Julia Koval <julia.koval@intel.com>
14086 * config/i386/i386.c (ix86_option_override_internal): Change flags type
14087 to wide_int_bitmask.
14088 * wide-int-bitmask.h: New.
14090 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
14093 * config/i386/i386.md: Replace Pmode with word_mode in
14094 builtin_setjmp_setup and builtin_longjmp to support x32.
14096 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
14100 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
14102 (rs6000_supported_cpu_names): New static variable.
14103 (linux_cpu_translation_table): Likewise.
14104 (elf_platform) <cpu>: Define new static variable and use it.
14105 Translate kernel AT_PLATFORM name to canonical name if needed.
14106 Error if platform name is unknown.
14108 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
14111 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
14113 2018-02-01 Jeff Law <law@redhat.com>
14116 * config/i386/i386.c (release_scratch_register_on_entry): Add new
14117 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
14118 the scratch if RELEASE_VIA_POP is false.
14119 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
14120 If we have to save a temporary register, decrement SIZE appropriately.
14121 Pass new arguments to release_scratch_register_on_entry.
14122 (ix86_adjust_stack_and_probe): Likewise.
14123 (ix86_emit_probe_stack_range): Pass new arguments to
14124 release_scratch_register_on_entry.
14126 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
14128 PR rtl-optimization/84157
14129 * combine.c (change_zero_ext): Use REG_P predicate in
14130 front of HARD_REGISTER_P predicate.
14132 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
14134 * config/avr/avr.c (avr_option_override): Move disabling of
14135 -fdelete-null-pointer-checks to...
14136 * common/config/avr/avr-common.c (avr_option_optimization_table):
14139 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14141 PR tree-optimization/81635
14142 * tree-data-ref.c (split_constant_offset_1): For types that
14143 wrap on overflow, try to use range info to prove that wrapping
14146 2018-02-01 Renlin Li <renlin.li@arm.com>
14149 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
14150 TAILCALL_ADDR_REGS.
14151 (aarch64_register_move_cost): Likewise.
14152 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
14153 TAILCALL_ADDR_REGS.
14154 (REG_CLASS_NAMES): Likewise.
14155 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
14156 TAILCALL_ADDR_REGS. Remove IP registers.
14157 * config/aarch64/aarch64.md (Ucs): Update register constraint.
14159 2018-02-01 Richard Biener <rguenther@suse.de>
14161 * domwalk.h (dom_walker::dom_walker): Add additional constructor
14162 for specifying RPO order and allow NULL for that.
14163 * domwalk.c (dom_walker::dom_walker): Likewise.
14164 (dom_walker::walk): Handle NULL RPO order.
14165 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
14167 (rewrite_update_dom_walker): Likewise.
14168 (mark_def_dom_walker): Likewise.
14170 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14172 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
14173 (aarch64_maybe_expand_sve_subreg_move): Declare.
14174 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
14175 * config/aarch64/predicates.md (aarch64_any_register_operand): New
14177 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
14178 that are semantically a reverse operation.
14179 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
14180 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
14181 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
14183 (aarch64_can_change_mode_class): For big-endian, forbid changes
14184 between two SVE modes if they have different element sizes.
14186 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14188 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
14189 the TImode handling for big-endian targets.
14191 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14193 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
14194 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
14196 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
14197 Remove BSWAP handing for big-endian targets and use the form of
14198 LD1RQ appropariate for the mode.
14200 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14202 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
14203 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
14204 duplicated element.
14206 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
14209 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
14210 check for operands that need to go through aarch64_sve_reload_be.
14212 2018-02-01 Jakub Jelinek <jakub@redhat.com>
14214 PR tree-optimization/81661
14215 PR tree-optimization/84117
14216 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
14217 * tree-eh.c: Include gimplify.h.
14218 (find_trapping_overflow, replace_trapping_overflow,
14219 rewrite_to_non_trapping_overflow): New functions.
14220 * tree-vect-loop.c: Include tree-eh.h.
14221 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
14222 * tree-data-ref.c: Include tree-eh.h.
14223 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
14225 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
14227 PR rtl-optimization/84123
14228 * combine.c (change_zero_ext): Check if hard register satisfies
14229 can_change_dest_mode before calling gen_lowpart_SUBREG.
14231 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
14234 * ira.c (ira_init_register_move_cost): Remove assert.
14236 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14238 PR rtl-optimization/84071
14239 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
14240 * doc/tm.texi: Regenerate.
14242 2018-01-31 Richard Biener <rguenther@suse.de>
14244 PR tree-optimization/84132
14245 * tree-data-ref.c (analyze_miv_subscript): Properly
14246 check whether evolution_function_is_affine_multivariate_p
14247 before calling gcd_of_steps_may_divide_p.
14249 2018-01-31 Julia Koval <julia.koval@intel.com>
14252 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
14253 * config/i386/i386.md (rdpid_rex64) New.
14254 (rdpid): Make 32bit only.
14256 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
14259 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
14260 an IDENTIFIER_NODE for FUNCTION_TYPE's.
14262 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14265 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14267 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14269 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
14271 PR rtl-optimization/84071
14272 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
14273 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
14275 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14277 * config/arc/arc.c (arc_handle_aux_attribute): New function.
14278 (arc_attribute_table): Add 'aux' attribute.
14279 (arc_in_small_data_p): Consider aux like variables.
14280 (arc_is_aux_reg_p): New function.
14281 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
14282 (arc_get_aux_arg): New function.
14283 (prepare_move_operands): Handle aux-register access.
14284 (arc_handle_aux_attribute): New function.
14285 * doc/extend.texi (ARC Variable attributes): Add subsection.
14287 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
14289 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
14290 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
14291 (arc_attribute_table): Add 'uncached' attribute.
14292 (arc_print_operand): Print '.di' flag for uncached memory
14294 (arc_in_small_data_p): Do not consider for small data the uncached
14296 (arc_is_uncached_mem_p): New function.
14297 * config/arc/predicates.md (compact_store_memory_operand): Check
14298 for uncached memory accesses.
14299 (nonvol_nonimm_operand): Likewise.
14300 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
14302 2018-01-31 Jakub Jelinek <jakub@redhat.com>
14305 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
14306 falign-loops=): Add Optimization flag.
14308 2018-01-30 Jeff Law <law@redhat.com>
14311 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
14312 INT_REGISTERS_SAVED. Check it prior to calling
14313 get_scratch_register_on_entry.
14314 (ix86_adjust_stack_and_probe): Similarly.
14315 (ix86_emit_probe_stack_range): Similarly.
14316 (ix86_expand_prologue): Corresponding changes.
14318 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14321 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
14322 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
14324 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
14327 * lra-constraints.c (curr_insn_transform): Process AND in the
14330 2018-01-30 Jakub Jelinek <jakub@redhat.com>
14332 PR rtl-optimization/83986
14333 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
14334 dependence against last_pending_memory_flush in addition to
14335 pending_jump_insns.
14337 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
14339 PR tree-optimization/81611
14340 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
14343 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14346 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
14349 2018-01-30 Richard Biener <rguenther@suse.de>
14350 Jakub Jelinek <jakub@redhat.com>
14352 PR tree-optimization/84111
14353 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
14354 inner loops added during recursion, as they don't have up-to-date
14357 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14360 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
14361 (can_inline_edge_by_limits_p): ... here.
14362 (can_early_inline_edge_p, check_callers,
14363 update_caller_keys, update_callee_keys, recursive_inlining,
14364 add_new_edges_to_heap, speculation_useful_p,
14365 inline_small_functions,
14366 inline_small_functions, flatten_function,
14367 inline_to_all_callers_1): Update.
14369 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
14371 * profile-count.c (profile_count::combine_with_ipa_count): Handle
14374 2018-01-30 Richard Biener <rguenther@suse.de>
14376 PR tree-optimization/83008
14377 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
14378 invariant and constant vector uses in stmts when they need
14379 more than one stmt.
14381 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14384 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
14385 * configure: Regenerate.
14387 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14389 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
14391 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
14392 Use gen_rtx_REG rather than gen_lowpart.
14394 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14396 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
14397 rather than 0 when creating partial subregs.
14399 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
14401 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
14404 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
14407 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
14408 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
14409 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
14410 flags. This restores the settings used before the 2017-07-24.
14411 Turning off pre increment/decrement/modify allows IVOPTS to
14412 optimize DF/SF loops where the index is an int.
14414 2018-01-29 Richard Biener <rguenther@suse.de>
14415 Kelvin Nilsen <kelvin@gcc.gnu.org>
14418 * tree-vect-stmts.c (vectorizable_call): Don't call
14419 targetm.vectorize_builtin_md_vectorized_function if callee is
14422 2018-01-22 Carl Love <cel@us.ibm.com>
14424 * doc/extend.tex: Fix typo in second arg in
14425 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
14427 2018-01-29 Richard Biener <rguenther@suse.de>
14429 PR tree-optimization/84086
14430 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
14431 (flush_ssaname_freelist): When SSA names were released reset
14432 the SCEV hash table.
14434 2018-01-29 Richard Biener <rguenther@suse.de>
14436 PR tree-optimization/84057
14437 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
14438 removed paths when removing edges.
14440 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
14442 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
14443 -mfunction-return=@var{choice}.
14445 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
14447 PR diagnostic/84034
14448 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
14449 Handle CR like TAB.
14450 (layout::print_source_line): Likewise.
14451 (test_get_line_width_without_trailing_whitespace): Add test cases.
14453 2018-01-27 Jakub Jelinek <jakub@redhat.com>
14455 PR middle-end/84040
14456 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
14459 2018-01-26 Jim Wilson <jimw@sifive.com>
14461 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
14463 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
14466 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14468 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
14469 and CMP + SUB-immediate -> SUBS.
14471 2018-01-26 Martin Sebor <msebor@redhat.com>
14473 PR tree-optimization/83896
14474 * tree-ssa-strlen.c (get_string_len): Rename...
14475 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
14476 Avoid assuming length is constant.
14477 (handle_char_store): Use HOST_WIDE_INT for string length.
14479 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
14482 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
14483 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
14485 2018-01-26 Richard Biener <rguenther@suse.de>
14487 PR rtl-optimization/84003
14488 * dse.c (record_store): Only record redundant stores when
14489 the earlier store aliases at least all accesses the later one does.
14491 2018-01-26 Jakub Jelinek <jakub@redhat.com>
14493 PR rtl-optimization/83985
14494 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
14495 REG_CFA_RESTORE insns.
14496 (delete_unmarked_insns): Don't ignore separate shrink wrapping
14497 REG_CFA_RESTORE insns here.
14500 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
14501 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
14503 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14505 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
14506 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
14507 (arc_init): Likewise.
14508 (arc_override_options): Likewise.
14509 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
14511 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
14513 * config/arc/arc.h (TARGET_DBNZ): Define.
14514 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
14515 properly set the tune attribute.
14516 (dbnz): Use TARGET_DBNZ guard.
14517 * config/arc/arc.opt (mtune): Add core3 option.
14519 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14521 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
14522 recognize new pic like addresses.
14523 (arc_delegitimize_address): Clean up.
14525 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14527 * config/arc/arc-arches.def: Option mrf16 valid for all
14529 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
14530 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
14531 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
14532 * config/arc/arc-tables.opt: Regenerate.
14533 * config/arc/arc.c (arc_conditional_register_usage): Handle
14534 reduced register file case.
14535 (arc_file_start): Set must have build attributes.
14536 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
14537 mrf16 option value.
14538 * config/arc/arc.opt (mrf16): Add new option.
14539 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
14540 * config/arc/genmultilib.awk: Handle new mrf16 option.
14541 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
14542 * config/arc/t-multilib: Regenerate.
14543 * doc/invoke.texi (ARC Options): Document mrf16 option.
14545 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14547 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
14548 * config/arc/arc.c (arc_handle_secure_attribute): New function.
14549 (arc_attribute_table): Add 'secure_call' attribute.
14550 (arc_print_operand): Print secure call operand.
14551 (arc_function_ok_for_sibcall): Don't optimize tail calls when
14553 (arc_is_secure_call_p): New function. * config/arc/arc.md
14554 (call_i): Add support for sjli instruction.
14555 (call_value_i): Likewise.
14556 * config/arc/constraints.md (Csc): New constraint.
14558 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
14559 John Eric Martin <John.Martin@emmicro-us.com>
14561 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
14562 * config/arc/arc.c (_arc_jli_section): New struct.
14563 (arc_jli_section): New type.
14564 (rc_jli_sections): New static variable.
14565 (arc_handle_jli_attribute): New function.
14566 (arc_attribute_table): Add jli_always and jli_fixed attribute.
14567 (arc_file_end): New function.
14568 (TARGET_ASM_FILE_END): Define.
14569 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
14570 (arc_add_jli_section): New function.
14571 (jli_call_scan): Likewise.
14572 (arc_reorg): Call jli_call_scan.
14573 (arc_output_addsi): Remove 'S' from printing asm operand.
14574 (arc_is_jli_call_p): New function.
14575 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
14577 (movhi_insn): Likewise.
14578 (movsi_insn): Likewise.
14579 (movsi_set_cc_insn): Likewise.
14580 (loadqi_update): Likewise.
14581 (load_zeroextendqisi_update): Likewise.
14582 (load_signextendqisi_update): Likewise.
14583 (loadhi_update): Likewise.
14584 (load_zeroextendhisi_update): Likewise.
14585 (load_signextendhisi_update): Likewise.
14586 (loadsi_update): Likewise.
14587 (loadsf_update): Likewise.
14588 (movsicc_insn): Likewise.
14589 (bset_insn): Likewise.
14590 (bxor_insn): Likewise.
14591 (bclr_insn): Likewise.
14592 (bmsk_insn): Likewise.
14593 (bicsi3_insn): Likewise.
14594 (cmpsi_cc_c_insn): Likewise.
14595 (movsi_ne): Likewise.
14596 (movsi_cond_exec): Likewise.
14597 (clrsbsi2): Likewise.
14598 (norm_f): Likewise.
14608 (call_i): Remove 'S' asm letter, add jli instruction.
14609 (call_value_i): Likewise.
14610 * config/arc/arc.op (mjli-always): New option.
14611 * config/arc/constraints.md (Cji): New constraint.
14612 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
14614 (subsf3_fpx): Likewise.
14615 (mulsf3_fpx): Likewise.
14616 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
14618 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
14619 function attrbutes.
14620 * doc/invoke.texi (ARC): Document mjli-always option.
14622 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
14624 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
14625 avoid addition with 0 and use incw and decw where possible.
14627 2018-01-26 Richard Biener <rguenther@suse.de>
14629 PR tree-optimization/81082
14630 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
14631 association if it requires casting to unsigned.
14632 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
14633 from fold_plusminus_mult_expr to catch important cases late when
14634 range info is available.
14636 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14638 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
14639 * configure.ac (hidden_linkonce): New test.
14640 * configure: Regenerate.
14641 * config.in: Regenerate.
14643 2018-01-26 Julia Koval <julia.koval@intel.com>
14645 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
14646 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
14647 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
14648 _mm_mask_bitshuffle_epi64_mask): Fix type.
14649 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
14650 USI_FTYPE_V4DI_V4DI_USI): Remove.
14651 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
14652 __builtin_ia32_vpshufbitqmb256_mask,
14653 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
14654 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
14655 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
14657 2018-01-26 Alan Modra <amodra@gmail.com>
14660 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
14661 UNSPEC_VBPERMQ. Sort other unspecs.
14663 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
14665 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
14667 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
14669 PR middle-end/83055
14670 * predict.c (drop_profile): Do not push/pop cfun; update also
14672 (handle_missing_profiles): Fix logic looking for zero profiles.
14674 2018-01-25 Jakub Jelinek <jakub@redhat.com>
14676 PR middle-end/83977
14677 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
14678 on functions with #pragma omp declare simd or functions with simd
14680 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
14681 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
14682 Remove trailing \n from warning_at calls.
14684 2018-01-25 Tom de Vries <tom@codesourcery.com>
14687 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14688 for neutered workers.
14690 2018-01-24 Joseph Myers <joseph@codesourcery.com>
14693 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
14694 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
14696 2018-01-24 Jeff Law <law@redhat.com>
14699 * i386.c (get_probe_interval): Move to earlier point.
14700 (ix86_compute_frame_layout): If -fstack-clash-protection and
14701 the frame is larger than the probe interval, then use pushes
14702 to save registers rather than reg->mem moves.
14703 (ix86_expand_prologue): Remove conditional for int_registers_saved
14706 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
14709 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
14710 min/max for never referenced object.
14712 2018-01-24 Jakub Jelinek <jakub@redhat.com>
14714 PR middle-end/83977
14715 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
14717 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
14718 attributes from DECL_ATTRIBUTES (decl) without affecting
14719 DECL_ATTRIBUTES (current_function_decl).
14720 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
14721 functions with non-NULL DECL_ABSTRACT_ORIGIN.
14723 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
14725 PR tree-optimization/83979
14726 * fold-const.c (fold_comparison): Use constant_boolean_node
14727 instead of boolean_{true,false}_node.
14729 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
14731 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
14734 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14736 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
14737 Simplify the clause that sets the length attribute.
14738 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
14739 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
14740 clause that sets the length attribute.
14741 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
14743 2018-01-24 Tom de Vries <tom@codesourcery.com>
14746 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
14747 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
14748 Add strict parameter.
14749 (prevent_branch_around_nothing): Insert dummy insn between branch to
14750 label and label with no ptx insn inbetween.
14751 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
14753 2018-01-24 Tom de Vries <tom@codesourcery.com>
14756 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
14757 for neutered threads in warp.
14758 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
14760 2018-01-24 Richard Biener <rguenther@suse.de>
14762 PR tree-optimization/83176
14763 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
14766 2018-01-24 Richard Biener <rguenther@suse.de>
14768 PR tree-optimization/82819
14769 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
14770 code generating pluses that are no-ops in the target precision.
14772 2018-01-24 Richard Biener <rguenther@suse.de>
14774 PR middle-end/84000
14775 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
14777 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14779 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
14780 to merge probabilities.
14781 * predict.c (probably_never_executed): Also mark as cold functions
14782 with global 0 profile and guessed local profile.
14783 * profile-count.c (profile_probability::combine_with_count): New
14785 * profile-count.h (profile_probability::operator*,
14786 profile_probability::operator*=, profile_probability::operator/,
14787 profile_probability::operator/=): Reduce precision to adjusted
14788 and set value to guessed on contradictory divisions.
14789 (profile_probability::combine_with_freq): Remove.
14790 (profile_probability::combine_wiht_count): Declare.
14791 (profile_count::force_nonzero):: Set to adjusted.
14792 (profile_count::probability_in):: Set quality to adjusted.
14793 * tree-ssa-tail-merge.c (replace_block_by): Use
14794 combine_with_count.
14796 2018-01-23 Andrew Waterman <andrew@sifive.com>
14797 Jim Wilson <jimw@sifive.com>
14799 * config/riscv/riscv.c (riscv_stack_boundary): New.
14800 (riscv_option_override): Set riscv_stack_boundary. Handle
14801 riscv_preferred_stack_boundary_arg.
14802 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
14803 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
14804 (STACK_BOUNDARY): Set to riscv_stack_boundary.
14805 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
14806 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
14807 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
14809 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
14812 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
14813 of struct ix86_frame.
14814 (ix86_expand_epilogue): Likewise. Add a local variable for
14815 the reg_save_offset field in struct ix86_frame.
14817 2018-01-23 Bin Cheng <bin.cheng@arm.com>
14819 PR tree-optimization/82604
14820 * tree-loop-distribution.c (enum partition_kind): New enum item
14821 PKIND_PARTIAL_MEMSET.
14822 (partition_builtin_p): Support above new enum item.
14823 (generate_code_for_partition): Ditto.
14824 (compute_access_range): Differentiate cases that equality can be
14825 proven at all loops, the innermost loops or no loops.
14826 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
14827 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
14828 (finalize_partitions, distribute_loop): Don't fuse partition of
14829 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
14830 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
14831 parloop is enabled.
14833 2018-01-23 Martin Liska <mliska@suse.cz>
14835 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
14836 order to ignore the predictor.
14837 (PRED_POLYMORPHIC_CALL): Likewise.
14838 (PRED_RECURSIVE_CALL): Likewise.
14840 2018-01-23 Martin Liska <mliska@suse.cz>
14842 * tree-profile.c (tree_profiling): Print function header to
14843 aware reader which function we are working on.
14844 * value-prof.c (gimple_find_values_to_profile): Do not print
14845 not interesting value histograms.
14847 2018-01-23 Martin Liska <mliska@suse.cz>
14849 * profile-count.h (enum profile_quality): Add
14850 profile_uninitialized as the first value. Do not number values
14851 as they are zero based.
14852 (profile_count::verify): Update sanity check.
14853 (profile_probability::verify): Likewise.
14855 2018-01-23 Nathan Sidwell <nathan@acm.org>
14857 * doc/invoke.texi (ffor-scope): Deprecate.
14859 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14861 PR tree-optimization/83510
14862 * domwalk.c (set_all_edges_as_executable): New function.
14863 (dom_walker::dom_walker): Convert bool param
14864 "skip_unreachable_blocks" to enum reachability. Move setup of
14865 edge flags to set_all_edges_as_executable and only do it when
14866 reachability is REACHABLE_BLOCKS.
14867 * domwalk.h (enum dom_walker::reachability): New enum.
14868 (dom_walker::dom_walker): Convert bool param
14869 "skip_unreachable_blocks" to enum reachability.
14870 (set_all_edges_as_executable): New decl.
14871 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
14872 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
14874 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
14875 but converting true to REACHABLE_BLOCKS.
14876 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
14878 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
14879 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
14880 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
14882 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
14883 if check_all_array_refs will be called.
14885 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14887 * tree.c (selftest::test_location_wrappers): Add more test
14890 2018-01-23 David Malcolm <dmalcolm@redhat.com>
14892 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
14893 (selftest::test_bit_in_range): Likewise.
14895 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14898 * doc/sourcebuild.texi (vect_float): Say that the selector
14899 only describes the situation when -funsafe-math-optimizations is on.
14900 (vect_float_strict): Document.
14902 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
14904 PR tree-optimization/83965
14905 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14906 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
14907 instead of checking only for a reduction.
14908 (vect_recog_widen_sum_pattern): Likewise.
14910 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14912 * predict.c (probably_never_executed): Only use precise profile info.
14913 (compute_function_frequency): Skip after inlining hack since we now
14914 have quality checking.
14916 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
14918 * profile-count.h (profile_probability::very_unlikely,
14919 profile_probability::unlikely, profile_probability::even): Set
14920 precision to guessed.
14922 2018-01-23 Richard Biener <rguenther@suse.de>
14924 PR tree-optimization/83963
14925 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
14926 Properly terminate dominator walk when crossing the exit edge not
14927 when visiting its source block.
14929 2018-01-23 Jakub Jelinek <jakub@redhat.com>
14932 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
14933 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
14935 2018-01-22 Jakub Jelinek <jakub@redhat.com>
14937 PR tree-optimization/83957
14938 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
14939 semicolon after for body surrounded by braces.
14941 PR tree-optimization/83081
14942 * profile-count.h (profile_probability::split): New method.
14943 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
14944 Use profile_probability::split.
14945 (do_compare_rtx_and_jump): Fix adjustment of probabilities
14946 when splitting a single conditional jump into 2.
14948 2018-01-22 David Malcolm <dmalcolm@redhat.com>
14950 PR tree-optimization/69452
14951 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
14954 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14956 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
14957 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
14958 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
14960 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14962 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
14964 * config/rl78/rl78.md (movdi): New define_expand.
14965 * config/rl78/rl78.c (rl78_split_movdi): New function.
14967 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
14970 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
14972 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
14973 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
14974 128-bit to produce an UNSPEC move to get the double word with the
14975 signbit and then a shift directly to do signbit.
14976 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
14977 implementation with a new version that just does either a direct
14978 move or a regular move. Move memory interface to separate insns.
14979 Move insns so they are next to the expander.
14980 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
14981 with signbit move. Split big and little endian case.
14982 (signbit<mode>2_dm_mem_le): Likewise.
14983 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
14984 (signbit<mode>2_dm2): Likewise.
14986 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14988 * config/rl78/rl78.md (anddi3): New define_expand.
14990 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14992 * config/rl78/rl78.md (umindi3): New define_expand.
14994 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
14996 * config/rl78/rl78.md (smindi3): New define_expand.
14998 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15000 * config/rl78/rl78.md (smaxdi3): New define_expand.
15002 2018-01-22 Carl Love <cel@us.ibm.com>
15004 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
15005 LVX_V1TI): Add macro expansion.
15006 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
15007 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
15008 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
15009 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
15010 Change check to determine if the instruction is a byte reversing
15011 entry. Fix typo in comment.
15012 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
15013 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
15014 Add def_builtin calls for new builtins.
15015 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
15016 Add define_insn expansion.
15018 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15020 * config/rl78/rl78.md (umaxdi3): New define_expand.
15022 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
15024 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
15025 for non-QImode registers.
15027 2018-01-22 Richard Biener <rguenther@suse.de>
15029 PR tree-optimization/83963
15030 * graphite-scop-detection.c (scop_detection::get_sese): Delay
15031 including the loop exit block.
15032 (scop_detection::merge_sese): Likewise.
15033 (scop_detection::add_scop): Do it here instead.
15035 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15037 * doc/sourcebuild.texi (arm_softfloat): Document.
15039 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
15042 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
15043 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
15044 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
15046 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15047 David Edelsohn <dje.gcc@gmail.com>
15050 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15051 Change "crset eq" to "crset 2".
15052 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15053 (*call_indirect_aix<mode>_nospec): Likewise.
15054 (*call_value_indirect_aix<mode>_nospec): Likewise.
15055 (*call_indirect_elfv2<mode>_nospec): Likewise.
15056 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
15057 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
15058 change assembly output from . to $.
15059 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15060 (indirect_jump<mode>_nospec): Change assembly output from . to $.
15061 (*tablejump<mode>_internal1_nospec): Likewise.
15063 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
15066 * config/sh/sh_optimize_sett_clrt.cc:
15067 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
15069 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
15071 PR tree-optimization/83940
15072 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
15073 offset_dt to vect_constant_def rather than vect_unknown_def_type.
15074 (vect_check_load_store_mask): Add a mask_dt_out parameter and
15075 use it to pass back the definition type.
15076 (vect_check_store_rhs): Likewise rhs_dt_out.
15077 (vect_build_gather_load_calls): Add a mask_dt argument and use
15078 it instead of a call to vect_is_simple_use.
15079 (vectorizable_store): Update calls to vect_check_load_store_mask
15080 and vect_check_store_rhs. Use the dt returned by the latter instead
15081 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
15082 instead of calls to vect_is_simple_use. Pass the scalar rather
15083 than the vector operand to vect_is_simple_use when handling
15084 second and subsequent copies of an rhs value.
15085 (vectorizable_load): Update calls to vect_check_load_store_mask
15086 and vect_build_gather_load_calls. Use the cached mask_dt and
15087 gs_info.offset_dt instead of calls to vect_is_simple_use.
15089 2018-01-20 Jakub Jelinek <jakub@redhat.com>
15091 PR middle-end/83945
15092 * tree-emutls.c: Include gimplify.h.
15093 (lower_emutls_2): New function.
15094 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
15095 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
15096 it before further processing.
15099 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
15100 UINTVAL (trueop1) instead of INTVAL (op1).
15102 2018-01-19 Jakub Jelinek <jakub@redhat.com>
15106 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
15107 INCOMING_FRAME_SP_OFFSET if not defined.
15108 (scan_trace): Add ENTRY argument. If true and
15109 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
15110 emit a note to adjust the CFA offset.
15111 (create_cfi_notes): Adjust scan_trace callers.
15112 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
15113 INCOMING_FRAME_SP_OFFSET in the CIE.
15114 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
15115 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
15117 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
15118 * doc/tm.texi: Regenerated.
15120 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
15122 PR rtl-optimization/83147
15123 * lra-constraints.c (remove_inheritance_pseudos): Use
15124 lra_substitute_pseudo_within_insn.
15126 2018-01-19 Tom de Vries <tom@codesourcery.com>
15127 Cesar Philippidis <cesar@codesourcery.com>
15130 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
15132 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
15135 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
15136 spaces for function labels.
15138 2018-01-19 Martin Liska <mliska@suse.cz>
15140 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
15141 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
15142 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
15143 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
15144 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
15145 (PRED_CONST_RETURN): Change from 69 to 65.
15146 (PRED_NULL_RETURN): Change from 91 to 71.
15147 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
15148 (PRED_LOOP_GUARD): Change from 66 to 73.
15150 2018-01-19 Martin Liska <mliska@suse.cz>
15152 * predict.c (predict_insn_def): Add new assert.
15153 (struct branch_predictor): Change type to signed integer.
15154 (test_prediction_value_range): Amend test to cover
15155 PROB_UNINITIALIZED.
15156 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
15157 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
15158 (PRED_LOOP_ITERATIONS_MAX): Likewise.
15159 (PRED_LOOP_IV_COMPARE): Likewise.
15160 * predict.h (PROB_UNINITIALIZED): Define new constant.
15162 2018-01-19 Martin Liska <mliska@suse.cz>
15164 * predict.c (dump_prediction): Add new format for
15165 analyze_brprob.py script which is enabled with -details
15167 * profile-count.h (precise_p): New function.
15169 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15171 PR tree-optimization/83922
15172 * tree-vect-loop.c (vect_verify_full_masking): Return false if
15173 there are no statements that need masking.
15174 (vect_active_double_reduction_p): New function.
15175 (vect_analyze_loop_operations): Use it when handling phis that
15176 are not in the loop header.
15178 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
15180 PR tree-optimization/83914
15181 * tree-vect-loop.c (vectorizable_induction): Don't convert
15182 init_expr or apply the peeling adjustment for inductions
15183 that are nested within the vectorized loop.
15185 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15187 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
15190 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15194 * function.h (gimplify_parameters): Add gimple_seq * argument.
15195 * function.c: Include gimple.h and options.h.
15196 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
15197 for the added local temporaries if needed.
15198 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
15199 if there are any parameter cleanups, wrap whole body into a
15200 try/finally with the cleanups.
15202 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
15205 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
15206 Use GET_MODE_CLASS for scalar floating point.
15208 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15212 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
15213 Fix call of call_cgraph_insertion_hooks.
15215 2018-01-18 Martin Sebor <msebor@redhat.com>
15217 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
15219 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
15222 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
15225 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
15228 * common.opt: (-ffile-prefix-map): New option.
15229 * opts.c (common_handle_option): Defer it.
15230 * opts-global.c (handle_common_deferred_options): Handle it.
15231 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
15232 * file-prefix-map.h: New file.
15233 (remap_debug_filename, add_debug_prefix_map): ...here.
15234 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
15235 * final.c (debug_prefix_map, add_debug_prefix_map
15236 remap_debug_filename): Move to...
15237 * file-prefix-map.c: New file.
15238 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
15239 generalize, get rid of alloca(), use strrchr() instead of strchr().
15240 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
15241 Implement in terms of add_prefix_map().
15242 (remap_macro_filename, remap_debug_filename): Implement in term of
15244 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
15245 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
15246 * dbxout.c: Include file-prefix-map.h.
15247 * varasm.c: Likewise.
15248 * vmsdbgout.c: Likewise.
15249 * xcoffout.c: Likewise.
15250 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
15251 * doc/cppopts.texi (-fmacro-prefix-map): Document.
15252 * doc/invoke.texi (-ffile-prefix-map): Document.
15253 (-fdebug-prefix-map): Update description.
15255 2018-01-18 Martin Liska <mliska@suse.cz>
15257 * config/i386/i386.c (indirect_thunk_name): Document that also
15259 (output_indirect_thunk): Document why both instructions
15260 (pause and lfence) are generated.
15262 2018-01-18 Richard Biener <rguenther@suse.de>
15264 PR tree-optimization/83887
15265 * graphite-scop-detection.c
15266 (scop_detection::get_nearest_dom_with_single_entry): Remove.
15267 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
15268 (scop_detection::merge_sese): Re-implement with a flood-fill
15269 algorithm that properly finds a SESE region if it exists.
15271 2018-01-18 Jakub Jelinek <jakub@redhat.com>
15274 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
15275 pointer_diff optimizations use view_convert instead of convert.
15277 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15279 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
15280 Generate different code for -mno-speculate-indirect-jumps.
15281 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
15282 (*call_indirect_aix<mode>): Disable for
15283 -mno-speculate-indirect-jumps.
15284 (*call_indirect_aix<mode>_nospec): New define_insn.
15285 (*call_value_indirect_aix<mode>): Disable for
15286 -mno-speculate-indirect-jumps.
15287 (*call_value_indirect_aix<mode>_nospec): New define_insn.
15288 (*sibcall_nonlocal_sysv<mode>): Generate different code for
15289 -mno-speculate-indirect-jumps.
15290 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
15292 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
15294 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
15295 long double type, set the flags for noting the default long double
15296 type, even if we don't pass or return a long double type.
15298 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
15301 * ipa-inline.c (flatten_function): Do not overwrite final inlining
15304 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
15306 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
15307 support for merge[hl].
15308 (fold_mergehl_helper): New helper function.
15309 (tree-vector-builder.h): New #include for tree_vector_builder usage.
15310 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
15311 (altivec_vmrglw_direct): Add xxmrglw insn.
15313 2018-01-17 Andrew Waterman <andrew@sifive.com>
15315 * config/riscv/riscv.c (riscv_conditional_register_usage): If
15316 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
15318 2018-01-17 David Malcolm <dmalcolm@redhat.com>
15321 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
15322 call the lto_location_cache before reading the
15323 DECL_SOURCE_LOCATION of the types.
15325 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
15326 Richard Sandiford <richard.sandiford@linaro.org>
15328 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
15329 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
15330 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
15331 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
15332 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
15334 * config/aarch64/constraints.md (aarch64_movti_operand):
15336 * config/aarch64/predicates.md (Uti): Add new constraint.
15338 2018-01-17 Carl Love <cel@us.ibm.com>
15340 * config/rs6000/vsx.md (define_expand xl_len_r,
15341 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
15342 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
15344 (define_expand, define_insn): Move the shift left from the
15345 define_insn to the define_expand for lxvl and stxvl instructions.
15346 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
15347 and XL_LEN_R definitions to PURE.
15349 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
15351 * config/i386/i386.c (indirect_thunk_name): Declare regno
15352 as unsigned int. Compare regno with INVALID_REGNUM.
15353 (output_indirect_thunk): Ditto.
15354 (output_indirect_thunk_function): Ditto.
15355 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
15356 in the call to output_indirect_thunk_function.
15358 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
15360 PR middle-end/83884
15361 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
15362 rather than the size of inner_type to determine the stack slot size
15363 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
15365 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
15368 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
15371 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
15373 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
15374 endian Linux systems to optionally enable multilibs for selecting
15375 the long double type if the user configured an explicit type.
15376 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
15377 have no long double multilibs if not defined.
15378 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
15379 warn if the user used -mabi={ieee,ibm}longdouble and we built
15380 multilibs for long double.
15381 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
15382 appropriate multilib option.
15383 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
15385 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
15386 for building long double multilibs.
15387 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
15389 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
15391 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
15394 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
15396 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
15399 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
15402 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
15405 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15407 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
15408 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
15410 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
15412 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
15413 different rtl trees depending on TARGET_64BIT.
15414 (rs6000_gen_lvx): Likewise.
15416 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
15418 * config/visium/visium.md (nop): Tweak comment.
15419 (hazard_nop): Likewise.
15421 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
15423 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
15424 -mspeculate-indirect-jumps.
15425 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
15426 for -mno-speculate-indirect-jumps.
15427 (*call_indirect_elfv2<mode>_nospec): New define_insn.
15428 (*call_value_indirect_elfv2<mode>): Disable for
15429 -mno-speculate-indirect-jumps.
15430 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
15431 (indirect_jump): Emit different RTL for
15432 -mno-speculate-indirect-jumps.
15433 (*indirect_jump<mode>): Disable for
15434 -mno-speculate-indirect-jumps.
15435 (*indirect_jump<mode>_nospec): New define_insn.
15436 (tablejump): Emit different RTL for
15437 -mno-speculate-indirect-jumps.
15438 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
15439 (tablejumpsi_nospec): New define_expand.
15440 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
15441 (tablejumpdi_nospec): New define_expand.
15442 (*tablejump<mode>_internal1): Disable for
15443 -mno-speculate-indirect-jumps.
15444 (*tablejump<mode>_internal1_nospec): New define_insn.
15445 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
15448 2018-01-16 Artyom Skrobov tyomitch@gmail.com
15450 * caller-save.c (insert_save): Drop unnecessary parameter. All
15453 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15454 Richard Biener <rguenth@suse.de>
15457 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
15458 return early, inline manually is_gimple_sizepos. Make sure if we
15459 call gimplify_expr we don't end up with a gimple constant.
15460 * tree.c (variably_modified_type_p): Don't return true for
15461 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
15462 * gimplify.h (is_gimple_sizepos): Remove.
15464 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15466 PR tree-optimization/83857
15467 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
15468 vectorizable_live_operation for pure SLP statements.
15469 (vectorizable_live_operation): Handle PHIs.
15471 2018-01-16 Richard Biener <rguenther@suse.de>
15473 PR tree-optimization/83867
15474 * tree-vect-stmts.c (vect_transform_stmt): Precompute
15475 nested_in_vect_loop_p since the scalar stmt may get invalidated.
15477 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15480 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
15481 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
15482 If off is not INTEGER_CST, issue a may not be aligned warning
15483 rather than isn't aligned. Use isn%'t rather than isn't.
15484 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
15486 <case MULT_EXPR>: Improve the case when bottom and one of the
15487 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
15488 operand, in that case check if the other operand is multiple of
15489 bottom divided by the INTEGER_CST operand.
15491 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15494 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
15495 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
15496 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
15497 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
15498 * config/pa/pa.c (pa_function_arg_advance): Likewise.
15499 (pa_function_arg, pa_arg_partial_bytes): Likewise.
15500 (pa_function_arg_size): New function.
15502 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15504 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
15505 in a separate statement.
15507 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
15509 PR tree-optimization/83847
15510 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
15511 group gathers and scatters.
15513 2018-01-16 Jakub Jelinek <jakub@redhat.com>
15515 PR rtl-optimization/86620
15516 * params.def (max-sched-ready-insns): Bump minimum value to 1.
15518 PR rtl-optimization/83213
15519 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
15520 to last if both are JUMP_INSNs.
15522 PR tree-optimization/83843
15523 * gimple-ssa-store-merging.c
15524 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
15525 store_immediate_info for bswap/nop orig_stores.
15527 2018-01-15 Andrew Waterman <andrew@sifive.com>
15529 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
15531 <UDIV>: Increase cost if !TARGET_DIV.
15533 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
15535 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
15536 (define_attr "cr_logical_3op"): New.
15537 (cceq_ior_compare): Adjust.
15538 (cceq_ior_compare_complement): Adjust.
15539 (*cceq_rev_compare): Adjust.
15540 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
15541 (is_cracked_insn): Adjust.
15542 (insn_must_be_first_in_group): Adjust.
15543 * config/rs6000/40x.md: Adjust.
15544 * config/rs6000/440.md: Adjust.
15545 * config/rs6000/476.md: Adjust.
15546 * config/rs6000/601.md: Adjust.
15547 * config/rs6000/603.md: Adjust.
15548 * config/rs6000/6xx.md: Adjust.
15549 * config/rs6000/7450.md: Adjust.
15550 * config/rs6000/7xx.md: Adjust.
15551 * config/rs6000/8540.md: Adjust.
15552 * config/rs6000/cell.md: Adjust.
15553 * config/rs6000/e300c2c3.md: Adjust.
15554 * config/rs6000/e500mc.md: Adjust.
15555 * config/rs6000/e500mc64.md: Adjust.
15556 * config/rs6000/e5500.md: Adjust.
15557 * config/rs6000/e6500.md: Adjust.
15558 * config/rs6000/mpc.md: Adjust.
15559 * config/rs6000/power4.md: Adjust.
15560 * config/rs6000/power5.md: Adjust.
15561 * config/rs6000/power6.md: Adjust.
15562 * config/rs6000/power7.md: Adjust.
15563 * config/rs6000/power8.md: Adjust.
15564 * config/rs6000/power9.md: Adjust.
15565 * config/rs6000/rs64.md: Adjust.
15566 * config/rs6000/titan.md: Adjust.
15568 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15570 * config/i386/predicates.md (indirect_branch_operand): Rewrite
15571 ix86_indirect_branch_register logic.
15573 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15575 * config/i386/constraints.md (Bs): Update
15576 ix86_indirect_branch_register check. Don't check
15577 ix86_indirect_branch_register with GOT_memory_operand.
15579 * config/i386/predicates.md (GOT_memory_operand): Don't check
15580 ix86_indirect_branch_register here.
15581 (GOT32_symbol_operand): Likewise.
15583 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15585 * config/i386/predicates.md (constant_call_address_operand):
15586 Rewrite ix86_indirect_branch_register logic.
15587 (sibcall_insn_operand): Likewise.
15589 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15591 * config/i386/constraints.md (Bs): Replace
15592 ix86_indirect_branch_thunk_register with
15593 ix86_indirect_branch_register.
15595 * config/i386/i386.md (indirect_jump): Likewise.
15596 (tablejump): Likewise.
15597 (*sibcall_memory): Likewise.
15598 (*sibcall_value_memory): Likewise.
15599 Peepholes of indirect call and jump via memory: Likewise.
15600 * config/i386/i386.opt: Likewise.
15601 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15602 (GOT_memory_operand): Likewise.
15603 (call_insn_operand): Likewise.
15604 (sibcall_insn_operand): Likewise.
15605 (GOT32_symbol_operand): Likewise.
15607 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15609 PR middle-end/83837
15610 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
15611 type rather than type addr's type points to.
15612 (expand_omp_atomic_mutex): Likewise.
15613 (expand_omp_atomic): Likewise.
15615 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
15618 * config/i386/i386.c (output_indirect_thunk_function): Use
15619 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
15620 for __x86_return_thunk.
15622 2018-01-15 Richard Biener <rguenther@suse.de>
15624 PR middle-end/83850
15625 * expmed.c (extract_bit_field_1): Fix typo.
15627 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15630 * config/arm/iterators.md (VF): New mode iterator.
15631 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
15632 Remove integer-related logic from pattern.
15633 (neon_vabd<mode>_3): Likewise.
15635 2018-01-15 Jakub Jelinek <jakub@redhat.com>
15637 PR middle-end/82694
15638 * common.opt (fstrict-overflow): No longer an alias.
15639 (fwrapv-pointer): New option.
15640 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
15641 also for pointer types based on flag_wrapv_pointer.
15642 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
15643 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
15644 opts->x_flag_wrapv got set.
15645 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
15646 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
15647 POINTER_TYPE_OVERFLOW_UNDEFINED.
15648 * match.pd: Likewise in address comparison pattern.
15649 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
15651 2018-01-15 Richard Biener <rguenther@suse.de>
15654 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
15655 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
15656 Reset type names to their identifier if their TYPE_DECL doesn't
15657 have linkage (and thus is used for ODR and devirt).
15658 (save_debug_info_for_decl): Remove.
15659 (save_debug_info_for_type): Likewise.
15660 (add_tree_to_fld_list): Adjust.
15661 * tree-pretty-print.c (dump_generic_node): Make dumping of
15662 type names more robust.
15664 2018-01-15 Richard Biener <rguenther@suse.de>
15666 * BASE-VER: Bump to 8.0.1.
15668 2018-01-14 Martin Sebor <msebor@redhat.com>
15671 * builtins.c (check_access): Avoid warning when the no-warning bit
15674 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
15676 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
15677 * ira-color (allocno_hard_regs_compare): Likewise.
15679 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
15682 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
15683 Use .pushsection/.popsection.
15685 2018-01-14 Martin Sebor <msebor@redhat.com>
15688 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
15690 2018-01-14 Jakub Jelinek <jakub@redhat.com>
15692 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
15693 entry from extra_headers.
15694 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
15695 extra_headers, make the list bitwise identical to the i?86-*-* one.
15697 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15699 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
15700 -mcmodel=large with -mindirect-branch=thunk,
15701 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
15702 -mfunction-return=thunk-extern.
15703 * doc/invoke.texi: Document -mcmodel=large is incompatible with
15704 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
15705 -mfunction-return=thunk and -mfunction-return=thunk-extern.
15707 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15709 * config/i386/i386.c (print_reg): Print the name of the full
15710 integer register without '%'.
15711 (ix86_print_operand): Handle 'V'.
15712 * doc/extend.texi: Document 'V' modifier.
15714 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15716 * config/i386/constraints.md (Bs): Disallow memory operand for
15717 -mindirect-branch-register.
15719 * config/i386/predicates.md (indirect_branch_operand): Likewise.
15720 (GOT_memory_operand): Likewise.
15721 (call_insn_operand): Likewise.
15722 (sibcall_insn_operand): Likewise.
15723 (GOT32_symbol_operand): Likewise.
15724 * config/i386/i386.md (indirect_jump): Call convert_memory_address
15725 for -mindirect-branch-register.
15726 (tablejump): Likewise.
15727 (*sibcall_memory): Likewise.
15728 (*sibcall_value_memory): Likewise.
15729 Disallow peepholes of indirect call and jump via memory for
15730 -mindirect-branch-register.
15731 (*call_pop): Replace m with Bw.
15732 (*call_value_pop): Likewise.
15733 (*sibcall_pop_memory): Replace m with Bs.
15734 * config/i386/i386.opt (mindirect-branch-register): New option.
15735 * doc/invoke.texi: Document -mindirect-branch-register option.
15737 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15739 * config/i386/i386-protos.h (ix86_output_function_return): New.
15740 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
15741 set function_return_type.
15742 (indirect_thunk_name): Add ret_p to indicate thunk for function
15744 (output_indirect_thunk_function): Pass false to
15745 indirect_thunk_name.
15746 (ix86_output_indirect_branch_via_reg): Likewise.
15747 (ix86_output_indirect_branch_via_push): Likewise.
15748 (output_indirect_thunk_function): Create alias for function
15749 return thunk if regno < 0.
15750 (ix86_output_function_return): New function.
15751 (ix86_handle_fndecl_attribute): Handle function_return.
15752 (ix86_attribute_table): Add function_return.
15753 * config/i386/i386.h (machine_function): Add
15754 function_return_type.
15755 * config/i386/i386.md (simple_return_internal): Use
15756 ix86_output_function_return.
15757 (simple_return_internal_long): Likewise.
15758 * config/i386/i386.opt (mfunction-return=): New option.
15759 (indirect_branch): Mention -mfunction-return=.
15760 * doc/extend.texi: Document function_return function attribute.
15761 * doc/invoke.texi: Document -mfunction-return= option.
15763 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
15765 * config/i386/i386-opts.h (indirect_branch): New.
15766 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
15767 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
15768 with local indirect jump when converting indirect call and jump.
15769 (ix86_set_indirect_branch_type): New.
15770 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
15771 (indirectlabelno): New.
15772 (indirect_thunk_needed): Likewise.
15773 (indirect_thunk_bnd_needed): Likewise.
15774 (indirect_thunks_used): Likewise.
15775 (indirect_thunks_bnd_used): Likewise.
15776 (INDIRECT_LABEL): Likewise.
15777 (indirect_thunk_name): Likewise.
15778 (output_indirect_thunk): Likewise.
15779 (output_indirect_thunk_function): Likewise.
15780 (ix86_output_indirect_branch_via_reg): Likewise.
15781 (ix86_output_indirect_branch_via_push): Likewise.
15782 (ix86_output_indirect_branch): Likewise.
15783 (ix86_output_indirect_jmp): Likewise.
15784 (ix86_code_end): Call output_indirect_thunk_function if needed.
15785 (ix86_output_call_insn): Call ix86_output_indirect_branch if
15787 (ix86_handle_fndecl_attribute): Handle indirect_branch.
15788 (ix86_attribute_table): Add indirect_branch.
15789 * config/i386/i386.h (machine_function): Add indirect_branch_type
15790 and has_local_indirect_jump.
15791 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
15793 (tablejump): Likewise.
15794 (*indirect_jump): Use ix86_output_indirect_jmp.
15795 (*tablejump_1): Likewise.
15796 (simple_return_indirect_internal): Likewise.
15797 * config/i386/i386.opt (mindirect-branch=): New option.
15798 (indirect_branch): New.
15801 (thunk-inline): Likewise.
15802 (thunk-extern): Likewise.
15803 * doc/extend.texi: Document indirect_branch function attribute.
15804 * doc/invoke.texi: Document -mindirect-branch= option.
15806 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
15809 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
15811 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
15813 * ipa-inline.c (want_inline_small_function_p): Return false if
15814 inlining has already failed with CIF_FINAL_ERROR.
15815 (update_caller_keys): Call want_inline_small_function_p before
15817 (update_callee_keys): Likewise.
15819 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
15821 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
15823 (rs6000_quadword_masked_address_p): Likewise.
15824 (quad_aligned_load_p): Likewise.
15825 (quad_aligned_store_p): Likewise.
15826 (const_load_sequence_p): Add comment to describe the outer-most loop.
15827 (mimic_memory_attributes_and_flags): New function.
15828 (rs6000_gen_stvx): Likewise.
15829 (replace_swapped_aligned_store): Likewise.
15830 (rs6000_gen_lvx): Likewise.
15831 (replace_swapped_aligned_load): Likewise.
15832 (replace_swapped_load_constant): Capitalize argument name in
15833 comment describing this function.
15834 (rs6000_analyze_swaps): Add a third pass to search for vector loads
15835 and stores that access quad-word aligned addresses and replace
15836 with stvx or lvx instructions when appropriate.
15837 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
15838 New function prototype.
15839 (rs6000_quadword_masked_address_p): Likewise.
15840 (rs6000_gen_lvx): Likewise.
15841 (rs6000_gen_stvx): Likewise.
15842 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
15843 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
15844 when memory address is aligned.
15845 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
15846 this split to select lvx instruction when memory address is aligned.
15847 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
15848 instruction when memory address is aligned.
15849 (*vsx_le_perm_load_v16qi): Likewise.
15850 (four unnamed splitters): Modify to select the stvx instruction
15851 when memory is aligned.
15853 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
15855 * predict.c (determine_unlikely_bbs): Handle correctly BBs
15856 which appears in the queue multiple times.
15858 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15859 Alan Hayward <alan.hayward@arm.com>
15860 David Sherwood <david.sherwood@arm.com>
15862 * tree-vectorizer.h (vec_lower_bound): New structure.
15863 (_loop_vec_info): Add check_nonzero and lower_bounds.
15864 (LOOP_VINFO_CHECK_NONZERO): New macro.
15865 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
15866 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
15867 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
15868 fields. Make seg_len the distance travelled, not including the
15870 (dr_direction_indicator): Declare.
15871 (dr_zero_step_indicator): Likewise.
15872 (dr_known_forward_stride_p): Likewise.
15873 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
15875 (runtime_alias_check_p): Allow runtime alias checks with
15877 (operator ==): Compare access_size and align.
15878 (prune_runtime_alias_test_list): Rework for new distinction between
15879 the access_size and seg_len.
15880 (create_intersect_range_checks_index): Likewise. Cope with polynomial
15882 (get_segment_min_max): New function.
15883 (create_intersect_range_checks): Use it.
15884 (dr_step_indicator): New function.
15885 (dr_direction_indicator): Likewise.
15886 (dr_zero_step_indicator): Likewise.
15887 (dr_known_forward_stride_p): Likewise.
15888 * tree-loop-distribution.c (data_ref_segment_size): Return
15889 DR_STEP * (niters - 1).
15890 (compute_alias_check_pairs): Update call to the dr_with_seg_len
15892 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
15893 (vect_preserves_scalar_order_p): New function, split out from...
15894 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
15895 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
15896 (vect_vfa_access_size): New function.
15897 (vect_vfa_align): Likewise.
15898 (vect_compile_time_alias): Take access_size_a and access_b arguments.
15899 (dump_lower_bound): New function.
15900 (vect_check_lower_bound): Likewise.
15901 (vect_small_gap_p): Likewise.
15902 (vectorizable_with_step_bound_p): Likewise.
15903 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
15904 depencies if the vectorization factor is 1. Convert the checks
15905 for nonzero steps into checks on the bounds of DR_STEP. Try using
15906 a bunds check for variable steps if the minimum required step is
15907 relatively small. Update calls to the dr_with_seg_len
15908 constructor and to vect_compile_time_alias.
15909 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
15911 (vect_loop_versioning): Call it.
15912 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
15914 (vect_estimate_min_profitable_iters): Account for any bounds checks.
15916 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15917 Alan Hayward <alan.hayward@arm.com>
15918 David Sherwood <david.sherwood@arm.com>
15920 * doc/sourcebuild.texi (vect_scatter_store): Document.
15921 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
15923 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
15925 * genopinit.c (main): Add supports_vec_scatter_store and
15926 supports_vec_scatter_store_cached to target_optabs.
15927 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
15928 IFN_MASK_SCATTER_STORE.
15929 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
15931 * internal-fn.h (internal_store_fn_p): Declare.
15932 (internal_fn_stored_value_index): Likewise.
15933 * internal-fn.c (scatter_store_direct): New macro.
15934 (expand_scatter_store_optab_fn): New function.
15935 (direct_scatter_store_optab_supported_p): New macro.
15936 (internal_store_fn_p): New function.
15937 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
15938 IFN_MASK_SCATTER_STORE.
15939 (internal_fn_mask_index): Likewise.
15940 (internal_fn_stored_value_index): New function.
15941 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
15942 for scatter stores.
15943 * optabs-query.h (supports_vec_scatter_store_p): Declare.
15944 * optabs-query.c (supports_vec_scatter_store_p): New function.
15945 * tree-vectorizer.h (vect_get_store_rhs): Declare.
15946 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
15947 true for scatter stores.
15948 (vect_gather_scatter_fn_p): Handle scatter stores too.
15949 (vect_check_gather_scatter): Consider using scatter stores if
15950 supports_vec_scatter_store_p.
15951 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
15952 scatter stores too.
15953 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
15954 internal_fn_stored_value_index.
15955 (check_load_store_masking): Handle scatter stores too.
15956 (vect_get_store_rhs): Make public.
15957 (vectorizable_call): Use internal_store_fn_p.
15958 (vectorizable_store): Handle scatter store internal functions.
15959 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
15960 when deciding whether the end of the group has been reached.
15961 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
15962 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
15963 (mask_scatter_store<mode>): New insns.
15965 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15966 Alan Hayward <alan.hayward@arm.com>
15967 David Sherwood <david.sherwood@arm.com>
15969 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
15970 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
15971 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
15973 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
15974 Use vect_truncate_gather_scatter_offset if we can't treat the
15975 operation as a normal gather load or scatter store.
15976 (get_group_load_store_type): Take the gather_scatter_info
15977 as argument. Try using a gather load or scatter store for
15978 single-element groups.
15979 (get_load_store_type): Update calls to get_group_load_store_type
15980 and vect_use_strided_gather_scatters_p.
15982 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
15983 Alan Hayward <alan.hayward@arm.com>
15984 David Sherwood <david.sherwood@arm.com>
15986 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
15987 optional tree argument.
15988 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
15990 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
15991 but continue to use the current value as a fallback.
15992 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
15993 to compare the updates.
15994 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
15995 (get_load_store_type): Use it when handling a strided access.
15996 (vect_get_strided_load_store_ops): New function.
15997 (vect_get_data_ptr_increment): Likewise.
15998 (vectorizable_load): Handle strided gather loads. Always pass
15999 a step to vect_create_data_ref_ptr and bump_vector_ptr.
16001 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16002 Alan Hayward <alan.hayward@arm.com>
16003 David Sherwood <david.sherwood@arm.com>
16005 * doc/md.texi (gather_load@var{m}): Document.
16006 (mask_gather_load@var{m}): Likewise.
16007 * genopinit.c (main): Add supports_vec_gather_load and
16008 supports_vec_gather_load_cached to target_optabs.
16009 * optabs-tree.c (init_tree_optimization_optabs): Use
16010 ggc_cleared_alloc to allocate target_optabs.
16011 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
16012 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
16014 * internal-fn.h (internal_load_fn_p): Declare.
16015 (internal_gather_scatter_fn_p): Likewise.
16016 (internal_fn_mask_index): Likewise.
16017 (internal_gather_scatter_fn_supported_p): Likewise.
16018 * internal-fn.c (gather_load_direct): New macro.
16019 (expand_gather_load_optab_fn): New function.
16020 (direct_gather_load_optab_supported_p): New macro.
16021 (direct_internal_fn_optab): New function.
16022 (internal_load_fn_p): Likewise.
16023 (internal_gather_scatter_fn_p): Likewise.
16024 (internal_fn_mask_index): Likewise.
16025 (internal_gather_scatter_fn_supported_p): Likewise.
16026 * optabs-query.c (supports_at_least_one_mode_p): New function.
16027 (supports_vec_gather_load_p): Likewise.
16028 * optabs-query.h (supports_vec_gather_load_p): Declare.
16029 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
16030 and memory_type field.
16031 (NUM_PATTERNS): Bump to 15.
16032 * tree-vect-data-refs.c: Include internal-fn.h.
16033 (vect_gather_scatter_fn_p): New function.
16034 (vect_describe_gather_scatter_call): Likewise.
16035 (vect_check_gather_scatter): Try using internal functions for
16036 gather loads. Recognize existing calls to a gather load function.
16037 (vect_analyze_data_refs): Consider using gather loads if
16038 supports_vec_gather_load_p.
16039 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
16040 (vect_get_gather_scatter_offset_type): Likewise.
16041 (vect_convert_mask_for_vectype): Likewise.
16042 (vect_add_conversion_to_patterm): Likewise.
16043 (vect_try_gather_scatter_pattern): Likewise.
16044 (vect_recog_gather_scatter_pattern): New pattern recognizer.
16045 (vect_vect_recog_func_ptrs): Add it.
16046 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
16047 internal_fn_mask_index and internal_gather_scatter_fn_p.
16048 (check_load_store_masking): Take the gather_scatter_info as an
16049 argument and handle gather loads.
16050 (vect_get_gather_scatter_ops): New function.
16051 (vectorizable_call): Check internal_load_fn_p.
16052 (vectorizable_load): Likewise. Handle gather load internal
16054 (vectorizable_store): Update call to check_load_store_masking.
16055 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
16056 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
16057 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
16058 (aarch64_gather_scale_operand_d): New predicates.
16059 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
16060 (mask_gather_load<mode>): New insns.
16062 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16063 Alan Hayward <alan.hayward@arm.com>
16064 David Sherwood <david.sherwood@arm.com>
16066 * optabs.def (fold_left_plus_optab): New optab.
16067 * doc/md.texi (fold_left_plus_@var{m}): Document.
16068 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
16069 * internal-fn.c (fold_left_direct): Define.
16070 (expand_fold_left_optab_fn): Likewise.
16071 (direct_fold_left_optab_supported_p): Likewise.
16072 * fold-const-call.c (fold_const_fold_left): New function.
16073 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
16074 * tree-parloops.c (valid_reduction_p): New function.
16075 (gather_scalar_reductions): Use it.
16076 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
16077 (vect_finish_replace_stmt): Declare.
16078 * tree-vect-loop.c (fold_left_reduction_fn): New function.
16079 (needs_fold_left_reduction_p): New function, split out from...
16080 (vect_is_simple_reduction): ...here. Accept reductions that
16081 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
16082 (vect_force_simple_reduction): Also store the reduction type in
16083 the assignment's STMT_VINFO_REDUC_TYPE.
16084 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
16085 (merge_with_identity): New function.
16086 (vect_expand_fold_left): Likewise.
16087 (vectorize_fold_left_reduction): Likewise.
16088 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
16089 scalar phi in place for it. Check for target support and reject
16090 cases that would reassociate the operation. Defer the transform
16091 phase to vectorize_fold_left_reduction.
16092 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
16093 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
16094 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
16096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16098 * tree-if-conv.c (predicate_mem_writes): Remove redundant
16099 call to ifc_temp_var.
16101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16102 Alan Hayward <alan.hayward@arm.com>
16103 David Sherwood <david.sherwood@arm.com>
16105 * target.def (legitimize_address_displacement): Take the original
16106 offset as a poly_int.
16107 * targhooks.h (default_legitimize_address_displacement): Update
16109 * targhooks.c (default_legitimize_address_displacement): Likewise.
16110 * doc/tm.texi: Regenerate.
16111 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
16112 as an argument, moving assert of ad->disp == ad->disp_term to...
16113 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
16114 Try calling targetm.legitimize_address_displacement before expanding
16115 the address rather than afterwards, and adjust for the new interface.
16116 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
16117 Match the new hook interface. Handle SVE addresses.
16118 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
16119 new hook interface.
16121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16123 * Makefile.in (OBJS): Add early-remat.o.
16124 * target.def (select_early_remat_modes): New hook.
16125 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
16126 * doc/tm.texi: Regenerate.
16127 * targhooks.h (default_select_early_remat_modes): Declare.
16128 * targhooks.c (default_select_early_remat_modes): New function.
16129 * timevar.def (TV_EARLY_REMAT): New timevar.
16130 * passes.def (pass_early_remat): New pass.
16131 * tree-pass.h (make_pass_early_remat): Declare.
16132 * early-remat.c: New file.
16133 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
16135 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
16137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16138 Alan Hayward <alan.hayward@arm.com>
16139 David Sherwood <david.sherwood@arm.com>
16141 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
16142 vfm1 with a bound_epilog parameter.
16143 (vect_do_peeling): Update calls accordingly, and move the prologue
16144 call earlier in the function. Treat the base bound_epilog as 0 for
16145 fully-masked loops and retain vf - 1 for other loops. Add 1 to
16146 this base when peeling for gaps.
16147 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
16148 with fully-masked loops.
16149 (vect_estimate_min_profitable_iters): Handle the single peeled
16150 iteration in that case.
16152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16153 Alan Hayward <alan.hayward@arm.com>
16154 David Sherwood <david.sherwood@arm.com>
16156 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
16157 single-element interleaving even if the size is not a power of 2.
16158 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
16159 accesses for single-element interleaving if the group size is
16162 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16163 Alan Hayward <alan.hayward@arm.com>
16164 David Sherwood <david.sherwood@arm.com>
16166 * doc/md.texi (fold_extract_last_@var{m}): Document.
16167 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
16168 * optabs.def (fold_extract_last_optab): New optab.
16169 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
16170 * internal-fn.c (fold_extract_direct): New macro.
16171 (expand_fold_extract_optab_fn): Likewise.
16172 (direct_fold_extract_optab_supported_p): Likewise.
16173 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
16174 * tree-vect-loop.c (vect_model_reduction_cost): Handle
16175 EXTRACT_LAST_REDUCTION.
16176 (get_initial_def_for_reduction): Do not create an initial vector
16177 for EXTRACT_LAST_REDUCTION reductions.
16178 (vectorizable_reduction): Leave the scalar phi in place for
16179 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
16180 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
16181 epilogue code for EXTRACT_LAST_REDUCTION and defer the
16182 transform phase to vectorizable_condition.
16183 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
16185 (vect_finish_stmt_generation): ...here.
16186 (vect_finish_replace_stmt): New function.
16187 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
16188 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
16190 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
16192 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16193 Alan Hayward <alan.hayward@arm.com>
16194 David Sherwood <david.sherwood@arm.com>
16196 * doc/md.texi (extract_last_@var{m}): Document.
16197 * optabs.def (extract_last_optab): New optab.
16198 * internal-fn.def (EXTRACT_LAST): New internal function.
16199 * internal-fn.c (cond_unary_direct): New macro.
16200 (expand_cond_unary_optab_fn): Likewise.
16201 (direct_cond_unary_optab_supported_p): Likewise.
16202 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
16203 loops using EXTRACT_LAST.
16204 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
16205 (extract_last_<mode>): ...this optab.
16206 (vec_extract<mode><Vel>): Update accordingly.
16208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16209 Alan Hayward <alan.hayward@arm.com>
16210 David Sherwood <david.sherwood@arm.com>
16212 * target.def (empty_mask_is_expensive): New hook.
16213 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
16214 * doc/tm.texi: Regenerate.
16215 * targhooks.h (default_empty_mask_is_expensive): Declare.
16216 * targhooks.c (default_empty_mask_is_expensive): New function.
16217 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
16218 if the target says that empty masks are expensive.
16219 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
16221 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
16223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16224 Alan Hayward <alan.hayward@arm.com>
16225 David Sherwood <david.sherwood@arm.com>
16227 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
16228 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
16229 (vect_use_loop_mask_for_alignment_p): New function.
16230 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
16231 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
16232 niters_skip argument. Make sure that the first niters_skip elements
16233 of the first iteration are inactive.
16234 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
16235 Update call to vect_set_loop_masks_directly.
16236 (get_misalign_in_elems): New function, split out from...
16237 (vect_gen_prolog_loop_niters): ...here.
16238 (vect_update_init_of_dr): Take a code argument that specifies whether
16239 the adjustment should be added or subtracted.
16240 (vect_update_init_of_drs): Likewise.
16241 (vect_prepare_for_masked_peels): New function.
16242 (vect_do_peeling): Skip prologue peeling if we're using a mask
16243 instead. Update call to vect_update_inits_of_drs.
16244 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16246 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
16247 alignment. Do not include the number of peeled iterations in
16248 the minimum threshold in that case.
16249 (vectorizable_induction): Adjust the start value down by
16250 LOOP_VINFO_MASK_SKIP_NITERS iterations.
16251 (vect_transform_loop): Call vect_prepare_for_masked_peels.
16252 Take the number of skipped iterations into account when calculating
16254 * tree-vect-stmts.c (vect_gen_while_not): New function.
16256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16257 Alan Hayward <alan.hayward@arm.com>
16258 David Sherwood <david.sherwood@arm.com>
16260 * doc/sourcebuild.texi (vect_fully_masked): Document.
16261 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
16262 default value to 0.
16263 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
16265 (vect_analyze_loop_2): ...here. Don't check the vectorization
16266 factor against the number of loop iterations if the loop is
16269 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16270 Alan Hayward <alan.hayward@arm.com>
16271 David Sherwood <david.sherwood@arm.com>
16273 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
16274 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
16275 (dump_groups): Update accordingly.
16276 (iv_use::mem_type): New member variable.
16277 (address_p): New function.
16278 (record_use): Add a mem_type argument and initialize the new
16280 (record_group_use): Add a mem_type argument. Use address_p.
16281 Remove obsolete null checks of base_object. Update call to record_use.
16282 (find_interesting_uses_op): Update call to record_group_use.
16283 (find_interesting_uses_cond): Likewise.
16284 (find_interesting_uses_address): Likewise.
16285 (get_mem_type_for_internal_fn): New function.
16286 (find_address_like_use): Likewise.
16287 (find_interesting_uses_stmt): Try find_address_like_use before
16288 calling find_interesting_uses_op.
16289 (addr_offset_valid_p): Use the iv mem_type field as the type
16290 of the addressed memory.
16291 (add_autoinc_candidates): Likewise.
16292 (get_address_cost): Likewise.
16293 (split_small_address_groups_p): Use address_p.
16294 (split_address_groups): Likewise.
16295 (add_iv_candidate_for_use): Likewise.
16296 (autoinc_possible_for_pair): Likewise.
16297 (rewrite_groups): Likewise.
16298 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
16299 (determine_group_iv_cost): Update after split of USE_ADDRESS.
16300 (get_alias_ptr_type_for_ptr_address): New function.
16301 (rewrite_use_address): Rewrite address uses in calls that were
16302 identified by find_address_like_use.
16304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16305 Alan Hayward <alan.hayward@arm.com>
16306 David Sherwood <david.sherwood@arm.com>
16308 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
16310 * gimple-expr.h (is_gimple_addressable: Likewise.
16311 * gimple-expr.c (is_gimple_address): Likewise.
16312 * internal-fn.c (expand_call_mem_ref): New function.
16313 (expand_mask_load_optab_fn): Use it.
16314 (expand_mask_store_optab_fn): Likewise.
16316 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16317 Alan Hayward <alan.hayward@arm.com>
16318 David Sherwood <david.sherwood@arm.com>
16320 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
16321 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
16322 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
16323 (cond_umax@var{mode}): Document.
16324 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
16325 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
16326 (cond_umin_optab, cond_umax_optab): New optabs.
16327 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
16328 (COND_IOR, COND_XOR): New internal functions.
16329 * internal-fn.h (get_conditional_internal_fn): Declare.
16330 * internal-fn.c (cond_binary_direct): New macro.
16331 (expand_cond_binary_optab_fn): Likewise.
16332 (direct_cond_binary_optab_supported_p): Likewise.
16333 (get_conditional_internal_fn): New function.
16334 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
16335 Cope with reduction statements that are vectorized as calls rather
16337 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
16338 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
16339 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
16340 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
16341 (UNSPEC_COND_EOR): New unspecs.
16342 (optab): Add mappings for them.
16343 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
16344 (sve_int_op, sve_fp_op): New int attributes.
16346 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16347 Alan Hayward <alan.hayward@arm.com>
16348 David Sherwood <david.sherwood@arm.com>
16350 * optabs.def (while_ult_optab): New optab.
16351 * doc/md.texi (while_ult@var{m}@var{n}): Document.
16352 * internal-fn.def (WHILE_ULT): New internal function.
16353 * internal-fn.h (direct_internal_fn_supported_p): New override
16354 that takes two types as argument.
16355 * internal-fn.c (while_direct): New macro.
16356 (expand_while_optab_fn): New function.
16357 (convert_optab_supported_p): Likewise.
16358 (direct_while_optab_supported_p): New macro.
16359 * wide-int.h (wi::udiv_ceil): New function.
16360 * tree-vectorizer.h (rgroup_masks): New structure.
16361 (vec_loop_masks): New typedef.
16362 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
16363 and fully_masked_p.
16364 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
16365 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
16366 (vect_max_vf): New function.
16367 (slpeel_make_loop_iterate_ntimes): Delete.
16368 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
16369 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
16370 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
16371 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
16372 internal-fn.h, stor-layout.h and optabs-query.h.
16373 (vect_set_loop_mask): New function.
16374 (add_preheader_seq): Likewise.
16375 (add_header_seq): Likewise.
16376 (interleave_supported_p): Likewise.
16377 (vect_maybe_permute_loop_masks): Likewise.
16378 (vect_set_loop_masks_directly): Likewise.
16379 (vect_set_loop_condition_masked): Likewise.
16380 (vect_set_loop_condition_unmasked): New function, split out from
16381 slpeel_make_loop_iterate_ntimes.
16382 (slpeel_make_loop_iterate_ntimes): Rename to..
16383 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
16384 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
16385 (vect_do_peeling): Update call accordingly.
16386 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
16388 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
16389 mask_compare_type, can_fully_mask_p and fully_masked_p.
16390 (release_vec_loop_masks): New function.
16391 (_loop_vec_info): Use it to free the loop masks.
16392 (can_produce_all_loop_masks_p): New function.
16393 (vect_get_max_nscalars_per_iter): Likewise.
16394 (vect_verify_full_masking): Likewise.
16395 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
16396 retries, and free the mask rgroups before retrying. Check loop-wide
16397 reasons for disallowing fully-masked loops. Make the final decision
16398 about whether use a fully-masked loop or not.
16399 (vect_estimate_min_profitable_iters): Do not assume that peeling
16400 for the number of iterations will be needed for fully-masked loops.
16401 (vectorizable_reduction): Disable fully-masked loops.
16402 (vectorizable_live_operation): Likewise.
16403 (vect_halve_mask_nunits): New function.
16404 (vect_double_mask_nunits): Likewise.
16405 (vect_record_loop_mask): Likewise.
16406 (vect_get_loop_mask): Likewise.
16407 (vect_transform_loop): Handle the case in which the final loop
16408 iteration might handle a partial vector. Call vect_set_loop_condition
16409 instead of slpeel_make_loop_iterate_ntimes.
16410 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
16411 (check_load_store_masking): New function.
16412 (prepare_load_store_mask): Likewise.
16413 (vectorizable_store): Handle fully-masked loops.
16414 (vectorizable_load): Likewise.
16415 (supportable_widening_operation): Use vect_halve_mask_nunits for
16417 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
16418 (vect_gen_while): New function.
16419 * config/aarch64/aarch64.md (umax<mode>3): New expander.
16420 (aarch64_uqdec<mode>): New insn.
16422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16423 Alan Hayward <alan.hayward@arm.com>
16424 David Sherwood <david.sherwood@arm.com>
16426 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
16427 (reduc_xor_scal_optab): New optabs.
16428 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
16429 (reduc_xor_scal_@var{m}): Document.
16430 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
16431 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
16432 internal functions.
16433 * fold-const-call.c (fold_const_call): Handle them.
16434 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
16435 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
16436 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
16437 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
16438 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
16439 (UNSPEC_XORV): New unspecs.
16440 (optab): Add entries for them.
16441 (BITWISEV): New int iterator.
16442 (bit_reduc_op): New int attributes.
16444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16445 Alan Hayward <alan.hayward@arm.com>
16446 David Sherwood <david.sherwood@arm.com>
16448 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
16449 * internal-fn.def (VEC_SHL_INSERT): New internal function.
16450 * optabs.def (vec_shl_insert_optab): New optab.
16451 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
16452 (duplicate_and_interleave): Likewise.
16453 * tree-vect-loop.c: Include internal-fn.h.
16454 (neutral_op_for_slp_reduction): New function, split out from
16455 get_initial_defs_for_reduction.
16456 (get_initial_def_for_reduction): Handle option 2 for variable-length
16457 vectors by loading the neutral value into a vector and then shifting
16458 the initial value into element 0.
16459 (get_initial_defs_for_reduction): Replace the code argument with
16460 the neutral value calculated by neutral_op_for_slp_reduction.
16461 Use gimple_build_vector for constant-length vectors.
16462 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
16463 but the first group_size elements have a neutral value.
16464 Use duplicate_and_interleave otherwise.
16465 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
16466 Update call to get_initial_defs_for_reduction. Handle SLP
16467 reductions for variable-length vectors by creating one vector
16468 result for each scalar result, with the elements associated
16469 with other scalar results stubbed out with the neutral value.
16470 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
16471 Require IFN_VEC_SHL_INSERT for double reductions on
16472 variable-length vectors, or SLP reductions that have
16473 a neutral value. Require can_duplicate_and_interleave_p
16474 support for variable-length unchained SLP reductions if there
16475 is no neutral value, such as for MIN/MAX reductions. Also require
16476 the number of vector elements to be a multiple of the number of
16477 SLP statements when doing variable-length unchained SLP reductions.
16478 Update call to vect_create_epilog_for_reduction.
16479 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
16480 and remove initial values.
16481 (duplicate_and_interleave): Make public.
16482 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
16483 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
16485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16486 Alan Hayward <alan.hayward@arm.com>
16487 David Sherwood <david.sherwood@arm.com>
16489 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
16490 (can_duplicate_and_interleave_p): New function.
16491 (vect_get_and_check_slp_defs): Take the vector of statements
16492 rather than just the current one. Remove excess parentheses.
16493 Restriction rejectinon of vect_constant_def and vect_external_def
16494 for variable-length vectors to boolean types, or types for which
16495 can_duplicate_and_interleave_p is false.
16496 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
16497 (duplicate_and_interleave): New function.
16498 (vect_get_constant_vectors): Use gimple_build_vector for
16499 constant-length vectors and suitable variable-length constant
16500 vectors. Use duplicate_and_interleave for other variable-length
16501 vectors. Don't defer the update when inserting new statements.
16503 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16504 Alan Hayward <alan.hayward@arm.com>
16505 David Sherwood <david.sherwood@arm.com>
16507 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
16508 min_profitable_iters doesn't go negative.
16510 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16511 Alan Hayward <alan.hayward@arm.com>
16512 David Sherwood <david.sherwood@arm.com>
16514 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
16515 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
16516 * optabs.def (vec_mask_load_lanes_optab): New optab.
16517 (vec_mask_store_lanes_optab): Likewise.
16518 * internal-fn.def (MASK_LOAD_LANES): New internal function.
16519 (MASK_STORE_LANES): Likewise.
16520 * internal-fn.c (mask_load_lanes_direct): New macro.
16521 (mask_store_lanes_direct): Likewise.
16522 (expand_mask_load_optab_fn): Handle masked operations.
16523 (expand_mask_load_lanes_optab_fn): New macro.
16524 (expand_mask_store_optab_fn): Handle masked operations.
16525 (expand_mask_store_lanes_optab_fn): New macro.
16526 (direct_mask_load_lanes_optab_supported_p): Likewise.
16527 (direct_mask_store_lanes_optab_supported_p): Likewise.
16528 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
16530 (vect_load_lanes_supported): Likewise.
16531 * tree-vect-data-refs.c (strip_conversion): New function.
16532 (can_group_stmts_p): Likewise.
16533 (vect_analyze_data_ref_accesses): Use it instead of checking
16534 for a pair of assignments.
16535 (vect_store_lanes_supported): Take a masked_p parameter.
16536 (vect_load_lanes_supported): Likewise.
16537 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
16538 vect_store_lanes_supported and vect_load_lanes_supported.
16539 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
16540 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
16541 parameter. Don't allow gaps for masked accesses.
16542 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
16543 and vect_load_lanes_supported.
16544 (get_load_store_type): Take a masked_p parameter and update
16545 call to get_group_load_store_type.
16546 (vectorizable_store): Update call to get_load_store_type.
16547 Handle IFN_MASK_STORE_LANES.
16548 (vectorizable_load): Update call to get_load_store_type.
16549 Handle IFN_MASK_LOAD_LANES.
16551 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16552 Alan Hayward <alan.hayward@arm.com>
16553 David Sherwood <david.sherwood@arm.com>
16555 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
16557 * config/aarch64/aarch64-protos.h
16558 (aarch64_sve_struct_memory_operand_p): Declare.
16559 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
16560 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
16561 (VPRED, vpred): Handle SVE structure modes.
16562 * config/aarch64/constraints.md (Utx): New constraint.
16563 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
16564 (aarch64_sve_struct_nonimmediate_operand): New predicates.
16565 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
16566 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
16567 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
16568 structure modes. Split into pieces after RA.
16569 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
16570 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
16572 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
16573 SVE structure modes.
16574 (aarch64_classify_address): Likewise.
16575 (sizetochar): Move earlier in file.
16576 (aarch64_print_operand): Handle SVE register lists.
16577 (aarch64_array_mode): New function.
16578 (aarch64_sve_struct_memory_operand_p): Likewise.
16579 (TARGET_ARRAY_MODE): Redefine.
16581 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16582 Alan Hayward <alan.hayward@arm.com>
16583 David Sherwood <david.sherwood@arm.com>
16585 * target.def (array_mode): New target hook.
16586 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
16587 * doc/tm.texi: Regenerate.
16588 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
16589 * hooks.c (hook_optmode_mode_uhwi_none): New function.
16590 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
16591 targetm.array_mode.
16592 * stor-layout.c (mode_for_array): Likewise. Support polynomial
16595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16596 Alan Hayward <alan.hayward@arm.com>
16597 David Sherwood <david.sherwood@arm.com>
16599 * fold-const.c (fold_binary_loc): Check the argument types
16600 rather than the result type when testing for a vector operation.
16602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16604 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
16605 * doc/tm.texi: Regenerate.
16607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16608 Alan Hayward <alan.hayward@arm.com>
16609 David Sherwood <david.sherwood@arm.com>
16611 * doc/invoke.texi (-msve-vector-bits=): Document new option.
16612 (sve): Document new AArch64 extension.
16613 * doc/md.texi (w): Extend the description of the AArch64
16614 constraint to include SVE vectors.
16615 (Upl, Upa): Document new AArch64 predicate constraints.
16616 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
16618 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
16619 (msve-vector-bits=): New option.
16620 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
16621 SVE when these are disabled.
16622 (sve): New extension.
16623 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
16624 modes. Adjust their number of units based on aarch64_sve_vg.
16625 (MAX_BITSIZE_MODE_ANY_MODE): Define.
16626 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
16627 aarch64_addr_query_type.
16628 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
16629 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
16630 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
16631 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
16632 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
16633 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
16634 (aarch64_simd_imm_zero_p): Delete.
16635 (aarch64_check_zero_based_sve_index_immediate): Declare.
16636 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16637 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16638 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16639 (aarch64_sve_float_mul_immediate_p): Likewise.
16640 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16641 rather than an rtx.
16642 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
16643 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
16644 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
16645 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
16646 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
16647 (aarch64_regmode_natural_size): Likewise.
16648 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
16649 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
16651 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
16652 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
16653 for VG and the SVE predicate registers.
16654 (V_ALIASES): Add a "z"-prefixed alias.
16655 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
16656 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
16657 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
16658 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
16659 (REG_CLASS_NAMES): Add entries for them.
16660 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
16661 and the predicate registers.
16662 (aarch64_sve_vg): Declare.
16663 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
16664 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
16665 (REGMODE_NATURAL_SIZE): Define.
16666 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
16668 * config/aarch64/aarch64.c: Include cfgrtl.h.
16669 (simd_immediate_info): Add a constructor for series vectors,
16670 and an associated step field.
16671 (aarch64_sve_vg): New variable.
16672 (aarch64_dbx_register_number): Handle VG and the predicate registers.
16673 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
16674 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
16675 (VEC_ANY_DATA, VEC_STRUCT): New constants.
16676 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
16677 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
16678 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
16679 (aarch64_get_mask_mode): New functions.
16680 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
16681 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
16682 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
16683 predicate modes and predicate registers. Explicitly restrict
16684 GPRs to modes of 16 bytes or smaller. Only allow FP registers
16685 to store a vector mode if it is recognized by
16686 aarch64_classify_vector_mode.
16687 (aarch64_regmode_natural_size): New function.
16688 (aarch64_hard_regno_caller_save_mode): Return the original mode
16690 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
16691 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
16692 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
16693 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
16695 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
16696 does not overlap dest if the function is frame-related. Handle
16698 (aarch64_split_add_offset): New function.
16699 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
16700 them aarch64_add_offset.
16701 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
16702 and update call to aarch64_sub_sp.
16703 (aarch64_add_cfa_expression): New function.
16704 (aarch64_expand_prologue): Pass extra temporary registers to the
16705 functions above. Handle the case in which we need to emit new
16706 DW_CFA_expressions for registers that were originally saved
16707 relative to the stack pointer, but now have to be expressed
16708 relative to the frame pointer.
16709 (aarch64_output_mi_thunk): Pass extra temporary registers to the
16711 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
16712 IP0 and IP1 values for SVE frames.
16713 (aarch64_expand_vec_series): New function.
16714 (aarch64_expand_sve_widened_duplicate): Likewise.
16715 (aarch64_expand_sve_const_vector): Likewise.
16716 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
16717 Handle SVE constants. Use emit_move_insn to move a force_const_mem
16718 into the register, rather than emitting a SET directly.
16719 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
16720 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
16721 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
16722 (offset_9bit_signed_scaled_p): New functions.
16723 (aarch64_replicate_bitmask_imm): New function.
16724 (aarch64_bitmask_imm): Use it.
16725 (aarch64_cannot_force_const_mem): Reject expressions involving
16726 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
16727 (aarch64_classify_index): Handle SVE indices, by requiring
16728 a plain register index with a scale that matches the element size.
16729 (aarch64_classify_address): Handle SVE addresses. Assert that
16730 the mode of the address is VOIDmode or an integer mode.
16731 Update call to aarch64_classify_symbol.
16732 (aarch64_classify_symbolic_expression): Update call to
16733 aarch64_classify_symbol.
16734 (aarch64_const_vec_all_in_range_p): New function.
16735 (aarch64_print_vector_float_operand): Likewise.
16736 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
16737 "vN" for FP registers with SVE modes. Handle (const ...) vectors
16738 and the FP immediates 1.0 and 0.5.
16739 (aarch64_print_address_internal): Handle SVE addresses.
16740 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
16741 (aarch64_regno_regclass): Handle predicate registers.
16742 (aarch64_secondary_reload): Handle big-endian reloads of SVE
16744 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
16745 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
16746 (aarch64_convert_sve_vector_bits): New function.
16747 (aarch64_override_options): Use it to handle -msve-vector-bits=.
16748 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
16749 rather than an rtx.
16750 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
16751 Handle SVE vector and predicate modes. Accept VL-based constants
16752 that need only one temporary register, and VL offsets that require
16753 no temporary registers.
16754 (aarch64_conditional_register_usage): Mark the predicate registers
16755 as fixed if SVE isn't available.
16756 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
16757 Return true for SVE vector and predicate modes.
16758 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
16759 rather than an unsigned int. Handle SVE modes.
16760 (aarch64_preferred_simd_mode): Update call accordingly. Handle
16762 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
16764 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
16765 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
16766 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
16767 (aarch64_sve_float_mul_immediate_p): New functions.
16768 (aarch64_sve_valid_immediate): New function.
16769 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
16770 Explicitly reject structure modes. Check for INDEX constants.
16771 Handle PTRUE and PFALSE constants.
16772 (aarch64_check_zero_based_sve_index_immediate): New function.
16773 (aarch64_simd_imm_zero_p): Delete.
16774 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
16775 vector modes. Accept constants in the range of CNT[BHWD].
16776 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
16777 ask for an Advanced SIMD mode.
16778 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
16779 (aarch64_simd_vector_alignment): Handle SVE predicates.
16780 (aarch64_vectorize_preferred_vector_alignment): New function.
16781 (aarch64_simd_vector_alignment_reachable): Use it instead of
16783 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
16784 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
16786 (MAX_VECT_LEN): Delete.
16787 (expand_vec_perm_d): Add a vec_flags field.
16788 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
16789 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
16790 (aarch64_evpc_ext): Don't apply a big-endian lane correction
16792 (aarch64_evpc_rev): Rename to...
16793 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
16794 (aarch64_evpc_rev_global): New function.
16795 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
16796 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
16798 (aarch64_evpc_sve_tbl): New function.
16799 (aarch64_expand_vec_perm_const_1): Update after rename of
16800 aarch64_evpc_rev. Handle SVE permutes too, trying
16801 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
16802 than aarch64_evpc_tbl.
16803 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
16804 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
16805 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
16806 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
16807 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
16808 (aarch64_expand_sve_vcond): New functions.
16809 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
16810 of aarch64_vector_mode_p.
16811 (aarch64_dwarf_poly_indeterminate_value): New function.
16812 (aarch64_compute_pressure_classes): Likewise.
16813 (aarch64_can_change_mode_class): Likewise.
16814 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
16815 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
16816 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
16817 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
16818 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
16819 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
16820 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
16821 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
16823 (Dn, Dl, Dr): Accept const as well as const_vector.
16824 (Dz): Likewise. Compare against CONST0_RTX.
16825 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
16826 of "vector" where appropriate.
16827 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
16828 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
16829 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
16830 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
16831 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
16832 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
16833 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
16834 (v_int_equiv): Extend to SVE modes.
16835 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
16837 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
16838 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
16839 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
16840 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
16841 (SVE_COND_FP_CMP): New int iterators.
16842 (perm_hilo): Handle the new unpack unspecs.
16843 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
16845 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
16846 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
16847 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
16848 (aarch64_equality_operator, aarch64_constant_vector_operand)
16849 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
16850 (aarch64_sve_nonimmediate_operand): Likewise.
16851 (aarch64_sve_general_operand): Likewise.
16852 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
16853 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
16854 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
16855 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
16856 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
16857 (aarch64_sve_float_arith_immediate): Likewise.
16858 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
16859 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
16860 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
16861 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
16862 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
16863 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
16864 (aarch64_sve_float_arith_operand): Likewise.
16865 (aarch64_sve_float_arith_with_sub_operand): Likewise.
16866 (aarch64_sve_float_mul_operand): Likewise.
16867 (aarch64_sve_vec_perm_operand): Likewise.
16868 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
16869 (aarch64_mov_operand): Accept const_poly_int and const_vector.
16870 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
16871 as well as const_vector.
16872 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
16873 in file. Use CONST0_RTX and CONSTM1_RTX.
16874 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
16875 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
16876 Use aarch64_simd_imm_zero.
16877 * config/aarch64/aarch64-sve.md: New file.
16878 * config/aarch64/aarch64.md: Include it.
16879 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
16880 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
16881 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
16882 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
16883 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
16884 (sve): New attribute.
16885 (enabled): Disable instructions with the sve attribute unless
16887 (movqi, movhi): Pass CONST_POLY_INT operaneds through
16888 aarch64_expand_mov_immediate.
16889 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
16890 CNT[BHSD] immediates.
16891 (movti): Split CONST_POLY_INT moves into two halves.
16892 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
16893 Split additions that need a temporary here if the destination
16894 is the stack pointer.
16895 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
16896 (*add<mode>3_poly_1): New instruction.
16897 (set_clobber_cc): New expander.
16899 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16901 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
16902 parameter and use it instead of GET_MODE_SIZE (innermode). Use
16903 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
16904 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
16905 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
16906 Change innermode from fixed_mode_size to machine_mode.
16907 (simplify_subreg): Update call accordingly. Handle a constant-sized
16908 subreg of a variable-length CONST_VECTOR.
16910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
16911 Alan Hayward <alan.hayward@arm.com>
16912 David Sherwood <david.sherwood@arm.com>
16914 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
16915 (add_offset_to_base): New function, split out from...
16916 (create_mem_ref): ...here. When handling a scale other than 1,
16917 check first whether the address is valid without the offset.
16918 Add it into the base if so, leaving the index and scale as-is.
16920 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16923 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
16924 fold_for_warn before checking if arg2 is INTEGER_CST.
16926 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
16928 * config/rs6000/predicates.md (load_multiple_operation): Delete.
16929 (store_multiple_operation): Delete.
16930 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
16931 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
16932 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
16933 guarded by TARGET_STRING.
16934 (rs6000_output_load_multiple): Delete.
16935 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
16936 OPTION_MASK_STRING / TARGET_STRING handling.
16937 (print_operand) <'N', 'O'>: Add comment that these are unused now.
16938 (const rs6000_opt_masks) <"string">: Change mask to 0.
16939 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
16940 (MASK_STRING): Delete.
16941 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
16943 (load_multiple): Delete.
16950 (store_multiple): Delete.
16957 (movmemsi_8reg): Delete.
16958 (corresponding unnamed define_insn): Delete.
16959 (movmemsi_6reg): Delete.
16960 (corresponding unnamed define_insn): Delete.
16961 (movmemsi_4reg): Delete.
16962 (corresponding unnamed define_insn): Delete.
16963 (movmemsi_2reg): Delete.
16964 (corresponding unnamed define_insn): Delete.
16965 (movmemsi_1reg): Delete.
16966 (corresponding unnamed define_insn): Delete.
16967 * config/rs6000/rs6000.opt (mno-string): New.
16968 (mstring): Replace by deprecation warning stub.
16969 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
16971 2018-01-12 Jakub Jelinek <jakub@redhat.com>
16973 * regrename.c (regrename_do_replace): If replacing the same
16974 reg multiple times, try to reuse last created gen_raw_REG.
16977 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
16978 main to workaround a bug in GDB.
16980 2018-01-12 Tom de Vries <tom@codesourcery.com>
16983 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
16985 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
16987 PR rtl-optimization/80481
16988 * ira-color.c (get_cap_member): New function.
16989 (allocnos_conflict_by_live_ranges_p): Use it.
16990 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
16991 (setup_slot_coalesced_allocno_live_ranges): Ditto.
16993 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
16996 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
16997 (*saddl_se_1): Ditto.
16998 (*ssubsi_1): Ditto.
16999 (*ssubl_se_1): Ditto.
17001 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17003 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
17004 rather than wi::to_widest for DR_INITs.
17005 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
17006 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
17007 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
17009 (vect_analyze_group_access_1): Note that here.
17011 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17013 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
17014 polynomial type sizes.
17016 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
17018 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
17019 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
17020 (gimple_add_tmp_var): Likewise.
17022 2018-01-12 Martin Liska <mliska@suse.cz>
17024 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
17025 (gimple_alloc_sizes): Likewise.
17026 (dump_gimple_statistics): Use PRIu64 in printf format.
17027 * gimple.h: Change uint64_t to int.
17029 2018-01-12 Martin Liska <mliska@suse.cz>
17031 * tree-core.h: Use uint64_t instead of int.
17032 * tree.c (tree_node_counts): Likewise.
17033 (tree_node_sizes): Likewise.
17034 (dump_tree_statistics): Use PRIu64 in printf format.
17036 2018-01-12 Martin Liska <mliska@suse.cz>
17038 * Makefile.in: As qsort_chk is implemented in vec.c, add
17039 vec.o to linkage of gencfn-macros.
17040 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
17041 passing the info to record_node_allocation_statistics.
17042 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
17044 * ggc-common.c (struct ggc_usage): Add operator== and use
17045 it in operator< and compare function.
17046 * mem-stats.h (struct mem_usage): Likewise.
17047 * vec.c (struct vec_usage): Remove operator< and compare
17048 function. Can be simply inherited.
17050 2018-01-12 Martin Jambor <mjambor@suse.cz>
17053 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
17054 * tree-ssa-math-opts.c: Include domwalk.h.
17055 (convert_mult_to_fma_1): New function.
17056 (fma_transformation_info): New type.
17057 (fma_deferring_state): Likewise.
17058 (cancel_fma_deferring): New function.
17059 (result_of_phi): Likewise.
17060 (last_fma_candidate_feeds_initial_phi): Likewise.
17061 (convert_mult_to_fma): Added deferring logic, split actual
17062 transformation to convert_mult_to_fma_1.
17063 (math_opts_dom_walker): New type.
17064 (math_opts_dom_walker::after_dom_children): New method, body moved
17065 here from pass_optimize_widening_mul::execute, added deferring logic
17067 (pass_optimize_widening_mul::execute): Moved most of code to
17068 math_opts_dom_walker::after_dom_children.
17069 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
17070 * config/i386/i386.c (ix86_option_override_internal): Added
17071 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
17073 2018-01-12 Richard Biener <rguenther@suse.de>
17076 * dwarf2out.c (gen_variable_die): Do not reset old_die for
17077 inline instance vars.
17079 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
17082 * config/rx/rx.c (rx_is_restricted_memory_address):
17083 Handle SUBREG case.
17085 2018-01-12 Richard Biener <rguenther@suse.de>
17087 PR tree-optimization/80846
17088 * target.def (split_reduction): New target hook.
17089 * targhooks.c (default_split_reduction): New function.
17090 * targhooks.h (default_split_reduction): Declare.
17091 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
17092 target requests first reduce vectors by combining low and high
17094 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
17095 (get_vectype_for_scalar_type_and_size): Export.
17096 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
17097 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
17098 * doc/tm.texi: Regenerate.
17099 * config/i386/i386.c (ix86_split_reduction): Implement
17100 TARGET_VECTORIZE_SPLIT_REDUCTION.
17102 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17105 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
17106 in PIC mode except for TARGET_VXWORKS_RTP.
17107 * config/sparc/sparc.c: Include cfgrtl.h.
17108 (TARGET_INIT_PIC_REG): Define.
17109 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
17110 (sparc_pic_register_p): New predicate.
17111 (sparc_legitimate_address_p): Use it.
17112 (sparc_legitimize_pic_address): Likewise.
17113 (sparc_delegitimize_address): Likewise.
17114 (sparc_mode_dependent_address_p): Likewise.
17115 (gen_load_pcrel_sym): Remove 4th parameter.
17116 (load_got_register): Adjust call to above. Remove obsolete stuff.
17117 (sparc_expand_prologue): Do not call load_got_register here.
17118 (sparc_flat_expand_prologue): Likewise.
17119 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
17120 (sparc_use_pseudo_pic_reg): New function.
17121 (sparc_init_pic_reg): Likewise.
17122 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
17123 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
17125 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
17127 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
17128 Add item for branch_cost.
17130 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
17132 PR rtl-optimization/83565
17133 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
17134 not extend the result to a larger mode for rotate operations.
17135 (num_sign_bit_copies1): Likewise.
17137 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17140 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
17142 Use values-Xc.o for -pedantic.
17143 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
17145 2018-01-12 Martin Liska <mliska@suse.cz>
17148 * ipa-devirt.c (final_warning_record::grow_type_warnings):
17150 (possible_polymorphic_call_targets): Use it.
17151 (ipa_devirt): Likewise.
17153 2018-01-12 Martin Liska <mliska@suse.cz>
17155 * profile-count.h (enum profile_quality): Use 0 as invalid
17156 enum value of profile_quality.
17158 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
17160 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
17161 -mext-string options.
17163 2018-01-12 Richard Biener <rguenther@suse.de>
17165 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
17166 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
17167 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
17169 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
17171 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
17173 * configure.ac (--with-long-double-format): Add support for the
17174 configuration option to change the default long double format on
17176 * config.gcc (powerpc*-linux*-*): Likewise.
17177 * configure: Regenerate.
17178 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
17179 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
17180 used without modification.
17182 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17184 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
17185 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
17186 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
17187 MISC_BUILTIN_SPEC_BARRIER.
17188 (rs6000_init_builtins): Likewise.
17189 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
17191 (speculation_barrier): New define_insn.
17192 * doc/extend.texi: Document __builtin_speculation_barrier.
17194 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17197 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
17198 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
17199 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
17201 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
17202 integral modes instead of "ss" and "sd".
17203 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
17204 vectors with 32-bit and 64-bit elements.
17205 (vecdupssescalarmodesuffix): New mode attribute.
17206 (vec_dup<mode>): Use it.
17208 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
17211 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
17212 frame if argument is passed on stack.
17214 2018-01-11 Jakub Jelinek <jakub@redhat.com>
17217 * ree.c (combine_reaching_defs): Optimize also
17218 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
17219 reg2=any_extend(exp); reg1=reg2;, formatting fix.
17221 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17223 PR middle-end/83189
17224 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
17226 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
17228 PR middle-end/83718
17229 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
17230 after they are computed.
17232 2018-01-11 Bin Cheng <bin.cheng@arm.com>
17234 PR tree-optimization/83695
17235 * gimple-loop-linterchange.cc
17236 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
17237 reset cached scev information after interchange.
17238 (pass_linterchange::execute): Remove call to scev_reset_htab.
17240 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17242 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
17243 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
17244 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
17245 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
17246 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
17247 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
17248 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
17249 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
17250 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
17251 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
17252 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
17253 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
17254 (V_lane_reg): Likewise.
17255 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
17257 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
17258 (vfmal_lane_low<mode>_intrinsic,
17259 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
17260 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
17261 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
17262 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
17263 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
17264 vfmsl_lane_high<mode>_intrinsic): New define_insns.
17266 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17268 * config/arm/arm-cpus.in (fp16fml): New feature.
17269 (ALL_SIMD): Add fp16fml.
17270 (armv8.2-a): Add fp16fml as an option.
17271 (armv8.3-a): Likewise.
17272 (armv8.4-a): Add fp16fml as part of fp16.
17273 * config/arm/arm.h (TARGET_FP16FML): Define.
17274 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
17276 * config/arm/arm-modes.def (V2HF): Define.
17277 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
17278 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
17279 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
17280 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
17281 vfmsl_low, vfmsl_high): New set of builtins.
17282 * config/arm/iterators.md (PLUSMINUS): New code iterator.
17283 (vfml_op): New code attribute.
17284 (VFMLHALVES): New int iterator.
17285 (VFML, VFMLSEL): New mode attributes.
17286 (V_reg): Define mapping for V2HF.
17287 (V_hi, V_lo): New mode attributes.
17288 (VF_constraint): Likewise.
17289 (vfml_half, vfml_half_selector): New int attributes.
17290 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
17292 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
17293 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
17295 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
17296 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
17297 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
17298 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
17300 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
17301 Document new effective target and option set.
17303 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17305 * config/arm/arm-cpus.in (armv8_4): New feature.
17306 (ARMv8_4a): New fgroup.
17307 (armv8.4-a): New arch.
17308 * config/arm/arm-tables.opt: Regenerate.
17309 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
17310 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
17311 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
17312 Add matching rules for -march=armv8.4-a and extensions.
17313 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
17315 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
17318 * config/rx/rx.md (BW): New mode attribute.
17319 (sync_lock_test_and_setsi): Add mode suffix to insn output.
17321 2018-01-11 Richard Biener <rguenther@suse.de>
17323 PR tree-optimization/83435
17324 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
17325 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
17326 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
17328 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17329 Alan Hayward <alan.hayward@arm.com>
17330 David Sherwood <david.sherwood@arm.com>
17332 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
17334 (aarch64_classify_address): Initialize it. Track polynomial offsets.
17335 (aarch64_print_address_internal): Use it to check for a zero offset.
17337 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17338 Alan Hayward <alan.hayward@arm.com>
17339 David Sherwood <david.sherwood@arm.com>
17341 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
17342 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
17343 Return a poly_int64 rather than a HOST_WIDE_INT.
17344 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
17345 rather than a HOST_WIDE_INT.
17346 * config/aarch64/aarch64.h (aarch64_frame): Protect with
17347 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
17348 hard_fp_offset, frame_size, initial_adjust, callee_offset and
17349 final_offset from HOST_WIDE_INT to poly_int64.
17350 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
17351 to_constant when getting the number of units in an Advanced SIMD
17353 (aarch64_builtin_vectorized_function): Check for a constant number
17355 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
17357 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
17358 attribute instead of GET_MODE_NUNITS.
17359 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
17360 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
17361 GET_MODE_SIZE for fixed-size registers.
17362 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
17363 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
17364 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
17365 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
17366 (aarch64_print_operand, aarch64_print_address_internal)
17367 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
17368 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
17369 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
17370 Handle polynomial GET_MODE_SIZE.
17371 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
17372 wider than SImode without modification.
17373 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
17374 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
17375 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
17376 passing and returning SVE modes.
17377 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
17378 rather than GEN_INT.
17379 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
17380 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
17381 (aarch64_allocate_and_probe_stack_space): Likewise.
17382 (aarch64_layout_frame): Cope with polynomial offsets.
17383 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
17384 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
17385 polynomial offsets.
17386 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
17387 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
17388 poly_int64 rather than a HOST_WIDE_INT.
17389 (aarch64_get_separate_components, aarch64_process_components)
17390 (aarch64_expand_prologue, aarch64_expand_epilogue)
17391 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
17392 (aarch64_anchor_offset): New function, split out from...
17393 (aarch64_legitimize_address): ...here.
17394 (aarch64_builtin_vectorization_cost): Handle polynomial
17395 TYPE_VECTOR_SUBPARTS.
17396 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
17398 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
17399 number of elements from the PARALLEL rather than the mode.
17400 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
17401 rather than GET_MODE_BITSIZE.
17402 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
17403 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
17404 (aarch64_expand_vec_perm_const_1): Handle polynomial
17405 d->perm.length () and d->perm elements.
17406 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
17407 Apply to_constant to d->perm elements.
17408 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
17409 polynomial CONST_VECTOR_NUNITS.
17410 (aarch64_move_pointer): Take amount as a poly_int64 rather
17412 (aarch64_progress_pointer): Avoid temporary variable.
17413 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
17414 the mode attribute instead of GET_MODE.
17416 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17417 Alan Hayward <alan.hayward@arm.com>
17418 David Sherwood <david.sherwood@arm.com>
17420 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
17421 x exists before using it.
17422 (aarch64_add_constant_internal): Rename to...
17423 (aarch64_add_offset_1): ...this. Replace regnum with separate
17424 src and dest rtxes. Handle the case in which they're different,
17425 including when the offset is zero. Replace scratchreg with an rtx.
17426 Use 2 additions if there is no spare register into which we can
17427 move a 16-bit constant.
17428 (aarch64_add_constant): Delete.
17429 (aarch64_add_offset): Replace reg with separate src and dest
17430 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
17431 Use aarch64_add_offset_1.
17432 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
17433 an rtx rather than an int. Take the delta as a poly_int64
17434 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
17435 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
17436 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
17437 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
17438 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
17439 and aarch64_add_sp.
17440 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
17441 aarch64_add_constant.
17443 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17445 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
17446 Use scalar_float_mode.
17448 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17450 * config/aarch64/aarch64-simd.md
17451 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
17452 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
17453 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
17454 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
17455 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
17456 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
17457 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
17458 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
17459 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
17460 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
17462 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17465 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
17466 targ_options->x_arm_arch_string is non NULL.
17468 2018-01-11 Tamar Christina <tamar.christina@arm.com>
17470 * config/aarch64/aarch64.h
17471 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
17473 2018-01-11 Sudakshina Das <sudi.das@arm.com>
17476 * expmed.c (emit_store_flag_force): Swap if const op0
17477 and change VOIDmode to mode of op0.
17479 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
17481 PR rtl-optimization/83761
17482 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
17483 than bytes to mode_for_size.
17485 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17487 PR middle-end/83189
17488 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
17489 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
17492 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17494 PR middle-end/83575
17495 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
17496 when in layout mode.
17497 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
17498 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
17501 2018-01-10 Michael Collison <michael.collison@arm.com>
17503 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
17504 * config/aarch64/aarch64-option-extension.def: Add
17505 AARCH64_OPT_EXTENSION of 'fp16fml'.
17506 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17507 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
17508 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
17509 * config/aarch64/constraints.md (Ui7): New constraint.
17510 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
17511 (VFMLA_SEL_W): Ditto.
17514 (VFMLA16_LOW): New int iterator.
17515 (VFMLA16_HIGH): Ditto.
17516 (UNSPEC_FMLAL): New unspec.
17517 (UNSPEC_FMLSL): Ditto.
17518 (UNSPEC_FMLAL2): Ditto.
17519 (UNSPEC_FMLSL2): Ditto.
17520 (f16mac): New code attribute.
17521 * config/aarch64/aarch64-simd-builtins.def
17522 (aarch64_fmlal_lowv2sf): Ditto.
17523 (aarch64_fmlsl_lowv2sf): Ditto.
17524 (aarch64_fmlalq_lowv4sf): Ditto.
17525 (aarch64_fmlslq_lowv4sf): Ditto.
17526 (aarch64_fmlal_highv2sf): Ditto.
17527 (aarch64_fmlsl_highv2sf): Ditto.
17528 (aarch64_fmlalq_highv4sf): Ditto.
17529 (aarch64_fmlslq_highv4sf): Ditto.
17530 (aarch64_fmlal_lane_lowv2sf): Ditto.
17531 (aarch64_fmlsl_lane_lowv2sf): Ditto.
17532 (aarch64_fmlal_laneq_lowv2sf): Ditto.
17533 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
17534 (aarch64_fmlalq_lane_lowv4sf): Ditto.
17535 (aarch64_fmlsl_lane_lowv4sf): Ditto.
17536 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
17537 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
17538 (aarch64_fmlal_lane_highv2sf): Ditto.
17539 (aarch64_fmlsl_lane_highv2sf): Ditto.
17540 (aarch64_fmlal_laneq_highv2sf): Ditto.
17541 (aarch64_fmlsl_laneq_highv2sf): Ditto.
17542 (aarch64_fmlalq_lane_highv4sf): Ditto.
17543 (aarch64_fmlsl_lane_highv4sf): Ditto.
17544 (aarch64_fmlalq_laneq_highv4sf): Ditto.
17545 (aarch64_fmlsl_laneq_highv4sf): Ditto.
17546 * config/aarch64/aarch64-simd.md:
17547 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
17548 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17549 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
17550 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
17551 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
17552 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
17553 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
17554 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
17555 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
17556 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
17557 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
17558 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
17559 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
17560 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
17561 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
17562 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
17563 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
17564 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
17565 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
17566 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
17567 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
17568 (vfmlsl_low_u32): Ditto.
17569 (vfmlalq_low_u32): Ditto.
17570 (vfmlslq_low_u32): Ditto.
17571 (vfmlal_high_u32): Ditto.
17572 (vfmlsl_high_u32): Ditto.
17573 (vfmlalq_high_u32): Ditto.
17574 (vfmlslq_high_u32): Ditto.
17575 (vfmlal_lane_low_u32): Ditto.
17576 (vfmlsl_lane_low_u32): Ditto.
17577 (vfmlal_laneq_low_u32): Ditto.
17578 (vfmlsl_laneq_low_u32): Ditto.
17579 (vfmlalq_lane_low_u32): Ditto.
17580 (vfmlslq_lane_low_u32): Ditto.
17581 (vfmlalq_laneq_low_u32): Ditto.
17582 (vfmlslq_laneq_low_u32): Ditto.
17583 (vfmlal_lane_high_u32): Ditto.
17584 (vfmlsl_lane_high_u32): Ditto.
17585 (vfmlal_laneq_high_u32): Ditto.
17586 (vfmlsl_laneq_high_u32): Ditto.
17587 (vfmlalq_lane_high_u32): Ditto.
17588 (vfmlslq_lane_high_u32): Ditto.
17589 (vfmlalq_laneq_high_u32): Ditto.
17590 (vfmlslq_laneq_high_u32): Ditto.
17591 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
17592 (AARCH64_FL_FOR_ARCH8_4): New.
17593 (AARCH64_ISA_F16FML): New ISA flag.
17594 (TARGET_F16FML): New feature flag for fp16fml.
17595 (doc/invoke.texi): Document new fp16fml option.
17597 2018-01-10 Michael Collison <michael.collison@arm.com>
17599 * config/aarch64/aarch64-builtins.c:
17600 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
17601 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17602 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
17603 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
17604 (AARCH64_ISA_SHA3): New ISA flag.
17605 (TARGET_SHA3): New feature flag for sha3.
17606 * config/aarch64/iterators.md (sha512_op): New int attribute.
17607 (CRYPTO_SHA512): New int iterator.
17608 (UNSPEC_SHA512H): New unspec.
17609 (UNSPEC_SHA512H2): Ditto.
17610 (UNSPEC_SHA512SU0): Ditto.
17611 (UNSPEC_SHA512SU1): Ditto.
17612 * config/aarch64/aarch64-simd-builtins.def
17613 (aarch64_crypto_sha512hqv2di): New builtin.
17614 (aarch64_crypto_sha512h2qv2di): Ditto.
17615 (aarch64_crypto_sha512su0qv2di): Ditto.
17616 (aarch64_crypto_sha512su1qv2di): Ditto.
17617 (aarch64_eor3qv8hi): Ditto.
17618 (aarch64_rax1qv2di): Ditto.
17619 (aarch64_xarqv2di): Ditto.
17620 (aarch64_bcaxqv8hi): Ditto.
17621 * config/aarch64/aarch64-simd.md:
17622 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
17623 (aarch64_crypto_sha512su0qv2di): Ditto.
17624 (aarch64_crypto_sha512su1qv2di): Ditto.
17625 (aarch64_eor3qv8hi): Ditto.
17626 (aarch64_rax1qv2di): Ditto.
17627 (aarch64_xarqv2di): Ditto.
17628 (aarch64_bcaxqv8hi): Ditto.
17629 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
17630 (vsha512h2q_u64): Ditto.
17631 (vsha512su0q_u64): Ditto.
17632 (vsha512su1q_u64): Ditto.
17633 (veor3q_u16): Ditto.
17634 (vrax1q_u64): Ditto.
17635 (vxarq_u64): Ditto.
17636 (vbcaxq_u16): Ditto.
17637 * config/arm/types.md (crypto_sha512): New type attribute.
17638 (crypto_sha3): Ditto.
17639 (doc/invoke.texi): Document new sha3 option.
17641 2018-01-10 Michael Collison <michael.collison@arm.com>
17643 * config/aarch64/aarch64-builtins.c:
17644 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
17645 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17646 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
17647 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
17648 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
17649 (AARCH64_ISA_SM4): New ISA flag.
17650 (TARGET_SM4): New feature flag for sm4.
17651 * config/aarch64/aarch64-simd-builtins.def
17652 (aarch64_sm3ss1qv4si): Ditto.
17653 (aarch64_sm3tt1aq4si): Ditto.
17654 (aarch64_sm3tt1bq4si): Ditto.
17655 (aarch64_sm3tt2aq4si): Ditto.
17656 (aarch64_sm3tt2bq4si): Ditto.
17657 (aarch64_sm3partw1qv4si): Ditto.
17658 (aarch64_sm3partw2qv4si): Ditto.
17659 (aarch64_sm4eqv4si): Ditto.
17660 (aarch64_sm4ekeyqv4si): Ditto.
17661 * config/aarch64/aarch64-simd.md:
17662 (aarch64_sm3ss1qv4si): Ditto.
17663 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
17664 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
17665 (aarch64_sm4eqv4si): Ditto.
17666 (aarch64_sm4ekeyqv4si): Ditto.
17667 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
17668 (sm3part_op): Ditto.
17669 (CRYPTO_SM3TT): Ditto.
17670 (CRYPTO_SM3PART): Ditto.
17671 (UNSPEC_SM3SS1): New unspec.
17672 (UNSPEC_SM3TT1A): Ditto.
17673 (UNSPEC_SM3TT1B): Ditto.
17674 (UNSPEC_SM3TT2A): Ditto.
17675 (UNSPEC_SM3TT2B): Ditto.
17676 (UNSPEC_SM3PARTW1): Ditto.
17677 (UNSPEC_SM3PARTW2): Ditto.
17678 (UNSPEC_SM4E): Ditto.
17679 (UNSPEC_SM4EKEY): Ditto.
17680 * config/aarch64/constraints.md (Ui2): New constraint.
17681 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
17682 * config/arm/types.md (crypto_sm3): New type attribute.
17683 (crypto_sm4): Ditto.
17684 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
17685 (vsm3tt1aq_u32): Ditto.
17686 (vsm3tt1bq_u32): Ditto.
17687 (vsm3tt2aq_u32): Ditto.
17688 (vsm3tt2bq_u32): Ditto.
17689 (vsm3partw1q_u32): Ditto.
17690 (vsm3partw2q_u32): Ditto.
17691 (vsm4eq_u32): Ditto.
17692 (vsm4ekeyq_u32): Ditto.
17693 (doc/invoke.texi): Document new sm4 option.
17695 2018-01-10 Michael Collison <michael.collison@arm.com>
17697 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
17698 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
17699 (AARCH64_FL_FOR_ARCH8_4): New.
17700 (AARCH64_FL_V8_4): New flag.
17701 (doc/invoke.texi): Document new armv8.4-a option.
17703 2018-01-10 Michael Collison <michael.collison@arm.com>
17705 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
17706 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
17707 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
17708 * config/aarch64/aarch64-option-extension.def: Add
17709 AARCH64_OPT_EXTENSION of 'sha2'.
17710 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
17711 (crypto): Disable sha2 and aes if crypto disabled.
17712 (crypto): Enable aes and sha2 if enabled.
17713 (simd): Disable sha2 and aes if simd disabled.
17714 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
17716 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
17717 (TARGET_SHA2): New feature flag for sha2.
17718 (TARGET_AES): New feature flag for aes.
17719 * config/aarch64/aarch64-simd.md:
17720 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
17721 conditional on TARGET_AES.
17722 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
17723 (aarch64_crypto_sha1hsi): Make pattern conditional
17725 (aarch64_crypto_sha1hv4si): Ditto.
17726 (aarch64_be_crypto_sha1hv4si): Ditto.
17727 (aarch64_crypto_sha1su1v4si): Ditto.
17728 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
17729 (aarch64_crypto_sha1su0v4si): Ditto.
17730 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
17731 (aarch64_crypto_sha256su0v4si): Ditto.
17732 (aarch64_crypto_sha256su1v4si): Ditto.
17733 (doc/invoke.texi): Document new aes and sha2 options.
17735 2018-01-10 Martin Sebor <msebor@redhat.com>
17737 PR tree-optimization/83781
17738 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
17741 2018-01-11 Martin Sebor <msebor@gmail.com>
17742 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
17744 PR tree-optimization/83501
17745 PR tree-optimization/81703
17747 * tree-ssa-strlen.c (get_string_cst): Rename...
17748 (get_string_len): ...to this. Handle global constants.
17749 (handle_char_store): Adjust.
17751 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
17752 Jim Wilson <jimw@sifive.com>
17754 * config/riscv/riscv-protos.h (riscv_output_return): New.
17755 * config/riscv/riscv.c (struct machine_function): New naked_p field.
17756 (riscv_attribute_table, riscv_output_return),
17757 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
17758 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
17759 (riscv_compute_frame_info): Only compute frame->mask if not a naked
17761 (riscv_expand_prologue): Add early return for naked function.
17762 (riscv_expand_epilogue): Likewise.
17763 (riscv_function_ok_for_sibcall): Return false for naked function.
17764 (riscv_set_current_function): New.
17765 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
17766 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
17767 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
17768 * doc/extend.texi (RISC-V Function Attributes): New.
17770 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
17772 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
17773 check for 128-bit long double before checking TCmode.
17774 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
17775 128-bit long doubles before checking TFmode or TCmode.
17776 (FLOAT128_IBM_P): Likewise.
17778 2018-01-10 Martin Sebor <msebor@redhat.com>
17780 PR tree-optimization/83671
17781 * builtins.c (c_strlen): Unconditionally return zero for the empty
17783 Use -Warray-bounds for warnings.
17784 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
17785 for non-constant array indices with COMPONENT_REF, arrays of
17786 arrays, and pointers to arrays.
17787 (gimple_fold_builtin_strlen): Determine and set length range for
17788 non-constant character arrays.
17790 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
17792 PR middle-end/81897
17793 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
17796 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
17798 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
17800 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17803 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
17804 VECTOR_MEM_ALTIVEC_OR_VSX_P.
17805 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
17806 indexed_or_indirect_operand predicate.
17807 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
17808 (*vsx_le_perm_load_v8hi): Likewise.
17809 (*vsx_le_perm_load_v16qi): Likewise.
17810 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
17811 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
17812 (*vsx_le_perm_store_v8hi): Likewise.
17813 (*vsx_le_perm_store_v16qi): Likewise.
17814 (eight unnamed splitters): Likewise.
17816 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
17818 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
17819 * config/rs6000/emmintrin.h: Likewise.
17820 * config/rs6000/mmintrin.h: Likewise.
17821 * config/rs6000/xmmintrin.h: Likewise.
17823 2018-01-10 David Malcolm <dmalcolm@redhat.com>
17826 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
17828 * tree.c (tree_nop_conversion): Return true for location wrapper
17830 (maybe_wrap_with_location): New function.
17831 (selftest::check_strip_nops): New function.
17832 (selftest::test_location_wrappers): New function.
17833 (selftest::tree_c_tests): Call it.
17834 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
17835 (maybe_wrap_with_location): New decl.
17836 (EXPR_LOCATION_WRAPPER_P): New macro.
17837 (location_wrapper_p): New inline function.
17838 (tree_strip_any_location_wrapper): New inline function.
17840 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
17843 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
17844 stack_realign_offset for the largest alignment of stack slot
17846 (ix86_find_max_used_stack_alignment): New function.
17847 (ix86_finalize_stack_frame_flags): Use it. Set
17848 max_used_stack_alignment if we don't realign stack.
17849 * config/i386/i386.h (machine_function): Add
17850 max_used_stack_alignment.
17852 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
17854 * config/arm/arm.opt (-mbranch-cost): New option.
17855 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
17858 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
17861 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
17862 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
17864 2018-01-10 Richard Biener <rguenther@suse.de>
17867 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
17868 early out so it also covers the case where we have a non-NULL
17871 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17873 PR tree-optimization/83753
17874 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
17875 for non-strided grouped accesses if the number of elements is 1.
17877 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
17880 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
17881 * i386.h (TARGET_USE_GATHER): Define.
17882 * x86-tune.def (X86_TUNE_USE_GATHER): New.
17884 2018-01-10 Martin Liska <mliska@suse.cz>
17887 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
17888 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
17890 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
17891 CLEANUP_NO_PARTITIONING is not set.
17893 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
17895 * doc/rtl.texi: Remove documentation of (const ...) wrappers
17896 for vectors, as a partial revert of r254296.
17897 * rtl.h (const_vec_p): Delete.
17898 (const_vec_duplicate_p): Don't test for vector CONSTs.
17899 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
17900 * expmed.c (make_tree): Likewise.
17903 * common.md (E, F): Use CONSTANT_P instead of checking for
17905 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
17906 checking for CONST_VECTOR.
17908 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
17910 PR middle-end/83575
17911 * predict.c (force_edge_cold): Handle in more sane way edges
17912 with no prediction.
17914 2018-01-09 Carl Love <cel@us.ibm.com>
17916 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
17918 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
17919 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
17920 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
17921 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
17922 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
17923 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
17924 * config/rs6000/rs6000-protos.h: Add extern defition for
17925 rs6000_generate_float2_double_code.
17926 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
17928 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
17929 (float2_v2df): Add define_expand.
17931 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
17934 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
17935 op_mode in the force_to_mode call.
17937 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17939 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
17940 instead of checking each element individually.
17941 (aarch64_evpc_uzp): Likewise.
17942 (aarch64_evpc_zip): Likewise.
17943 (aarch64_evpc_ext): Likewise.
17944 (aarch64_evpc_rev): Likewise.
17945 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
17946 instead of checking each element individually. Return true without
17948 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
17949 whether all selected elements come from the same input, instead of
17950 checking each element individually. Remove calls to gen_rtx_REG,
17951 start_sequence and end_sequence and instead assert that no rtl is
17954 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17956 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
17957 order of HIGH and CONST checks.
17959 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
17961 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
17962 if the destination isn't an SSA_NAME.
17964 2018-01-09 Richard Biener <rguenther@suse.de>
17966 PR tree-optimization/83668
17967 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
17969 (canonicalize_loop_form): ... here, renamed from ...
17970 (canonicalize_loop_closed_ssa_form): ... this and amended to
17971 swap successor edges for loop exit blocks to make us use
17972 the RPO order we need for initial schedule generation.
17974 2018-01-09 Joseph Myers <joseph@codesourcery.com>
17976 PR tree-optimization/64811
17977 * match.pd: When optimizing comparisons with Inf, avoid
17978 introducing or losing exceptions from comparisons with NaN.
17980 2018-01-09 Martin Liska <mliska@suse.cz>
17983 * asan.c (shadow_mem_size): Add gcc_assert.
17985 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
17987 Don't save registers in main().
17990 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
17991 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
17992 * config/avr/avr.c (avr_set_current_function): Don't error if
17993 naked, OS_task or OS_main are specified at the same time.
17994 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
17996 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
17998 * common/config/avr/avr-common.c (avr_option_optimization_table):
17999 Switch on -mmain-is-OS_task for optimizing compilations.
18001 2018-01-09 Richard Biener <rguenther@suse.de>
18003 PR tree-optimization/83572
18004 * graphite.c: Include cfganal.h.
18005 (graphite_transform_loops): Connect infinite loops to exit
18006 and remove fake edges at the end.
18008 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18010 * ipa-inline.c (edge_badness): Revert accidental checkin.
18012 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
18015 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
18016 symbols; not inline clones.
18018 2018-01-09 Jakub Jelinek <jakub@redhat.com>
18021 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
18022 hard registers. Formatting fixes.
18024 PR preprocessor/83722
18025 * gcc.c (try_generate_repro): Pass
18026 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
18027 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
18030 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
18031 Kito Cheng <kito.cheng@gmail.com>
18033 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
18034 (riscv_leaf_function_p): Delete.
18035 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
18037 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18039 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
18041 (do_ifelse): New function.
18042 (do_isel): New function.
18043 (do_sub3): New function.
18044 (do_add3): New function.
18045 (do_load_mask_compare): New function.
18046 (do_overlap_load_compare): New function.
18047 (expand_compare_loop): New function.
18048 (expand_block_compare): Call expand_compare_loop() when appropriate.
18049 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
18050 option description.
18051 (-mblock-compare-inline-loop-limit): New option.
18053 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
18056 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
18057 Reverse order of second and third operands in first alternative.
18058 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
18059 of first and second elements in UNSPEC_VPERMR vector.
18060 (altivec_expand_vec_perm_le): Likewise.
18062 2018-01-08 Jeff Law <law@redhat.com>
18064 PR rtl-optimizatin/81308
18065 * tree-switch-conversion.c (cfg_altered): New file scoped static.
18066 (process_switch): If group_case_labels makes a change, then set
18068 (pass_convert_switch::execute): If a switch is converted, then
18069 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
18071 PR rtl-optimization/81308
18072 * recog.c (split_all_insns): Conditionally cleanup the CFG after
18075 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
18077 PR target/83663 - Revert r255946
18078 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
18079 generation for cases where splatting a value is not useful.
18080 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
18081 across a vec_duplicate and a paradoxical subreg forming a vector
18082 mode to a vec_concat.
18084 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18086 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
18087 -march=armv8.3-a variants.
18088 * config/arm/t-multilib: Likewise.
18089 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
18091 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18093 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
18095 (cceq_ior_compare_complement): Give it a name so I can use it, and
18096 change boolean_or_operator predicate to boolean_operator so it can
18097 be used to generate a crand.
18098 (eqne): New code iterator.
18099 (bd/bd_neg): New code_attrs.
18100 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
18101 a single define_insn.
18102 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
18103 decrement (bdnzt/bdnzf/bdzt/bdzf).
18104 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
18105 with the new names of the branch decrement patterns, and added the
18106 names of the branch decrement conditional patterns.
18108 2018-01-08 Richard Biener <rguenther@suse.de>
18110 PR tree-optimization/83563
18111 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
18114 2018-01-08 Richard Biener <rguenther@suse.de>
18116 PR middle-end/83713
18117 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
18119 2018-01-08 Richard Biener <rguenther@suse.de>
18121 PR tree-optimization/83685
18122 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
18123 references to abnormals.
18125 2018-01-08 Richard Biener <rguenther@suse.de>
18128 * dwarf2out.c (output_indirect_strings): Handle empty
18129 skeleton_debug_str_hash.
18130 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
18132 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18134 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
18135 (emit_store_direct): Likewise.
18136 (arc_trampoline_adjust_address): Likewise.
18137 (arc_asm_trampoline_template): New function.
18138 (arc_initialize_trampoline): Use asm_trampoline_template.
18139 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
18140 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
18141 * config/arc/arc.md (flush_icache): Delete pattern.
18143 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
18145 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
18146 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
18149 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18152 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
18153 by not USED_FOR_TARGET.
18154 (make_pass_resolve_sw_modes): Likewise.
18156 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
18158 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
18161 2018-01-08 Richard Biener <rguenther@suse.de>
18163 PR middle-end/83580
18164 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
18166 2018-01-08 Richard Biener <rguenther@suse.de>
18168 PR middle-end/83517
18169 * match.pd ((t * 2) / 2) -> t): Add missing :c.
18171 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
18173 PR middle-end/81897
18174 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
18175 basic blocks with a small number of successors.
18176 (convert_control_dep_chain_into_preds): Improve handling of
18178 (dump_predicates): Split apart into...
18179 (dump_pred_chain): ...here...
18180 (dump_pred_info): ...and here.
18181 (can_one_predicate_be_invalidated_p): Add debugging printfs.
18182 (can_chain_union_be_invalidated_p): Improve check for invalidation
18184 (uninit_uses_cannot_happen): Avoid unnecessary if
18185 convert_control_dep_chain_into_preds yielded nothing.
18187 2018-01-06 Martin Sebor <msebor@redhat.com>
18189 PR tree-optimization/83640
18190 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
18191 subtracting negative offset from size.
18192 (builtin_access::overlap): Adjust offset bounds of the access to fall
18193 within the size of the object if possible.
18195 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
18197 PR rtl-optimization/83699
18198 * expmed.c (extract_bit_field_1): Restrict the vector usage of
18199 extract_bit_field_as_subreg to cases in which the extracted
18200 value is also a vector.
18202 * lra-constraints.c (process_alt_operands): Test for the equivalence
18203 substitutions when detecting a possible reload cycle.
18205 2018-01-06 Jakub Jelinek <jakub@redhat.com>
18208 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
18209 by default if flag_selective_schedling{,2}. Formatting fixes.
18211 PR rtl-optimization/83682
18212 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
18213 if it has non-VECTOR_MODE element mode.
18214 (vec_duplicate_p): Likewise.
18216 PR middle-end/83694
18217 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
18218 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
18220 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18223 * config/i386/i386-builtin.def
18224 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
18225 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
18226 Require also OPTION_MASK_ISA_AVX512F in addition to
18227 OPTION_MASK_ISA_GFNI.
18228 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
18229 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
18230 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
18231 to OPTION_MASK_ISA_GFNI.
18232 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
18233 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
18234 OPTION_MASK_ISA_AVX512BW.
18235 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
18236 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
18237 addition to OPTION_MASK_ISA_GFNI.
18238 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
18239 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
18240 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
18241 to OPTION_MASK_ISA_GFNI.
18242 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
18243 a requirement for all ISAs rather than any of them with a few
18245 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
18247 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
18248 bitmasks to be enabled with 3 exceptions, instead of requiring any
18249 enabled ISA with lots of exceptions.
18250 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
18251 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
18252 Change avx512bw in isa attribute to avx512f.
18253 * config/i386/sgxintrin.h: Add license boilerplate.
18254 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
18255 to __AVX512F__ and __AVX512VL to __AVX512VL__.
18256 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
18257 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
18259 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
18260 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
18261 temporarily sse2 rather than sse if not enabled already.
18264 * config/i386/sse.md (VI248_VLBW): Rename to ...
18265 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
18266 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
18267 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
18268 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
18269 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
18270 mode iterator instead of VI248_VLBW.
18272 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
18274 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
18275 (record_modified): Skip clobbers; add debug output.
18276 (param_change_prob): Use sreal frequencies.
18278 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18280 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
18281 punt for user-aligned variables.
18283 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
18285 * tree-chrec.c (chrec_contains_symbols): Return true for
18288 2018-01-05 Sudakshina Das <sudi.das@arm.com>
18291 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
18292 of (x|y) == x for BICS pattern.
18294 2018-01-05 Jakub Jelinek <jakub@redhat.com>
18296 PR tree-optimization/83605
18297 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
18298 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
18301 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
18303 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
18304 * config/epiphany/rtems.h: New file.
18306 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18307 Uros Bizjak <ubizjak@gmail.com>
18310 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
18311 QIreg_operand instead of register_operand predicate.
18312 * config/i386/i386.c (ix86_rop_should_change_byte_p,
18313 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
18314 comments instead of -fmitigate[-_]rop.
18316 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18319 * cgraphunit.c (symbol_table::compile): Switch to text_section
18320 before calling assembly_start debug hook.
18321 * run-rtl-passes.c (run_rtl_passes): Likewise.
18324 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18326 * tree-vrp.c (extract_range_from_binary_expr_1): Check
18327 range_int_cst_p rather than !symbolic_range_p before calling
18328 extract_range_from_multiplicative_op_1.
18330 2018-01-04 Jeff Law <law@redhat.com>
18332 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
18333 redundant test in assertion.
18335 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18337 * doc/rtl.texi: Document machine_mode wrapper classes.
18339 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18341 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
18342 using tree_to_uhwi.
18344 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18346 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
18347 the VEC_PERM_EXPR fold to fail.
18349 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18352 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
18353 to switched_sections.
18355 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18358 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
18359 test for d.testing.
18361 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
18364 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
18365 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
18367 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18370 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
18371 is BLKmode and bitpos not zero or mode change is needed.
18373 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
18376 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
18379 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
18382 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
18383 instead of MULT rtx. Update all corresponding splitters.
18384 (*saddl_se): Ditto.
18385 (*ssub<modesuffix>): Ditto.
18386 (*ssubl_se): Ditto.
18387 (*cmp_sadd_di): Update split patterns.
18388 (*cmp_sadd_si): Ditto.
18389 (*cmp_sadd_sidi): Ditto.
18390 (*cmp_ssub_di): Ditto.
18391 (*cmp_ssub_si): Ditto.
18392 (*cmp_ssub_sidi): Ditto.
18393 * config/alpha/predicates.md (const23_operand): New predicate.
18394 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
18395 Look for ASHIFT, not MULT inner operand.
18396 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
18398 2018-01-04 Martin Liska <mliska@suse.cz>
18400 PR gcov-profile/83669
18401 * gcov.c (output_intermediate_file): Add version to intermediate
18403 * doc/gcov.texi: Document new field 'version' in intermediate
18404 file format. Fix location of '-k' option of gcov command.
18406 2018-01-04 Martin Liska <mliska@suse.cz>
18409 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
18411 2018-01-04 Jakub Jelinek <jakub@redhat.com>
18413 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
18415 2018-01-03 Martin Sebor <msebor@redhat.com>
18417 PR tree-optimization/83655
18418 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
18419 checking calls with invalid arguments.
18421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18423 * tree-vect-stmts.c (vect_get_store_rhs): New function.
18424 (vectorizable_mask_load_store): Delete.
18425 (vectorizable_call): Return false for masked loads and stores.
18426 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
18427 instead of gimple_assign_rhs1.
18428 (vectorizable_load): Handle IFN_MASK_LOAD.
18429 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
18431 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18433 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
18435 (vectorizable_mask_load_store): ...here.
18436 (vectorizable_load): ...and here.
18438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18440 * tree-vect-stmts.c (vect_build_all_ones_mask)
18441 (vect_build_zero_merge_argument): New functions, split out from...
18442 (vectorizable_load): ...here.
18444 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18446 * tree-vect-stmts.c (vect_check_store_rhs): New function,
18448 (vectorizable_mask_load_store): ...here.
18449 (vectorizable_store): ...and here.
18451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18453 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
18455 (vectorizable_mask_load_store): ...here.
18457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18459 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
18460 (vect_model_store_cost): Take a vec_load_store_type instead of a
18462 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
18463 (vect_model_store_cost): Take a vec_load_store_type instead of a
18465 (vectorizable_mask_load_store): Update accordingly.
18466 (vectorizable_store): Likewise.
18467 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
18469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18471 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
18472 IFN_MASK_LOAD calls here rather than...
18473 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
18475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18476 Alan Hayward <alan.hayward@arm.com>
18477 David Sherwood <david.sherwood@arm.com>
18479 * expmed.c (extract_bit_field_1): For vector extracts,
18480 fall back to extract_bit_field_as_subreg if vec_extract
18483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18484 Alan Hayward <alan.hayward@arm.com>
18485 David Sherwood <david.sherwood@arm.com>
18487 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
18488 they are variable or constant sized.
18489 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
18490 slots for constant-sized data.
18492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18493 Alan Hayward <alan.hayward@arm.com>
18494 David Sherwood <david.sherwood@arm.com>
18496 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
18497 handling COND_EXPRs with boolean comparisons, try to find a better
18498 basis for the mask type than the boolean itself.
18500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18502 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
18503 is calculated and how it can be overridden.
18504 * genmodes.c (max_bitsize_mode_any_mode): New variable.
18505 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
18507 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
18510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18511 Alan Hayward <alan.hayward@arm.com>
18512 David Sherwood <david.sherwood@arm.com>
18514 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
18515 Remove the mode argument.
18516 (aarch64_simd_valid_immediate): Remove the mode and inverse
18518 * config/aarch64/iterators.md (bitsize): New iterator.
18519 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
18520 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
18521 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
18522 aarch64_simd_valid_immediate.
18523 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
18524 (aarch64_reg_or_bic_imm): Likewise.
18525 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
18526 with an insn_type enum and msl with a modifier_type enum.
18527 Replace element_width with a scalar_mode. Change the shift
18528 to unsigned int. Add constructors for scalar_float_mode and
18529 scalar_int_mode elements.
18530 (aarch64_vect_float_const_representable_p): Delete.
18531 (aarch64_can_const_movi_rtx_p)
18532 (aarch64_simd_scalar_immediate_valid_for_move)
18533 (aarch64_simd_make_constant): Update call to
18534 aarch64_simd_valid_immediate.
18535 (aarch64_advsimd_valid_immediate_hs): New function.
18536 (aarch64_advsimd_valid_immediate): Likewise.
18537 (aarch64_simd_valid_immediate): Remove mode and inverse
18538 arguments. Rewrite to use the above. Use const_vec_duplicate_p
18539 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
18540 and aarch64_float_const_representable_p on the result.
18541 (aarch64_output_simd_mov_immediate): Remove mode argument.
18542 Update call to aarch64_simd_valid_immediate and use of
18543 simd_immediate_info.
18544 (aarch64_output_scalar_simd_mov_immediate): Update call
18547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18548 Alan Hayward <alan.hayward@arm.com>
18549 David Sherwood <david.sherwood@arm.com>
18551 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
18552 (mode_nunits): Likewise CONST_MODE_NUNITS.
18553 * machmode.def (ADJUST_NUNITS): Document.
18554 * genmodes.c (mode_data::need_nunits_adj): New field.
18555 (blank_mode): Update accordingly.
18556 (adj_nunits): New variable.
18557 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
18559 (emit_mode_size_inline): Set need_bytesize_adj for all modes
18560 listed in adj_nunits.
18561 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
18562 listed in adj_nunits. Don't emit case statements for such modes.
18563 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
18564 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
18565 nothing if adj_nunits is nonnull.
18566 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
18567 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
18568 (emit_mode_fbit): Update use of print_maybe_const_decl.
18569 (emit_move_size): Likewise. Treat the array as non-const
18571 (emit_mode_adjustments): Handle adj_nunits.
18573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18575 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
18576 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
18577 (VECTOR_MODES): Use it.
18578 (make_vector_modes): Take the prefix as an argument.
18580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18581 Alan Hayward <alan.hayward@arm.com>
18582 David Sherwood <david.sherwood@arm.com>
18584 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
18585 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
18586 for MODE_VECTOR_BOOL.
18587 * machmode.def (VECTOR_BOOL_MODE): Document.
18588 * genmodes.c (VECTOR_BOOL_MODE): New macro.
18589 (make_vector_bool_mode): New function.
18590 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
18592 * lto-streamer-in.c (lto_input_mode_table): Likewise.
18593 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
18595 * stor-layout.c (int_mode_for_mode): Likewise.
18596 * tree.c (build_vector_type_for_mode): Likewise.
18597 * varasm.c (output_constant_pool_2): Likewise.
18598 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
18599 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
18600 for MODE_VECTOR_BOOL.
18601 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
18602 of mode class checks.
18603 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
18604 instead of a list of mode class checks.
18605 (expand_vector_scalar_condition): Likewise.
18606 (type_for_widest_vector_mode): Handle BImode as an inner mode.
18608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18609 Alan Hayward <alan.hayward@arm.com>
18610 David Sherwood <david.sherwood@arm.com>
18612 * machmode.h (mode_size): Change from unsigned short to
18614 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
18615 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18616 or if measurement_type is not polynomial.
18617 (fixed_size_mode::includes_p): Check for constant-sized modes.
18618 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
18619 return a poly_uint16 rather than an unsigned short.
18620 (emit_mode_size): Change the type of mode_size from unsigned short
18621 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
18622 (emit_mode_adjustments): Cope with polynomial vector sizes.
18623 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18625 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18627 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
18628 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
18629 * caller-save.c (setup_save_areas): Likewise.
18630 (replace_reg_with_saved_mem): Likewise.
18631 * calls.c (emit_library_call_value_1): Likewise.
18632 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
18633 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
18634 (gen_lowpart_for_combine): Likewise.
18635 * convert.c (convert_to_integer_1): Likewise.
18636 * cse.c (equiv_constant, cse_insn): Likewise.
18637 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
18638 (cselib_subst_to_values): Likewise.
18639 * dce.c (word_dce_process_block): Likewise.
18640 * df-problems.c (df_word_lr_mark_ref): Likewise.
18641 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
18642 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
18643 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
18644 (rtl_for_decl_location): Likewise.
18645 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
18646 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
18647 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
18648 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
18649 (expand_expr_real_1): Likewise.
18650 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
18651 (pad_below): Likewise.
18652 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18653 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
18654 * ira.c (get_subreg_tracking_sizes): Likewise.
18655 * ira-build.c (ira_create_allocno_objects): Likewise.
18656 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
18657 (ira_sort_regnos_for_alter_reg): Likewise.
18658 * ira-costs.c (record_operand_costs): Likewise.
18659 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
18660 (resolve_simple_move): Likewise.
18661 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
18662 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
18663 (lra_constraints): Likewise.
18664 (CONST_POOL_OK_P): Reject variable-sized modes.
18665 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
18666 (add_pseudo_to_slot, lra_spill): Likewise.
18667 * omp-low.c (omp_clause_aligned_alignment): Likewise.
18668 * optabs-query.c (get_best_extraction_insn): Likewise.
18669 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18670 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
18671 (expand_mult_highpart, valid_multiword_target_p): Likewise.
18672 * recog.c (offsettable_address_addr_space_p): Likewise.
18673 * regcprop.c (maybe_mode_change): Likewise.
18674 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
18675 * regrename.c (build_def_use): Likewise.
18676 * regstat.c (dump_reg_info): Likewise.
18677 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
18678 (find_reloads, find_reloads_subreg_address): Likewise.
18679 * reload1.c (eliminate_regs_1): Likewise.
18680 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
18681 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
18682 (simplify_binary_operation_1, simplify_subreg): Likewise.
18683 * targhooks.c (default_function_arg_padding): Likewise.
18684 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
18685 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
18686 (verify_gimple_assign_ternary): Likewise.
18687 * tree-inline.c (estimate_move_cost): Likewise.
18688 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18689 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
18690 (get_address_cost_ainc): Likewise.
18691 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
18692 (vect_supportable_dr_alignment): Likewise.
18693 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18694 (vectorizable_reduction): Likewise.
18695 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
18696 (vectorizable_operation, vectorizable_load): Likewise.
18697 * tree.c (build_same_sized_truth_vector_type): Likewise.
18698 * valtrack.c (cleanup_auto_inc_dec): Likewise.
18699 * var-tracking.c (emit_note_insn_var_location): Likewise.
18700 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
18701 (ADDR_VEC_ALIGN): Likewise.
18703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18704 Alan Hayward <alan.hayward@arm.com>
18705 David Sherwood <david.sherwood@arm.com>
18707 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
18709 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
18710 or if measurement_type is polynomial.
18711 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
18712 * combine.c (make_extraction): Likewise.
18713 * dse.c (find_shift_sequence): Likewise.
18714 * dwarf2out.c (mem_loc_descriptor): Likewise.
18715 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
18716 (extract_bit_field, extract_low_bits): Likewise.
18717 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
18718 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
18719 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
18720 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
18721 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
18722 * reload.c (find_reloads): Likewise.
18723 * reload1.c (alter_reg): Likewise.
18724 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
18725 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
18726 * tree-if-conv.c (predicate_mem_writes): Likewise.
18727 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
18728 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
18729 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
18730 * valtrack.c (dead_debug_insert_temp): Likewise.
18731 * varasm.c (mergeable_constant_section): Likewise.
18732 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
18734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18735 Alan Hayward <alan.hayward@arm.com>
18736 David Sherwood <david.sherwood@arm.com>
18738 * expr.c (expand_assignment): Cope with polynomial mode sizes
18739 when assigning to a CONCAT.
18741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18742 Alan Hayward <alan.hayward@arm.com>
18743 David Sherwood <david.sherwood@arm.com>
18745 * machmode.h (mode_precision): Change from unsigned short to
18747 (mode_to_precision): Return a poly_uint16 rather than an unsigned
18749 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
18750 or if measurement_type is not polynomial.
18751 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
18752 in which the mode is already known to be a scalar_int_mode.
18753 * genmodes.c (emit_mode_precision): Change the type of mode_precision
18754 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
18756 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18757 for GET_MODE_PRECISION.
18758 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18759 for GET_MODE_PRECISION.
18760 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
18762 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
18763 (expand_field_assignment, make_extraction): Likewise.
18764 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
18765 (get_last_value): Likewise.
18766 * convert.c (convert_to_integer_1): Likewise.
18767 * cse.c (cse_insn): Likewise.
18768 * expr.c (expand_expr_real_1): Likewise.
18769 * lra-constraints.c (simplify_operand_subreg): Likewise.
18770 * optabs-query.c (can_atomic_load_p): Likewise.
18771 * optabs.c (expand_atomic_load): Likewise.
18772 (expand_atomic_store): Likewise.
18773 * ree.c (combine_reaching_defs): Likewise.
18774 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
18775 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
18776 * tree.h (type_has_mode_precision_p): Likewise.
18777 * ubsan.c (instrument_si_overflow): Likewise.
18779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18780 Alan Hayward <alan.hayward@arm.com>
18781 David Sherwood <david.sherwood@arm.com>
18783 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
18784 polynomial numbers of units.
18785 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
18786 (valid_vector_subparts_p): New function.
18787 (build_vector_type): Remove temporary shim and take the number
18788 of units as a poly_uint64 rather than an int.
18789 (build_opaque_vector_type): Take the number of units as a
18790 poly_uint64 rather than an int.
18791 * tree.c (build_vector_from_ctor): Handle polynomial
18792 TYPE_VECTOR_SUBPARTS.
18793 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
18794 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
18795 (build_vector_from_val): If the number of units is variable,
18796 use build_vec_duplicate_cst for constant operands and
18797 VEC_DUPLICATE_EXPR otherwise.
18798 (make_vector_type): Remove temporary is_constant ().
18799 (build_vector_type, build_opaque_vector_type): Take the number of
18800 units as a poly_uint64 rather than an int.
18801 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
18803 * cfgexpand.c (expand_debug_expr): Likewise.
18804 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
18805 (store_constructor, expand_expr_real_1): Likewise.
18806 (const_scalar_mask_from_tree): Likewise.
18807 * fold-const-call.c (fold_const_reduction): Likewise.
18808 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
18809 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
18810 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
18811 (fold_relational_const): Likewise.
18812 (native_interpret_vector): Likewise. Change the size from an
18813 int to an unsigned int.
18814 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
18815 TYPE_VECTOR_SUBPARTS.
18816 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
18817 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
18818 duplicating a non-constant operand into a variable-length vector.
18819 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
18820 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
18821 * ipa-icf.c (sem_variable::equals): Likewise.
18822 * match.pd: Likewise.
18823 * omp-simd-clone.c (simd_clone_subparts): Likewise.
18824 * print-tree.c (print_node): Likewise.
18825 * stor-layout.c (layout_type): Likewise.
18826 * targhooks.c (default_builtin_vectorization_cost): Likewise.
18827 * tree-cfg.c (verify_gimple_comparison): Likewise.
18828 (verify_gimple_assign_binary): Likewise.
18829 (verify_gimple_assign_ternary): Likewise.
18830 (verify_gimple_assign_single): Likewise.
18831 * tree-pretty-print.c (dump_generic_node): Likewise.
18832 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
18833 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
18834 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
18835 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
18836 (vect_shift_permute_load_chain): Likewise.
18837 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
18838 (expand_vector_condition, optimize_vector_constructor): Likewise.
18839 (lower_vec_perm, get_compute_type): Likewise.
18840 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
18841 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
18842 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
18843 (vect_recog_mask_conversion_pattern): Likewise.
18844 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
18845 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
18846 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
18847 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
18848 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
18849 (vectorizable_shift, vectorizable_operation, vectorizable_store)
18850 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
18851 (supportable_widening_operation): Likewise.
18852 (supportable_narrowing_operation): Likewise.
18853 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
18855 * varasm.c (output_constant): Likewise.
18857 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18858 Alan Hayward <alan.hayward@arm.com>
18859 David Sherwood <david.sherwood@arm.com>
18861 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
18862 so that both the length == 3 and length != 3 cases set up their
18863 own permute vectors. Add comments explaining why we know the
18864 number of elements is constant.
18865 (vect_permute_load_chain): Likewise.
18867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18868 Alan Hayward <alan.hayward@arm.com>
18869 David Sherwood <david.sherwood@arm.com>
18871 * machmode.h (mode_nunits): Change from unsigned char to
18873 (ONLY_FIXED_SIZE_MODES): New macro.
18874 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
18875 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
18876 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
18878 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
18879 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
18880 or if measurement_type is not polynomial.
18881 * genmodes.c (ZERO_COEFFS): New macro.
18882 (emit_mode_nunits_inline): Make mode_nunits_inline return a
18884 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
18885 Use ZERO_COEFFS when emitting initializers.
18886 * data-streamer.h (bp_pack_poly_value): New function.
18887 (bp_unpack_poly_value): Likewise.
18888 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
18889 for GET_MODE_NUNITS.
18890 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
18891 for GET_MODE_NUNITS.
18892 * tree.c (make_vector_type): Remove temporary shim and make
18893 the real function take the number of units as a poly_uint64
18894 rather than an int.
18895 (build_vector_type_for_mode): Handle polynomial nunits.
18896 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
18897 * emit-rtl.c (const_vec_series_p_1): Likewise.
18898 (gen_rtx_CONST_VECTOR): Likewise.
18899 * fold-const.c (test_vec_duplicate_folding): Likewise.
18900 * genrecog.c (validate_pattern): Likewise.
18901 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
18902 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
18903 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
18904 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
18905 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
18906 * rtlanal.c (subreg_get_info): Likewise.
18907 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
18908 (vect_grouped_load_supported): Likewise.
18909 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
18910 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
18911 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18912 (simplify_const_unary_operation, simplify_binary_operation_1)
18913 (simplify_const_binary_operation, simplify_ternary_operation)
18914 (test_vector_ops_duplicate, test_vector_ops): Likewise.
18915 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
18916 instead of CONST_VECTOR_NUNITS.
18917 * varasm.c (output_constant_pool_2): Likewise.
18918 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
18919 explicit-encoded elements in the XVEC for variable-length vectors.
18921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18923 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
18925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18926 Alan Hayward <alan.hayward@arm.com>
18927 David Sherwood <david.sherwood@arm.com>
18929 * coretypes.h (fixed_size_mode): Declare.
18930 (fixed_size_mode_pod): New typedef.
18931 * builtins.h (target_builtins::x_apply_args_mode)
18932 (target_builtins::x_apply_result_mode): Change type to
18933 fixed_size_mode_pod.
18934 * builtins.c (apply_args_size, apply_result_size, result_vector)
18935 (expand_builtin_apply_args_1, expand_builtin_apply)
18936 (expand_builtin_return): Update accordingly.
18938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
18940 * cse.c (hash_rtx_cb): Hash only the encoded elements.
18941 * cselib.c (cselib_hash_rtx): Likewise.
18942 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
18943 CONST_VECTOR encoding.
18945 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18946 Jeff Law <law@redhat.com>
18949 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
18950 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
18951 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
18952 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
18955 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
18956 explicitly probe *sp in a noreturn function if there were any callee
18957 register saves or frame pointer is needed.
18959 2018-01-03 Jakub Jelinek <jakub@redhat.com>
18962 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
18963 BLKmode for ternary, binary or unary expressions.
18966 * var-tracking.c (delete_vta_debug_insn): New inline function.
18967 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
18968 insns from get_insns () to NULL instead of each bb separately.
18969 Use delete_vta_debug_insn. No longer static.
18970 (vt_debug_insns_local, variable_tracking_main_1): Adjust
18971 delete_vta_debug_insns callers.
18972 * rtl.h (delete_vta_debug_insns): Declare.
18973 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
18974 instead of variable_tracking_main.
18976 2018-01-03 Martin Sebor <msebor@redhat.com>
18978 PR tree-optimization/83603
18979 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
18980 arguments past the endof the argument list in functions declared
18981 without a prototype.
18982 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
18983 Avoid checking when arguments are null.
18985 2018-01-03 Martin Sebor <msebor@redhat.com>
18988 * doc/extend.texi (attribute const): Fix a typo.
18989 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
18990 issuing -Wsuggest-attribute for void functions.
18992 2018-01-03 Martin Sebor <msebor@redhat.com>
18994 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
18995 offset_int::from instead of wide_int::to_shwi.
18996 (maybe_diag_overlap): Remove assertion.
18997 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
18998 * gimple-ssa-sprintf.c (format_directive): Same.
18999 (parse_directive): Same.
19000 (sprintf_dom_walker::compute_format_length): Same.
19001 (try_substitute_return_value): Same.
19003 2018-01-03 Jeff Law <law@redhat.com>
19005 PR middle-end/83654
19006 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
19007 non-constant residual for zero at runtime and avoid probing in
19008 that case. Reorganize code for trailing problem to mirror handling
19011 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19013 PR tree-optimization/83501
19014 * tree-ssa-strlen.c (get_string_cst): New.
19015 (handle_char_store): Call get_string_cst.
19017 2018-01-03 Martin Liska <mliska@suse.cz>
19019 PR tree-optimization/83593
19020 * tree-ssa-strlen.c: Include tree-cfg.h.
19021 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
19022 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
19023 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
19025 (strlen_dom_walker::before_dom_children): Call
19026 gimple_purge_dead_eh_edges. Dump tranformation with details
19028 (strlen_dom_walker::before_dom_children): Update call by adding
19029 new argument cleanup_eh.
19030 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
19032 2018-01-03 Martin Liska <mliska@suse.cz>
19035 * cif-code.def (VARIADIC_THUNK): New enum value.
19036 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
19039 2018-01-03 Jan Beulich <jbeulich@suse.com>
19041 * sse.md (mov<mode>_internal): Tighten condition for when to use
19042 vmovdqu<ssescalarsize> for TI and OI modes.
19044 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19046 Update copyright years.
19048 2018-01-03 Martin Liska <mliska@suse.cz>
19051 * ipa-visibility.c (function_and_variable_visibility): Skip
19052 functions with noipa attribure.
19054 2018-01-03 Jakub Jelinek <jakub@redhat.com>
19056 * gcc.c (process_command): Update copyright notice dates.
19057 * gcov-dump.c (print_version): Ditto.
19058 * gcov.c (print_version): Ditto.
19059 * gcov-tool.c (print_version): Ditto.
19060 * gengtype.c (create_file): Ditto.
19061 * doc/cpp.texi: Bump @copying's copyright year.
19062 * doc/cppinternals.texi: Ditto.
19063 * doc/gcc.texi: Ditto.
19064 * doc/gccint.texi: Ditto.
19065 * doc/gcov.texi: Ditto.
19066 * doc/install.texi: Ditto.
19067 * doc/invoke.texi: Ditto.
19069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19071 * vector-builder.h (vector_builder::m_full_nelts): Change from
19072 unsigned int to poly_uint64.
19073 (vector_builder::full_nelts): Update prototype accordingly.
19074 (vector_builder::new_vector): Likewise.
19075 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
19076 (vector_builder::operator ==): Likewise.
19077 (vector_builder::finalize): Likewise.
19078 * int-vector-builder.h (int_vector_builder::int_vector_builder):
19079 Take the number of elements as a poly_uint64 rather than an
19081 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
19082 from unsigned int to poly_uint64.
19083 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
19084 (vec_perm_indices::new_vector): Likewise.
19085 (vec_perm_indices::length): Likewise.
19086 (vec_perm_indices::nelts_per_input): Likewise.
19087 (vec_perm_indices::input_nelts): Likewise.
19088 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
19089 number of elements per input as a poly_uint64 rather than an
19090 unsigned int. Use the original encoding for variable-length
19091 vectors, rather than clamping each individual element.
19092 For the second and subsequent elements in each pattern,
19093 clamp the step and base before clamping their sum.
19094 (vec_perm_indices::series_p): Handle polynomial element counts.
19095 (vec_perm_indices::all_in_range_p): Likewise.
19096 (vec_perm_indices_to_tree): Likewise.
19097 (vec_perm_indices_to_rtx): Likewise.
19098 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19099 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
19100 (tree_vector_builder::new_binary_operation): Handle polynomial
19101 element counts. Return false if we need to know the number
19102 of elements at compile time.
19103 * fold-const.c (fold_vec_perm): Punt if the number of elements
19104 isn't known at compile time.
19106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19108 * vec-perm-indices.h (vec_perm_builder): Change element type
19109 from HOST_WIDE_INT to poly_int64.
19110 (vec_perm_indices::element_type): Update accordingly.
19111 (vec_perm_indices::clamp): Handle polynomial element_types.
19112 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19113 (vec_perm_indices::all_in_range_p): Likewise.
19114 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
19116 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
19117 polynomial vec_perm_indices element types.
19118 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
19119 * fold-const.c (fold_vec_perm): Likewise.
19120 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
19121 * tree-vect-generic.c (lower_vec_perm): Likewise.
19122 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19123 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
19124 element type to HOST_WIDE_INT.
19126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19127 Alan Hayward <alan.hayward@arm.com>
19128 David Sherwood <david.sherwood@arm.com>
19130 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
19131 rather than an int. Use plus_constant.
19132 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
19133 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
19135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19136 Alan Hayward <alan.hayward@arm.com>
19137 David Sherwood <david.sherwood@arm.com>
19139 * calls.c (emit_call_1, expand_call): Change struct_value_size from
19140 a HOST_WIDE_INT to a poly_int64.
19142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19143 Alan Hayward <alan.hayward@arm.com>
19144 David Sherwood <david.sherwood@arm.com>
19146 * calls.c (load_register_parameters): Cope with polynomial
19147 mode sizes. Require a constant size for BLKmode parameters
19148 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
19149 forces a parameter to be padded at the lsb end in order to
19150 fill a complete number of words, require the parameter size
19151 to be ordered wrt UNITS_PER_WORD.
19153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19154 Alan Hayward <alan.hayward@arm.com>
19155 David Sherwood <david.sherwood@arm.com>
19157 * reload1.c (spill_stack_slot_width): Change element type
19158 from unsigned int to poly_uint64_pod.
19159 (alter_reg): Treat mode sizes as polynomial.
19161 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19162 Alan Hayward <alan.hayward@arm.com>
19163 David Sherwood <david.sherwood@arm.com>
19165 * reload.c (complex_word_subreg_p): New function.
19166 (reload_inner_reg_of_subreg, push_reload): Use it.
19168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19169 Alan Hayward <alan.hayward@arm.com>
19170 David Sherwood <david.sherwood@arm.com>
19172 * lra-constraints.c (process_alt_operands): Reject matched
19173 operands whose sizes aren't ordered.
19174 (match_reload): Refer to this check here.
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 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
19181 that the mode size is in the set {1, 2, 4, 8, 16}.
19183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19184 Alan Hayward <alan.hayward@arm.com>
19185 David Sherwood <david.sherwood@arm.com>
19187 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
19188 Use plus_constant instead of gen_rtx_PLUS.
19190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19191 Alan Hayward <alan.hayward@arm.com>
19192 David Sherwood <david.sherwood@arm.com>
19194 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
19195 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
19196 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
19197 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
19198 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
19199 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
19200 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
19201 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
19202 * config/i386/i386.c (ix86_push_rounding): ...this new function.
19203 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
19205 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
19206 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
19207 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
19208 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
19209 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
19210 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
19211 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
19212 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
19213 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
19214 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
19216 * expr.c (emit_move_resolve_push): Treat the input and result
19217 of PUSH_ROUNDING as a poly_int64.
19218 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
19219 (emit_push_insn): Likewise.
19220 * lra-eliminations.c (mark_not_eliminable): Likewise.
19221 * recog.c (push_operand): Likewise.
19222 * reload1.c (elimination_effects): Likewise.
19223 * rtlanal.c (nonzero_bits1): Likewise.
19224 * calls.c (store_one_arg): Likewise. Require the padding to be
19225 known at compile time.
19227 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19228 Alan Hayward <alan.hayward@arm.com>
19229 David Sherwood <david.sherwood@arm.com>
19231 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
19232 Use plus_constant instead of gen_rtx_PLUS.
19234 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19235 Alan Hayward <alan.hayward@arm.com>
19236 David Sherwood <david.sherwood@arm.com>
19238 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
19239 rather than an int.
19241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19242 Alan Hayward <alan.hayward@arm.com>
19243 David Sherwood <david.sherwood@arm.com>
19245 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
19246 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
19247 via stack temporaries. Treat the mode size as polynomial too.
19249 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19250 Alan Hayward <alan.hayward@arm.com>
19251 David Sherwood <david.sherwood@arm.com>
19253 * expr.c (expand_expr_real_2): When handling conversions involving
19254 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
19255 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
19256 as a poly_uint64 too.
19258 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19259 Alan Hayward <alan.hayward@arm.com>
19260 David Sherwood <david.sherwood@arm.com>
19262 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
19264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19265 Alan Hayward <alan.hayward@arm.com>
19266 David Sherwood <david.sherwood@arm.com>
19268 * combine.c (can_change_dest_mode): Handle polynomial
19269 REGMODE_NATURAL_SIZE.
19270 * expmed.c (store_bit_field_1): Likewise.
19271 * expr.c (store_constructor): Likewise.
19272 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
19273 and polynomial REGMODE_NATURAL_SIZE.
19274 (gen_lowpart_common): Likewise.
19275 * reginfo.c (record_subregs_of_mode): Likewise.
19276 * rtlanal.c (read_modify_subreg_p): Likewise.
19278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19279 Alan Hayward <alan.hayward@arm.com>
19280 David Sherwood <david.sherwood@arm.com>
19282 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
19283 numbers of elements.
19285 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19286 Alan Hayward <alan.hayward@arm.com>
19287 David Sherwood <david.sherwood@arm.com>
19289 * match.pd: Cope with polynomial numbers of vector elements.
19291 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19292 Alan Hayward <alan.hayward@arm.com>
19293 David Sherwood <david.sherwood@arm.com>
19295 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19296 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
19303 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
19304 (ipa_simd_modify_function_body): Likewise.
19306 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19307 Alan Hayward <alan.hayward@arm.com>
19308 David Sherwood <david.sherwood@arm.com>
19310 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
19311 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
19312 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
19313 (expand_vector_condition, vector_element): Likewise.
19314 (subparts_gt): New function.
19315 (get_compute_type): Use subparts_gt.
19316 (count_type_subparts): Delete.
19317 (expand_vector_operations_1): Use subparts_gt instead of
19318 count_type_subparts.
19320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19321 Alan Hayward <alan.hayward@arm.com>
19322 David Sherwood <david.sherwood@arm.com>
19324 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
19325 (vect_compile_time_alias): ...this new function. Do the calculation
19326 on poly_ints rather than trees.
19327 (vect_prune_runtime_alias_test_list): Update call accordingly.
19329 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19330 Alan Hayward <alan.hayward@arm.com>
19331 David Sherwood <david.sherwood@arm.com>
19333 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
19335 (vect_schedule_slp_instance): Likewise.
19337 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19338 Alan Hayward <alan.hayward@arm.com>
19339 David Sherwood <david.sherwood@arm.com>
19341 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
19342 constant and extern definitions for variable-length vectors.
19343 (vect_get_constant_vectors): Note that the number of units
19344 is known to be constant.
19346 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19347 Alan Hayward <alan.hayward@arm.com>
19348 David Sherwood <david.sherwood@arm.com>
19350 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
19351 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
19359 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
19361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19362 Alan Hayward <alan.hayward@arm.com>
19363 David Sherwood <david.sherwood@arm.com>
19365 * tree-vect-stmts.c (vectorizable_call): Treat the number of
19366 vectors as polynomial. Use build_index_vector for
19367 IFN_GOMP_SIMD_LANE.
19369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19370 Alan Hayward <alan.hayward@arm.com>
19371 David Sherwood <david.sherwood@arm.com>
19373 * tree-vect-stmts.c (get_load_store_type): Treat the number of
19374 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
19375 for variable-length vectors.
19376 (vectorizable_mask_load_store): Treat the number of units as
19377 polynomial, asserting that it is constant if the condition has
19378 already been enforced.
19379 (vectorizable_store, vectorizable_load): Likewise.
19381 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19382 Alan Hayward <alan.hayward@arm.com>
19383 David Sherwood <david.sherwood@arm.com>
19385 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
19386 of units as polynomial. Punt if we can't tell at compile time
19387 which vector contains the final result.
19389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19390 Alan Hayward <alan.hayward@arm.com>
19391 David Sherwood <david.sherwood@arm.com>
19393 * tree-vect-loop.c (vectorizable_induction): Treat the number
19394 of units as polynomial. Punt on SLP inductions. Use an integer
19395 VEC_SERIES_EXPR for variable-length integer reductions. Use a
19396 cast of such a series for variable-length floating-point
19399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19400 Alan Hayward <alan.hayward@arm.com>
19401 David Sherwood <david.sherwood@arm.com>
19403 * tree.h (build_index_vector): Declare.
19404 * tree.c (build_index_vector): New function.
19405 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
19406 of units as polynomial, forcibly converting it to a constant if
19407 vectorizable_reduction has already enforced the condition.
19408 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
19409 to create a {1,2,3,...} vector.
19410 (vectorizable_reduction): Treat the number of units as polynomial.
19411 Choose vectype_in based on the largest scalar element size rather
19412 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
19420 number of units as polynomial.
19422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19423 Alan Hayward <alan.hayward@arm.com>
19424 David Sherwood <david.sherwood@arm.com>
19426 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
19427 * target.def (autovectorize_vector_sizes): Return the vector sizes
19428 by pointer, using vector_sizes rather than a bitmask.
19429 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
19430 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
19431 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
19433 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
19434 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
19435 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
19436 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
19437 * omp-general.c (omp_max_vf): Likewise.
19438 * omp-low.c (omp_clause_aligned_alignment): Likewise.
19439 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
19440 * tree-vect-loop.c (vect_analyze_loop): Likewise.
19441 * tree-vect-slp.c (vect_slp_bb): Likewise.
19442 * doc/tm.texi: Regenerate.
19443 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
19445 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
19446 the vector size as a poly_uint64 rather than an unsigned int.
19447 (current_vector_size): Change from an unsigned int to a poly_uint64.
19448 (get_vectype_for_scalar_type): Update accordingly.
19449 * tree.h (build_truth_vector_type): Take the size and number of
19450 units as a poly_uint64 rather than an unsigned int.
19451 (build_vector_type): Add a temporary overload that takes
19452 the number of units as a poly_uint64 rather than an unsigned int.
19453 * tree.c (make_vector_type): Likewise.
19454 (build_truth_vector_type): Take the number of units as a poly_uint64
19455 rather than an unsigned int.
19457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19458 Alan Hayward <alan.hayward@arm.com>
19459 David Sherwood <david.sherwood@arm.com>
19461 * target.def (get_mask_mode): Take the number of units and length
19462 as poly_uint64s rather than unsigned ints.
19463 * targhooks.h (default_get_mask_mode): Update accordingly.
19464 * targhooks.c (default_get_mask_mode): Likewise.
19465 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19466 * doc/tm.texi: Regenerate.
19468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19469 Alan Hayward <alan.hayward@arm.com>
19470 David Sherwood <david.sherwood@arm.com>
19472 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
19473 * omp-general.c (omp_max_vf): Likewise.
19474 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
19475 (expand_omp_simd): Handle polynomial safelen.
19476 * omp-low.c (omplow_simd_context): Add a default constructor.
19477 (omplow_simd_context::max_vf): Change from int to poly_uint64.
19478 (lower_rec_simd_input_clauses): Update accordingly.
19479 (lower_rec_input_clauses): Likewise.
19481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19482 Alan Hayward <alan.hayward@arm.com>
19483 David Sherwood <david.sherwood@arm.com>
19485 * tree-vectorizer.h (vect_nunits_for_cost): New function.
19486 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
19487 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
19488 (vect_analyze_slp_cost): Likewise.
19489 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
19490 (vect_model_load_cost): Likewise.
19492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19493 Alan Hayward <alan.hayward@arm.com>
19494 David Sherwood <david.sherwood@arm.com>
19496 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
19497 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
19498 from an unsigned int * to a poly_uint64_pod *.
19499 (calculate_unrolling_factor): New function.
19500 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
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 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
19507 from an unsigned int to a poly_uint64.
19508 (_loop_vec_info::slp_unrolling_factor): Likewise.
19509 (_loop_vec_info::vectorization_factor): Change from an int
19511 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
19512 (vect_get_num_vectors): New function.
19513 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
19514 (vect_get_num_copies): Use vect_get_num_vectors.
19515 (vect_analyze_data_ref_dependences): Change max_vf from an int *
19516 to an unsigned int *.
19517 (vect_analyze_data_refs): Change min_vf from an int * to a
19519 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19520 than an unsigned HOST_WIDE_INT.
19521 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
19522 (vect_analyze_data_ref_dependence): Change max_vf from an int *
19523 to an unsigned int *.
19524 (vect_analyze_data_ref_dependences): Likewise.
19525 (vect_compute_data_ref_alignment): Handle polynomial vf.
19526 (vect_enhance_data_refs_alignment): Likewise.
19527 (vect_prune_runtime_alias_test_list): Likewise.
19528 (vect_shift_permute_load_chain): Likewise.
19529 (vect_supportable_dr_alignment): Likewise.
19530 (dependence_distance_ge_vf): Take the vectorization factor as a
19531 poly_uint64 rather than an unsigned HOST_WIDE_INT.
19532 (vect_analyze_data_refs): Change min_vf from an int * to a
19534 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
19535 vfm1 as a poly_uint64 rather than an int. Make the same change
19536 for the returned bound_scalar.
19537 (vect_gen_vector_loop_niters): Handle polynomial vf.
19538 (vect_do_peeling): Likewise. Update call to
19539 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
19540 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
19542 * tree-vect-loop.c (vect_determine_vectorization_factor)
19543 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
19544 (vect_get_known_peeling_cost): Likewise.
19545 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
19546 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
19547 (vect_transform_loop): Likewise. Use the lowest possible VF when
19548 updating the upper bounds of the loop.
19549 (vect_min_worthwhile_factor): Make static. Return an unsigned int
19550 rather than an int.
19551 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
19552 polynomial unroll factors.
19553 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
19554 (vect_make_slp_decision): Likewise.
19555 (vect_supported_load_permutation_p): Likewise, and polynomial
19557 (vect_analyze_slp_cost): Handle polynomial vf.
19558 (vect_slp_analyze_node_operations): Likewise.
19559 (vect_slp_analyze_bb_1): Likewise.
19560 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
19561 than an unsigned HOST_WIDE_INT.
19562 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
19563 (vectorizable_load): Handle polynomial vf.
19564 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
19566 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
19568 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19569 Alan Hayward <alan.hayward@arm.com>
19570 David Sherwood <david.sherwood@arm.com>
19572 * match.pd: Handle bit operations involving three constants
19573 and try to fold one pair.
19575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
19577 * tree-vect-loop-manip.c: Include gimple-fold.h.
19578 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
19579 niters_maybe_zero parameters. Handle other cases besides a step of 1.
19580 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
19581 Add a path that uses a step of VF instead of 1, but disable it
19583 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
19584 and niters_no_overflow parameters. Update calls to
19585 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
19586 Create a new SSA name if the latter choses to use a ste other
19587 than zero, and return it via niters_vector_mult_vf_var.
19588 * tree-vect-loop.c (vect_transform_loop): Update calls to
19589 vect_do_peeling, vect_gen_vector_loop_niters and
19590 slpeel_make_loop_iterate_ntimes.
19591 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
19592 (vect_gen_vector_loop_niters): Update declarations after above changes.
19594 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
19596 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
19597 128-bit round to integer instructions.
19598 (ceil<mode>2): Likewise.
19599 (btrunc<mode>2): Likewise.
19600 (round<mode>2): Likewise.
19602 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
19604 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
19605 unaligned VSX load/store on P8/P9.
19606 (expand_block_clear): Allow the use of unaligned VSX
19607 load/store on P8/P9.
19609 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19611 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
19613 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
19614 swap associated with both a load and a store.
19616 2018-01-02 Andrew Waterman <andrew@sifive.com>
19618 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
19619 * config/riscv/riscv.md (clear_cache): Use it.
19621 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
19623 * web.c: Remove out-of-date comment.
19625 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19627 * expr.c (fixup_args_size_notes): Check that any existing
19628 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
19629 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
19630 (emit_single_push_insn): ...here.
19632 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19634 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
19635 (const_vector_encoded_nelts): New function.
19636 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
19637 (const_vector_int_elt, const_vector_elt): Declare.
19638 * emit-rtl.c (const_vector_int_elt_1): New function.
19639 (const_vector_elt): Likewise.
19640 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
19641 of CONST_VECTOR_ELT.
19643 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19645 * expr.c: Include rtx-vector-builder.h.
19646 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
19647 directly on the tree encoding.
19648 (const_vector_from_tree): Likewise.
19649 * optabs.c: Include rtx-vector-builder.h.
19650 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
19651 sequence of "u" values.
19652 * vec-perm-indices.c: Include rtx-vector-builder.h.
19653 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
19654 directly on the vec_perm_indices encoding.
19656 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19658 * doc/rtl.texi (const_vector): Describe new encoding scheme.
19659 * Makefile.in (OBJS): Add rtx-vector-builder.o.
19660 * rtx-vector-builder.h: New file.
19661 * rtx-vector-builder.c: Likewise.
19662 * rtl.h (rtx_def::u2): Add a const_vector field.
19663 (CONST_VECTOR_NPATTERNS): New macro.
19664 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
19665 (CONST_VECTOR_DUPLICATE_P): Likewise.
19666 (CONST_VECTOR_STEPPED_P): Likewise.
19667 (CONST_VECTOR_ENCODED_ELT): Likewise.
19668 (const_vec_duplicate_p): Check for a duplicated vector encoding.
19669 (unwrap_const_vec_duplicate): Likewise.
19670 (const_vec_series_p): Check for a non-duplicated vector encoding.
19671 Say that the function only returns true for integer vectors.
19672 * emit-rtl.c: Include rtx-vector-builder.h.
19673 (gen_const_vec_duplicate_1): Delete.
19674 (gen_const_vector): Call gen_const_vec_duplicate instead of
19675 gen_const_vec_duplicate_1.
19676 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
19677 (gen_const_vec_duplicate): Use rtx_vector_builder.
19678 (gen_const_vec_series): Likewise.
19679 (gen_rtx_CONST_VECTOR): Likewise.
19680 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
19681 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19682 Build a new vector rather than modifying a CONST_VECTOR in-place.
19683 (handle_special_swappables): Update call accordingly.
19684 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
19685 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
19686 Build a new vector rather than modifying a CONST_VECTOR in-place.
19687 (handle_special_swappables): Update call accordingly.
19689 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19691 * simplify-rtx.c (simplify_const_binary_operation): Use
19692 CONST_VECTOR_ELT instead of XVECEXP.
19694 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19696 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
19697 the selector elements to be different from the data elements
19698 if the selector is a VECTOR_CST.
19699 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
19700 ssizetype for the selector.
19702 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19704 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
19705 before testing each element individually.
19706 * tree-vect-generic.c (lower_vec_perm): Likewise.
19708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19710 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
19711 * selftest-run-tests.c (selftest::run_tests): Call it.
19712 * vector-builder.h (vector_builder::operator ==): New function.
19713 (vector_builder::operator !=): Likewise.
19714 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
19715 (vec_perm_indices::all_from_input_p): New function.
19716 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
19717 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
19718 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
19719 instead of reading the VECTOR_CST directly. Detect whether both
19720 vector inputs are the same before constructing the vec_perm_indices,
19721 and update the number of inputs argument accordingly. Use the
19722 utility functions added above. Only construct sel2 if we need to.
19724 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19726 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
19727 the broadcast of the low byte.
19728 (expand_mult_highpart): Use an explicit encoding for the permutes.
19729 * optabs-query.c (can_mult_highpart_p): Likewise.
19730 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
19731 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19732 (vectorizable_bswap): Likewise.
19733 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
19734 explicit encoding for the power-of-2 permutes.
19735 (vect_permute_store_chain): Likewise.
19736 (vect_grouped_load_supported): Likewise.
19737 (vect_permute_load_chain): Likewise.
19739 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19741 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
19742 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
19743 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
19744 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19745 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
19746 (vect_gen_perm_mask_any): Likewise.
19748 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19750 * int-vector-builder.h: New file.
19751 * vec-perm-indices.h: Include int-vector-builder.h.
19752 (vec_perm_indices): Redefine as an int_vector_builder.
19753 (auto_vec_perm_indices): Delete.
19754 (vec_perm_builder): Redefine as a stand-alone class.
19755 (vec_perm_indices::vec_perm_indices): New function.
19756 (vec_perm_indices::clamp): Likewise.
19757 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
19758 (vec_perm_indices::new_vector): New function.
19759 (vec_perm_indices::new_expanded_vector): Update for new
19760 vec_perm_indices class.
19761 (vec_perm_indices::rotate_inputs): New function.
19762 (vec_perm_indices::all_in_range_p): Operate directly on the
19763 encoded form, without computing elided elements.
19764 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
19765 encoding. Update for new vec_perm_indices class.
19766 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
19767 the given vec_perm_builder.
19768 (expand_vec_perm_var): Update vec_perm_builder constructor.
19769 (expand_mult_highpart): Use vec_perm_builder instead of
19770 auto_vec_perm_indices.
19771 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
19772 vec_perm_indices instead of auto_vec_perm_indices. Use a single
19773 or double series encoding as appropriate.
19774 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
19775 vec_perm_indices instead of auto_vec_perm_indices.
19776 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19777 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19778 (vect_permute_store_chain): Likewise.
19779 (vect_grouped_load_supported): Likewise.
19780 (vect_permute_load_chain): Likewise.
19781 (vect_shift_permute_load_chain): Likewise.
19782 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19783 (vect_transform_slp_perm_load): Likewise.
19784 (vect_schedule_slp_instance): Likewise.
19785 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19786 (vectorizable_mask_load_store): Likewise.
19787 (vectorizable_bswap): Likewise.
19788 (vectorizable_store): Likewise.
19789 (vectorizable_load): Likewise.
19790 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
19791 vec_perm_indices instead of auto_vec_perm_indices. Use
19792 tree_to_vec_perm_builder to read the vector from a tree.
19793 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
19794 vec_perm_builder instead of a vec_perm_indices.
19795 (have_whole_vector_shift): Use vec_perm_builder and
19796 vec_perm_indices instead of auto_vec_perm_indices. Leave the
19797 truncation to calc_vec_perm_mask_for_shift.
19798 (vect_create_epilog_for_reduction): Likewise.
19799 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
19800 from auto_vec_perm_indices to vec_perm_indices.
19801 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19802 instead of changing individual elements.
19803 (aarch64_vectorize_vec_perm_const): Use new_vector to install
19804 the vector in d.perm.
19805 * config/arm/arm.c (expand_vec_perm_d::perm): Change
19806 from auto_vec_perm_indices to vec_perm_indices.
19807 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
19808 instead of changing individual elements.
19809 (arm_vectorize_vec_perm_const): Use new_vector to install
19810 the vector in d.perm.
19811 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
19812 Update vec_perm_builder constructor.
19813 (rs6000_expand_interleave): Likewise.
19814 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
19815 (rs6000_expand_interleave): Likewise.
19817 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19819 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
19820 to qimode could truncate the indices.
19821 * optabs.c (expand_vec_perm_var): Likewise.
19823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19825 * Makefile.in (OBJS): Add vec-perm-indices.o.
19826 * vec-perm-indices.h: New file.
19827 * vec-perm-indices.c: Likewise.
19828 * target.h (vec_perm_indices): Replace with a forward class
19830 (auto_vec_perm_indices): Move to vec-perm-indices.h.
19831 * optabs.h: Include vec-perm-indices.h.
19832 (expand_vec_perm): Delete.
19833 (selector_fits_mode_p, expand_vec_perm_var): Declare.
19834 (expand_vec_perm_const): Declare.
19835 * target.def (vec_perm_const_ok): Replace with...
19836 (vec_perm_const): ...this new hook.
19837 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
19838 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
19839 * doc/tm.texi: Regenerate.
19840 * optabs.def (vec_perm_const): Delete.
19841 * doc/md.texi (vec_perm_const): Likewise.
19842 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
19843 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
19844 expand_vec_perm for constant permutation vectors. Assert that
19845 the mode of variable permutation vectors is the integer equivalent
19846 of the mode that is being permuted.
19847 * optabs-query.h (selector_fits_mode_p): Declare.
19848 * optabs-query.c: Include vec-perm-indices.h.
19849 (selector_fits_mode_p): New function.
19850 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
19851 is defined, instead of checking whether the vec_perm_const_optab
19852 exists. Use targetm.vectorize.vec_perm_const instead of
19853 targetm.vectorize.vec_perm_const_ok. Check whether the indices
19854 fit in the vector mode before using a variable permute.
19855 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
19856 vec_perm_indices instead of an rtx.
19857 (expand_vec_perm): Replace with...
19858 (expand_vec_perm_const): ...this new function. Take the selector
19859 as a vec_perm_indices rather than an rtx. Also take the mode of
19860 the selector. Update call to shift_amt_for_vec_perm_mask.
19861 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
19862 Use vec_perm_indices::new_expanded_vector to expand the original
19863 selector into bytes. Check whether the indices fit in the vector
19864 mode before using a variable permute.
19865 (expand_vec_perm_var): Make global.
19866 (expand_mult_highpart): Use expand_vec_perm_const.
19867 * fold-const.c: Includes vec-perm-indices.h.
19868 * tree-ssa-forwprop.c: Likewise.
19869 * tree-vect-data-refs.c: Likewise.
19870 * tree-vect-generic.c: Likewise.
19871 * tree-vect-loop.c: Likewise.
19872 * tree-vect-slp.c: Likewise.
19873 * tree-vect-stmts.c: Likewise.
19874 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
19876 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
19877 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
19878 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
19879 (aarch64_vectorize_vec_perm_const): ...this new function.
19880 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19881 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19882 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
19883 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
19884 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19885 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19886 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
19888 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
19889 check for NEON modes.
19890 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
19891 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
19892 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
19893 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
19895 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
19896 the old VEC_PERM_CONST conditions.
19897 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
19898 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
19899 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
19900 (ia64_vectorize_vec_perm_const_ok): Merge into...
19901 (ia64_vectorize_vec_perm_const): ...this new function.
19902 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
19903 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
19904 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
19905 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
19906 * config/mips/mips.c (mips_expand_vec_perm_const)
19907 (mips_vectorize_vec_perm_const_ok): Merge into...
19908 (mips_vectorize_vec_perm_const): ...this new function.
19909 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
19910 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
19911 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
19912 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
19913 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
19914 (rs6000_expand_vec_perm_const): Delete.
19915 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
19917 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19918 (altivec_expand_vec_perm_const_le): Take each operand individually.
19919 Operate on constant selectors rather than rtxes.
19920 (altivec_expand_vec_perm_const): Likewise. Update call to
19921 altivec_expand_vec_perm_const_le.
19922 (rs6000_expand_vec_perm_const): Delete.
19923 (rs6000_vectorize_vec_perm_const_ok): Delete.
19924 (rs6000_vectorize_vec_perm_const): New function.
19925 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19926 an element count and rtx array.
19927 (rs6000_expand_extract_even): Update call accordingly.
19928 (rs6000_expand_interleave): Likewise.
19929 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
19930 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
19931 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
19932 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
19933 (rs6000_expand_vec_perm_const): Delete.
19934 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
19935 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19936 (altivec_expand_vec_perm_const_le): Take each operand individually.
19937 Operate on constant selectors rather than rtxes.
19938 (altivec_expand_vec_perm_const): Likewise. Update call to
19939 altivec_expand_vec_perm_const_le.
19940 (rs6000_expand_vec_perm_const): Delete.
19941 (rs6000_vectorize_vec_perm_const_ok): Delete.
19942 (rs6000_vectorize_vec_perm_const): New function. Remove stray
19943 reference to the SPE evmerge intructions.
19944 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
19945 an element count and rtx array.
19946 (rs6000_expand_extract_even): Update call accordingly.
19947 (rs6000_expand_interleave): Likewise.
19948 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
19949 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
19951 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
19953 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19955 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
19956 vector mode and that that mode matches the mode of the data
19958 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
19959 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
19960 directly using expand_vec_perm_1 when forcing selectors into
19962 (expand_vec_perm_var): New function, split out from expand_vec_perm.
19964 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19966 * optabs-query.h (can_vec_perm_p): Delete.
19967 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
19968 * optabs-query.c (can_vec_perm_p): Split into...
19969 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
19970 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
19971 particular selector is valid.
19972 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
19973 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
19974 (vect_grouped_load_supported): Likewise.
19975 (vect_shift_permute_load_chain): Likewise.
19976 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
19977 (vect_transform_slp_perm_load): Likewise.
19978 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
19979 (vectorizable_bswap): Likewise.
19980 (vect_gen_perm_mask_checked): Likewise.
19981 * fold-const.c (fold_ternary_loc): Likewise. Don't take
19982 implementations of variable permutation vectors into account
19983 when deciding which selector to use.
19984 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
19985 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
19986 with a false third argument.
19987 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
19988 to test whether the constant selector is valid and can_vec_perm_var_p
19989 to test whether a variable selector is valid.
19991 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
19993 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
19994 * optabs-query.c (can_vec_perm_p): Likewise.
19995 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
19996 instead of vec_perm_indices.
19997 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
19998 (vect_gen_perm_mask_checked): Likewise,
19999 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
20000 (vect_gen_perm_mask_checked): Likewise,
20002 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
20004 * optabs-query.h (qimode_for_vec_perm): Declare.
20005 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
20006 (qimode_for_vec_perm): ...this new function.
20007 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
20009 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
20011 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
20012 does not have a conditional at the top.
20014 2018-01-02 Richard Biener <rguenther@suse.de>
20016 * ipa-inline.c (big_speedup_p): Fix expression.
20018 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20021 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
20024 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
20028 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
20029 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
20030 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
20031 cond_taken_branch_cost 3->4.
20033 2018-01-01 Jakub Jelinek <jakub@redhat.com>
20035 PR tree-optimization/83581
20036 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
20037 TODO_cleanup_cfg if any changes have been made.
20039 PR middle-end/83608
20040 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
20041 convert_modes if target mode has the right side, but different mode
20044 PR middle-end/83609
20045 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
20046 last argument when extracting from CONCAT. If either from_real or
20047 from_imag is NULL, use expansion through memory. If result is not
20048 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
20049 the parts directly to inner mode, if even that fails, use expansion
20052 PR middle-end/83623
20053 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
20054 check for bswap in mode rather than HImode and use that in expand_unop
20057 Copyright (C) 2018 Free Software Foundation, Inc.
20059 Copying and distribution of this file, with or without modification,
20060 are permitted in any medium without royalty provided the copyright
20061 notice and this notice are preserved.