1 2018-10-03 Martin Liska <mliska@suse.cz>
4 * coverage.c (coverage_begin_function): Do not
5 mark lambdas as artificial.
6 * tree-core.h (struct GTY): Remove tm_clone_flag
7 and introduce new lambda_function.
8 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
10 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
13 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
14 P8_VECTOR and VSX are enabled.
16 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
18 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
19 0x3907 as CPU model number.
21 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
23 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
24 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
25 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
26 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
27 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
28 * config/s390/s390.md: Likewise. Rename also the cpu attribute
29 value from arch12 to z14.
31 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
33 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
35 (isinf<mode>2): Ditto.
37 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
39 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
40 before emitting fxam. Perform calculations in XFmode.
42 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
44 * match.pd (((X /[ex] A) +- B) * A): New transformation.
46 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
49 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
50 * tree.c (signed_or_unsigned_type_for): Handle complex.
52 2018-10-02 Jeff Law <law@redhat.com>
54 * gimple-fold.c (get_range_strlen): Remove dead code.
56 2018-10-02 Martin Sebor <msebor@redhat.com>
57 Jeff Law <law@redhat.com>
59 * builtins.c (unterminated_array): Add new arguments.
60 If argument is not terminated, bubble up size and exact
62 (expand_builtin_strnlen): Detect, avoid expanding
63 and diagnose unterminated arrays.
64 (c_strlen): Fill in offset of start of unterminated strings.
65 * builtins.h (unterminated_array): Update prototype.
67 2018-10-02 Richard Biener <rguenther@suse.de>
69 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
70 of haddv4df, first reduce to SSE width and exploit the fact
71 that we only need element zero with the reduction result.
72 (reduc_plus_scal_v2df): Likewise.
74 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
76 * dojump.h (do_jump): Delete.
77 (do_jump_1): Likewise.
78 (split_comparison): Move around.
79 * dojump.c (do_jump): Make static.
80 (do_jump_1): Likewise.
81 (jumpifnot): Move around.
82 (jumpifnot_1): Likewise.
85 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
87 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
89 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
90 insns in the delay slot and add_insn_after for the jump insn.
92 2018-10-02 Richard Biener <rguenther@suse.de>
94 * tree-inline.c (expand_call_inline): Use the location of
95 the callee declaration for the inline-entry marker.
96 * final.c (notice_source_line): Remove special-casing of
97 NOTE_INSN_INLINE_ENTRY.
99 2018-10-01 Carl Love <cel@us.ibm.com>
102 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
103 (__builtin_mtfsb0): New.
104 (__builtin_mtfsb1): New.
105 ( __builtin_set_fpscr_rn): New.
106 (__builtin_set_fpscr_drn): New.
107 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
108 (rs6000_expand_set_fpscr_rn_builtin): Add.
109 (rs6000_expand_set_fpscr_drn_builtin): Add.
110 (rs6000_expand_builtin): Add case statement entries for
111 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
112 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
113 RS6000_BUILTIN_MFFSL.
114 (rs6000_init_builtins): Add ftype initialization and def_builtin
115 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
116 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
117 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
118 rs6000_mffscdrn): Add define_insn.
119 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
120 * doc/extend.texi: Add documentation for the builtins.
122 2018-10-01 Richard Biener <rguenther@suse.de>
124 PR tree-optimization/87465
125 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
126 causing branch miscounts.
128 2018-10-01 Tamar Christina <tamar.christina@arm.com>
130 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
131 aarch64_option_default_param): New.
133 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
134 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
135 stack-clash protection validation code.
137 2018-10-01 Tamar Christina <tamar.christina@arm.com>
139 * params.c (validate_param): New.
140 (add_params): Use it.
141 (set_param_value): Refactor param validation into validate_param.
142 (diagnostic.h): Include.
143 * diagnostic.h (diagnostic_ready_p): New.
145 2018-10-01 Tamar Christina <tamar.christina@arm.com>
147 * params.c (set_param_value):
148 Add index of parameter being validated.
149 * common/common-target.def (option_validate_param): New.
150 * common/common-targhooks.h (default_option_validate_param): New.
151 * common/common-targhooks.c (default_option_validate_param): New.
152 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
153 * doc/tm.texi: Regenerate.
155 2018-10-01 Tamar Christina <tamar.christina@arm.com>
158 * config/aarch64/aarch64.c (aarch64_override_options_internal):
159 Add validation for stack-clash parameters and set defaults.
161 2018-10-01 Tamar Christina <tamar.christina@arm.com>
164 * configure.ac: Add stack-clash-protection-guard-size.
165 * doc/install.texi: Document it.
166 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
167 * params.def: Update comment for guard-size.
168 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
169 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
170 * configure: Regenerate.
172 2018-10-01 Tamar Christina <tamar.christina@arm.com>
175 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
176 STACK_DYNAMIC_OFFSET): New.
177 * config/aarch64/aarch64.c (aarch64_layout_frame):
178 Update outgoing args size.
179 (aarch64_stack_clash_protection_alloca_probe_range,
180 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
182 2018-10-01 Tamar Christina <tamar.christina@arm.com>
185 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
187 * target.def (stack_clash_protection_alloca_probe_range): New.
188 (stack_clash_protection_final_dynamic_probe): Remove.
189 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
190 (default_stack_clash_protection_final_dynamic_probe): Remove.
191 * targhooks.c: Likewise.
192 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
193 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
194 * doc/tm.texi: Regenerate.
196 2018-10-01 Tamar Christina <tamar.christina@arm.com>
199 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
200 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
201 aarch64_clamp_to_uimm12_shift): New.
202 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
203 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
205 2018-10-01 Tamar Christina <tamar.christina@arm.com>
208 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
210 2018-10-01 Jeff Law <law@redhat.com>
211 Richard Sandiford <richard.sandiford@linaro.org>
212 Tamar Christina <tamar.christina@arm.com>
215 * config/aarch64/aarch64.md
216 (probe_stack_range): Add k (SP) constraint.
217 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
218 STACK_CLASH_MAX_UNROLL_PAGES): New.
219 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
220 stack probes for stack clash.
221 (aarch64_allocate_and_probe_stack_space): New.
222 (aarch64_expand_prologue): Use it.
223 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
224 (aarch64_sub_sp): Add emit_move_imm optional param.
226 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
228 PR tree-optimization/87261
229 * match.pd: Remove trailing whitespace.
230 Add (x & y) | ~(x | y) -> ~(x ^ y),
231 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
233 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
235 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
237 (ashlsi3_insn): Update instruction constraints.
238 (ashrsi3_insn): Likewise.
240 (add_shift): Likewise.
241 * config/arc/constraints.md (Csz): New 32 bit constraint. It
242 avoids placing in the limm field small constants which, otherwise,
243 could end into a small instruction.
245 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
247 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
248 destination register is not odd-even.
249 (umaddsidi4_split): Likewise.
251 2018-10-01 Richard Biener <rguenther@suse.de>
253 * tree-inline.c (expand_call_inline): Store origin of fn
254 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
255 * tree.c (block_ultimate_origin): Simplify and do some
258 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
260 * config/i386/mmx.md (EMMS): New int iterator.
261 (emms): New int attribute.
262 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
263 EMMS int iterator. Explicitly declare clobbers.
264 (mmx_emms): Remove expander.
266 * config/i386/predicates.md (emms_operation): Remove predicate.
267 (vzeroall_pattern): New predicate.
268 (vzeroupper_pattern): Rename from vzeroupper_operation.
269 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
270 vzeroupper_pattern and vzeroall_pattern predicates.
272 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
274 PR rtl-optimization/86939
275 * ira-lives.c (make_hard_regno_born): Rename from this...
276 (make_hard_regno_live): ... to this. Remove update to conflict
277 information. Update function comment.
278 (make_hard_regno_dead): Add conflict information update. Update
280 (make_object_born): Rename from this...
281 (make_object_live): ... to this. Remove update to conflict information.
282 Update function comment.
283 (make_object_dead): Add conflict information update. Update function
285 (mark_pseudo_regno_live): Call make_object_live.
286 (mark_pseudo_regno_subword_live): Likewise.
287 (mark_hard_reg_dead): Update function comment.
288 (mark_hard_reg_live): Call make_hard_regno_live.
289 (process_bb_node_lives): Likewise.
290 * lra-lives.c (make_hard_regno_born): Rename from this...
291 (make_hard_regno_live): ... to this. Remove update to conflict
292 information. Remove now uneeded check_pic_pseudo_p argument.
293 Update function comment.
294 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
295 to conflict information. Update function comment.
296 (mark_pseudo_live): Remove update to conflict information. Update
298 (mark_pseudo_dead): Add conflict information update.
299 (mark_regno_live): Call make_hard_regno_live.
300 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
301 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
303 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
306 * config/i386/i386.c (construct_container): Use TImode for
307 BLKmode values in 2 integer registers.
309 2018-09-29 Jeff Law <law@redhat.com>
311 * builtins.c (unterminated_array): Pass in c_strlen_data * to
312 c_strlen rather than just a tree *.
313 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
314 Update recursive calls appropriately. If caller did not provide a
315 suitable data pointer, create a local one. When a non-terminated
316 string is discovered, bubble up information about the string via the
317 c_strlen_data object.
318 * builtins.h (c_strlen): Update prototype.
319 (c_strlen_data): New structure.
320 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
321 For a type 2 call, if c_strlen indicates a non-terminated string
322 use the length of the non-terminated string.
323 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
325 2018-09-29 Jakub Jelinek <jakub@redhat.com>
328 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
329 __m512d type for __A argument rather than __m512.
331 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
333 * match.pd (simple_comparison): Don't optimize if either operand is
334 a function pointer when target needs function pointer canonicalization.
336 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
338 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
339 power5 .. power9 to remove indirection.
340 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
341 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
342 ASM_CPU_476_SPEC): Delete.
343 (ASM_CPU_SPEC): Adjust.
344 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
345 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
347 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
349 * config.in: Delete HAVE_AS_DCI.
350 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
351 * config/rs6000/rs6000.h: Ditto.
352 * configure.ac: Delete HAVE_AS_DCI.
353 * configure: Regenerate.
355 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
357 * config.in (HAVE_AS_LWSYNC): Delete.
358 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
359 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
361 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
362 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
364 * configure.ac: Delete HAVE_AS_LWSYNC.
365 * configure: Regenerate.
367 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
368 Pierre-Marie de Rodat <derodat@adacore.com>
370 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
371 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
372 (cgraph_node::create_thunk): Add indirect_offset parameter.
373 (thunk_adjust): Likewise.
374 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
375 and initialize the corresponding field with it.
376 (cgraph_node::dump): Dump indirect_offset field.
377 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
378 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
379 (thunk_adjust): Add indirect_offset parameter and deal with it.
380 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
381 pass it to thunk_adjust. Do not call the target hook if it's non-zero
382 or if the thunk is external or local. Fix formatting. Do not chain
383 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
384 if any, in the GIMPLE representation.
385 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
386 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
387 (input_node): Read indirect_offset field.
388 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
389 call to thunk_adjust.
390 * tree-nested.c (struct nesting_info): Add thunk_p field.
391 (create_nesting_tree): Set it.
392 (convert_all_function_calls): Copy static chain from targets to thunks.
393 (finalize_nesting_tree_1): Return early for thunks.
394 (unnest_nesting_tree_1): Do not finalize thunks.
395 (gimplify_all_functions): Do not gimplify thunks.
397 2018-09-28 David Malcolm <dmalcolm@redhat.com>
399 * opt-suggestions.c (option_proposer::build_option_suggestions):
400 Release "option_values".
402 2018-09-28 David Malcolm <dmalcolm@redhat.com>
404 * coverage.c (get_coverage_counts): Convert problem-reporting dump
405 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
406 * dumpfile.c (kind_as_string): New function.
407 (dump_loc): Rather than a hardcoded prefix of "note: ", use
408 kind_as_string to vary the prefix based on dump_kind.
409 (selftest::test_capture_of_dump_calls): Update for above.
411 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
413 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
414 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
416 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
418 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
419 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
420 INVALID_REGNUM instead of FPSR_REG.
421 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
422 * config/i386/i386.md: Update comment of FP compares.
423 (fldenv): Do not clobber FPSR_REG.
425 2018-09-28 Richard Biener <rguenther@suse.de>
427 * tree.h (BLOCK_ORIGIN): New.
428 * omp-expand.c (grid_expand_target_grid_body): Assign
429 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
430 * tree-inline.c (remap_block): Likewise.
431 * auto-profile.c (get_function_decl_from_block): Simplify
432 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
433 * langhooks.c (lhd_print_error_function): Likewise.
434 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
436 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
437 * tree.c (block_nonartificial_location): Likewise.
438 (block_ultimate_origin): Likewise.
439 * tree-pretty-print.c (percent_K_format): Likewise. Remove
440 no longer needed LTO case.
442 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
443 Jan Hubicka <jh@suse.cz>
444 Martin Jambor <mjambor@suse.cz>
446 * simplify-rtx.c (simplify_merge_mask): New function.
447 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
448 same masks are used in op1 or op2.
449 (test_vec_merge): New function.
450 (test_vector_ops): Call test_vec_merge.
452 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
454 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
455 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
456 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
457 (cypress_costs): Set it.
458 (supersparc_costs): Likewise.
459 (hypersparc_costs): Likewise.
460 (leon_cost): Likewise.
461 (leon3_costs): Likewise.
462 (sparclet_costs): Likewise.
463 (ultrasparc_costs): Likewise.
464 (ultrasparc_costs): Likewise.
465 (niagara_costs): Likewise.
466 (niagara2_costs): Likewise.
467 (niagara3_costs): Likewise.
468 (niagara4_costs): Likewise.
469 (niagara7_costs): Likewise.
470 (m8_costs): Likewise.
471 (TARGET_CAN_FOLLOW_JUMP): Define.
472 (pass_work_around_errata::gate): Minor tweak.
473 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
474 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
475 Automaitcally clear MASK_FSMULD mask for V7 processors.
476 (sparc_can_follow_jump): New static function.
477 (output_ubranch): Deal with CROSSING_JUMP_P.
478 (sparc_use_sched_lookahead): Rewrite using switch statement.
479 (sparc_issue_rate): Reorder.
480 (sparc_branch_cost): New function.
482 2018-09-27 Martin Sebor <msebor@redhat.com>
484 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
485 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
486 (int_fits_type_p): Same.
488 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
490 * config/i386/i386.md (FPCR_REG): Remove.
491 (UNSPEC_FLDCW): Remove.
492 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
493 (x86_fldcw_1): Remove insn pattern.
494 (fnstenv): Do not clobber FPCR_REG.
496 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
497 (CALL_USED_REGISTERS): Ditto.
498 (REG_ALLOC_ORDER): Ditto.
499 (REG_CLASS_CONTENTS): Ditto.
500 (HI_REGISTER_NAMES): Ditto.
501 (ADDITIONAL_REGISTER_NAMES): Use defines instead
502 of numerical constants.
503 * config/i386/i386.c (regclass_map): Remove fpsr register.
504 (dbx_register_map): Ditto.
505 (dbx64_register_map): Ditto.
506 (svr4_dbx_register_map): Ditto.
507 (print_reg): Do not handle FPCR_REG.
509 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
512 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
513 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
514 Delete, always treat as true.
515 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
516 Ditto. Simplify remaining code.
517 * config/powerpcspe/powerpcspe.h: Ditto.
518 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
519 Simplify remaining code.
520 (rs6000_expand_builtin): Ditto.
521 * config/rs6000/rs6000.h: Ditto.
522 * configure.ac: Ditto.
523 * configure: Regenerate.
525 2018-09-27 Martin Liska <mliska@suse.cz>
527 * coverage.c (get_coverage_counts): Revert the formatting
528 of missing profile opt info.
530 2018-09-27 Richard Biener <rguenther@suse.de>
534 * dwarf2out.c (set_block_origin_self): Do not mark outermost
535 block as we do not output that.
536 (gen_inlined_subroutine_die): Elide the originally outermost
537 block, matching what we do for concrete instances.
538 (decls_for_scope): Add parameter specifying whether to recurse
541 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
542 Tom de Vries <tom@codesourcery.com>
546 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
547 STORE_FLAG_VALUE == 1.
549 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
551 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
552 constant definitions.
553 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
554 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
555 ("speculation_barrier"): New expander definition.
557 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
559 PR gcov-profile/86957
560 * common.opt: New warning option -Wmissing-profile.
561 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
562 * doc/invoke.texi: Document -Wmissing-profile.
564 2018-09-26 Jim Wilson <jimw@sifive.com>
566 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
567 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
569 2018-09-26 Martin Sebor <msebor@redhat.com>
571 * tree.c (zerop): Change return type to bool.
572 (integer_zerop, integer_onep, integer_each_onep): Same.
573 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
574 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
575 (real_onep, real_minus_onep, chain_index): Same.
576 (print_type_hash_statistics, type_list_equal): Same.
577 * tree.h (zerop): Same.
578 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
579 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
580 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
581 (real_onep, real_minus_onep, chain_index): Same.
582 (print_type_hash_statistics, type_list_equal): Same.
584 2018-09-26 Jim Wilson <jimw@sifive.com>
586 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
588 2018-09-26 Jakub Jelinek <jakub@redhat.com>
591 * config/i386/i386.c: Include debug.h and dwarf2out.h.
592 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
595 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
597 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
599 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
601 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
602 and FP_SECOND_SSE_REGS.
603 (REG_CLASS_NAMES): Ditto.
604 (REG_CLASS_CONTENTS): Ditto.
605 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
606 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
607 (ix86_preferred_output_reload_class): Ditto.
608 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
609 clobber constraint to "=&f".
610 (fix_truncdi_i387): Ditto.
612 (fistdi2_<rounding>): Ditto.
613 (fpremxf4_i387): Change "=u" constraint to "=f".
614 (fprem1xf4_i387): Ditto.
616 (fptanxf4_i387): Ditto.
617 (fxtractxf3_i387): Ditto.
618 (fscalexf4_i387): Ditto.
619 (atan2xf3): Change "u" constraint to "f".
620 (fyl2xxf3_i387): Ditto.
621 (fyl2xp1xf3_i387): Ditto.
623 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
626 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
627 for removed I387_MASK_PM entity.
630 2018-09-26 Jeff Law <law@redhat.com>
632 2018-09-26 Alexey Neyman <stilor@att.net>
634 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
635 headers are no longer pulled in by <isl/val.h>.
637 2018-09-26 Richard Biener <rguenther@suse.de>
640 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
641 or concrete instance DIE to the tree. Create abstract origin
642 attributes also for concrete instances.
644 2018-09-26 Alexey Neyman <stilor@att.net>
646 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
647 headers are no longer pulled in by <isl/val.h>.
649 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
651 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
652 Use new helper functions.
653 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
654 Use new helper functions.
655 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
656 aarch_mm_needs_release): New declarations.
657 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
658 aarch_mm_needs_release): New.
660 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
662 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
663 (arm32_output_mi_thunk): Deal with long calls.
665 2018-09-26 Richard Biener <rguenther@suse.de>
669 * tree-inline.c (expand_call_inline): When the location
670 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
671 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
672 the inserted BLOCK to make inlined_function_outer_scope_p
674 * dwarf2out.c (add_call_src_coords_attributes): Do not add
675 coords for reserved locations.
677 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
679 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
680 (*call_indirect_nonlocal_sysv<mode>): Ditto.
681 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
682 (*sibcall_nonlocal_sysv<mode>): Ditto.
683 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
684 (<bd>_<mode>): Ditto.
685 (<bd>tf_<mode>): Ditto.
687 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
689 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
690 control string as a list of templates instead of as C code.
691 (*altivec_movti): Ditto.
692 * config/rs6000/darwin.md (movdf_low_di): Ditto.
694 2018-09-25 Jim Wilson <jimw@sifive.com>
696 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
697 when target symbol is weak.
699 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
702 * builtins.c (unterminated_array): Simplify.
703 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
704 where pointer arithmetic is safe.
706 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
709 * config/rs6000/altivec.md (altivec_vspltb): Use
710 const_0_to_15_operand instead of u5bit_cint_operand.
711 (*altivec_vspltb_internal): Ditto.
712 (altivec_vspltb_direct): Ditto.
713 (altivec_vsplth): Use const_0_to_7_operand instead of
715 (*altivec_vsplth_internal): Ditto.
716 (altivec_vsplth_direct): Ditto.
717 (altivec_vspltw): Use const_0_to_3_operand instead of
719 (*altivec_vspltw_internal): Ditto.
720 (altivec_vspltw_direct): Ditto.
721 (altivec_vspltsf): Ditto.
722 (*altivec_vspltsf_internal): Ditto.
723 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
724 various splats with the proper size immediate. Reorder the various
725 cases by ascending size of immediate, and put all such together.
727 2018-09-25 Richard Biener <rguenther@suse.de>
730 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
732 (maybe_create_die_with_external_ref): Do not create
733 DW_TAG_imported_unit here.
734 (add_abstract_origin_attribute): Handle external BLOCK refs.
735 (dwarf2out_abstract_function): Simplify LTO case.
736 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
737 rather than using maybe_create_die_with_external_ref.
739 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
742 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
743 (frndintxf2_mask_pm_i387): Ditto.
744 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
745 Enable for !flag_trapping_math.
746 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
747 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
748 Change operand 1 predicate to nonimmediate_operand.
749 (attr "i387_cw"): Remove mask_pm.
750 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
751 (enum ix86_entity): Remove I387_MASK_PM.
752 * config/i386/i386.c (ix86_i387_mode_needed): Do not
754 (ix86_mode_needed): Ditto.
755 (ix86_mode_after): Ditto.
756 (ix86_mode_entry): Ditto.
757 (ix86_mode_exit): Ditto.
758 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
760 2018-09-25 Jakub Jelinek <jakub@redhat.com>
762 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
763 to_update_switch_stmts to vNULL instead of calling create on them
766 2018-09-25 Richard Biener <rguenther@suse.de>
768 PR tree-optimization/87402
769 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
770 (visit_phi): Re-instantiate handling of supposed to be VARYING
771 but non-VARYING backedge value.
773 2018-09-25 Richard Biener <rguenther@suse.de>
776 * dwarf2out.c (struct sym_off_pair): New.
777 (external_die_map): New global.
778 (lookup_decl_die): When in LTO create DIEs lazily from the
780 (lookup_block_die): New function, create DIEs lazily in LTO.
781 (equate_block_to_die): New function.
782 (dwarf2out_die_ref_for_decl): During WPA get the association
783 from the external DIE map.
784 (dwarf2out_register_external_die): Record mapping into the
786 (maybe_create_die_with_external_ref): New function split out from
787 DIE generation part of old dwarf2out_register_external_die.
788 (add_abstract_origin_attribute): Do not return the DIE. When
789 in LTO reference externals directly.
790 (dwarf2out_abstract_function): When in LTO ignore calls for
791 decls with external DIEs (already present abstract instances).
792 (gen_call_site_die): Adjust.
793 (add_high_low_attributes): Likewise.
794 (gen_lexical_block_die): Likewise.
795 (gen_inlined_subroutine_die): Likewie.
796 (gen_block_die): Likewise.
797 (dwarf2out_inline_entry): Likewise.
798 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
801 2018-09-25 Martin Liska <mliska@suse.cz>
803 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
804 integers and not by a float value.
806 2018-09-25 Martin Liska <mliska@suse.cz>
809 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
811 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
812 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
813 * godump.c (go_format_type): Remove extra parenthesis.
815 2018-09-25 Martin Liska <mliska@suse.cz>
817 * alias.c (set_dest_equal_p): Remove unused function.
818 * config/i386/i386.c (def_builtin_pure2): Likewise.
819 * diagnostic-show-locus.c (class layout): Remove
821 (layout::layout): Likewise here.
822 * dump-context.h (class temp_dump_context): Likewise.
823 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
824 (add_AT_loclistsptr): Likewise.
825 (add_AT_offset): Likewise.
826 (get_AT_hi_pc): Likewise.
827 (is_comdat_die): Likewise.
828 (type_is_enum): Likewise.
830 (add_AT_vms_delta): Likewise.
831 (is_class_die): Likewise.
832 * edit-context.c (class line_event): Remove unused field.
833 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
835 * ipa-cp.c (ipa_get_vr_lat): Likewise.
836 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
837 (ok_for_base_p_nonstrict): Likewise.
838 * tree-chrec.c (is_not_constant_evolution): Likewise.
839 (chrec_fold_poly_cst): Likewise.
840 * tree-if-conv.c (has_pred_critical_p): Likewise.
841 * tree-ssa-coalesce.c (print_exprs): Likewise.
842 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
843 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
844 * tree-vrp.c (value_ranges_intersect_p): Likewise.
845 (value_range_nonnegative_p): Likewise.
847 2018-09-25 Martin Liska <mliska@suse.cz>
849 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
850 Do not handle "GNU Pascal".
851 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
853 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
854 from documentation. Likewise.
855 * dbxout.c (dbxout_range_type): Likewise.
856 * doc/cpp.texi: Likewise.
857 * doc/extend.texi: Likewise.
858 * doc/frontends.texi: Likewise.
859 * doc/invoke.texi: Remove Pascal entry.
860 * tree.def (CLEANUP_POINT_EXPR): Likewise.
861 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
863 2018-09-25 Martin Liska <mliska@suse.cz>
866 * doc/invoke.texi: Document all parameters and remove default
869 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
872 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
873 contains HOST_WIDE_INTs when computing its size.
875 2018-09-24 Jim Wilson <jimw@sifive.com>
878 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
880 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
882 2018-09-24 Andrew Pinski <apinski@marvell.com>
884 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
885 access prev before checking it for NULLness in the
886 AARCH64_FUSE_CMP_BRANCH case.
888 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
891 * config/i386/i386.c (rest_of_insert_endbranch): Set
892 endbr_queued_at_entrance to true and don't insert ENDBR if
893 x86_function_profiler will be called.
894 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
896 * config/i386/i386.h (machine_function): Add
897 endbr_queued_at_entrance.
899 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
901 * genattrtab.c (mk_attr_alt): Use alternative_mask.
902 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
904 (check_attr_test): Use alternative_mask.
905 (get_attr_value): Likewise.
906 (compute_alternative_mask): Use alternative_mask and XWINT.
907 (make_alternative_compare): Use alternative_mask.
908 (attr_alt_subset_p): Use XWINT.
909 (attr_alt_subset_of_compl_p): Likewise.
910 (attr_alt_intersection): Use alternative_mask and XWINT.
911 (attr_alt_union): Likewise.
912 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
913 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
914 (simplify_test_exp): Use alternative_mask and XWINT.
915 (write_test_expr): Use alternative_mask and XWINT, adjust bit
916 number calculation to support 64 bits. Generate code that
918 (main): Use alternative_mask.
919 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
921 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
924 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
925 RETURN+USE when returning via %r14.
927 2018-09-24 Martin Liska <mliska@suse.cz>
929 * gcov.c (output_lines): Print colorization legend
930 for both flag_use_colors and flag_use_hotness_colors.
933 2018-09-24 Martin Liska <mliska@suse.cz>
935 * coverage.c (get_coverage_counts): Use warning_at
936 with current_function_decl location. Use %qD in warning
939 2018-09-24 Martin Liska <mliska@suse.cz>
941 * memory-block.h (memory_block_pool::release): Annotate with
942 valgrind that the memory is not accessible.
944 2018-09-24 Martin Liska <mliska@suse.cz>
947 * asan.c: Make asan_handled_variables extern.
949 * cfgexpand.c (expand_stack_vars): Make sure
950 a representative is unpoison if another
951 variable in the partition is handled by
952 use-after-scope sanitization.
954 2018-09-24 Richard Biener <rguenther@suse.de>
956 PR tree-optimization/63155
957 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
958 the worklist when the edge of the respective argument isn't
961 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
963 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
964 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
965 (MASK_CLASS_P): Update for rename.
966 (MAYBE_MASK_CLASS_P): Ditto.
967 (REG_CLASS_NAMES): Update.
968 (REG_CLASS_CONTENT): Update.
969 * config/i386/i386.c (regclass_map): Update for MASK_REG
970 and ALL_MASK_REGS rename.
971 * config/i386/constraints.md (Yk): Update for rename.
974 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
976 * config/i386/i386.h (enum reg_class): Remove
977 EVEX_SSE_REGS and MOD4_SSE_REGS.
978 (REG_CLASS_NAMES): Update.
979 (REG_CLASS_CONTENT): Update.
980 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
981 registers as ALL_SSE_REGS.
982 (ix86_additional_allocno_class_p): Remove.
983 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
984 (ix86_register_priority): Lower priority of EVEX SSE registers.
985 Use IN_RANGE macro where appropriate.
986 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
987 AVX-5124VNNIW checks.
988 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
989 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
990 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
991 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
992 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
993 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
994 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
995 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
996 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
997 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
998 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
999 * config/i386/constraints.md (Yh): Remove.
1001 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
1003 * config/i386/i386.c (regclass_map): Declare integer REX registers
1006 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
1008 * doc/service.texi (Service): Switch the fsf.org link to https.
1010 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
1013 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
1014 Define to speculation_safe_value_not_needed.
1016 2018-09-21 Florian Weimer <fweimer@redhat.com>
1019 * doc/extend.texi (Common Function Attributes): Mention that
1020 noreturn suppresses tail call optimization.
1022 2018-09-21 Jeff Law <law@redhat.com>
1024 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
1025 vr_values::cleanup_edges_and_switches.
1026 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
1028 (identify_jump_threads): Remove EDGE_IGNORE handling.
1029 (execute_vrp): Move handling of to_remove_edges and
1030 to_update_switch_stmts into vr_values class member functions.
1031 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
1032 (to_update_switch_stmts): Likewise.
1033 * vr-values.c: Include cfghooks.h.
1034 (vr_values::vr_values): Initialize to_remove_edges and
1035 to_update_switch_stmts.
1036 (vr_values::~vr_values): Verify to_remove_edges and
1037 to_update_switch_stmts are empty.
1038 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
1039 (vr_values::cleanup_edges_and_switches): New member function.
1040 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
1041 function. Add new data members.
1043 2018-09-21 David Malcolm <dmalcolm@redhat.com>
1045 PR tree-optimization/87309
1046 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
1047 calls with pflags and alt_flags.
1048 (selftest::test_capture_of_dump_calls): Add test of interaction of
1049 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
1051 2018-09-21 Olivier Hainque <hainque@adacore.com>
1053 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
1055 2018-09-21 Olivier Hainque <hainque@adacore.com>
1057 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
1059 2018-09-21 Olivier Hainque <hainque@adacore.com>
1061 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
1062 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
1063 (VXWORKS_LIBS_RTP): Minor reordering.
1065 2018-09-21 Olivier Hainque <hainque@adacore.com>
1067 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
1068 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
1070 2018-09-21 Olivier Hainque <hainque@adacore.com>
1072 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
1073 (PTRDIFF_TYPE): Likewise.
1075 2018-09-21 Olivier Hainque <hainque@adacore.com>
1077 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
1078 triplet, similar to support for VxWorks7.
1079 * config/vxworks-dummy.h: Provide a default definition
1080 of TARGET_VXWORKS64 to 0.
1082 2018-09-21 Olivier Hainque <hainque@adacore.com>
1084 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
1085 * config/vxworks-dummy.h: here.
1087 2018-09-21 Olivier Hainque <hainque@adacore.com>
1089 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
1091 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
1092 Bo Zhou <zbo.zhou@hisilicon.com>
1094 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
1095 * config/aarch64/aarch64-tune.md: Regenerated.
1096 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
1097 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1098 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1100 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
1101 Julian Brown <julian@codesourcery.com>
1103 * builtins.c (get_builtin_sync_mem): Handle address spaces.
1105 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1107 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1108 if the call takes a static chain.
1110 2018-09-21 Martin Liska <mliska@suse.cz>
1112 * auto-profile.c (autofdo_source_profile::read): Do not
1114 (read_profile): Do not add working sets.
1115 (read_autofdo_file): Remove sum_all.
1116 (afdo_callsite_hot_enough_for_early_inline): Remove const
1118 * coverage.c (struct counts_entry): Remove gcov_summary.
1119 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1120 do not support GCOV_TAG_PROGRAM_SUMMARY.
1121 (get_coverage_counts): Remove summary and expected
1123 * coverage.h (get_coverage_counts): Likewise.
1124 * doc/gcov-dump.texi: Remove -w option.
1125 * gcov-dump.c (dump_working_sets): Remove.
1126 (main): Do not support '-w' option.
1127 (print_usage): Likewise.
1128 (tag_summary): Likewise.
1129 * gcov-io.c (gcov_write_summary): Do not dump
1131 (gcov_read_summary): Likewise.
1132 (gcov_histo_index): Remove.
1133 (gcov_histogram_merge): Likewise.
1134 (compute_working_sets): Likewise.
1135 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1137 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1138 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1139 (GCOV_HISTOGRAM_SIZE): Remove.
1140 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1141 (struct gcov_summary): Simplify rapidly just
1142 to runs and sum_max fields.
1143 (gcov_histo_index): Remove.
1144 (NUM_GCOV_WORKING_SETS): Likewise.
1145 (compute_working_sets): Likewise.
1146 * gcov-tool.c (print_overlap_usage_message): Remove
1147 trailing empty line.
1148 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1149 (output_lines): Remove program related line.
1150 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1151 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1153 (input_profile_summary): Do not read it.
1154 (merge_profile_summaries): And do not merge it.
1155 (input_symtab): Do not call removed function.
1156 * modulo-sched.c (sms_schedule): Do not print sum_max.
1157 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1158 removed when histogram method was invented.
1159 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1161 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1163 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1164 and dump selected value.
1165 * profile.c (add_working_set): Remove.
1166 (get_working_sets): Likewise.
1167 (find_working_set): Likewise.
1168 (get_exec_counts): Do not work with working sets.
1169 (read_profile_edge_counts): Do not inform as sum_max is removed.
1170 (compute_branch_probabilities): Likewise.
1171 (compute_value_histograms): Remove argument for call of
1172 get_coverage_counts.
1173 * profile.h: Do not make gcov_summary const.
1175 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
1177 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1179 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
1181 PR tree-optimization/86990
1182 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1183 Check that the entire merged store group is made of constants only for
1186 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
1188 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1189 (VTABLE_VERIFICATION_SPEC): Likewise.
1190 (SANITIZER_EARLY_SPEC): Likewise.
1191 (SANITIZER_SPEC): Likewise.
1192 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1193 * doc/invoke.texi (Link Options): Document -r.
1195 2018-09-20 Richard Biener <rguenther@suse.de>
1198 * gimplify.c (gimplify_expr): Retain alignment of
1199 addressable lvalue in dereference.
1201 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
1204 * configure.ac: Check for .loc is_stmt support.
1205 * configure, config.in: Rebuilt.
1206 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1209 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1211 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1212 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1215 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
1217 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1219 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1222 2018-09-20 Marek Polacek <polacek@redhat.com>
1224 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1226 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1228 PR tree-optimization/87288
1229 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1230 into account when determining PEELING_FOR_NITERS.
1232 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
1234 PR tree-optimization/86877
1235 * tree-vect-loop.c (vect_analyze_loop_2): Call
1236 vect_verify_datarefs_alignment.
1238 2018-09-19 Marek Polacek <polacek@redhat.com>
1240 * doc/invoke.texi: Document -Wclass-conversion.
1242 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
1244 * config/pa/pa.c (pa_adjust_priority): Delete.
1245 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
1247 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
1248 (atomic_storehi): Likewise.
1249 (atomic_storesi): Likewise.
1250 (atomic_loaddi): Restore compare and swap exchange loop code.
1252 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
1254 PR rtl-optimization/86902
1255 * combine.c (try_combine): When changing the CC mode used, don't change
1256 an unrelated mode in other_insn to that new CC mode.
1258 2018-09-19 David Malcolm <dmalcolm@redhat.com>
1260 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
1261 with %T in place of calls to dump_generic_expr.
1262 (prune_runtime_alias_test_list): Likewise.
1263 (create_runtime_alias_checks): Likewise.
1264 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
1265 (vect_analyze_data_ref_dependence): Likewise.
1266 (vect_slp_analyze_data_ref_dependence): Likewise.
1267 (vect_record_base_alignment): Likewise. Use %G in place of call
1268 to dump_gimple_stmt.
1269 (vect_compute_data_ref_alignment): Likewise.
1270 (verify_data_ref_alignment): Likewise.
1271 (vect_find_same_alignment_drs): Likewise.
1272 (vect_analyze_group_access_1): Likewise.
1273 (vect_analyze_data_ref_accesses): Likewise.
1274 (dependence_distance_ge_vf): Likewise.
1275 (dump_lower_bound): Likewise.
1276 (vect_prune_runtime_alias_test_list): Likewise.
1277 (vect_find_stmt_data_reference): Likewise.
1278 (vect_analyze_data_refs): Likewise.
1279 (vect_create_addr_base_for_vector_ref): Likewise.
1280 (vect_create_data_ref_ptr): Likewise.
1281 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
1282 (vect_can_advance_ivs_p): Likewise.
1283 (vect_update_ivs_after_vectorizer): Likewise.
1284 (vect_gen_prolog_loop_niters): Likewise.
1285 (vect_prepare_for_masked_peels): Likewise.
1286 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1287 (vect_determine_vectorization_factor): Likewise.
1288 (vect_is_simple_iv_evolution): Likewise.
1289 (vect_analyze_scalar_cycles_1): Likewise.
1290 (vect_analyze_loop_operations): Likewise.
1291 (report_vect_op): Likewise.
1292 (vect_is_slp_reduction): Likewise.
1293 (check_reduction_path): Likewise.
1294 (vect_is_simple_reduction): Likewise.
1295 (vect_create_epilog_for_reduction): Likewise.
1296 (vect_finalize_reduction:): Likewise.
1297 (vectorizable_induction): Likewise.
1298 (vect_transform_loop_stmt): Likewise.
1299 (vect_transform_loop): Likewise.
1300 (optimize_mask_stores): Likewise.
1301 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
1302 (vect_split_statement): Likewise.
1303 (vect_recog_over_widening_pattern): Likewise.
1304 (vect_recog_average_pattern): Likewise.
1305 (vect_determine_min_output_precision_1): Likewise.
1306 (vect_determine_precisions_from_range): Likewise.
1307 (vect_determine_precisions_from_users): Likewise.
1308 (vect_mark_pattern_stmts): Likewise.
1309 (vect_pattern_recog_1): Likewise.
1310 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1311 (vect_record_max_nunits): Likewise.
1312 (vect_build_slp_tree_1): Likewise.
1313 (vect_build_slp_tree_2): Likewise.
1314 (vect_print_slp_tree): Likewise.
1315 (vect_analyze_slp_instance): Likewise.
1316 (vect_detect_hybrid_slp_stmts): Likewise.
1317 (vect_detect_hybrid_slp_1): Likewise.
1318 (vect_slp_analyze_operations): Likewise.
1319 (vect_slp_analyze_bb_1): Likewise.
1320 (vect_transform_slp_perm_load): Likewise.
1321 (vect_schedule_slp_instance): Likewise.
1322 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1323 (vect_mark_stmts_to_be_vectorized): Likewise.
1324 (vect_init_vector_1): Likewise.
1325 (vect_get_vec_def_for_operand): Likewise.
1326 (vect_finish_stmt_generation_1): Likewise.
1327 (vect_check_load_store_mask): Likewise.
1328 (vectorizable_call): Likewise.
1329 (vectorizable_conversion): Likewise.
1330 (vectorizable_operation): Likewise.
1331 (vectorizable_load): Likewise.
1332 (vect_analyze_stmt): Likewise.
1333 (vect_is_simple_use): Likewise.
1334 (vect_get_vector_types_for_stmt): Likewise.
1335 (vect_get_mask_type_for_stmt): Likewise.
1336 * tree-vectorizer.c (increase_alignment): Likewise.
1338 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
1340 * doc/rtl.texi: Adjust vec_select description.
1341 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
1342 non-constant selectors.
1344 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
1346 * config/aarch64/aarch64-protos.h
1347 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
1348 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
1349 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
1350 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
1351 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
1352 (AARCH64_FL_PROFILE): Move index so flags are ordered.
1353 (AARCH64_ISA_RCPC8_4): New flag.
1354 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
1355 to aarch64_offset_9bit_signed_unscaled_p.
1356 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
1358 * config/aarch64/constraints.md (Ust): New constraint.
1359 * config/aarch64/predicates.md.
1360 (aarch64_9bit_offset_memory_operand): New predicate.
1361 (aarch64_rcpc_memory_operand): New predicate.
1363 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
1365 PR rtl-optimization/87361
1366 * rtlanal.c (nonzero_bits1): Revert accidental change.
1368 2018-09-19 Richard Biener <rguenther@suse.de>
1370 PR tree-optimization/87349
1371 PR tree-optimization/87342
1372 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
1374 2018-09-18 Marek Polacek <polacek@redhat.com>
1376 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
1377 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
1379 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
1381 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
1383 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
1385 PR rtl-optimization/86882
1386 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
1388 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
1390 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
1391 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
1393 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
1396 * doc/invoke.texi (Link Options): Fix formatting and grammar.
1398 2018-09-18 Richard Biener <rguenther@suse.de>
1401 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
1402 (compute_samebase_partition_bases): Likewise.
1403 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
1404 (gimple_can_coalesce_p): Simplify.
1406 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
1408 Handle a library implementation of ffs calling __builtin_ffs.
1409 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
1410 (mmix_init_libfuncs): New function: make __builtin_ffs expand
1413 2018-09-17 David Malcolm <dmalcolm@redhat.com>
1415 * diagnostic-show-locus.c (class layout_range): Add field
1417 (layout_range::layout_range): Add "original_idx" param and use it
1418 to initialize new field.
1419 (make_range): Use 0 for original_idx.
1420 (layout::layout): Pass in index to calls to
1421 maybe_add_location_range.
1422 (layout::maybe_add_location_range): Add param "original_idx" and
1423 pass it on to layout_range.
1424 (layout::print_any_labels): Pass on range->m_original_idx to
1426 (gcc_rich_location::add_location_if_nearby): Use 0 for
1428 * gcc-rich-location.h (text_range_label::get_text): Update for new
1430 (range_label_for_type_mismatch::get_text): Likewise.
1432 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
1434 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
1436 2018-09-17 David Malcolm <dmalcolm@redhat.com>
1438 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
1439 format_string_diagnostic_t.
1440 (fmtwarn_n): Likewise.
1441 * substring-locations.c
1442 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
1443 (format_warning_n_va): Convert to...
1444 (format_string_diagnostic_t::emit_warning_n_va): ...this.
1445 (format_warning_va): Convert to...
1446 (format_string_diagnostic_t::emit_warning_va): ...this.
1447 (format_warning_at_substring): Convert to...
1448 (format_string_diagnostic_t::emit_warning): ...this.
1449 (format_warning_at_substring_n): Convert to...
1450 (format_string_diagnostic_t::emit_warning_n): ...this.
1451 * substring-locations.h (class format_string_diagnostic_t): New
1453 (format_warning_va): Convert to
1454 format_string_diagnostic_t::emit_warning_va.
1455 (format_warning_n_va): Convert to
1456 format_string_diagnostic_t::emit_warning_n_va.
1457 (format_warning_at_substring): Convert to
1458 format_string_diagnostic_t::emit_warning.
1459 (format_warning_at_substring_n): Convert to
1460 format_string_diagnostic_t::emit_warning_n.
1462 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
1463 Bernd Schmidt <bernds_cb1@t-online.de>
1465 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
1468 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
1470 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
1471 operand 0 predicate to nonimmediate operand.
1472 (rint<mode>2_frndint): Remove insn pattern.
1473 (rint<mode>2): Change operand 1 predicate to general_operand.
1474 Extend operand 1 to XFmode and generate rintxf2 insn.
1475 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
1476 Do not use X87MODEF mode macro.
1477 (frndintxf2_<rounding>_i387): Rename from
1478 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
1479 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
1480 to XFmode and generate significandxf3 insn.
1482 2018-09-17 Richard Biener <rguenther@suse.de>
1484 PR tree-optimization/87328
1485 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
1486 visiting unexecutable backedges when not iterating.
1487 (do_rpo_vn): Mark all edges not executable even when not
1490 2018-09-17 Martin Jambor <mjambor@suse.cz>
1493 * doc/invoke.texi (Warning Options): Likewise.
1495 2018-09-17 Richard Biener <rguenther@suse.de>
1497 PR tree-optimization/87301
1498 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
1499 clean EH info from leftover copy assignments.
1501 2018-09-17 Martin Liska <mliska@suse.cz>
1503 PR gcov-profile/85871
1504 * gcov.c (output_intermediate_file): Fix out of bounds
1507 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
1509 * config/arc/arc.c: Object attributes for core4 not reflected
1511 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
1514 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
1516 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
1518 2018-09-17 Martin Liska <mliska@suse.cz>
1520 * doc/gcov.texi: Document new option --use-hotness-colors.
1521 * gcov.c (struct source_info): Declare new field.
1522 (source_info::source_info): Set default for maximum_count.
1523 (print_usage): Add new -q option.
1524 (process_args): Process it.
1525 (accumulate_line_info): Save src->maximum_count.
1526 (output_line_beginning): Make color line number if
1527 flag_use_hotness_colors is set.
1528 (output_line_details): Pass default argument value.
1529 (output_lines): Pass src->maximum_count.
1531 2018-09-17 Martin Liska <mliska@suse.cz>
1533 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
1534 Use processor_names table.
1535 * config/i386/i386.c (ix86_default_align): Use
1536 processor_cost_table for alignment values.
1537 (ix86_option_override_internal): Use processor_names.
1538 (ix86_function_specific_print): Likewise.
1539 * config/i386/i386.h (struct processor_costs):
1540 Add alignment values.
1541 (struct ptt): Remove and replace with const char *.
1542 * config/i386/x86-tune-costs.h (struct processor_costs):
1543 Declare default alignments for all costs.
1545 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
1547 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
1548 symbolics or VR_VARYING ranges for ABS_EXPR.
1549 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
1550 when range will wrap.
1552 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
1555 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
1556 before and after a JUMP_TABLE_DATA.
1558 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
1561 * dojump.c (do_compare_and_jump): Canonicalize function pointers
1562 when one operand is a function pointer. Use POINTER_TYPE_P and
1563 FUNC_OR_METHOD_TYPE_P.
1564 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
1565 * fold-const.c (build_range_check): Likewise.
1566 * match.pd (simple_comparison): Likewise.
1568 2018-09-14 David Malcolm <dmalcolm@redhat.com>
1571 * spellcheck.c (get_edit_distance_cutoff): New function.
1572 (selftest::test_edit_distance_unit_test_oneway): Rename to...
1573 (selftest::test_get_edit_distance_one_way): ...this.
1574 (selftest::test_get_edit_distance_unit): Rename to...
1575 (selftest::test_get_edit_distance_both_ways): ...this.
1576 (selftest::test_edit_distances): Move tests to this new function,
1577 and test some more pairs of strings. Update for above renaming.
1578 (selftest::get_old_cutoff): New function.
1579 (selftest::test_get_edit_distance_cutoff): New function.
1580 (selftest::assert_suggested_for): New function.
1581 (ASSERT_SUGGESTED_FOR): New macro.
1582 (selftest::assert_not_suggested_for): New function.
1583 (ASSERT_NOT_SUGGESTED_FOR): New macro.
1584 (selftest::test_suggestions): New function.
1585 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
1586 tests to selftest::test_edit_distances and call it. Add calls to
1587 selftest::test_get_edit_distance_cutoff and
1588 selftest::test_suggestions.
1589 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
1590 (best_match::consider): Replace hard-coded cutoff calculation with
1592 (best_match::get_cutoff): New declaration.
1593 (best_match::get_best_meaningful_candidate): Likewise.
1595 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
1597 * builtins.c (fold_builtin_strlen): Remove TODO comment.
1599 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
1602 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
1604 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1605 terminated string literal.
1607 2018-09-14 Martin Sebor <msebor@redhat.com>
1609 * builtins.c (unterminated_array): Handle ARRAY_REF.
1610 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
1611 * builtins.h (unterminated_array): Declare extern.
1612 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
1614 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
1617 2018-09-14 Martin Sebor <msebor@redhat.com>
1618 Jeff Law <law@redhat.com>
1620 * builtins.c (unterminated_array): New.
1621 (expand_builtin_strcpy): Adjust.
1622 (expand_builtin_strcpy_args): Detect unterminated arrays.
1623 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
1624 unterminated arrays.
1625 * gimple-fold.h (get_maxval_strlen): Add argument.
1626 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
1628 * gimple-fold.c (get_range_strlen): Add argument.
1629 (get_maxval_strlen): Adjust.
1630 * gimple-fold.h (get_range_strlen): Add argument.
1632 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
1634 * config/i386/movdirintrin.h: Fix copyright year.
1636 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
1638 * reg-stack.c: Include regs.h.
1639 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
1640 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
1641 FIRST_STACK_REG, not DFmode.
1642 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
1643 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
1644 (change stack): Default register mode to the reg_raw_mode of
1645 FIRST_STACK_REG, not DFmode.
1646 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
1647 (*swapxf): Rename from swapxf.
1649 2018-09-14 Carl Love <cel@us.ibm.com>
1651 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
1652 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
1654 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
1657 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
1660 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
1663 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
1665 2018-09-14 Jason Merrill <jason@redhat.com>
1667 Fix --enable-gather-detailed-mem-stats.
1668 * hash-table.c (hash_table_usage): Change from variable to function.
1669 * hash-table.h: Adjust.
1670 * Makefile.in: Add missing dependencies on hash-table.h.
1672 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1674 PR tree-optimization/87259
1676 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
1677 execute_cse_reciprocals_1 has tried transforming.
1679 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
1681 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
1682 VR_VARYING for PLUS/MINUS_EXPR.
1684 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
1686 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
1689 2018-09-14 Richard Biener <rguenther@suse.de>
1692 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
1693 bits for the merged partition.
1695 2018-09-13 Martin Sebor <msebor@redhat.com>
1696 Bernd Edlinger <bernd.edlinger@hotmail.de>
1698 * builtins.h (c_srlen): Add argument.
1699 * builtins.c (warn_string_no_nul): New function.
1700 (c_strlen): Add argument and use it. Update recursive calls.
1701 Pass DECL argument to string_constant to get info on non
1702 terminated strings. Update *NONSTR as needed.
1703 (fold_builtin_strlen): Add argument to calls to c_strlen.
1704 Warn for unterminated arrays.
1705 (warn_string_no_null): Add prototype.
1706 * expr.c (string_constant): Update arguments. Update recursive
1707 calls appropriately. Detect missing NUL terminator and outermost
1708 declaration its missing in.
1709 Improve checks for arrays with nonzero lower bound or elements
1710 that are not a single byte. Simplify offset computation.
1711 Simplify checks for non-NUL terminated strings.
1712 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
1713 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
1715 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
1717 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
1719 * fold-const.c (c_getstr): Fix function comment. Remove unused third
1720 argument. Fix range checks.
1721 * fold-const.h (c_getstr): Adjust protoype.
1722 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
1723 string is constant but contains no NUL byte.
1725 * expr.c (string_constant): Adjust function comment.
1726 Remove bogus check for zero termination.
1728 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
1730 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
1731 (get_constant_size): Don't make STRING_CSTs larger than they are.
1732 (check_string_literal): New check function for STRING_CSTs.
1733 (output_constant): Use it.
1735 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
1738 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
1740 2018-09-13 Richard Biener <rguenther@suse.de>
1742 PR tree-optimization/87263
1743 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
1744 (struct unwind_state): Add max_rpo field.
1745 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
1746 Compute max_rpo, the max RPO number a block can be backwards reached
1747 from. Re-write non-iterating mode to a RPO ordered worklist approach,
1748 separating it from the iterating mode.
1750 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1752 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
1753 (rfs_decision): New scheduling decision.
1755 2018-09-13 Richard Biener <rguenther@suse.de>
1758 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
1759 (vn_nary_op_insert_pieces_predicated): Do not write useless
1760 valid_dominated_by_p entry outside of the allocated storage.
1762 2018-09-13 Omar Sandoval <osandov@osandov.com>
1763 Tom de Vries <tdevries@suse.de>
1766 * dwarf2out.c (is_c): New function.
1767 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
1769 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
1772 * config/aarch64/aarch64.md (*aarch64_bfxil):
1774 * config/aarch64/constraints.md (Ulc): Define.
1775 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
1777 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
1780 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
1782 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
1783 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
1784 aarch64_layout_frame call.
1785 (aarch64_expand_epilogue): Likewise.
1786 (aarch64_initial_elimination_offset): Likewise.
1787 (aarch64_get_separate_components): Likewise.
1788 (aarch64_use_return_insn_p): Likewise.
1789 (aarch64_layout_frame): Remove unneeded check.
1791 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1793 * configure.ac: Only append
1794 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
1795 gcc_config_arguments if it was never reconfigured or last reconfigure
1796 was with different arguments.
1797 * configure: Regenerated.
1799 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1800 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1803 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
1804 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
1806 2018-09-13 Jakub Jelinek <jakub@redhat.com>
1808 PR tree-optimization/87287
1809 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
1810 X % C == 0 to X % (unsigned) C == 0 optimization to ...
1811 * match.pd (X % C == 0): ... here. New optimization.
1813 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1816 * expr.h (maybe_optimize_mod_cmp): Declare.
1817 * expr.c (mod_inv): New function.
1818 (maybe_optimize_mod_cmp): New function.
1819 (do_store_flag): Use it.
1820 * cfgexpand.c (expand_gimple_cond): Likewise.
1822 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
1823 Julian Brown <julian@codesourcery.com>
1826 * gimplify.c (gimplify_scan_omp_clauses): Set
1827 target_firstprivatize_array_bases in OpenACC parallel and kernels
1828 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
1829 OpenACC data regions.
1831 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
1833 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
1834 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
1836 2018-09-12 Richard Biener <rguenther@suse.de>
1838 PR tree-optimization/87280
1839 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
1840 edge but unreachable target.
1841 (do_rpo_vn): For conservatively handling a PHI only mark
1842 the backedge executable but not the block reachable.
1844 2018-09-12 Richard Biener <rguenther@suse.de>
1846 PR tree-optimization/87266
1847 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
1850 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
1852 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
1854 ("trunc<BFP:mode><DFP_ALL:mode>2")
1855 ("trunc<DFP_ALL:mode><BFP:mode>2")
1856 ("extend<BFP:mode><DFP_ALL:mode>2")
1857 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
1858 according to the target operand type.
1860 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1861 Andreas Krebbel <krebbel@linux.ibm.com>
1863 PR tree-optimization/86844
1864 * gimple-ssa-store-merging.c
1865 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
1866 there are any overlapping stores in between them, make sure they are
1867 also coalesced or we give up completely.
1869 2018-09-12 Jakub Jelinek <jakub@redhat.com>
1872 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
1873 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
1875 2018-09-12 Tom de Vries <tdevries@suse.de>
1877 * common.opt (gdescribe-dies): Add option.
1878 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
1879 attribute for artifical and nameless decls.
1880 (dwarf2out_register_external_die): Add description attribute to
1881 external reference die.
1882 (add_desc_attribute): New functions.
1883 (gen_subprogram_die): Add description attribute to
1884 DW_TAG_call_site_parameter.
1885 * tree-pretty-print.c (print_generic_expr_to_str): New function.
1886 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
1887 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
1889 (@item -gdescribe-dies): Add.
1891 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1893 * tree-vrp.c (vrp_shift_undefined_p): Remove.
1894 (extract_range_from_binary_expr_1: Call
1895 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
1896 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
1899 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
1901 * gimple-ssa-warn-alloca.c
1902 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
1903 field for ALLOCA_BOUND_*_LARGE.
1905 2018-09-11 Nathan Sidwell <nathan@acm.org>
1907 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
1909 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1911 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
1912 for clobbers. Remove obsolete comment.
1914 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1916 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
1917 mpxchk, mpxld and mpxst types.
1918 (define_attr length_immediate): Remove all processing of mpx types.
1919 (define_attr prefix_0f): Ditto.
1920 (define_attr memory): Ditto.
1922 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
1924 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
1925 (log<mode>2): Change operand 1 predicate to general_operand.
1926 Extend operand 1 to XFmode and generate logxf3 insn.
1927 (log10<mode>2): Change operand 1 predicate to general_operand.
1928 Extend operand 1 to XFmode and generate log10xf3 insn.
1929 (log2<mode>2): Change operand 1 predicate to general_operand.
1930 Extend operand 1 to XFmode and generate log2xf3 insn.
1931 (fyl2xp1_extend<mode>xf3_i387): Remove.
1932 (log1p<mode>2): Change operand 1 predicate to general_operand.
1933 Extend operand 1 to XFmode and generate log1pxf3 insn.
1934 (fxtract_extend<mode>xf3_i387): Remove.
1935 (logb<mode>2): Change operand 1 predicate to general_operand.
1936 Extend operand 1 to XFmode and generate logbxf3 insn.
1937 (ilogb<mode>2): Change operand 1 predicate to general_operand.
1938 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
1939 (significand<mode>2): Change operand 1 predicate to general_operand.
1940 Extend operand 1 to XFmode and generate significandxf3 insn.
1942 2018-09-11 Nathan Sidwell <nathan@acm.org>
1944 * gcc.c (perror_with_name, pfatal_with_name): Delete.
1945 (load_specs): Use fatal_error.
1946 (DELETE_IF_ORDINARY, process_command): Use error.
1947 (execute, run_attempt): Use fatal_error.
1949 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
1951 * diagnostic-core.h (sorry_at): New prototype.
1952 * diagnostic.c (sorry_at): New function.
1954 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
1956 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
1957 by zero as VR_UNDEFINED.
1959 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
1961 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
1962 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
1963 (<sincos>mode2): New expander.
1964 (sincos_extend<mode>xf3_i387): Remove insn pattern.
1965 (sincos -> sin, cos splitters): Remove splitter patterns.
1966 (sincos<mode>3): Change operand 2 predicate to general_operand.
1967 Extend operand 2 to XFmode and generate sincosxf3 insn.
1968 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
1969 Change operand 3 predicate to const1_operand.
1970 (fptan_extend<mode>xf4_i387): Remove insn pattern.
1971 (tanxf2): Update operands in the call to fptanxf4_i387.
1972 (tan<mode>2): Change operand 1 predicate to general_operand.
1973 Extend operand 1 to XFmode and generate tanxf3 insn.
1974 (atan2xf3): Rename from *fpatanxf3_i387.
1975 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
1976 (atan2xf3): Remove expander.
1977 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
1978 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
1979 (atan<mode>2): Change operand 1 predicate to general_operand.
1980 Extend operand 1 to XFmode and generate atanxf3 insn.
1982 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
1984 * config/i386/i386.md (x87/SSE constant load splitter): Use
1985 memory_operand instead of nonimmediate_operand for input operand
1988 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
1990 * config/i386/i386.md (float partial SSE register stall splitter): Move
1991 splitter near its instruction pattern.
1992 (float_extend partial SSE register stall splitter): Ditto.
1993 (float_truncate partial SSE register stall splitter): Ditto.
1995 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
1998 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1999 to speculation_safe_value_not_needed.
2002 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
2003 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
2004 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
2005 leaf_function_p, instead use has_hard_reg_initial_val.
2007 2018-09-09 Nathan Sidwell <nathan@acm.org>
2009 * gcc.h (pfatal_with_name): Don't declare here.
2010 * gcc.c (pfatal_with_name): Make static.
2012 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
2014 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
2017 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
2019 PR rtl-optimization/85458
2020 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
2021 priority hook to reduce the priority of EXPR.
2023 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
2025 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
2026 DImode for x87 on 32bit targets. Conditionally disable x87 modes
2027 with X87_ENABLE_FLOAT. Remove preparation code.
2028 (*float<SWI48:mode><MODEF:mode>2): Rename from
2029 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
2030 math using "enabled" attribute.
2031 (*floatdi<MODEF:mode>2_i387): Rename from
2032 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
2033 enable for 32bit targets only.
2034 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
2036 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
2037 as operand 1 predicate. Rewrite as define_insn_and_split.
2038 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
2040 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2042 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
2043 to fallthru to FLOAT case.
2045 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2048 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
2049 around folding of vec_sl to handle out of range shift values.
2051 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
2053 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
2054 Update callers to gen_fix_trunc<mode>_i387_fisttp
2055 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
2056 nonimmediate_operand.
2057 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
2058 and corresponding splitters.
2059 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
2060 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
2061 (fix_truncdi_i387_with_temp): Remove insn pattern
2062 and corresponding splitters.
2063 (fix_trunc<mode>_i387): Change operand 0 predicate to
2064 nonimmediate_operand.
2065 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
2066 and corresponding splitters.
2067 (*fistdi2_1): Remove.
2069 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
2070 (lrintxfdi2): Remove expander. Reimplement as define_insn.
2071 (*fist<mode>2_1): Remove.
2072 (fist<mode>2): Ditto.
2073 (fist<mode>2_with_temp): Remove insn pattern and corresponding
2075 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
2076 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
2077 (fistdi2_<rounding>): Change operand 0 predicate to
2078 nonimmediate_operand.
2079 (fistdi2_<rounding>_with_temp): Remove insn pattern
2080 and corresponding splitters.
2081 (fist<mode>2_<rounding>): Change operand 0 predicate to
2082 nonimmediate_operand.
2083 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
2084 and corresponding splitters.
2086 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
2088 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
2090 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
2093 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2095 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
2096 early gimple folding of vec_splat().
2097 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2098 * gimple-fold.h: Add an extern define for tree_vec_extract().
2100 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
2102 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2103 wrappers around TREE_TYPE comparisons.
2105 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2108 * config/s390/predicates.md: Add nonsym_memory_operand.
2109 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2110 contains a SYMBOL_REF, load it into an intermediate pseudo.
2111 (s390_emit_compare_and_swap): Legitimize operand.
2112 * config/s390/s390.md: Use the new nonsym_memory_operand
2113 with UNSPECV_CAS patterns.
2115 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2118 * config/s390/s390-passes.def: New file.
2119 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2121 (make_pass_s390_early_mach): Add declaration.
2122 * config/s390/s390.c (make_pass_s390_early_mach):
2123 (s390_option_override): Remove dynamic registration.
2124 * config/s390/t-s390: Add s390-passes.def.
2126 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
2128 * config/s390/s390.c (s390_decompose_constant_pool_ref):
2129 Remove UNSPEC_LTREL_BASE check.
2130 (annotate_constant_pool_refs): Likewise.
2131 (find_constant_pool_ref): Likewise.
2132 (find_ltrel_base): Removed.
2133 (replace_ltrel_base): Removed.
2134 (s390_mainpool_finish): Remove replace_ltrel_base call.
2135 (s390_chunkify_start): Remove pending LTREL_BASE logic.
2136 (s390_chunkify_finish): Remove replace_ltrel_base call.
2137 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2139 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
2142 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2143 to speculation_safe_value_not_needed.
2145 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
2146 Bernd Schmidt <bernds_cb1@t-online.de>
2148 * config/nvptx/nvptx-opts.h: New file.
2149 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2150 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2152 (TARGET_SM35): New macro.
2153 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2155 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2157 (misa=): New option.
2158 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2160 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
2162 * config/i386/i386.md (truncdfsf2): Remove expander.
2163 (truncdfsf2_with_temp): Ditto.
2164 (truncxf<mode>2): Ditto.
2165 (*truncdfsf_fast_mixed): Remove insn pattern.
2166 (*truncdfsf_fast_i387): Ditto.
2167 (*truncdfsf_mixed): Ditto.
2168 (*truncdfsf_i387): Ditto.
2169 (*truncdfsf2_i387_1): Ditto.
2170 (*truncxfsf2_mixed): Ditto.
2171 (*truncxfdf2_mixed): Ditto.
2172 (*truncxf<mode>2_i387_noop): Ditto. Update callers
2173 to call gen_truncxf<mode>2 instead.
2174 (*truncxf<mode>2_i387): Remove.
2175 (reg->reg splitters): Remove splitter pattern.
2176 (reg->mem splitters): Ditto.
2178 (truncdfsf2): New insn pattern.
2179 (truncxf<mode>2): Ditto.
2181 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2183 * tree-ssa-math-opts.c (is_mult_by): New function.
2184 (is_square_of): Use the above.
2185 (optimize_recip_sqrt): New function.
2186 (pass_cse_reciprocals::execute): Use the above.
2188 2018-09-05 Richard Biener <rguenther@suse.de>
2191 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2192 to zero-init the emplaced vec.
2194 2018-09-05 Martin Liska <mliska@suse.cz>
2196 PR tree-optimization/87205
2197 * tree-switch-conversion.c (pass_lower_switch::execute):
2198 Group cases for switch statements.
2200 2018-09-05 Richard Biener <rguenther@suse.de>
2202 PR tree-optimization/87217
2203 * tree-ssa-sccvn.c (vuse_valueize): New.
2204 (vn_reference_lookup_pieces): Use it.
2205 (vn_reference_lookup): Likewise.
2207 2018-09-05 Nathan Sidwell <nathan@acm.org>
2210 * stor-layout.c (place_field): Scan forwards to check last
2211 bitfield when ms_bitfield_placement is in effect.
2213 2018-09-05 Richard Biener <rguenther@suse.de>
2216 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2219 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
2220 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
2222 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2223 * config.gcc (extra_objs): Build it.
2224 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2226 * config/aarch64/aarch64-passes.def
2227 (pass_tag_collision_avoidance): New pass.
2228 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2229 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2230 (aarch64_classify_address): Remove static qualifier.
2231 (aarch64_address_info, aarch64_address_type): Move to...
2232 * config/aarch64/aarch64-protos.h: ... here.
2233 (make_pass_tag_collision_avoidance): New function.
2234 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
2237 2018-09-05 Martin Liska <mliska@suse.cz>
2239 * doc/gcov.texi: Update documentation of humar
2241 * gcov.c (format_count): Print one decimal place, it provides
2242 more fine number of situations like '1G' vs. '1.4G'.
2244 2018-09-05 Martin Liska <mliska@suse.cz>
2247 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
2248 * optc-gen.awk: Allow 'Var' for Deprecated options in order
2249 to generate a MASK value.
2251 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
2254 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
2255 if frame pointer isn't used.
2256 (compute_frame_pointer_to_fb_displacement): Likewise.
2258 2018-09-04 Jakub Jelinek <jakub@redhat.com>
2261 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2262 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2263 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2264 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2265 and OPTION_MASK_ISA_XSAVEC_UNSET.
2267 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
2269 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2270 XOR operations in NAND case.
2272 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
2274 * wide-int-range.cc (wide_int_range_convert): New.
2275 * wide-int-range.h (wide_int_range_convert): New.
2276 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
2277 code into wide_int_range_convert.
2278 (extract_range_into_wide_ints): Do not munge anti range constants
2279 into the entire domain. Just return the range back.
2281 2018-09-04 Martin Liska <mliska@suse.cz>
2283 * genmatch.c (output_line_directive): Add new argument
2285 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
2287 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2289 * doc/invoke.texi (Option Summary): Add whitespace.
2291 * doc/invoke.texi (Option Summary): Add -Waligned-new.
2293 2018-09-04 Richard Biener <rguenther@suse.de>
2295 PR tree-optimization/87211
2296 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
2297 backedge value we're supposed to treat as VARYING also number
2298 the PHI to VARYING in case it got a different value-number already.
2300 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
2302 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
2303 (extract_range_from_binary_expr_1): Do not call
2304 vrp_can_optimize_bit_op.
2305 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
2307 (wide_int_range_get_mask_and_bounds): New.
2308 (wide_int_range_optimize_bit_op): New.
2309 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
2310 (wide_int_range_bit_and): Same.
2311 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
2312 (wide_int_range_optimize_bit_op): New.
2313 (wide_int_range_get_mask_and_bounds): New.
2315 2018-09-04 Richard Biener <rguenther@suse.de>
2317 PR tree-optimization/87176
2318 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
2319 variable. When value-numbering a virtual PHI node make sure
2320 to not value-number to the backedge value.
2322 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
2324 * doc/extend.texi (Long Long, Hex Floats): Document support for
2325 long long and hex floats in more recent versions of ISO C++.
2327 2018-09-03 Richard Biener <rguenther@suse.de>
2329 PR tree-optimization/87177
2330 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
2333 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2335 * bb-reorder.c (edge_order): Convert to C-qsort-style
2336 tri-state comparator.
2337 (reorder_basic_blocks_simple): Change std::stable_sort to
2340 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2342 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
2343 tri-state comparator.
2344 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
2346 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2348 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
2349 (mergesort): ... here as maximum count for using netsort.
2350 (gcc_qsort): Set nlim to 3 if stable sort is requested.
2351 (gcc_stablesort): New.
2352 * system.h (gcc_stablesort): Declare.
2354 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
2356 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
2357 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
2358 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
2360 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
2362 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
2363 lxsdx and stxsdx alternatives.
2364 (*mov<mode>_hardfloat64): Ditto.
2365 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
2367 2018-09-03 Richard Biener <rguenther@suse.de>
2369 PR tree-optimization/87200
2370 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
2373 2018-09-03 Martin Liska <mliska@suse.cz>
2375 PR tree-optimization/87201
2376 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2377 Fix parenthesis in an expression.
2379 2018-09-03 Richard Biener <rguenther@suse.de>
2381 PR tree-optimization/87197
2382 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
2383 visited. CSE the VN_INFO hashtable lookup.
2385 PR tree-optimization/87169
2386 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
2387 iterating make sure there's no extra backedges from irreducible
2388 regions feeding the header. Mark the destination block
2391 2018-09-03 Martin Liska <mliska@suse.cz>
2394 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
2395 * common/common-targhooks.c (default_get_valid_option_values):
2397 * common/common-targhooks.h (default_get_valid_option_values):
2399 * common/config/i386/i386-common.c: Move processor_target_table
2401 (ix86_get_valid_option_values): New function.
2402 (TARGET_GET_VALID_OPTION_VALUES): New macro.
2403 * config/i386/i386.c (struct ptt): Move to i386-common.c.
2404 (PTA_*): Move all defined masks into i386-common.c.
2405 (ix86_function_specific_restore): Use new processor_cost_table.
2406 * config/i386/i386.h (struct ptt): Moved from i386.c.
2407 (struct pta): Likewise.
2408 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
2409 * doc/tm.texi.in: Likewise.
2410 * opt-suggestions.c (option_proposer::suggest_option):
2411 Pass prefix to build_option_suggestions.
2412 (option_proposer::get_completions): Likewise.
2413 (option_proposer::build_option_suggestions): Use the new target
2415 * opts.c (struct option_help_tuple): New struct.
2416 (print_filtered_help): Use the new target hook.
2418 2018-09-03 Martin Liska <mliska@suse.cz>
2421 * predict.c (set_even_probabilities): Add likely_edges
2422 argument and handle cases where we have precisely one
2424 (combine_predictions_for_bb): Catch also likely_edges.
2425 (tree_predict_by_opcode): Handle gswitch statements.
2426 * tree-cfg.h (find_case_label_for_value): New declaration.
2427 (find_taken_edge_switch_expr): Likewise.
2428 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2429 Find pivot in decision tree based on probabily, not by number of
2432 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
2434 * doc/standards.texi (Standards): Update Objective-C reference.
2436 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2438 * doc/install.texi (Prerequisites): Update link for MPC.
2440 2018-09-01 Michael Matz <matz@suse.de>
2442 PR tree-optimization/87074
2443 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
2444 PHIs for outer-loop uses.
2446 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2448 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
2449 * doc/invoke.texi (C Dialect Options): Ditto.
2451 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
2453 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
2455 2018-08-31 Richard Biener <rguenther@suse.de>
2457 PR tree-optimization/87168
2458 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
2459 (rpo_elim::eliminate_avail): When OP was not visited it must
2462 2018-08-31 David Malcolm <dmalcolm@redhat.com>
2464 * tree-vrp.c (copy_value_range): Convert param "from" from
2465 "value_range *" to "const value_range *".
2466 (range_is_null): Likewise for param "vr".
2467 (range_int_cst_p): Likewise.
2468 (range_int_cst_singleton_p): Likewise.
2469 (symbolic_range_p): Likewise.
2470 (value_ranges_intersect_p): Likewise for both params.
2471 (value_range_nonnegative_p): Likewise for param "vr".
2472 (value_range_constant_singleton): Likewise.
2473 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
2474 (extract_range_into_wide_ints): Likewise for param "vr".
2475 (extract_range_from_multiplicative_op): Likewise for params "vr0"
2477 (vrp_can_optimize_bit_op): Likewise.
2478 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
2480 (extract_range_from_unary_expr): Likewise.
2481 (debug_value_range): Likewise for param "vr".
2482 (value_range::dump): Add "const" qualifier.
2483 (vrp_prop::check_array_ref): Convert local "vr" from
2484 "value_range *" to "const value_range *".
2485 (vrp_prop::check_mem_ref): Likewise.
2486 (vrp_prop::visit_stmt): Likewise for local "old_vr".
2487 (vrp_intersect_ranges_1): Likewise for param "vr_1".
2488 (vrp_intersect_ranges): Likewise.
2489 (simplify_stmt_for_jump_threading): Likewise for local "vr".
2490 (vrp_prop::vrp_finalize): Likewise.
2491 * tree-vrp.h (value_range::dump): Add "const" qualifier.
2492 (vrp_intersect_ranges): Add "const" qualifier to params as above.
2493 (extract_range_from_unary_expr): Likewise.
2494 (value_range_constant_singleton): Likewise.
2495 (symbolic_range_p): Likewise.
2496 (copy_value_range): Likewise.
2497 (extract_range_from_binary_expr_1): Likewise.
2498 (range_int_cst_p): Likewise.
2499 (vrp_set_zero_nonzero_bits): Likewise.
2500 (range_int_cst_singleton_p): Likewise.
2502 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
2504 * config/aarch64/arm_neon.h (vabsd_s64): New.
2505 (vnegd_s64): Likewise.
2507 2018-08-31 Martin Jambor <mjambor@suse.cz>
2509 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
2511 2018-08-31 Martin Liska <mliska@suse.cz>
2513 * ipa-icf.c (sem_item::add_type): Use
2514 sem_item::m_type_hash_cache.
2515 * ipa-icf.h: Move the cache from sem_item_optimizer
2518 2018-08-31 Nathan Sidwell <nathan@acm.org>
2520 * doc/extend.texi (Backwards Compatibility): Remove implicit
2521 extern C leeway of () being (...).
2523 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2525 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
2527 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
2531 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
2533 2018-08-31 Jakub Jelinek <jakub@redhat.com>
2536 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
2537 gen_int_mode. Formatting fixes.
2539 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
2541 * target.def (custom_function_descriptors): Improve documentation.
2542 * doc/tm.texi.in (Trampolines): Expand discussion of function
2543 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
2544 beginning of the section.
2545 * doc/tm.texi: Regenerated.
2547 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
2549 * cfg.h (class auto_edge_flag): Spell out the template-id of the
2550 base class in the initializer list. This is a workaround for
2551 building with older GCC.
2552 (class auto_bb_flag): Likewise.
2554 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
2556 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
2557 (altivec_vcmpequ<VI_char>_p): Remove star.
2558 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
2560 (expand_strncmp_vec_sequence): New function.
2561 (emit_final_str_compare_vec): New function.
2562 (expand_strn_compare): Add support for vector strncmp.
2563 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
2564 length specification to bytes.
2565 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
2566 (vcmpnezb_p): New pattern.
2567 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
2568 for option -mstring-compare-inline-limit.
2570 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
2572 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
2573 (PTA_SKYLAKE): Add PTA_AES.
2574 (PTA_GOLDMONT): Likewise.
2576 2018-08-29 Jan Hubicka <jh@suse.cz>
2579 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
2580 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
2582 2018-08-29 Jan Hubicka <jh@suse.cz>
2584 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
2586 (hash_tree): Do not visit TYPE_STUB_DECL.
2587 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
2588 stream TYPE_STUB_DECL.
2589 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
2590 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
2591 after free_lang_data.
2592 (type_in_anonymous_namespace_p): Likewise.
2594 2018-08-29 Jan Hubicka <jh@suse.cz>
2596 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2597 comment that it has to be even number.
2598 (class sreal): Change m_sig type to int32_t.
2599 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2600 int64_t for temporary calculations.
2601 (sreal_verify_basics): Drop one bit from minimum and maximum.
2603 2018-08-30 Richard Biener <rguenther@suse.de>
2605 PR tree-optimization/87147
2606 * tree-ssa-sccvn.c (SSA_VISITED): New function.
2607 (visit_phi): When the degenerate result is from the backedge and
2608 we didn't visit its definition yet drop to VARYING.
2609 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
2611 2018-08-29 Jan Hubicka <jh@suse.cz>
2613 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
2615 (hash_tree): Likewise.
2617 2018-08-29 Jan Hubicka <jh@suse.cz>
2619 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
2620 and TYPE_NEXT_REF_TO.
2622 2018-08-29 Jan Hubicka <jh@suse.cz>
2624 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2625 comment that it has to be even number.
2626 (class sreal): Change m_sig type to int32_t.
2627 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2628 int64_t for temporary calculations.
2629 (sreal_verify_basics): Drop one bit from minimum and maximum.
2631 2018-08-30 Tamar Christina <tamar.christina@arm.com>
2633 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
2635 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
2638 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
2639 if to_add is negative.
2641 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
2644 * builtins.c (c_strlen): Improve range checks.
2646 2018-08-29 Martin Sebor <msebor@redhat.com>
2647 Jeff Law <law@redhat.com>
2649 PR tree-optimization/86714
2650 PR tree-optimization/86711
2651 * builtins.c (c_strlen): Add arguments to call to string_constant.
2652 * expr.c (string_constant): Add argument. Detect missing nul
2653 terminator and outermost declaration it's missing in.
2654 * expr.h (string_constant): Add argument.
2655 * fold-const.c (read_from_constant_string): Add arguments to call to
2657 (c_getstr): Likewise.
2658 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2660 * tree-ssa-strlen.c (get_stridx): Likewise.
2662 2018-08-29 Jan Hubicka <jh@suse.cz>
2664 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
2665 Do not stream DECL_VINDEX.
2666 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
2667 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
2668 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
2670 2018-08-29 Richard Biener <rguenther@suse.de>
2672 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
2673 virtual operands that are not default defs to honor region
2675 (rpo_vn_valueize): Remove ineffective code here.
2677 2018-08-29 Richard Biener <rguenther@suse.de>
2679 PR tree-optimization/87132
2680 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
2681 when skipping defs reachable over backedges.
2683 2018-08-29 Richard Biener <rguenther@suse.de>
2685 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
2686 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
2687 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
2688 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
2689 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
2690 (vn_reference_lookup_3): Remove use of const_parms.
2691 (free_rpo_vn): Do not free const_parms.
2692 (do_rpo_vn): Do not call init_const_parms.
2693 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
2694 SSA_NAME_POINTS_TO_READONLY_MEMORY.
2695 (call_may_clobber_ref_p_1): Likewise.
2697 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
2700 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
2702 (-ftree-scev-cprop): Document.
2704 2018-08-29 Jan Hubicka <jh@suse.cz>
2706 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
2708 (sreal constructor): Update.
2709 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
2710 sreal:operator/): Update.
2712 2018-08-29 Martin Liska <mliska@suse.cz>
2714 * tree-switch-conversion.c (switch_conversion::expand):
2715 Strenghten assumption about gswitch statements.
2717 2018-08-29 Richard Biener <rguenther@suse.de>
2719 PR tree-optimization/87117
2720 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
2721 re-value-number released SSA VDEFs.
2723 2018-08-29 Richard Biener <rguenther@suse.de>
2725 PR tree-optimization/87126
2726 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
2728 2018-08-28 Jim Wilson <jimw@sifive.com>
2730 * config/riscv/pic.md: Rewrite.
2731 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
2733 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
2734 (SOFTF, default_load, softload, softstore): New.
2736 2018-08-28 Jeff Law <law@redhat.com>
2738 * fold-const.c (fold_binary_loc): Remove recently added assert.
2740 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
2742 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
2743 to OP parmeter of generated function.
2745 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
2747 PR tree-optimization/87009
2748 * match.pd: Add boolean optimizations.
2750 2018-08-28 Martin Sebor <msebor@redhat.com>
2753 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
2754 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
2755 (pass_walloca::gate): Use it.
2756 (alloca_call_type): Same.
2757 (pass_walloca::execute): Same.
2758 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
2760 2018-08-28 David Malcolm <dmalcolm@redhat.com>
2762 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
2763 GCC_VERSION for usage of "__gcc_dump_printf__" format from
2766 2018-08-28 Richard Biener <rguenther@suse.de>
2768 PR tree-optimization/87124
2769 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
2770 constants before looking up avail.
2772 2018-08-28 Jakub Jelinek <jakub@redhat.com>
2775 * calls.c (maybe_warn_nonstring_arg): Punt early if
2776 warn_stringop_overflow is zero. Don't call get_range_strlen
2777 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2778 Swap comparison operands to have constants on rhs. Only use
2779 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
2780 increment lenrng[0].
2782 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
2784 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
2785 use of tree_to_shwi. Remove duplicated test for the size being
2786 a whole number of bytes.
2788 2018-08-28 Richard Biener <rguenther@suse.de>
2790 PR tree-optimization/87117
2791 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
2792 Handle removed stmt without LHS (GIMPLE_NOP).
2794 2018-08-28 Richard Biener <rguenther@suse.de>
2796 PR tree-optimization/87117
2797 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
2798 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
2800 2018-08-28 Richard Biener <rguenther@suse.de>
2802 PR tree-optimization/87117
2803 * tree-ssa-pre.c (compute_avail): Do not make expressions
2804 with predicated values available.
2805 (get_expr_value_id): Assert we do not run into predicated value
2808 2018-08-28 Richard Biener <rguenther@suse.de>
2810 PR tree-optimization/87117
2811 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
2812 get virtual operands.
2813 (get_expr_operands): Handle STRING_CST like other decls.
2815 2018-08-28 Martin Liska <mliska@suse.cz>
2817 * tree.h: Update documentation of fndecl_built_in_p
2821 2018-08-27 Jeff Law <law@redhat.com>
2822 PR tree-optimization/87110
2823 * tree-ssa-dse.c (compute_trims): Handle non-constant
2826 2018-08-27 Martin Sebor <msebor@redhat.com>
2828 PR tree-optimization/86914
2829 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2831 2018-08-27 Martin Sebor <msebor@redhat.com>
2833 PR tree-optimization/87112
2834 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
2835 the type of the bound argument.
2837 2018-08-27 Jeff Law <law@redhat.com>
2839 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
2840 type does not have a TYPE_SIZE_UNIT.
2842 2018-08-27 Steve Ellcey <sellcey@cavium.com>
2844 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
2845 with include of backend.h.
2847 2018-08-27 Richard Biener <rguenther@suse.de>
2849 PR tree-optimization/86927
2850 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2851 use const cond reduction code.
2853 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
2855 PR tree-optimization/85758
2856 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
2858 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2861 * diagnostic-show-locus.c (class layout_range): Update for
2862 conversion of show_caret_p to a tri-state.
2863 (layout_range::layout_range): Likewise.
2864 (make_range): Likewise.
2865 (layout::maybe_add_location_range): Likewise.
2866 (layout::should_print_annotation_line_p): Don't show annotation
2867 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
2868 (layout::get_state_at_point): Update for conversion of
2869 show_caret_p to a tri-state. Bail out early for
2870 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
2871 underlining or source colorization.
2872 (gcc_rich_location::add_location_if_nearby): Update for conversion
2873 of show_caret_p to a tri-state.
2874 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
2875 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
2877 (selftest::test_one_liner_labels): Likewise.
2878 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
2879 conversion of show_caret_p to a tri-state.
2880 * pretty-print.c (text_info::set_location): Likewise.
2881 * pretty-print.h (text_info::set_location): Likewise.
2882 * substring-locations.c (format_warning_n_va): Likewise.
2883 * tree-diagnostic.c (default_tree_printer): Likewise.
2884 * tree-pretty-print.c (newline_and_indent): Likewise.
2886 2018-08-27 David Malcolm <dmalcolm@redhat.com>
2889 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
2890 line above for line-insertion fix-it hints.
2891 (selftest::test_fixit_insert_containing_newline): Update the
2892 expected results, and add a test with line-numbering enabled.
2894 2018-08-27 Martin Liska <mliska@suse.cz>
2897 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
2898 params with DECL_HAS_VALUE_EXPR_P.
2900 2018-08-27 Martin Liska <mliska@suse.cz>
2902 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
2903 selected expansion strategy.
2905 2018-08-27 Martin Liska <mliska@suse.cz>
2907 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
2908 * builtins.c (is_builtin_fn): Likewise.
2909 * attribs.c (diag_attr_exclusions): Use new function
2910 fndecl_built_in_p and remove check for FUNCTION_DECL if
2912 (builtin_mathfn_code): Likewise.
2913 (fold_builtin_expect): Likewise.
2914 (fold_call_expr): Likewise.
2915 (fold_builtin_call_array): Likewise.
2916 (fold_call_stmt): Likewise.
2917 (set_builtin_user_assembler_name): Likewise.
2918 (is_simple_builtin): Likewise.
2919 * calls.c (gimple_alloca_call_p): Likewise.
2920 (maybe_warn_nonstring_arg): Likewise.
2921 * cfgexpand.c (expand_call_stmt): Likewise.
2922 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
2923 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
2924 (cgraph_node::verify_node): Likewise.
2925 * cgraphclones.c (build_function_decl_skip_args): Likewise.
2926 (cgraph_node::create_clone): Likewise.
2927 * config/arm/arm.c (arm_insert_attributes): Likewise.
2928 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
2929 * dse.c (scan_insn): Likewise.
2930 * expr.c (expand_expr_real_1): Likewise.
2931 * fold-const.c (operand_equal_p): Likewise.
2932 (fold_binary_loc): Likewise.
2933 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
2934 * gimple-low.c (lower_stmt): Likewise.
2935 * gimple-pretty-print.c (dump_gimple_call): Likewise.
2936 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
2938 * gimple.c (gimple_build_call_from_tree): Likewise.
2939 (gimple_call_builtin_p): Likewise.
2940 (gimple_call_combined_fn): Likewise.
2941 * gimplify.c (gimplify_call_expr): Likewise.
2942 (gimple_boolify): Likewise.
2943 (gimplify_modify_expr): Likewise.
2944 (gimplify_addr_expr): Likewise.
2945 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
2946 * ipa-cp.c (determine_versionability): Likewise.
2947 * ipa-fnsummary.c (compute_fn_summary): Likewise.
2948 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
2949 * ipa-split.c (visit_bb): Likewise.
2950 (split_function): Likewise.
2951 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2952 * lto-cgraph.c (input_node): Likewise.
2953 * lto-streamer-out.c (write_symbol): Likewise.
2954 * omp-low.c (setjmp_or_longjmp_p): Likewise.
2955 (lower_omp_1): Likewise.
2956 * predict.c (strip_predict_hints): Likewise.
2957 * print-tree.c (print_node): Likewise.
2958 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
2959 * trans-mem.c (is_tm_irrevocable): Likewise.
2960 (is_tm_load): Likewise.
2961 (is_tm_simple_load): Likewise.
2962 (is_tm_store): Likewise.
2963 (is_tm_simple_store): Likewise.
2964 (is_tm_abort): Likewise.
2965 (tm_region_init_1): Likewise.
2966 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
2967 * tree-cfg.c (verify_gimple_call): Likewise.
2968 (move_stmt_r): Likewise.
2969 (stmt_can_terminate_bb_p): Likewise.
2970 * tree-eh.c (lower_eh_constructs_2): Likewise.
2971 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2972 * tree-inline.c (remap_gimple_stmt): Likewise.
2973 (copy_bb): Likewise.
2974 (estimate_num_insns): Likewise.
2975 (fold_marked_statements): Likewise.
2976 * tree-sra.c (scan_function): Likewise.
2977 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
2978 (optimize_stack_restore): Likewise.
2979 (pass_fold_builtins::execute): Likewise.
2980 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
2981 (mark_all_reaching_defs_necessary_1): Likewise.
2982 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
2983 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2984 (pass_forwprop::execute): Likewise.
2985 * tree-ssa-loop-im.c (stmt_cost): Likewise.
2986 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2987 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
2988 * tree-ssa-strlen.c (get_string_length): Likewise.
2989 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
2990 (find_func_aliases_for_call): Likewise.
2991 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
2992 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
2993 * tree-tailcall.c (find_tail_calls): Likewise.
2994 * tree.c (need_assembler_name_p): Likewise.
2995 (free_lang_data_in_decl): Likewise.
2996 (get_call_combined_fn): Likewise.
2997 * ubsan.c (is_ubsan_builtin_p): Likewise.
2998 * varasm.c (incorporeal_function_p): Likewise.
2999 * tree.h (DECL_BUILT_IN): Remove and replace with
3001 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
3002 (fndecl_built_in_p): New.
3004 2018-08-27 Martin Liska <mliska@suse.cz>
3006 PR tree-optimization/86847
3007 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
3008 Dump also subtree probability.
3009 (switch_decision_tree::do_jump_if_equal): New function.
3010 (switch_decision_tree::emit_case_nodes): Handle special
3011 situations in balanced tree that can be emitted much simpler.
3012 Fix calculation of probabilities that happen in tree expansion.
3013 * tree-switch-conversion.h (struct cluster): Add
3015 (struct simple_cluster): Likewise.
3016 (struct case_tree_node): Add new function has_child.
3017 (do_jump_if_equal): New.
3019 2018-08-27 Martin Liska <mliska@suse.cz>
3021 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
3022 Add new argument to bit_test_cluster constructor.
3023 (bit_test_cluster::emit): Set bits really number of values
3025 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
3026 probability argument.
3027 * tree-switch-conversion.h (struct bit_test_cluster):
3028 Add m_handles_entire_switch.
3030 2018-08-27 Martin Liska <mliska@suse.cz>
3032 PR tree-optimization/86702
3033 * tree-switch-conversion.c (jump_table_cluster::emit):
3034 Make probabilities even for values in jump table
3035 according to number of cases handled.
3036 (switch_decision_tree::compute_cases_per_edge): Pass
3037 argument to reset_out_edges_aux function.
3038 (switch_decision_tree::analyze_switch_statement): Likewise.
3039 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
3042 2018-08-27 Martin Liska <mliska@suse.cz>
3044 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
3045 cfun argument explicitly.
3046 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
3047 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
3048 function gimple_switch_default_bb.
3049 (convert_switch_statements):
3051 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
3052 * stmt.c (label_to_block_fn): Use label_to_block and pass
3053 cfun argument explicitly and use gimple_switch_label_bb.
3054 (expand_case): Likewise.
3055 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
3056 cfun argument explicitly. Likewise.
3057 (make_edges_bb): Likewise.
3058 (make_cond_expr_edges): Likewise.
3059 (get_cases_for_edge): Likewise.
3060 (make_gimple_switch_edges): Likewise.
3061 (label_to_block_fn): Likewise.
3062 (label_to_block): Likewise.
3063 (make_goto_expr_edges): Likewise.
3064 (make_gimple_asm_edges): Likewise.
3065 (main_block_label): Likewise.
3066 (group_case_labels_stmt): Likewise.
3067 (find_taken_edge_computed_goto): Likewise.
3068 (find_taken_edge_switch_expr): Likewise.
3069 (gimple_verify_flow_info): Likewise.
3070 (gimple_redirect_edge_and_branch): Likewise.
3071 (gimple_switch_label_bb): New function.
3072 (gimple_switch_default_bb): Likewise.
3073 (gimple_switch_edge): Likewise.
3074 (gimple_switch_default_edge): Likewise.
3075 * tree-cfg.h (label_to_block_fn): Remove and replace ...
3076 (label_to_block): ... with this.
3077 (gimple_switch_label_bb): New.
3078 (gimple_switch_default_bb): Likewise.
3079 (gimple_switch_edge): Likewise.
3080 (gimple_switch_default_edge): Likewise.
3081 * tree-cfgcleanup.c (convert_single_case_switch): Use
3082 new gimple functions and pass new argument to label_to_block.
3083 (cleanup_control_flow_bb):
3084 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
3085 cfun argument explicitly.
3086 (make_eh_edges): Likewise.
3087 (redirect_eh_dispatch_edge): Likewise.
3088 (lower_resx): Likewise.
3089 (lower_eh_dispatch): Likewise.
3090 (maybe_remove_unreachable_handlers): Likewise.
3091 (unsplit_eh): Likewise.
3092 (cleanup_empty_eh): Likewise.
3093 (verify_eh_edges): Likewise.
3094 (verify_eh_dispatch_edge): Likewise.
3095 * tree-ssa-dom.c (record_edge_info): Likewise.
3096 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
3097 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3098 (thread_through_normal_block): Likewise.
3099 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3100 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3101 * tree-switch-conversion.c (switch_conversion::collect): Use new
3103 (switch_conversion::check_final_bb): Likewise.
3104 (switch_conversion::gather_default_values): Pass new argument
3106 (switch_conversion::build_constructors): Likewise.
3107 (switch_decision_tree::compute_cases_per_edge): Use new
3108 gimple_switch_edge function.
3109 (switch_decision_tree::analyze_switch_statement): Pass new argument
3111 (switch_decision_tree::try_switch_expansion): Use
3112 gimple_switch_default_edge.
3113 * tree-vrp.c (find_switch_asserts): Pass new argument
3115 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3116 (vr_values::simplify_switch_using_ranges): Likewise.
3118 2018-08-27 Richard Biener <rguenther@suse.de>
3120 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3121 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3123 * tree-ssa-sccvn.h (struct vn_pval): New structure.
3124 (struct vn_nary_op_s): Add unwind_to member. Add
3125 predicated_values flag and put result into a union together
3126 with a linked list of vn_pval.
3127 (struct vn_ssa_aux): Add name member to make maintaining
3128 a map of SSA name to vn_ssa_aux possible. Remove no longer
3129 needed info, dfsnum, low, visited, on_sccstack, use_processed
3130 and range_info_anti_range_p members.
3131 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3132 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3134 (vn_valueize): New global.
3135 (vn_context_bb): Likewise.
3136 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3137 VN_INFO_PTR_INFO): Remove.
3138 * tree-ssa-sccvn.c: ... (rewrite)
3139 (pass_fre::execute): For -O2+ initialize loops and run
3140 RPO VN in optimistic mode (iterating). For -O1 and -Og
3141 run RPO VN in non-optimistic mode.
3142 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3143 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3144 * doc/invoke.texi (sccvn-max-scc-size): Remove.
3145 (rpo-vn-max-loop-depth): Document.
3146 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3147 when valuezing the VUSE signals we walked out of the region.
3148 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3149 (phi_translate): Set VN context block to use for availability
3151 (compute_avail): Likewise.
3152 (pre_valueize): New function.
3153 (pass_pre::execute): Adjust to the RPO VN API.
3155 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3156 (propagate_constants_for_unrolling): Remove.
3157 (tree_unroll_loops_completely): Perform value-numbering
3158 on the unrolled bodies loop parent.
3160 2018-08-27 Richard Biener <rguenther@suse.de>
3162 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3163 for partial antic compute.
3165 2018-08-27 Jakub Jelinek <jakub@redhat.com>
3167 PR rtl-optimization/87065
3168 * combine.c (simplify_if_then_else): Formatting fix.
3169 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3171 (known_cond): Don't return const_true_rtx for vector modes. Use
3172 CONST0_RTX instead of const0_rtx. Formatting fixes.
3174 2018-08-27 Martin Liska <mliska@suse.cz>
3176 PR gcov-profile/87069
3177 * gcov.c (process_file): Record files already processed
3178 and warn about a file being processed multiple times.
3180 2018-08-27 Martin Liska <mliska@suse.cz>
3183 * config/aarch64/aarch64.c (aarch64_override_options_internal):
3184 Set default values for x_aarch64_*_string strings.
3185 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3186 prefix. For -mabi do not print '=ABI' in help and use
3187 <option_value> format for -msve-vector-bits and -moverride
3190 2018-08-26 Jeff Law <law@redhat.com>
3192 * config/mips/frame-header-opt.c: Include "backend.h" rather than
3195 2018-08-26 Marek Polacek <polacek@redhat.com>
3197 PR c++/87029, Implement -Wredundant-move.
3198 * doc/invoke.texi: Document -Wredundant-move.
3200 2018-08-25 Martin Sebor <msebor@redhat.com>
3202 PR tree-optimization/87059
3203 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3204 to the same type as the other.
3205 * fold-const.c (fold_binary_loc): Assert expectation.
3207 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
3209 * config/darwin.c (machopic_legitimize_pic_address): Clean up
3210 extraneous parentheses, dead code section and formatting.
3212 2018-08-24 David Malcolm <dmalcolm@redhat.com>
3215 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3216 wide enough for jumps in the line-numbering to be visible.
3217 (layout::print_gap_in_line_numbering): New member function.
3218 (layout::calculate_line_spans): When using line numbering, merge
3219 line spans that are only 1 line apart.
3220 (diagnostic_show_locus): When printing line numbers, show gaps in
3221 line numbering directly, rather than printing headers.
3222 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3223 line-numbering with multiple line spans.
3224 (selftest::test_fixit_insert_containing_newline_2): Add test of
3225 line-numbering, in which the spans are close enough to be merged.
3227 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3229 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3230 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
3231 * tree-vrp.c (range_is_nonnull): Remove.
3232 (range_includes_zero_p): Accept value_range instead of min/max.
3233 (extract_range_from_binary_expr_1): Do not early bail on
3235 Use range_includes_zero_p instead of range_is_nonnull.
3236 (extract_range_from_unary_expr): Use range_includes_zero_p instead
3237 of range_is_nonnull.
3238 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
3239 special case VR_ANTI_RANGE.
3240 (vrp_finalize): Same.
3241 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
3243 (range_is_nonnull): Remove.
3244 * vr-values.c (vrp_stmt_computes_nonzero): Use
3245 range_includes_zero_p instead of range_is_nonnull.
3246 (extract_range_basic): Pass value_range to range_includes_zero_p
3247 instead of range_is_nonnull.
3249 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
3251 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
3252 * emit-rtl.h (rtl_data): Remove return_bnd.
3253 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
3254 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
3255 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
3256 (POINTER_BOUNDS_MODE): Remove definition.
3257 (make_pointer_bounds_mode): Remove.
3258 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
3259 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
3260 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
3261 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
3262 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
3263 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
3264 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
3266 * config/i386/i386-modes.def (BND32, BND64): Remove.
3267 * config/i386/i386.c (dbx_register_map): Remove bound registers.
3268 (dbx64_register_map): Ditto.
3269 (svr4_dbx_register_map): Ditto.
3270 (indirect_thunk_bnd_needed): Remove.
3271 (indirect_thunks_bnd_used): Ditto.
3272 (indirect_return_bnd_needed): Ditto.
3273 (indirect_return_via_cx_bnd): Ditto.
3274 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
3275 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
3276 (output_indirect_thunk): Ditto. Remove need_prefix argument.
3277 (output_indirect_thunk_function): Remove handling of
3278 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
3279 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
3280 (ix86_save_reg): Remove handling of crtl->return_bnd.
3281 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
3282 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
3283 and UNSPEC_BNDLX_ADDR.
3284 (ix86_output_indirect_branch_via_reg): Remove handling of
3285 indirect_thunk_prefix_bnd.
3286 (ix86_output_indirect_branch_via_push): Ditto.
3287 (ix86_output_function_return): Ditto.
3288 (ix86_output_indirect_function_return): Ditto.
3289 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
3290 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
3291 (CALL_USED_REGISTERS): Ditto.
3292 (REG_ALLOC_ORDER): Update for removal of bound registers.
3293 (HI_REGISTER_NAMES): Ditto.
3294 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
3295 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
3296 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
3297 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
3298 (FIRST_PSEUDO_REG): Update.
3299 (BND): Remove mode iterator.
3300 * config/i386/predicates.md (bnd_mem_operator): Remove.
3302 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
3304 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
3307 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
3309 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
3310 the case in which the permute needs only a single element and
3311 repeats for every vector of the result. Extend that case to
3312 handle variable-length vectors.
3313 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3315 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
3318 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
3319 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
3321 2018-08-24 Richard Biener <rguenther@suse.de>
3323 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
3324 bb_flags_allocated members.
3325 (auto_flag): New RAII class for allocating flags.
3326 (auto_edge_flag): New RAII class for allocating edge flags.
3327 (auto_bb_flag): New RAII class for allocating bb flags.
3328 * cfgloop.c (verify_loop_structure): Allocate temporary edge
3330 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
3331 in favor of temporarily allocated BB flag.
3332 * hsa-brig.c: Re-order includes.
3333 * hsa-dump.c: Likewise.
3334 * hsa-regalloc.c: Likewise.
3335 * print-rtl.c: Likewise.
3336 * profile-count.c: Likewise.
3338 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
3341 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
3344 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
3347 * wide-int-range.cc (wide_int_range_div): Do not ignore result
3348 from wide_int_range_multiplicative_op.
3350 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3352 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
3355 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
3357 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
3360 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
3362 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
3363 full GENERIC expression used for replacement.
3365 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
3367 * tree-vrp.c (abs_extent_range): Remove.
3368 (extract_range_into_wide_ints): Pass wide ints by reference.
3369 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
3370 Pass wide ints by reference in all calls to
3371 extract_range_into_wide_ints.
3372 * wide-int-range.cc (wide_int_range_div): New.
3373 * wide-int-range.h (wide_int_range_div): New.
3374 (wide_int_range_includes_zero_p): New.
3375 (wide_int_range_zero_p): New.
3377 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
3379 * config/aarch64/aarch64.md (arches): New enum.
3380 (arch): New enum attr.
3381 (arch_enabled): New attr.
3382 (enabled): Now uses arch_enabled only.
3383 (simd, sve, fp16): Removed attribute.
3384 (fp): Attr now defined in terms of 'arch'.
3385 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
3386 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
3387 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
3388 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
3389 attributes into 'arch'.
3390 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
3391 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
3392 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
3393 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
3394 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
3395 'simd' attribute into 'arch'.
3396 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
3397 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
3398 Convert use of 'fp' attribute to 'arch'.
3399 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
3400 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
3402 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
3403 (different modes) Merge 'fp' and 'simd' into 'arch'.
3404 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
3407 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
3409 PR rtl-optimization/87026
3410 * expmed.c (canonicalize_comparison): If we can no longer create
3411 pseudoregisters, don't.
3413 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
3416 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
3418 * config/arm/arm.c (speculation_barrier_libfunc): New static
3420 (arm_init_libfuncs): Initialize it.
3421 (arm_emit_speculation_barrier): New function.
3422 * config/arm/arm.md (speculation_barrier): Call
3423 arm_emit_speculation_barrier for architectures that do not have
3425 (speculation_barrier_insn): Only match on Armv7 or later.
3427 2018-08-23 Richard Biener <rguenther@suse.de>
3430 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
3433 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
3435 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
3436 of single-vector TBLs.
3437 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
3440 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
3443 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
3444 aarch64_evpc_tbl guard.
3446 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
3448 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
3451 2018-08-22 Martin Sebor <msebor@redhat.com>
3454 * tree-pretty-print.c (pretty_print_string): Add argument.
3455 (dump_generic_node): Call to pretty_print_string with string size.
3457 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
3459 PR rtl-optimization/86771
3460 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
3461 of two SETs into those two SETs, one to be placed at i2, if that SETs
3462 destination is modified between i2 and i3.
3464 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3466 PR tree-optimization/86725
3467 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
3469 (vect_analyze_scalar_cycles_1): Check it.
3471 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3473 PR tree-optimization/86725
3474 * tree-vect-loop.c (vect_is_simple_reduction): When treating
3475 an outer loop phi as a double reduction, make sure that the
3476 single user of the phi result is an inner loop phi.
3478 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3480 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
3481 grouped stores with gaps to a strided group.
3483 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
3485 * tree-vect-stmts.c (get_group_load_store_type)
3486 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
3487 first statement in a group.
3489 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3491 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
3492 the sequence used in gcc/gcc.c.
3494 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3497 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
3500 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3502 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
3503 Darwin10-specific unwinder-shim.
3504 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
3505 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
3506 New to cater for Darwin10 Rosetta.
3508 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3510 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
3511 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
3513 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
3518 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
3519 (dwarf2out_switch_text_section): Generate a local label for the
3520 second function sub-section and apply it as the second FDE start
3522 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
3523 second sub-section start.
3525 2018-08-22 Richard Biener <rguenther@suse.de>
3527 PR tree-optimization/86988
3528 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
3530 2018-08-22 Richard Biener <rguenther@suse.de>
3532 PR tree-optimization/86945
3533 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
3535 2018-08-22 Alexandre Oliva <oliva@adacore.com>
3537 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
3538 a comment about how uses of r2 for .sdata2 come about.
3540 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
3542 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
3544 2018-08-21 Marek Polacek <polacek@redhat.com>
3546 PR c++/86981, Implement -Wpessimizing-move.
3547 * doc/invoke.texi: Document -Wpessimizing-move.
3549 2018-08-21 Jan Hubicka <jh@suse.cz>
3551 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
3552 * tree.h (is_redundant_typedef): Remove.
3553 * dwarf2out.c (is_redundant_typedef): Turn into static function.
3555 2018-08-21 Jan Hubicka <jh@suse.cz>
3557 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
3560 2018-08-21 Tamar Christina <tamar.christina@arm.com>
3562 * expmed.c (extract_low_bits): Reject invalid subregs early.
3564 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
3567 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
3570 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3572 * config/vxworks.h: Guard vxworks_asm_out_constructor and
3573 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
3574 * config/vxworks.c: Likewise.
3576 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3578 * config/vxworks.c: Set targetm.have_ctors_dtors
3579 if HAVE_INITFINI_ARRAY_SUPPORT.
3580 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
3581 if HAVE_INITFINI_ARRAY_SUPPORT.
3583 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
3585 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
3586 default search path for VxWorks < 7.
3588 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
3590 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
3591 (get_format_string): Refer to c_getstr.
3593 2018-08-21 Tom de Vries <tdevries@suse.de>
3595 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3596 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3597 (debuginfo_early_stop): Declare.
3598 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3599 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3600 (debuginfo_early_stop): New function.
3601 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
3602 and debuginfo_early_stop.
3603 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
3604 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
3605 (general_init): Call debuginfo_early_init.
3606 (finalize): Call debuginfo_fini.
3607 (do_compile): Call debuginfo_init.
3608 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
3610 (@item -fdump-debug, @item -fdump-earlydebug): Add.
3612 2018-08-21 Tom de Vries <tdevries@suse.de>
3614 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
3615 flag_dump_noaddr and flag_dump_unnumbered.
3617 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
3619 * wide-int-range.cc (wide_int_range_abs): New.
3620 (wide_int_range_order_set): Rename from wide_int_range_min_max.
3621 * wide-int-range.h (wide_int_range_abs): New.
3622 (wide_int_range_min_max): New.
3623 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
3624 case to call wide_int_range_abs.
3625 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
3626 (extract_range_from_abs_expr): Delete.
3628 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
3631 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
3632 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
3635 2018-08-20 Nathan Sidwell <nathan@acm.org>
3636 Jeff Law <law@redhat.com>
3638 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
3639 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
3641 2018-08-20 David Malcolm <dmalcolm@redhat.com>
3644 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
3645 (decl_attributes): Likewise.
3646 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
3648 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3649 * diagnostic-core.h (class auto_diagnostic_group): New class.
3650 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
3651 (diagnostic_report_diagnostic): Handle the first diagnostics within
3653 (emit_diagnostic): Add auto_diagnostic_group instance.
3655 (inform_n): Likewise.
3656 (warning): Likewise.
3657 (warning_at): Likewise.
3658 (warning_n): Likewise.
3659 (pedwarn): Likewise.
3660 (permerror): Likewise.
3662 (error_n): Likewise.
3663 (error_at): Likewise.
3665 (fatal_error): Likewise.
3666 (internal_error): Likewise.
3667 (internal_error_no_backtrace): Likewise.
3668 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
3669 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
3670 * diagnostic.h (struct diagnostic_context): Add fields
3671 "diagnostic_group_nesting_depth",
3672 "diagnostic_group_emission_count", "begin_group_cb",
3674 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
3675 Add auto_diagnostic_group instance(s).
3676 (find_explicit_erroneous_behavior): Likewise.
3677 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
3678 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
3679 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
3680 (gimplify_va_arg_expr): Likewise.
3681 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
3682 (HSA_SORRY_AT): Likewise.
3683 * ipa-devirt.c (compare_virtual_tables): Likewise.
3684 (warn_odr): Likewise.
3685 * multiple_target.c (expand_target_clones): Likewise.
3686 * opts-common.c (cmdline_handle_error): Likewise.
3687 * reginfo.c (globalize_reg): Likewise.
3688 * substring-locations.c (format_warning_n_va): Likewise.
3689 * tree-inline.c (expand_call_inline): Likewise.
3690 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
3691 * tree-ssa-loop-niter.c
3692 (do_warn_aggressive_loop_optimizations): Likewise.
3693 * tree-ssa-uninit.c (warn_uninit): Likewise.
3694 * tree.c (warn_deprecated_use): Likewise.
3696 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
3699 * config/i386/i386.md (eh_return): Always update EH return
3700 address in word_mode.
3702 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
3704 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
3705 TARGET_SPLIT_COMPLEX_ARG is defined.
3707 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3709 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
3711 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
3714 * expr.c (expand_assignment): Assert that bitpos is positive.
3715 (store_field): Likewise
3716 (expand_expr_real_1): Make sure that bitpos is positive.
3717 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
3720 2018-08-20 Nathan Sidwell <nathan@acm.org>
3722 * Makefile.in (CPP_ID_DATA_H): Delete.
3723 (CPP_INTERNAL_H): Don't add it.
3724 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
3725 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
3727 2018-08-20 Richard Biener <rguenther@suse.de>
3729 PR tree-optimization/78655
3730 * tree-vrp.c (extract_range_from_binary_expr_1): Make
3731 pointer + offset nonnull if either operand is nonnull work.
3733 2018-08-20 Tom de Vries <tdevries@suse.de>
3735 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
3736 unless the referenced die describes the added property using
3737 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
3738 Otherwise, add a DW_AT_location to the referenced die.
3740 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
3743 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
3744 register_operand when calling ix86_set_reg_reg_cost.
3745 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
3746 Set *total to 0 for operands that satisfy x86_64_immediate_operand
3747 predicate and to 1 otherwise.
3749 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
3751 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
3752 emit a diagnostic that it is not supported and reset the option.
3753 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
3754 supported and consume the option. (ASM_FINAL_SPEC): New.
3756 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3758 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
3761 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
3763 C-SKY port: Documentation
3765 * doc/extend.texi (C-SKY Function Attributes): New section.
3766 * doc/invoke.texi (Option Summary): Add C-SKY options.
3767 (C-SKY Options): New section.
3768 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
3770 2018-08-17 Jojo <jijie_rong@c-sky.com>
3771 Huibin Wang <huibin_wang@c-sky.com>
3772 Sandra Loosemore <sandra@codesourcery.com>
3773 Chung-Lin Tang <cltang@codesourcery.com>
3775 C-SKY port: Backend implementation
3777 * config/csky/*: New.
3778 * common/config/csky/*: New.
3780 2018-08-17 Jojo <jijie_rong@c-sky.com>
3781 Huibin Wang <huibin_wang@c-sky.com>
3782 Sandra Loosemore <sandra@codesourcery.com>
3783 Chung-Lin Tang <cltang@codesourcery.com>
3784 Andrew Jenner <andrew@codesourcery.com>
3786 C-SKY port: Configury
3788 * config.gcc (csky-*-*): New.
3789 * configure.ac: Add csky to targets for dwarf2 debug_line support.
3790 * configure: Regenerated.
3792 2018-08-17 David Malcolm <dmalcolm@redhat.com>
3794 * dump-context.h: Include "dumpfile.h".
3795 (dump_context::dump_printf_va): Convert final param from va_list
3796 to va_list *. Convert from ATTRIBUTE_PRINTF to
3797 ATTRIBUTE_GCC_DUMP_PRINTF.
3798 (dump_context::dump_printf_loc_va): Likewise.
3799 * dumpfile.c: Include "stringpool.h".
3800 (make_item_for_dump_printf_va): Delete.
3801 (make_item_for_dump_printf): Delete.
3802 (class dump_pretty_printer): New class.
3803 (dump_pretty_printer::dump_pretty_printer): New ctor.
3804 (dump_pretty_printer::emit_items): New member function.
3805 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
3807 (dump_pretty_printer::emit_item): New member function.
3808 (dump_pretty_printer::stash_item): New member function.
3809 (dump_pretty_printer::format_decoder_cb): New member function.
3810 (dump_pretty_printer::decode_format): New member function.
3811 (dump_context::dump_printf_va): Reimplement in terms of
3812 dump_pretty_printer.
3813 (dump_context::dump_printf_loc_va): Convert final param from va_list
3815 (dump_context::begin_scope): Reimplement call to
3816 make_item_for_dump_printf.
3817 (dump_printf): Update for change to dump_printf_va.
3818 (dump_printf_loc): Likewise.
3819 (selftest::test_capture_of_dump_calls): Convert "stmt" from
3820 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
3821 with %T, %E, and %G.
3822 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
3823 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
3824 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
3825 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
3826 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
3827 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
3828 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
3829 within a dump_printf_loc call to "%wu".
3830 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
3831 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
3832 missing space after "=".
3833 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
3834 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
3835 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
3836 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
3837 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
3838 duplicate "vectorized" from message.
3840 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
3842 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
3843 polyNxK_t element's TYPE_STRING_FLAG.
3845 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
3847 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
3848 (they were unnamed before). Fix comments.
3850 2018-08-17 Nathan Sidwell <nathan@acm.org>
3852 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
3854 2018-08-17 Richard Biener <rguenther@suse.de>
3856 PR tree-optimization/86841
3857 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
3859 2018-08-17 Martin Liska <mliska@suse.cz>
3861 * common.opt: Remove Warn, Init and Report for options with
3862 Ignore/Deprecated flag. Warning is done automatically for
3864 * config/i386/i386.opt: Likewise.
3865 * config/ia64/ia64.opt: Likewise.
3866 * config/rs6000/rs6000.opt: Likewise.
3867 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
3868 Remove usage of flag_check_pointer_bounds.
3869 * lto-wrapper.c (merge_and_complain): Do not handle
3870 OPT_fcheck_pointer_bounds.
3871 (append_compiler_options): Likewise.
3872 * opt-functions.awk: Do not handle Deprecated.
3873 * optc-gen.awk: Check that Var, Report and Init are not
3874 used for an option with Ignore/Deprecated flag.
3875 * opts-common.c (decode_cmdline_option): Do not report
3877 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
3879 * opts.h (struct cl_option): Remove cl_deprecated flag.
3880 (CL_ERR_DEPRECATED): Remove error enum value.
3882 2018-08-17 Richard Biener <rguenther@suse.de>
3885 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
3886 across a va-arg-pack using call adjust its return value accordingly.
3888 2018-08-16 Martin Sebor <msebor@redhat.com>
3890 PR tree-optimization/86853
3891 * gimple-ssa-sprintf.c (struct format_result): Rename member.
3892 (struct fmtresult): Add member and initialize it in ctors.
3893 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
3894 (format_string): Handle %S the same as %ls. Set MAYFAIL.
3895 (format_directive): Set POSUNDER4K when MAYFAIL is set.
3896 (parse_directive): Handle %C same as %c.
3897 (sprintf_dom_walker::compute_format_length): Adjust.
3898 (is_call_safe): Adjust.
3900 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
3902 * builtins.c (c_strlen): Add new parameter eltsize. Use it
3903 for determining how to count the elements.
3904 * builtins.h (c_strlen): Adjust prototype.
3905 * expr.c (string_constant): Add new parameter mem_size.
3906 Set *mem_size appropriately.
3907 * expr.h (string_constant): Adjust protoype.
3908 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
3909 * gimple-fold.h (get_range_strlen): Adjust prototype.
3910 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
3911 (format_string): Call get_string_length with eltsize.
3913 2018-08-16 David Malcolm <dmalcolm@redhat.com>
3915 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
3916 to emit the span, rather than setting it as the prefix.
3918 2018-08-16 David Malcolm <dmalcolm@redhat.com>
3920 * diagnostic-show-locus.c (layout::start_annotation_line): Add
3921 "margin_char" parameter, defaulting to space. Use it in place
3922 of pp_space for the initial part of the margin.
3923 (layout::print_leading_fixits): Use '+' when filling the margin
3924 of line-insertion fix-it hints.
3926 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3928 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
3931 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3933 * config/rs6000/altivec.md: Don't set length attribute to the default
3935 * config/rs6000/darwin.md: Ditto.
3936 * config/rs6000/dfp.md: Ditto.
3937 * config/rs6000/htm.md: Ditto.
3938 * config/rs6000/rs6000.md: Ditto.
3939 * config/rs6000/sync.md: Ditto.
3940 * config/rs6000/vsx.md: Ditto.
3942 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3944 * config/rs6000/altivec.md: Don't set length attribute to the default
3945 value, for branch instructions.
3946 * config/rs6000/darwin.md: Ditto.
3947 * config/rs6000/rs6000.md: Ditto.
3949 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
3951 * config/rs6000/rs6000.md (length): Always define as const_int 4.
3952 (unnamed conditional branch define_insn): Set length to 4 or 8
3953 depending on offset.
3954 (<bd>_<mode>): Similar, for alternative 0.
3955 (<bd>tf_<mode>): Ditto.
3957 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3959 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
3961 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
3963 * doc/rtl.texi: Replace old RTX class names with new names.
3966 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
3968 * expmed.h (canonicalize_comparison): New declaration.
3969 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
3970 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
3971 * optabs.c (prepare_cmp_insn): Likewise.
3972 * rtl.h (unsigned_condition_p): New function which checks if a
3973 comparison operator is unsigned.
3975 2018-08-16 Nathan Sidwell <nathan@acm.org>
3977 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
3978 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
3980 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3983 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
3984 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
3985 (mov<mov>): ..this and enable unconditionally.
3987 2018-08-16 Tamar Christina <tamar.christina@arm.com>
3989 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
3991 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
3993 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
3994 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
3995 "Common" with "Target".
3997 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
3999 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
4000 * doc/invoke.texi (mmitigate-rop): Remove.
4001 * config/i386/i386.c: Do not include "regrename.h".
4002 (ix86_rop_should_change_byte_p, reg_encoded_number)
4003 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
4005 (ix86_reorg): Remove call to ix86_mitigate_rop.
4006 * config/i386/i386.md (attr "modrm_class"): Remove.
4007 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
4008 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
4009 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
4011 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
4013 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
4014 allow folding of mergeh() and mergel() for the float and double types.
4015 (fold_mergehl_helper): Rework to handle building a permute tree
4018 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
4020 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
4023 2018-08-15 David Malcolm <dmalcolm@redhat.com>
4025 * common.opt (fdiagnostics-show-labels): New option.
4026 * diagnostic-show-locus.c (class layout_range): Add field
4028 (class layout): Add field "m_show_labels_p".
4029 (layout_range::layout_range): Add param "label" and use it to
4031 (make_range): Pass in NULL for new "label" param of layout_range's
4033 (layout::layout): Initialize m_show_labels_p.
4034 (layout::maybe_add_location_range): Pass in loc_range->m_label
4035 when constructing layout_range instances.
4036 (struct line_label): New struct.
4037 (layout::print_any_labels): New member function.
4038 (layout::print_line): Call it if label-printing is enabled.
4039 (selftest::test_one_liner_labels): New test.
4040 (selftest::test_diagnostic_show_locus_one_liner): Call it.
4041 * diagnostic.c (diagnostic_initialize): Initialize
4042 context->show_labels_p.
4043 * diagnostic.h (struct diagnostic_context): Add field
4045 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4046 -fno-diagnostics-show-labels.
4047 * dwarf2out.c (gen_producer_string): Add
4048 OPT_fdiagnostics_show_labels to the ignored options.
4049 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
4051 (gcc_rich_location::maybe_add_expr): Likewise.
4052 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
4053 label" param, defaulting to NULL.
4054 (gcc_rich_location::add_expr): Add "label" param.
4055 (gcc_rich_location::maybe_add_expr): Likewise.
4056 (class text_range_label): New class.
4057 (class range_label_for_type_mismatch): New class.
4058 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
4059 of format_warning_va.
4060 (fmtwarn_n): Likewise for new params of format_warning_n_va.
4061 * lto-wrapper.c (merge_and_complain): Add
4062 OPT_fdiagnostics_show_labels to the "pick one setting" options.
4063 (append_compiler_options): Likewise to the dropped options.
4064 (append_diag_options): Likewise to the passed-on options.
4065 * opts.c (common_handle_option): Handle the new option.
4066 * selftest-diagnostic.c
4067 (test_diagnostic_context::test_diagnostic_context): Enable
4069 * substring-locations.c: Include "gcc-rich-location.h".
4070 (format_warning_n_va): Add "fmt_label" and "param_label" params
4071 and use them as appropriate.
4072 (format_warning_va): Add "fmt_label" and "param_label" params,
4073 passing them on to format_warning_n_va.
4074 (format_warning_at_substring): Likewise.
4075 (format_warning_at_substring_n): Likewise.
4076 * substring-locations.h (format_warning_va): Add "fmt_label" and
4077 "param_label" params.
4078 (format_warning_n_va): Likewise.
4079 (format_warning_at_substring): Likewise.
4080 (format_warning_at_substring_n): Likewise.
4081 * toplev.c (general_init): Initialize global_dc->show_labels_p.
4083 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
4086 * builtins.c (expand_builtin_memcmp): Do not expand the call
4087 when overflow is detected.
4089 2018-08-15 Martin Sebor <msebor@redhat.com>
4091 PR tree-optimization/71625
4092 * config/aarch64/aarch64-builtins.c
4093 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
4095 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
4097 * config/s390/s390.c (s390_reorg): Remove loop.
4099 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4102 (darwin_function_switched_text_sections): Delete.
4103 * gcc/config/darwin.h
4104 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4106 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
4109 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4110 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
4111 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4113 2018-08-15 Martin Liska <mliska@suse.cz>
4115 PR tree-optimization/86925
4116 * predict.c (expr_expected_value_1): When taking
4117 later predictor, assign also probability.
4118 Use fold_build2_initializer_loc in order to fold
4119 the expression in -frounding-math.
4121 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
4123 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4125 (expand_vec_perm_1): Try the new method.
4127 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
4130 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4131 Check whether lra_live_max_point is 0 before dividing.
4133 2018-08-14 Martin Sebor <msebor@redhat.com>
4135 PR tree-optimization/86650
4136 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4137 (vrp_prop::check_mem_ref): Same.
4139 2018-08-13 Liu Hao <lh_mouse@126.com>
4141 * pretty-print.c (eat_esc_sequence): Swap the foreground and
4142 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4143 and clear it thereafter, as it only works for DBCS.
4145 2018-08-13 Liu Hao <lh_mouse@126.com>
4147 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4148 handle returned by _get_osf_handle().
4150 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4152 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4153 for folding vec_perm.
4155 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4157 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4158 Add support for gimple-folding of vec_pack() and vec_unpack()
4161 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
4163 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4164 vec_xst variants to the list.
4165 (rs6000_gimple_fold_builtin): Add support for folding unaligned
4166 vector loads and stores.
4168 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
4170 * config.gcc (rs6000-ibm-aix4.x): Delete.
4171 (rs6000-ibm-aix5.1): Delete.
4172 (rs6000-ibm-aix5.2): Delete.
4173 (rs6000-ibm-aix5.3): Delete.
4174 * config/rs6000/aix43.h: Delete.
4175 * config/rs6000/aix51.h: Delete.
4176 * config/rs6000/aix52.h: Delete.
4177 * config/rs6000/t-aix43: Delete.
4179 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
4181 * config/s390/s390.c (s390_decompose_constant_pool_ref):
4183 (s390_decompose_address): Factor out constant pool ref
4186 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4188 * config/nds32/nds32-predicates.c
4189 (nds32_can_use_bclr_p): Change return type as bool.
4190 (nds32_can_use_bset_p): Ditto.
4191 (nds32_can_use_btgl_p): Ditto.
4192 (nds32_can_use_bitci_p): Ditto.
4193 * config/nds32/nds32-protos.h
4194 (nds32_can_use_bclr_p): Change declaration.
4195 (nds32_can_use_bset_p): Ditto.
4196 (nds32_can_use_btgl_p): Ditto.
4197 (nds32_can_use_bitci_p): Ditto.
4199 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4201 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4202 Support -msched-prolog-epilog option.
4203 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4205 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4207 * common/config/nds32/nds32-common.c
4208 (nds32_option_optimization_table): Enalbe -malways-align.
4210 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
4212 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4214 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4215 OPT_misr_secure_ case.
4216 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4217 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4218 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4220 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4221 (nds32_isr_info): New field security_level.
4222 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4223 * config/nds32/nds32.md (return_internal): Consider critical attribute.
4224 * config/nds32/nds32.opt (misr-secure): New option.
4225 * config/nds32/nds32_init.inc: New file.
4226 * config/nds32/nds32_isr.h: New file.
4228 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
4230 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4231 Update comment for atomic instructions.
4232 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4233 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
4235 (atomic_loaddi): Revise fence expansion to only emit fence prior to
4236 load for __ATOMIC_SEQ_CST model.
4237 (atomic_loaddi_1): Remove float register target.
4238 (atomic_storedi): Handle CONST_INT values.
4239 (atomic_storedi_1): Remove float register source. Add special case
4241 (memory_barrier): New expander and insn.
4243 2018-08-11 Jakub Jelinek <jakub@redhat.com>
4245 PR tree-optimization/86835
4246 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
4247 new_stmt after def_gsi, make sure to insert new_square_stmt after
4248 that stmt, not 2 stmts before it.
4250 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
4253 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
4256 2018-08-10 Martin Liska <mliska@suse.cz>
4259 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
4261 * builtins.c (expand_builtin_expect_with_probability):
4263 (expand_builtin_expect_with_probability): New function.
4264 (build_builtin_expect_predicate): Add new argumnet probability
4265 for BUILT_IN_EXPECT_WITH_PROBABILITY.
4266 (fold_builtin_expect):
4269 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
4270 * builtins.h (fold_builtin_expect): Set new argument.
4271 * doc/extend.texi: Document __builtin_expect_with_probability.
4272 * doc/invoke.texi: Likewise.
4273 * gimple-fold.c (gimple_fold_call): Pass new argument.
4274 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
4275 also BUILT_IN_EXPECT_WITH_PROBABILITY.
4276 * predict.c (get_predictor_value): New function.
4277 (expr_expected_value): Add new argument probability. Assume
4278 that predictor and probability are always non-null.
4279 (expr_expected_value_1): Likewise. For __builtin_expect and
4280 __builtin_expect_with_probability set probability. Handle
4281 combination in binary expressions.
4282 (tree_predict_by_opcode): Simplify code by simply calling
4283 get_predictor_value.
4284 (pass_strip_predict_hints::execute): Add handling of
4285 BUILT_IN_EXPECT_WITH_PROBABILITY.
4286 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
4288 * tree.h (DECL_BUILT_IN_P): New function.
4290 2018-08-10 Martin Liska <mliska@suse.cz>
4292 PR tree-optimization/85799
4293 * passes.def: Add argument for pass_strip_predict_hints.
4294 * predict.c (class pass_strip_predict_hints): Add new argument
4296 (strip_predictor_early): New function.
4297 (pass_strip_predict_hints::execute): Call the function to
4299 (strip_predict_hints): New function.
4300 * predict.def: Fix comment.
4302 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
4304 * Makefile.in: Clarify which tm.texi to copy over to assert the
4305 right to grant a GFDL license for all.
4307 2018-08-09 Jeff Law <law@redhat.com>
4309 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
4312 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
4314 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
4317 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4319 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
4320 reductions for variable-length vectors.
4322 2018-08-09 David Malcolm <dmalcolm@redhat.com>
4325 * common.opt (fdiagnostics-show-line-numbers): New option.
4326 * diagnostic-show-locus.c (class layout): Add fields
4327 "m_show_line_numbers_p" and "m_linenum_width";
4328 (num_digits): New function.
4329 (test_num_digits): New function.
4330 (layout::layout): Initialize new fields. Update m_x_offset
4331 logic to handle any left margin.
4332 (layout::print_source_line): Print line number when requested.
4333 (layout::start_annotation_line): New member function.
4334 (layout::print_annotation_line): Call it.
4335 (layout::print_leading_fixits): Likewise.
4336 (layout::print_trailing_fixits): Likewise. Update calls to
4337 move_to_column for new parameter.
4338 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
4339 it to potentially call start_annotation_line.
4340 (layout::show_ruler): Call start_annotation_line.
4341 (selftest::test_line_numbers_multiline_range): New selftest.
4342 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
4343 and selftest::test_line_numbers_multiline_range.
4344 * diagnostic.c (diagnostic_initialize): Initialize
4345 show_line_numbers_p.
4346 * diagnostic.h (struct diagnostic_context): Add field
4347 "show_line_numbers_p".
4348 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4349 -fno-diagnostics-show-line-numbers.
4350 * dwarf2out.c (gen_producer_string): Add
4351 OPT_fdiagnostics_show_line_numbers to the ignored options.
4352 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
4353 one setting" options.
4354 (append_compiler_options): Likewise to the dropped options.
4355 (append_diag_options): Likewise to the passed-on options.
4356 * opts.c (common_handle_option): Handle the new option.
4357 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
4359 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
4361 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
4362 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
4363 third argument of type "const signed char" to descriptions of
4364 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
4365 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
4366 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
4367 __builtin_bcdsub_ov functions.
4369 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4371 PR tree-optimization/86858
4372 * tree-vect-loop.c (vect_is_simple_reduction): Restore
4373 flow_bb_inside_loop_p calls.
4375 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
4377 PR tree-optimization/86871
4378 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
4379 instead of gimple_assign_lhs.
4381 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
4384 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
4385 register constraint to operand 0.
4386 (add<mode>3_carryinC): Likewise.
4387 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
4389 2018-08-09 Martin Liska <mliska@suse.cz>
4392 * common.opt: Remove extra line.
4394 2018-08-09 Martin Liska <mliska@suse.cz>
4396 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
4397 at the end of a line, make first letter capital and end up
4398 a sentence with a dot.
4399 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
4400 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
4401 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
4402 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
4403 (PARAM_MAX_ISL_OPERATIONS): Likewise.
4404 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
4405 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
4406 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
4407 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
4408 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
4409 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
4410 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
4411 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
4412 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
4413 (PARAM_TREE_REASSOC_WIDTH): Likewise.
4414 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
4415 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
4416 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
4418 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
4421 * config/s390/s390.c (s390_option_override_internal): Reduce the
4422 stack-clash-protection-probe-interval param if it would be too big
4425 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
4428 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
4429 * config/m68k/m68k.c (handle_move_double): Don't call
4430 m68k_final_prescan_insn.
4431 (m68k_adjust_decorated_operand): Renamed from
4432 m68k_final_prescan_insn, remove first and third operand and
4434 (print_operand): Call it.
4435 (print_operand_address): Call it.
4437 2018-08-08 Nathan Sidwell <nathan@acm.org>
4439 * diagnostic.c (diagnostic_report_current_module): Use
4440 linemap_included_from & linemap_included_from_linemap.
4442 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
4444 * config/aarch64/aarch64-cores.def: Add phecda core.
4445 * config/aarch64/aarch64-tune.md: Regenerate.
4446 * doc/invoke.texi: Add phecda core.
4448 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
4451 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
4453 * config/s390/s390.md ("movti"): Add more alternatives for
4454 constant to GPR copies.
4456 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
4458 * config/s390/s390.c: Fix whitespace damage throughout the file.
4459 * config/s390/s390.h: Likewise.
4460 * config/s390/tpf.h: Likewise.
4462 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
4464 * config/s390/s390.c (s390_loadrelative_operand_p):
4465 Remove TARGET_CPU_ZARCH usages.
4466 (s390_rtx_costs): Likewise.
4467 (s390_legitimate_constant_p): Likewise.
4468 (s390_cannot_force_const_mem): Likewise.
4469 (legitimate_reload_constant_p): Likewise.
4470 (s390_preferred_reload_class): Likewise.
4471 (legitimize_pic_address): Likewise.
4472 (legitimize_tls_address): Likewise.
4473 (s390_split_branches): Removed.
4474 (s390_add_execute): Removed.
4475 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
4476 (s390_mainpool_start): Likewise.
4477 (s390_mainpool_finish): Likewise.
4478 (s390_mainpool_cancel): Removed.
4479 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
4480 (s390_chunkify_cancel): Likewise.
4481 (s390_return_addr_rtx): Likewise.
4482 (s390_register_info): Remove split_branches_pending_p uages.
4483 (s390_optimize_register_info): Likewise.
4484 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
4485 split_branches_pending_p usages.
4486 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
4487 (s390_load_got): Likewise.
4488 (s390_expand_split_stack_prologue): Likewise.
4489 (output_asm_nops): Likewise.
4490 (s390_function_profiler): Likewise.
4491 (s390_emit_call): Likewise.
4492 (s390_conditional_register_usage): Likewise.
4493 (s390_optimize_prologue): Likewise.
4494 (s390_reorg): Remove TARGET_CPU_ZARCH and
4495 split_branches_pending_p usages.
4496 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
4498 (s390_output_indirect_thunk_function): Likewise.
4499 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
4500 (TARGET_CPU_ZARCH_P): Removed.
4501 (struct machine_function): Remove split_branches_pending_p.
4502 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
4504 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
4506 * common/config/s390/s390-common.c (processor_flags_table):
4508 * config.gcc: Remove with_arch/with_tune support.
4509 * config/s390/2064.md: Remove cpu attribute comparisons.
4510 * config/s390/driver-native.c (s390_host_detect_local_cpu):
4512 * config/s390/linux.h (ASM_SPEC):
4513 Remove -march support.
4514 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
4515 Use a table to get an arch level.
4516 * config/s390/s390-opts.h (enum processor_type):
4518 * config/s390/s390.c
4519 (processor_table): Remove entries, add arch_level values.
4520 (s390_issue_rate): Remove cases.
4521 (s390_option_override): Adjust
4522 s390_option_override_internal() call.
4523 (s390_option_override_internal): Remove deprecation warning.
4524 (s390_valid_target_attribute_tree): Adjust
4525 s390_option_override_internal() call.
4526 * config/s390/s390.h (struct s390_processor):
4527 Share with s390-c.c, add arch_level field.
4528 * config/s390/s390.md:
4529 Remove occurrences in cpu attribute.
4530 * config/s390/s390.opt: Remove -march/-mtune support.
4531 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
4532 * doc/invoke.texi: Remove deprecation warning.
4534 2018-08-08 Luis Machado <luis.machado@linaro.org>
4536 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
4538 (qdf24xx_tunings): Set vector cost structure to
4539 qdf24xx_vector_cost.
4541 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
4542 <register_sextend>: Set to 3.
4544 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
4547 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
4548 * config/aarch64/aarch64-simd.md
4549 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
4550 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
4551 (aarch64_frecpx<mode>): ...this new pattern.
4552 * config/aarch64/aarch64-simd-builtins.def: Remove comment
4553 about aarch64_frecp<FRECP:frecp_suffix><mode>.
4555 2018-08-07 Martin Liska <mliska@suse.cz>
4558 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
4559 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
4560 * predict.def (PRED_MALLOC_NONNULL): New predictor.
4561 * doc/extend.texi: Document that malloc attribute adds
4564 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
4567 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4568 Define to speculation_safe_value_not_needed.
4570 2018-08-06 Jeff Law <law@redhat.com>
4572 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
4573 the vr_values instance to cprop_into_stmt.
4574 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
4575 (cprop_operand): Also query EVRP to determine if OP is a constant.
4577 2018-08-06 Nathan Sidwell <nathan@acm.org>
4579 * diagnostic.c (diagnostic_report_current_module): Reroll
4580 included-at loop. Translate text.
4582 2018-08-06 David Malcolm <dmalcolm@redhat.com>
4584 * function-tests.c (selftest::test_expansion_to_rtl): Call
4585 free_after_compilation.
4587 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4589 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
4591 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
4593 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
4594 loops with memory block operations from getting unrolled.
4596 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
4599 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4600 Define to speculation_safe_value_not_needed.
4602 2018-08-06 Jeff Law <law@redhat.com>
4604 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
4607 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
4610 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
4611 with all enabled __intN types.
4613 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
4615 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4617 * alias.c (record_set): Check for clobber high.
4618 * cfgexpand.c (expand_gimple_stmt): Likewise.
4619 * combine-stack-adj.c (single_set_for_csa): Likewise.
4620 * combine.c (find_single_use_1): Likewise.
4621 (set_nonzero_bits_and_sign_copies): Likewise.
4622 (get_combine_src_dest): Likewise.
4623 (is_parallel_of_n_reg_sets): Likewise.
4624 (try_combine): Likewise.
4625 (record_dead_and_set_regs_1): Likewise.
4626 (reg_dead_at_p_1): Likewise.
4627 (reg_dead_at_p): Likewise.
4628 * dce.c (deletable_insn_p): Likewise.
4629 (mark_nonreg_stores_1): Likewise.
4630 (mark_nonreg_stores_2): Likewise.
4631 * df-scan.c (df_find_hard_reg_defs): Likewise.
4632 (df_uses_record): Likewise.
4633 (df_get_call_refs): Likewise.
4634 * dwarf2out.c (mem_loc_descriptor): Likewise.
4635 * haifa-sched.c (haifa_classify_rtx): Likewise.
4636 * ira-build.c (create_insn_allocnos): Likewise.
4637 * ira-costs.c (scan_one_insn): Likewise.
4638 * ira.c (equiv_init_movable_p): Likewise.
4639 (rtx_moveable_p): Likewise.
4640 (interesting_dest_for_shprep): Likewise.
4641 * jump.c (mark_jump_label_1): Likewise.
4642 * postreload-gcse.c (record_opr_changes): Likewise.
4643 * postreload.c (reload_cse_simplify): Likewise.
4644 (struct reg_use): Add source expr.
4645 (reload_combine): Check for clobber high.
4646 (reload_combine_note_use): Likewise.
4647 (reload_cse_move2add): Likewise.
4648 (move2add_note_store): Likewise.
4649 * print-rtl.c (print_pattern): Likewise.
4650 * recog.c (decode_asm_operands): Likewise.
4651 (store_data_bypass_p): Likewise.
4652 (if_test_bypass_p): Likewise.
4653 * regcprop.c (kill_clobbered_value): Likewise.
4654 (kill_set_value): Likewise.
4655 * reginfo.c (reg_scan_mark_refs): Likewise.
4656 * reload1.c (maybe_fix_stack_asms): Likewise.
4657 (eliminate_regs_1): Likewise.
4658 (elimination_effects): Likewise.
4659 (mark_not_eliminable): Likewise.
4660 (scan_paradoxical_subregs): Likewise.
4661 (forget_old_reloads_1): Likewise.
4662 * reorg.c (find_end_label): Likewise.
4663 (try_merge_delay_insns): Likewise.
4664 (redundant_insn): Likewise.
4665 (own_thread_p): Likewise.
4666 (fill_simple_delay_slots): Likewise.
4667 (fill_slots_from_thread): Likewise.
4668 (dbr_schedule): Likewise.
4669 * resource.c (update_live_status): Likewise.
4670 (mark_referenced_resources): Likewise.
4671 (mark_set_resources): Likewise.
4672 * rtl.c (copy_rtx): Likewise.
4673 * rtlanal.c (reg_referenced_p): Likewise.
4674 (single_set_2): Likewise.
4675 (noop_move_p): Likewise.
4676 (note_stores): Likewise.
4677 * sched-deps.c (sched_analyze_reg): Likewise.
4678 (sched_analyze_insn): Likewise.
4680 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4682 * cse.c (invalidate_reg): New function extracted from...
4683 (invalidate): ...here.
4684 (canonicalize_insn): Check for clobber high.
4685 (invalidate_from_clobbers): invalidate clobber highs.
4686 (invalidate_from_sets_and_clobbers): Likewise.
4687 (count_reg_usage): Check for clobber high.
4688 (insn_live_p): Likewise.
4689 * cselib.c (cselib_expand_value_rtx_1):Likewise.
4690 (cselib_invalidate_regno): Check for clobber in setter.
4691 (cselib_invalidate_rtx): Pass through setter.
4692 (cselib_invalidate_rtx_note_stores):
4693 (cselib_process_insn): Check for clobber high.
4694 * cselib.h (cselib_invalidate_rtx): Add operand.
4696 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4698 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
4699 (mark_not_eliminable): Likewise.
4700 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
4701 * lra-lives.c (process_bb_lives): Check for clobber high.
4702 * lra.c (new_insn_reg): Remember clobber highs.
4703 (collect_non_operand_hard_regs): Check for clobber high.
4704 (lra_set_insn_recog_data): Likewise.
4705 (add_regs_to_insn_regno_info): Likewise.
4706 (lra_update_insn_regno_info): Likewise.
4708 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4710 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
4711 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
4713 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4715 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
4716 (copy_insn_1): Likewise.
4717 (gen_hard_reg_clobber_high): New gen function.
4718 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
4719 * genemit.c (gen_exp): Likewise.
4720 (gen_emit_seq): Pass through info.
4721 (gen_insn): Check for CLOBBER_HIGH.
4722 (gen_expand): Pass through info.
4723 (gen_split): Likewise.
4724 (output_add_clobbers): Likewise.
4725 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
4726 (remove_clobbers): Likewise.
4727 * rtl.h (gen_hard_reg_clobber_high): New declaration.
4729 2018-08-06 Alan Hayward <alan.hayward@arm.com>
4731 * doc/rtl.texi (clobber_high): Add.
4732 (parallel): Add in clobber high
4733 * rtl.c (rtl_check_failed_code3): Add function.
4734 * rtl.def (CLOBBER_HIGH): Add expression.
4735 * rtl.h (RTL_CHECKC3): Add macro.
4736 (rtl_check_failed_code3): Add declaration.
4737 (XC3EXP): Add macro.
4739 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
4742 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
4743 cfun->machine->max_used_stack_alignment if needed.
4745 2018-08-04 Martin Sebor <msebor@redhat.com>
4747 PR tree-optimization/86571
4748 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
4751 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
4754 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4757 2018-08-03 Jeff Law <law@redhat.com>
4760 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4761 Define to speculation_safe_value_not_needed.
4763 2018-08-03 David Malcolm <dmalcolm@redhat.com>
4765 * doc/gcov.texi (-x): Remove duplicate "to".
4766 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
4767 (-Wif-not-aligned): Remove duplicate "is".
4768 (-flto): Remove duplicate "the".
4769 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
4770 duplicate "v5.00.b".
4771 (MSP430 Options): Remove duplicate "and" from the description
4772 of "-mgprel-sec=regexp".
4773 (x86 Options): Remove duplicate copies of "vmldLog102" and
4774 vmlsLog104 from description of "-mveclibabi=type".
4776 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
4778 * internal-fn.h (first_commutative_argument): Declare.
4779 * internal-fn.c (first_commutative_argument): New function.
4780 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
4781 restrictions for pattern statements. Use first_commutative_argument
4782 to look for commutative operands in calls to internal functions.
4784 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
4786 * Makefile.in (wide-int-range.o): New.
4787 * tree-vrp.c: Move all the wide_int_* functions to...
4788 * wide-int-range.cc: ...here.
4789 * tree-vrp.h: Move all the wide_int_* prototypes to...
4790 * wide-int-range.h: ...here.
4792 2018-08-03 Tom de Vries <tdevries@suse.de>
4794 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
4796 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
4797 * except.c (output_function_exception_table): Do early exit if
4798 targetm_common.except_unwind_info (&global_options) == UI_NONE.
4800 2018-08-03 Martin Liska <mliska@suse.cz>
4802 * predict.c (dump_prediction): Change to 2 digits
4805 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
4807 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
4809 (falkor_am_1_gtov_gtov): ... a new insn reservation.
4811 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
4813 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
4814 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
4815 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
4816 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
4817 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
4819 2018-08-02 David Malcolm <dmalcolm@redhat.com>
4821 * diagnostic-show-locus.c (diagnostic_show_locus): Use
4822 pp_take_prefix when saving the existing prefix.
4823 * diagnostic.c (diagnostic_append_note): Likewise.
4824 * langhooks.c (lhd_print_error_function): Likewise.
4825 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
4826 param's type. Free the existing prefix.
4827 (pp_take_prefix): New function.
4828 (pretty_printer::pretty_printer): Drop the prefix parameter.
4829 Rename the length parameter to match the comment.
4830 (pretty_printer::~pretty_printer): Free the prefix.
4831 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
4833 (struct pretty_printer): Drop the "const" from "prefix" field's
4834 type and clarify memory management.
4835 (pp_set_prefix): Drop the "const" from the 2nd param.
4836 (pp_take_prefix): New decl.
4838 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
4840 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
4841 for word_mode_ok here instead of passing as argument.
4842 (expand_block_compare): Change select_block_compare_mode() call.
4843 (expand_strncmp_gpr_sequence): New function.
4844 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
4846 2018-08-02 Jeff Law <law@redhat.com>
4849 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4850 Define to speculation_safe_value_not_needed.
4853 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4854 Define to speculation_safe_value_not_needed.
4856 2018-08-02 Tom de Vries <tdevries@suse.de>
4859 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
4860 function. Return UI_TARGET unconditionally.
4861 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
4862 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
4864 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4866 * genemit.c (print_overload_test): Fix typo.
4868 2018-08-02 Richard Biener <rguenther@suse.de>
4870 PR tree-optimization/86816
4871 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
4872 which checks for value availability before querying it.
4873 (gvn_uses_equal): Use it.
4874 (same_succ_hash): Likewise.
4875 (gimple_equal_p): Likewise.
4877 2018-08-02 Nick Clifton <nickc@redhat.com>
4880 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4881 Define to speculation_safe_value_not_needed.
4884 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4885 Define to speculation_safe_value_not_needed.
4888 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4889 Define to speculation_safe_value_not_needed.
4892 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4893 Define to speculation_safe_value_not_needed.
4896 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4897 Define to speculation_safe_value_not_needed.
4900 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4901 Define to speculation_safe_value_not_needed.
4904 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4905 Define to speculation_safe_value_not_needed.
4908 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4909 Define to speculation_safe_value_not_needed.
4912 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4913 speculation_safe_value_not_needed.
4916 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
4917 to speculation_safe_value_not_needed.
4919 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4921 * doc/md.texi: Expand the documentation of instruction names
4922 to mention port-local uses. Document '@' in pattern names.
4923 * read-md.h (overloaded_instance, overloaded_name): New structs.
4925 (md_reader::handle_overloaded_name): New member function.
4926 (md_reader::get_overloads): Likewise.
4927 (md_reader::m_first_overload): New member variable.
4928 (md_reader::m_next_overload_ptr): Likewise.
4929 (md_reader::m_overloads_htab): Likewise.
4930 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
4931 m_next_overload_ptr and m_overloads_htab.
4932 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
4933 (get_mode_token, get_code_token, get_int_token): New functions.
4934 (map_attr_string): Add an optional argument that passes back
4935 the associated iterator.
4936 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
4937 (md_reader::handle_overloaded_name, add_overload_instance): New
4939 (apply_iterators): Handle '@' names. Report an error if '@'
4940 is used without iterators.
4941 (initialize_iterators): Initialize the new iterator_group fields.
4942 * genopinit.c (handle_overloaded_code_for)
4943 (handle_overloaded_gen): New functions.
4944 (main): Use them to print declarations of maybe_code_for_* and
4945 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
4946 * genemit.c (print_overload_arguments, print_overload_test)
4947 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
4948 (main): Use it to print definitions of maybe_code_for_* and
4949 maybe_gen_* functions.
4950 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
4951 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
4952 instead of explicit mode checks.
4953 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
4954 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
4955 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
4956 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
4957 (aarch64_expand_compare_and_swap): Likewise
4958 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
4959 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
4960 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
4961 (aarch64_constant_pool_reload_icode): Delete.
4962 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
4963 instead of aarch64_constant_pool_reload_icode. Use
4964 code_for_aarch64_reload_mov instead of explicit mode checks.
4965 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
4966 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
4967 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
4968 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
4969 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
4970 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
4971 (aarch64_atomic_load_op_code): Delete.
4972 (aarch64_emit_atomic_load_op): Likewise.
4973 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
4974 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
4975 instead of aarch64_emit_atomic_load_op.
4976 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
4977 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
4978 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
4979 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
4980 character before the pattern name.
4981 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
4982 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
4983 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
4984 (aarch64_frecps<mode>): Likewise.
4985 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
4986 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
4987 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
4988 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
4989 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
4991 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
4993 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
4994 Allow HFmode constants if TARGET_FP_F16INST.
4996 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
4999 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
5000 No longer check last store for clobber of address register.
5002 2018-08-02 Martin Liska <mliska@suse.cz>
5004 PR gcov-profile/86817
5005 * gcov.c (process_all_functions): New function.
5007 (process_file): Move functions processing to
5008 process_all_functions.
5010 2018-08-02 David Malcolm <dmalcolm@redhat.com>
5012 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
5013 "const" to the "gimple *" and "rtx_insn *" parameters.
5014 * dumpfile.h (dump_user_location_t::dump_user_location_t):
5016 (dump_location_t::dump_location_t): Likewise.
5018 2018-08-01 Martin Sebor <msebor@redhat.com>
5020 PR tree-optimization/86650
5021 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
5022 rather than a "gcall *". Directly pass the data of interest
5023 to percent_K_format, rather than building a temporary CALL_EXPR
5025 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
5026 (gimple_fold_builtin_strncat): Adjust.
5027 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
5028 gcall* argument with gimple*.
5029 * gimple-ssa-warn-restrict.c (check_call): Same.
5030 (wrestrict_dom_walker::before_dom_children): Same.
5031 (builtin_access::builtin_access): Same.
5032 (check_bounds_or_overlap): Same
5033 (maybe_diag_overlap): Same.
5034 (maybe_diag_offset_bounds): Same.
5035 * tree-diagnostic.c (default_tree_printer): Move usage of
5036 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
5038 * tree-pretty-print.c (percent_K_format): Add argument.
5039 * tree-pretty-print.h: Add argument.
5040 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
5041 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
5042 (maybe_diag_stxncpy_trunc): Same.
5043 (handle_builtin_stxncpy): Same.
5044 (handle_builtin_strcat): Same.
5046 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5048 * match.pd: Optimise pointer range checks.
5050 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5052 PR tree-optimization/86758
5053 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
5054 to remove pattern statements.
5056 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5058 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
5059 result of dfs_enumerate_from when constructing stmt_vec_infos,
5060 instead of additionally calling get_loop_body.
5062 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5064 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
5066 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
5067 When creating an iv, assert that the step is not known to be zero.
5068 (vect_setup_realignment): Update call accordingly.
5069 * tree-vect-stmts.c (vectorizable_store): Likewise.
5070 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
5072 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5074 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
5075 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
5076 (vectorizable_reduction): Likewise.
5077 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5078 (vect_detect_hybrid_slp_stmts): Likewise.
5079 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
5081 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
5083 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
5084 (wide_int_set_zero_nonzero_bits): ...this.
5085 (zero_nonzero_bits_from_vr): Rename to...
5086 (vrp_set_zero_nonzero_bits): ...this.
5087 (extract_range_from_multiplicative_op_1): Abstract wide int
5089 (wide_int_range_multiplicative_op): ...here.
5090 (extract_range_from_binary_expr_1): Extract wide int binary
5091 operations into their own functions.
5092 (wide_int_range_lshift): New.
5093 (wide_int_range_can_optimize_bit_op): New.
5094 (wide_int_range_shift_undefined_p): New.
5095 (wide_int_range_bit_xor): New.
5096 (wide_int_range_bit_ior): New.
5097 (wide_int_range_bit_and): New.
5098 (wide_int_range_trunc_mod): New.
5099 (extract_range_into_wide_ints): New.
5100 (vrp_shift_undefined_p): New.
5101 (extract_range_from_multiplicative_op): New.
5102 (vrp_can_optimize_bit_op): New.
5103 * tree-vrp.h (value_range::dump): New.
5104 (wide_int_range_multiplicative_op): New.
5105 (wide_int_range_lshift):New.
5106 (wide_int_range_shift_undefined_p): New.
5107 (wide_int_range_bit_xor): New.
5108 (wide_int_range_bit_ior): New.
5109 (wide_int_range_bit_and): New.
5110 (wide_int_range_trunc_mod): New.
5111 (zero_nonzero_bits_from_bounds): Rename to...
5112 (wide_int_set_zero_nonzero_bits): ...this.
5113 (zero_nonzero_bits_from_vr): Rename to...
5114 (vrp_set_zero_nonzero_bits): ...this.
5115 (range_easy_mask_min_max): Rename to...
5116 (wide_int_range_can_optimize_bit_op): this.
5117 * vr-values.c (simplify_bit_ops_using_ranges): Rename
5118 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5120 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5122 * tree-vectorizer.h (vect_orig_stmt): New function.
5123 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5124 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5125 (vect_create_epilog_for_reduction): Likewise.
5126 (vectorizable_live_operation): Likewise.
5127 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5128 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5129 * tree-vect-stmts.c (vectorizable_call): Likewise.
5130 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5132 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5134 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5136 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5137 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5138 (vect_transform_loop): Likewise.
5139 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5141 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5143 * tree-vectorizer.h (vect_schedule_slp): Return void.
5144 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5145 (vect_schedule_slp): Likewise.
5147 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5149 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5151 (vect_transform_loop): Update calls accordingly. Schedule SLP
5152 instances before the main loop, if any exist.
5154 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
5156 PR tree-optimization/86749
5157 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5158 If the lhs is used in a COND_EXPR, check that it is being used
5159 as the "then" or "else" value.
5161 2018-08-01 Tom de Vries <tdevries@suse.de>
5164 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5165 speculation_safe_value_not_needed.
5167 2018-08-01 Richard Biener <rguenther@suse.de>
5169 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5172 2018-08-01 Martin Liska <mliska@suse.cz>
5174 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5175 format how successful transformation is dumped.
5176 (gimple_mod_pow2_value_transform): Likewise.
5177 (gimple_mod_subtract_transform): Likewise.
5178 (gimple_stringops_transform): Likewise.
5180 2018-08-01 Martin Liska <mliska@suse.cz>
5183 * value-prof.c (interesting_stringop_to_profile_p):
5184 Simplify the code and add BUILT_IN_MEMMOVE.
5185 (gimple_stringops_transform): Likewise.
5187 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
5189 * config/aarch64/aarch64-simd.md
5190 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5191 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5192 use GPI iterator instead of SI mode.
5194 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5196 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5197 rs6000_speculation_barrier.
5198 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5199 new barrier pattern name.
5201 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
5203 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5204 (speculation_barrier): New insn.
5206 2018-08-01 Richard Biener <rguenther@suse.de>
5209 * graphite.h: Include isl/id.h and isl/space.h to allow build
5212 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
5215 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5216 mode for COFF targets.
5217 * defaults.h (TARGET_COFF): Define.
5218 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5219 TARGET_COFF): Define.
5220 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5221 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5222 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5224 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5226 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5227 * gimple-streamer-out.c (output_bb): Save it.
5228 * lto-streamer-in.c (input_struct_function_base): Restore
5229 instance discriminator if available. Create map on demand.
5230 * lto-streamer-out.c (output_struct_function_base): Save it if
5232 * final.c (decl_to_instance_map): Document LTO strategy.
5234 2018-07-31 Alexandre Oliva <oliva@adacore.com>
5235 Olivier Hainque <hainque@adacore.com>
5237 * debug.h (decl_to_instance_map_t): New type.
5238 (decl_to_instance_map): Declare.
5239 (maybe_create_decl_to_instance_map): New inline function.
5240 * final.c (bb_discriminator, last_bb_discriminator): New statics,
5241 to track basic block discriminators.
5242 (final_start_function_1): Initialize them.
5243 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
5245 (decl_to_instance_map): New variable.
5246 (map_decl_to_instance, maybe_set_discriminator): New functions.
5247 (notice_source_line): Set discriminator.
5249 2018-07-31 Ian Lance Taylor <iant@golang.org>
5251 * targhooks.c (default_have_speculation_safe_value): Add
5254 2018-07-31 David Malcolm <dmalcolm@redhat.com>
5256 * dump-context.h: Include "pretty-print.h".
5257 (dump_context::refresh_dumps_are_enabled): New decl.
5258 (dump_context::emit_item): New decl.
5259 (class dump_context): Add fields "m_test_pp" and
5261 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5262 (temp_dump_context::get_dumped_text): New decl.
5263 (class temp_dump_context): Add field "m_pp".
5264 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
5265 (dump_context::refresh_dumps_are_enabled): ...and add a test for
5267 (set_dump_file): Update for above change.
5268 (set_alt_dump_file): Likewise.
5269 (dump_loc): New overload, taking a pretty_printer *.
5270 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
5271 to any test pretty-printer.
5272 (make_item_for_dump_gimple_stmt): New function, adapted from
5273 optinfo::add_gimple_stmt.
5274 (dump_context::dump_gimple_stmt): Call it, and use the result,
5275 eliminating the direct usage of dump_file and alt_dump_file in
5276 favor of indirectly using them via emit_item.
5277 (make_item_for_dump_gimple_expr): New function, adapted from
5278 optinfo::add_gimple_expr.
5279 (dump_context::dump_gimple_expr): Call it, and use the result,
5280 eliminating the direct usage of dump_file and alt_dump_file in
5281 favor of indirectly using them via emit_item.
5282 (make_item_for_dump_generic_expr): New function, adapted from
5284 (dump_context::dump_generic_expr): Call it, and use the result,
5285 eliminating the direct usage of dump_file and alt_dump_file in
5286 favor of indirectly using them via emit_item.
5287 (make_item_for_dump_printf_va): New function, adapted from
5288 optinfo::add_printf_va.
5289 (make_item_for_dump_printf): New function.
5290 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
5291 and use the result, eliminating the direct usage of dump_file and
5292 alt_dump_file in favor of indirectly using them via emit_item.
5293 (make_item_for_dump_dec): New function.
5294 (dump_context::dump_dec): Call it, and use the result,
5295 eliminating the direct usage of dump_file and alt_dump_file in
5296 favor of indirectly using them via emit_item.
5297 (make_item_for_dump_symtab_node): New function, adapted from
5298 optinfo::add_symtab_node.
5299 (dump_context::dump_symtab_node): Call it, and use the result,
5300 eliminating the direct usage of dump_file and alt_dump_file in
5301 favor of indirectly using them via emit_item.
5302 (dump_context::begin_scope): Reimplement, avoiding direct usage
5303 of dump_file and alt_dump_file in favor of indirectly using them
5305 (dump_context::emit_item): New member function.
5306 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5307 Set up test pretty-printer on the underlying context. Call
5308 refresh_dumps_are_enabled.
5309 (temp_dump_context::~temp_dump_context): Call
5310 refresh_dumps_are_enabled.
5311 (temp_dump_context::get_dumped_text): New member function.
5312 (selftest::verify_dumped_text): New function.
5313 (ASSERT_DUMPED_TEXT_EQ): New macro.
5314 (selftest::test_capture_of_dump_calls): Run all tests twice, with
5315 and then without optinfo enabled. Add uses of
5316 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
5317 * dumpfile.h: Update comment for the dump_* API.
5318 * optinfo-emit-json.cc
5319 (selftest::test_building_json_from_dump_calls): Update for new
5320 param for temp_dump_context ctor.
5321 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
5322 and "m_owned" field.
5323 (optinfo_item::~optinfo_item): Likewise.
5324 (optinfo::add_item): New member function.
5325 (optinfo::emit): Update comment.
5326 (optinfo::add_string): Delete.
5327 (optinfo::add_printf): Delete.
5328 (optinfo::add_printf_va): Delete.
5329 (optinfo::add_gimple_stmt): Delete.
5330 (optinfo::add_gimple_expr): Delete.
5331 (optinfo::add_tree): Delete.
5332 (optinfo::add_symtab_node): Delete.
5333 (optinfo::add_dec): Delete.
5334 * optinfo.h (class dump_context): New forward decl.
5335 (optinfo::add_item): New decl.
5336 (optinfo::add_string): Delete.
5337 (optinfo::add_printf): Delete.
5338 (optinfo::add_printf_va): Delete.
5339 (optinfo::add_gimple_stmt): Delete.
5340 (optinfo::add_gimple_expr): Delete.
5341 (optinfo::add_tree): Delete.
5342 (optinfo::add_symtab_node): Delete.
5343 (optinfo::add_dec): Delete.
5344 (optinfo::add_poly_int): Delete.
5345 (optinfo_item::optinfo_item): Remove "owned" param.
5346 (class optinfo_item): Remove field "m_owned".
5348 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
5351 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
5352 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
5353 requested variable alignment.
5354 (expand_one_ssa_partition): Likewise.
5355 (expand_one_var): Likewise.
5357 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5359 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5360 to speculation_safe_value_not_needed.
5362 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5364 * targhooks.h (speculation_safe_value_not_needed): New prototype.
5365 * targhooks.c (speculation_safe_value_not_needed): New function.
5366 * target.def (have_speculation_safe_value): Update documentation.
5367 * doc/tm.texi: Regenerated.
5369 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5371 * config/aarch64/iterators.md (ALLI_TI): New iterator.
5372 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
5374 (despeculate_copy<ALLI:mode>_insn): New insn.
5375 (despeculate_copyti_insn): New insn.
5376 (despeculate_simple<ALLI:mode>): New insn
5377 (despeculate_simpleti): New insn.
5378 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
5380 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
5381 aarch64_speculation_safe_value.
5382 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
5384 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5386 * config/aarch64/aarch64-speculation.cc: New file.
5387 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
5388 before pass_reorder_blocks.
5389 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
5391 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
5392 X14 and X15 when tracking speculation.
5393 * config/aarch64/aarch64.md (register name constants): Add
5394 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
5395 (unspec): Add UNSPEC_SPECULATION_TRACKER.
5396 (speculation_barrier): New insn attribute.
5397 (cmp<mode>): Allow SP in comparisons.
5398 (speculation_tracker): New insn.
5399 (speculation_barrier): Add speculation_barrier attribute.
5400 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
5401 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
5402 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
5404 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5406 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
5407 aarch64_track_speculation is true.
5408 (tb<optab><mode>1): Likewise.
5409 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
5410 generate CB[N]Z when tracking speculation.
5411 (aarch64_split_compare_and_swap): Likewise.
5412 (aarch64_split_atomic_op): Likewise.
5414 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5416 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
5418 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5420 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
5421 (speculation_barrier): New insn.
5423 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5425 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
5426 * config/arm/arm.md (speculation_barrier): New expand.
5427 (speculation_barrier_insn): New pattern.
5429 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
5431 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
5432 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
5433 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
5434 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
5436 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
5437 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
5438 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
5439 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
5440 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
5441 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
5442 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
5443 * builtins.c (expand_speculation_safe_value): New function.
5444 (expand_builtin): Call it.
5445 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
5446 * doc/extend.texi: Document __builtin_speculation_safe_value.
5447 * doc/md.texi: Document "speculation_barrier" pattern.
5448 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
5449 TARGET_HAVE_SPECULATION_SAFE_VALUE.
5450 * doc/tm.texi: Regenerated.
5451 * target.def (have_speculation_safe_value, speculation_safe_value): New
5453 * targhooks.c (default_have_speculation_safe_value): New function.
5454 (default_speculation_safe_value): New function.
5455 * targhooks.h (default_have_speculation_safe_value): Add prototype.
5456 (default_speculation_safe_value): Add prototype.
5458 2018-07-31 David Malcolm <dmalcolm@redhat.com>
5460 * dump-context.h (dump_context::dump_loc): New decl.
5461 * dumpfile.c (dump_context::dump_loc): New member function.
5462 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
5463 and dump_gimple_stmt.
5464 (dump_context::dump_gimple_expr_loc): Likewise, using
5466 (dump_context::dump_generic_expr_loc): Likewise, using
5468 (dump_context::dump_printf_loc_va): Likewise, using
5470 (dump_context::begin_scope): Explicitly using the global function
5471 "dump_loc", rather than the member function.
5473 2018-07-31 Martin Sebor <msebor@redhat.com>
5475 PR tree-optimization/86741
5476 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
5478 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
5480 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
5481 depend on whether prefetch instructions will be emitted or not.
5482 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
5483 will be emitted or not.
5484 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
5485 (TARGET_SETMEM_PFD): New macros.
5487 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5489 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
5490 (NULL_STMT_VEC_INFO): Delete.
5491 (stmt_vec_info::operator*): Likewise.
5492 (stmt_vec_info::operator gimple *): Likewise.
5493 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
5494 of NULL_STMT_VEC_INFO.
5495 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5496 (vect_reassociating_reduction_p): Likewise.
5497 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5498 (vectorizable_store): Likewise.
5499 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
5500 (vec_info::free_stmt_vec_infos): Likewise.
5502 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5504 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
5505 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
5507 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5509 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
5510 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5511 (vec_info::free_stmt_vec_info): New private member functions.
5512 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
5513 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
5514 * tree-parloops.c (gather_scalar_reductions): Remove calls to
5515 set_stmt_vec_info_vec and free_stmt_vec_infos.
5516 * tree-vect-loop.c (_loop_vec_info): Remove call to
5517 set_stmt_vec_info_vec.
5518 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
5519 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
5520 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
5521 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5522 (vec_info::free_stmt_vec_info): ...these new functions. Remove
5523 assignments in {vec_info::,}new_stmt_vec_info that are redundant
5524 with the clearing in the xcalloc.
5525 (stmt_vec_info_vec): Delete.
5526 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
5527 (vectorize_loops): Likewise.
5528 (vec_info::~vec_info): Remove argument from call to
5529 free_stmt_vec_infos.
5530 (vec_info::add_stmt): Remove vinfo argument from call to
5533 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5535 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
5536 rather than a gimple stmt.
5537 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
5538 information for pattern statements when passed the original
5539 statement; instead wait to be passed the pattern statement itself.
5540 Don't call set_vinfo_for_stmt here.
5541 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
5542 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
5543 stmt_vec_infos here.
5544 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
5545 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
5546 stmt_vec_infos entry.
5548 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5550 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
5551 * tree-vectorizer.c (vec_info::replace_stmt): New function.
5552 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
5553 * tree-vect-stmts.c (vectorizable_call): Likewise.
5554 (vectorizable_simd_clone_call): Likewise.
5556 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5558 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
5559 * tree-vectorizer.c (vec_info::remove_stmt): New function.
5560 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
5561 * tree-vect-loop.c (vect_transform_loop): Likewise.
5562 * tree-vect-slp.c (vect_schedule_slp): Likewise.
5563 * tree-vect-stmts.c (vect_remove_stores): Likewise.
5565 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5567 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
5568 (vect_dr_stmt): Delete.
5569 * tree-vectorizer.c (vec_info::lookup_dr): New function.
5570 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
5572 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
5573 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
5574 (vect_verify_datarefs_alignment, vect_peeling_supportable)
5575 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5576 (vect_analyze_data_refs): Likewise.
5577 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
5579 (vect_find_same_alignment_drs): Likewise.
5580 (vect_slp_analyze_node_dependences): Update calls accordingly.
5581 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
5582 instead of DR_VECT_AUX.
5583 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
5584 of a vector data references. Use vec_info::lookup_dr instead of
5586 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
5587 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
5588 instead of DR_VECT_AUX.
5590 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5592 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
5594 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5596 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5597 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5598 (vect_gen_prolog_loop_niters): Likewise.
5600 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5602 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
5603 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
5604 (vect_known_alignment_in_bytes, vect_dr_behavior)
5605 (vect_get_scalar_dr_size): Take references as dr_vec_infos
5606 instead of data_references. Update calls to other routines for
5607 which the same change has been made.
5608 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
5609 dr_vec_infos instead of stmt_vec_infos.
5610 (vect_analyze_data_ref_dependence): Update call accordingly.
5611 (vect_slp_analyze_data_ref_dependence)
5612 (vect_record_base_alignments): Use DR_VECT_AUX.
5613 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
5614 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
5615 (vector_alignment_reachable_p, vect_get_data_access_cost)
5616 (vect_peeling_supportable, vect_analyze_group_access_1)
5617 (vect_analyze_group_access, vect_analyze_data_ref_access)
5618 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
5619 (vect_compile_time_alias, vect_small_gap_p)
5620 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
5621 (vect_supportable_dr_alignment): Take references as dr_vec_infos
5622 instead of data_references. Update calls to other routines for
5623 which the same change has been made.
5624 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
5625 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5626 (vect_slp_analyze_and_verify_node_alignment)
5627 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5628 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5629 (vect_setup_realignment): Use dr_vec_infos. Update calls after
5631 (_vect_peel_info::dr): Replace with...
5632 (_vect_peel_info::dr_info): ...this new field.
5633 (vect_peeling_hash_get_most_frequent)
5634 (vect_peeling_hash_choose_best_peeling): Update accordingly.
5635 (vect_peeling_hash_get_lowest_cost):
5636 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
5637 routines for which the same change has been made.
5638 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
5640 * tree-vect-loop-manip.c (get_misalign_in_elems)
5641 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
5643 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5644 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
5645 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
5646 (get_group_load_store_type, get_negative_load_store_type)
5647 (vect_get_data_ptr_increment, vectorizable_store)
5648 (vectorizable_load): Likewise.
5649 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
5650 Update calls to other routines for which the same change has been made.
5652 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5654 * tree-vectorizer.h (vec_info::move_dr): New member function.
5655 (dataref_aux): Rename to...
5656 (dr_vec_info): ...this and add "dr" and "stmt" fields.
5657 (_stmt_vec_info::dr_aux): Update accordingly.
5658 (_stmt_vec_info::data_ref_info): Delete.
5659 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
5660 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
5661 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
5662 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
5664 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
5665 (STMT_VINFO_DR_INFO): New macro.
5666 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
5667 (set_dr_misalignment): Update after rename of dataref_aux.
5668 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
5669 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
5670 initialization of STMT_VINFO_DATA_REF.
5671 * tree-vectorizer.c (vec_info::move_dr): New function.
5672 * tree-vect-patterns.c (vect_recog_bool_pattern)
5673 (vect_recog_mask_conversion_pattern)
5674 (vect_recog_gather_scatter_pattern): Use it.
5675 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
5676 the "dr" and "stmt" fields of dr_vec_info instead of
5677 STMT_VINFO_DATA_REF.
5679 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5681 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
5682 (is_pattern_stmt_p): Use it.
5683 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
5684 on pattern statements.
5686 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5688 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
5689 original stmt as a stmt_vec_info rather than a gimple stmt.
5690 (vect_pattern_recog_1): Take the statement directly as a
5691 stmt_vec_info, rather than via a gimple_stmt_iterator.
5692 Update call to vect_mark_pattern_stmts.
5693 (vect_pattern_recog): Update calls accordingly.
5695 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5697 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
5698 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
5699 a vect_def_type for the first argument.
5700 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
5701 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
5702 operand if it isn't defined by a vectorized statement.
5703 (vect_build_gather_load_calls): Remove the mask_dt argument and
5704 update calls to vect_get_vec_def_for_stmt_copy.
5705 (vectorizable_bswap): Likewise the dt argument.
5706 (vectorizable_call): Update calls to vectorizable_bswap and
5707 vect_get_vec_def_for_stmt_copy.
5708 (vectorizable_simd_clone_call, vectorizable_assignment)
5709 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
5710 (vectorizable_comparison): Update calls to
5711 vect_get_vec_def_for_stmt_copy.
5712 (vectorizable_store): Likewise. Remove now-unnecessary calls to
5714 (vect_get_loop_based_defs): Remove dt argument and update call
5715 to vect_get_vec_def_for_stmt_copy.
5716 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
5717 and vect_get_vec_def_for_stmt_copy.
5718 (vectorizable_load): Update calls to vect_build_gather_load_calls
5719 and vect_get_vec_def_for_stmt_copy.
5720 * tree-vect-loop.c (vect_create_epilog_for_reduction)
5721 (vectorizable_reduction, vectorizable_live_operation): Update calls
5722 to vect_get_vec_def_for_stmt_copy.
5724 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5726 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
5727 and gimple stmt arguments with a stmt_vec_info.
5728 (vect_record_base_alignments): Update calls accordingly.
5729 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
5730 and gimple stmt arguments with a stmt_vec_info.
5731 (vect_build_slp_tree_1): Remove vinfo argument and update call
5732 to vect_record_max_nunits.
5733 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
5734 and vect_record_max_nunits.
5736 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5738 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
5739 file and take a stmt_vec_info instead of a gimple stmt.
5740 (supportable_widening_operation, vect_finish_replace_stmt)
5741 (vect_finish_stmt_generation, vect_get_store_rhs)
5742 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
5743 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
5744 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
5745 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
5746 (vect_create_data_ref_ptr, bump_vector_ptr)
5747 (vect_permute_store_chain, vect_setup_realignment)
5748 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
5749 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
5750 (vectorizable_reduction, vectorizable_induction)
5751 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
5752 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
5753 than gimple stmts as arguments.
5754 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
5755 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
5756 (can_group_stmts_p, vect_check_gather_scatter)
5757 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5758 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
5759 (vect_permute_load_chain, vect_shift_permute_load_chain)
5760 (vect_transform_grouped_load)
5761 (vect_record_grouped_load_vectors): Likewise.
5762 * tree-vect-loop.c (vect_fixup_reduc_chain)
5763 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
5764 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
5765 (vectorizable_reduction, vectorizable_induction)
5766 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
5767 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
5768 (vect_get_load_store_mask): Likewise.
5769 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
5770 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
5771 * tree-vect-stmts.c (vect_mark_relevant)
5772 (is_simple_and_all_uses_invariant)
5773 (exist_non_indexing_operands_for_use_p, process_use)
5774 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
5775 (vect_get_vec_def_for_operand, vect_get_vec_defs)
5776 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
5777 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
5778 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
5779 (get_negative_load_store_type, get_load_store_type)
5780 (vect_check_load_store_mask, vect_check_store_rhs)
5781 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
5782 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5783 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5784 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5785 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
5786 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
5787 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
5788 (supportable_widening_operation): Likewise.
5790 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5792 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
5793 a stmt_vec_info instead of a gcall.
5794 (vect_check_gather_scatter): Update call accordingly.
5795 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
5797 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
5798 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
5799 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
5800 instead of a gimple stmt.
5801 (vect_transform_loop): Update calls accordingly.
5802 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
5803 stmt_vec_infos instead of gimple stmts.
5804 (vect_analyze_slp_instance): Update use accordingly.
5805 * tree-vect-stmts.c (read_vector_array, write_vector_array)
5806 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
5807 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
5808 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
5809 (vect_gen_widened_results_half, vect_get_loop_based_defs)
5810 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
5811 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
5812 down to subroutines.
5814 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5816 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
5817 of the worklist from a vector of gimple stmts to a vector of
5819 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5820 (vect_mark_stmts_to_be_vectorized): Likewise
5822 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5824 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
5825 statement before passing it to vect_analyze_stmt.
5826 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
5827 the chain of phi vector definitions. Track the exit phi via its
5829 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
5830 STMT_VINFO_REDUC_DEF.
5831 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
5832 stmt_vec_infos to handle the statement chains.
5833 (vect_get_slp_defs): Record the first statement in the node
5834 using a stmt_vec_info.
5835 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
5836 statements here and pass their stmt_vec_info down to subroutines.
5837 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
5838 down to vect_finish_stmt_generation.
5839 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
5840 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
5841 stmt_vec_infos to subroutines.
5842 (vect_remove_stores): Use stmt_vec_infos to handle the statement
5845 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5847 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
5848 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
5849 (vect_permute_store_chain, vect_setup_realignment)
5850 (vect_permute_load_chain, vect_shift_permute_load_chain)
5851 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
5852 stmts internally, and when passing values to other vectorizer routines.
5853 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5854 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
5855 (vect_analyze_loop_operations, get_initial_def_for_reduction)
5856 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
5857 (vectorizable_reduction, vectorizable_induction)
5858 (vectorizable_live_operation, vect_transform_loop_stmt)
5859 (vect_transform_loop): Likewise.
5860 * tree-vect-patterns.c (vect_reassociating_reduction_p)
5861 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
5862 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
5863 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5864 (vect_slp_analyze_node_operations_1): Likewise.
5865 * tree-vect-stmts.c (vect_mark_relevant, process_use)
5866 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
5867 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
5868 (vect_finish_stmt_generation_1, get_group_load_store_type)
5869 (get_load_store_type, vect_build_gather_load_calls)
5870 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5871 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5872 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5873 (vectorizable_store, vectorizable_load, vectorizable_condition)
5874 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
5875 (supportable_widening_operation): Likewise.
5876 (vect_get_vector_types_for_stmt): Likewise.
5877 * tree-vectorizer.h (vect_dr_behavior): Likewise.
5879 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5881 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5882 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
5883 (vect_permute_store_chain, vect_permute_load_chain)
5884 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
5885 repeated stmt_vec_info lookups.
5886 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5887 (vect_update_ivs_after_vectorizer): Likewise.
5888 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
5889 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5890 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
5891 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5892 (vect_bb_slp_scalar_cost): Likewise.
5893 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
5895 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5897 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
5898 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
5899 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
5900 to get gassigns and gcalls, rather than operating on generc gimple
5902 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
5903 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
5904 (vectorizable_load, vect_analyze_stmt): Likewise.
5905 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
5907 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5909 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
5910 return stmt_vec_infos rather than gimple stmts. Do not accept
5912 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
5914 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5915 Update use of get_later_stmt.
5916 (vect_get_constant_vectors): Update call accordingly.
5917 (vect_schedule_slp_instance): Likewise
5918 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
5919 (vect_slp_analyze_instance_dependence): Likewise.
5920 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
5922 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5924 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
5925 (stmt_info_for_cost::stmt_info): ...this new field.
5926 (add_stmt_costs): Update accordingly.
5927 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5928 (vect_get_known_peeling_cost): Likewise.
5929 (vect_estimate_min_profitable_iters): Likewise.
5930 * tree-vect-stmts.c (record_stmt_cost): Likewise.
5932 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5934 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
5935 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5936 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5938 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
5940 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5942 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
5943 a gimple stmt to a stmt_vec_info.
5944 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5946 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5948 * tree-vectorizer.h (vec_info::grouped_stores): Change from
5949 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5950 (_loop_vec_info::reduction_chains): Likewise.
5951 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
5953 * tree-vect-slp.c (vect_analyze_slp): Likewise.
5955 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5957 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
5958 a gimple stmt to a stmt_vec_info.
5959 (_stmt_vec_info::next_element): Likewise.
5960 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
5961 (vect_slp_analyze_and_verify_node_alignment)
5962 (vect_analyze_group_access_1, vect_analyze_group_access)
5963 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
5964 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
5965 (vect_supportable_dr_alignment): Update accordingly.
5966 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
5967 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
5968 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
5969 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5970 * tree-vect-slp.c (vect_build_slp_tree_1)
5971 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
5972 (vect_split_slp_store_group, vect_analyze_slp_instance)
5973 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
5974 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
5975 (get_group_load_store_type, get_load_store_type)
5976 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
5977 (vect_transform_stmt, vect_remove_stores): Likewise.
5979 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
5981 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
5983 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5984 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
5985 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
5986 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
5987 (vector_alignment_reachable_p, vect_get_data_access_cost)
5988 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
5989 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
5990 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5991 (vect_analyze_group_access_1, vect_analyze_group_access)
5992 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
5993 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
5994 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
5995 result of vect_dr_stmt and use the stmt_vec_info instead of
5996 the associated gimple stmt.
5997 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5998 (vect_gen_prolog_loop_niters): Likewise.
5999 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
6001 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6003 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
6004 to a vec<stmt_vec_info>.
6005 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
6006 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
6008 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
6009 to a vec<stmt_vec_info>.
6010 (bst_traits::value_type, bst_traits::value_type): Likewise.
6011 (bst_traits::hash): Update accordingly.
6012 (vect_get_and_check_slp_defs): Change the stmts parameter from
6013 a vec<gimple *> to a vec<stmt_vec_info>.
6014 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
6015 (vect_build_slp_tree): Likewise.
6016 (vect_build_slp_tree_2): Likewise. Update uses of
6017 SLP_TREE_SCALAR_STMTS.
6018 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
6019 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
6020 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
6021 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
6022 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
6023 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
6024 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
6025 (vect_get_constant_vectors, vect_get_slp_defs)
6026 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
6027 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
6028 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
6029 instead of gimple stmts.
6030 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
6031 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
6032 (vect_slp_analyze_instance_dependence): Update uses of
6033 SLP_TREE_SCALAR_STMTS.
6034 (vect_slp_analyze_and_verify_node_alignment): Likewise.
6035 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
6036 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
6037 (get_initial_defs_for_reduction): Likewise.
6038 (vect_create_epilog_for_reduction): Likewise.
6039 (vectorize_fold_left_reduction): Likewise.
6040 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
6041 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
6042 (can_vectorize_live_stmts): Likewise.
6044 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6046 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
6047 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
6048 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
6050 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
6052 (gather_scalar_reductions): Update after above interface changes.
6053 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
6054 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
6056 (vect_force_simple_reduction): Likewise.
6057 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
6058 LOOP_VINFO_REDUCTIONS.
6059 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
6061 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6063 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
6064 a gimple stmt to a stmt_vec_info.
6065 * tree-vect-loop.c (vect_active_double_reduction_p)
6066 (vect_force_simple_reduction, vectorizable_reduction): Update
6069 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6071 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
6072 vec<gimple *> to a vec<stmt_vec_info>.
6073 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
6074 the reduction_phis argument from a vec<gimple *> to a
6076 (vectorizable_reduction): Likewise the phis local variable that
6077 is passed to vect_create_epilog_for_reduction. Update for new type
6078 of SLP_TREE_VEC_STMTS.
6079 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
6080 (vectorizable_live_operation): Likewise.
6081 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
6082 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
6084 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6086 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
6087 a gimple stmt to a stmt_vec_info.
6088 (vectorizable_condition, vectorizable_live_operation)
6089 (vectorizable_reduction, vectorizable_induction): Pass back the
6090 vectorized statement as a stmt_vec_info.
6091 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
6092 use of STMT_VINFO_VEC_STMT.
6093 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
6094 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
6095 as stmt_vec_infos rather than gimple stmts.
6096 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
6098 (vectorizable_live_operation): Likewise.
6099 (vectorizable_reduction, vectorizable_induction): Likewise,
6100 updating use of STMT_VINFO_VEC_STMT.
6101 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6102 of STMT_VINFO_VEC_STMT.
6103 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6104 (vectorizable_simd_clone_call, vectorizable_conversion)
6105 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6106 (vectorizable_store, vectorizable_load, vectorizable_condition)
6107 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6108 from a gimple stmt to a stmt_vec_info.
6109 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
6110 pointer to a stmt_vec_info to the vectorizable_* routines.
6112 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6114 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6115 a gimple stmt to a stmt_vec_info.
6116 (is_pattern_stmt_p): Update accordingly.
6117 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6118 (vect_record_grouped_load_vectors): Likewise.
6119 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6120 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6121 (vect_model_reduction_cost): Likewise.
6122 (vect_create_epilog_for_reduction): Likewise.
6123 (vectorizable_reduction, vectorizable_induction): Likewise.
6124 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6125 Return the stmt_vec_info for the pattern statement.
6126 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6127 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6128 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6129 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6130 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6131 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6132 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6133 (free_stmt_vec_info, vect_is_simple_use): Likewise.
6135 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6137 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6138 (vect_finish_stmt_generation): Likewise.
6139 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6140 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6141 (vect_build_gather_load_calls): Use the return value of the above
6142 functions instead of a separate call to vinfo_for_stmt. Use narrow
6143 scopes for the input gimple stmt and wider scopes for the associated
6144 stmt_vec_info. Use vec_info::lookup_def when setting these
6145 stmt_vec_infos from an SSA_NAME definition.
6146 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6147 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6148 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6149 (vectorizable_store, vectorizable_load, vectorizable_condition)
6150 (vectorizable_comparison): Likewise.
6151 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6152 (vectorizable_reduction): Likewise.
6154 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6156 * tree-vectorizer.h (vect_is_simple_use): Add an optional
6157 stmt_vec_info * parameter before the optional gimple **.
6158 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6159 (process_use, vect_get_vec_def_for_operand_1): Update callers.
6160 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6161 * tree-vect-loop.c (vectorizable_reduction): Likewise.
6162 (vectorizable_live_operation): Likewise.
6163 * tree-vect-patterns.c (type_conversion_p): Likewise.
6164 (vect_look_through_possible_promotion): Likewise.
6165 (vect_recog_rotate_pattern): Likewise.
6166 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6168 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6170 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6171 a typedef to a wrapper class.
6172 (NULL_STMT_VEC_INFO): New macro.
6173 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6174 (stmt_vec_info::operator*): New function.
6175 (stmt_vec_info::operator gimple *): Likewise.
6176 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6177 (add_stmt_costs): Likewise.
6178 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6179 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6180 (vect_get_known_peeling_cost): Likewise.
6181 (vect_estimate_min_profitable_iters): Likewise.
6182 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6183 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6184 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6185 (vectorizable_store, free_stmt_vec_infos): Likewise.
6186 (new_stmt_vec_info): Change return type of xcalloc to
6189 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6191 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6192 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6193 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6194 a single_imm_use-based sequence.
6195 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6197 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6199 * tree-vectorizer.h (vec_info::lookup_def): Declare.
6200 * tree-vectorizer.c (vec_info::lookup_def): New function.
6201 * tree-vect-patterns.c (vect_get_internal_def): Use it.
6202 (vect_widened_op_tree): Likewise.
6203 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6204 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6205 (vectorizable_reduction): Likewise.
6206 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6208 (vect_is_slp_reduction): Update calls accordingly. Use
6209 vec_info::lookup_def.
6210 (vect_is_simple_reduction): Likewise
6211 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6213 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6215 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6216 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6217 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6219 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6220 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6221 (vect_update_vf_for_slp, vect_analyze_loop_operations)
6222 (vect_is_slp_reduction, vectorizable_induction)
6223 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6224 * tree-vect-patterns.c (vect_init_pattern_stmt):
6225 (vect_determine_min_output_precision_1, vect_determine_precisions)
6226 (vect_pattern_recog): Likewise.
6227 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6228 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6229 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6230 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6231 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6232 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
6233 info field from a loop to a loop_vec_info.
6235 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6237 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
6238 (vec_info::add_stmt): Declare.
6239 * tree-vectorizer.c (vec_info::add_stmt): New function.
6240 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
6241 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
6242 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6243 (vectorizable_induction): Likewise.
6244 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
6245 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6246 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
6247 (vectorizable_load): Likewise.
6248 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6249 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
6250 (vect_recog_gather_scatter_pattern): Likewise.
6251 (append_pattern_def_seq): Likewise. Remove a check that is
6252 performed by add_stmt itself.
6254 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6256 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
6257 which make_ssa_name was called with new_stmt before new_stmt
6260 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6262 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
6264 (vect_is_slp_reduction): ...here...
6265 (vect_is_simple_reduction): ...and here. Remove repetition of tests
6266 that are already known to be false.
6268 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6270 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
6271 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
6272 STMT_VINFO_NUM_SLP_USES when it's true.
6273 (vect_free_slp_instance): Add a final_p parameter and pass it to
6275 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
6276 (vect_analyze_slp_instance): Likewise.
6277 (vect_slp_analyze_operations): Likewise.
6278 (vect_slp_analyze_bb_1): Likewise.
6279 * tree-vectorizer.c (vec_info): Likewise.
6280 * tree-vect-loop.c (vect_transform_loop): Likewise.
6282 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6284 * tree-vect-loop.c (vectorizable_reduction): Assert that the
6285 function is not called for second and subsequent members of
6288 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
6290 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
6291 cases for nested loops from here to ...
6292 (vect_create_epilog_for_reduction): ...here. Only call
6293 vect_is_simple_use for inner-loop reductions.
6295 2018-07-31 Martin Liska <mliska@suse.cz>
6297 PR gcov-profile/85338
6298 PR gcov-profile/85350
6299 PR gcov-profile/85372
6300 * profile.c (struct location_triplet): New.
6301 (struct location_triplet_hash): Likewise.
6302 (output_location): Do not output a BB that
6303 is already recorded for a line.
6304 (branch_prob): Use streamed_locations.
6306 2018-07-31 Martin Liska <mliska@suse.cz>
6308 PR gcov-profile/85370
6309 * coverage.c (coverage_begin_function): Do not mark target
6310 clones as artificial functions.
6312 2018-07-31 Martin Liska <mliska@suse.cz>
6314 PR gcov-profile/83813
6315 PR gcov-profile/84758
6316 PR gcov-profile/85217
6317 PR gcov-profile/85332
6318 * profile.c (branch_prob): Do not record GOTO expressions
6319 for GIMPLE statements which locations are already streamed.
6321 2018-07-31 Olivier Hainque <hainque@adacore.com>
6323 * gcc.c (handle_spec_function): Accept a soft_matched_part
6324 argument, as do_spec_1. Pass it down to ...
6325 (eval_spec_function): Accept a soft_matched_part argument,
6326 and pass it down to ...
6327 (do_spec_2): Accept a soft_matched_part argument, and pass
6328 it down to do_spec_1.
6329 (do_spec_1): Pass soft_matched_part to handle_spec_function.
6330 (handle_braces): Update call to handle_spec_function.
6331 (driver::set_up_specs): Update calls to do_spec_2.
6332 (compare_debug_dump_opt_spec_function): Likewise.
6333 (compare_debug_self_opt_spec_function): Likewise.
6335 2018-07-31 Olivier Hainque <hainque@adacore.com>
6337 * common.opt (nolibc): New option.
6338 * doc/invoke.texi (Link Options): Document it.
6339 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
6340 * config/alpha/linux.h: Likewise.
6341 * config/arc/elf.h: Likewise.
6342 * config/arm/uclinux-elf.h: Likewise.
6343 * config/arm/unknown-elf.h: Likewise.
6344 * config/avr/avrlibc.h: Likewise.
6345 * config/bfin/bfin.h: Likewise.
6346 * config/bfin/linux.h: Likewise.
6347 * config/bfin/uclinux.h: Likewise.
6348 * config/darwin.h: Likewise.
6349 * config/darwin10.h: Likewise.
6350 * config/darwin12.h: Likewise.
6351 * config/gnu-user.h: Likewise.
6352 * config/lm32/uclinux-elf.h: Likewise.
6353 * config/pa/pa-hpux11.h: Likewise.
6354 * config/pa/pa64-hpux.h: Likewise.
6355 * config/sparc/sparc.h: Likewise.
6357 2018-07-31 Olivier Hainque <hainque@adacore.com>
6359 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
6360 undefined variables.
6362 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
6365 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
6368 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
6370 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6371 terminated string literal.
6373 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
6375 PR rtl-optimization/85160
6376 * combine.c (is_just_move): New function.
6377 (try_combine): Allow combining two instructions into two if neither of
6378 the original instructions was a move.
6380 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
6383 * doc/extend.texi (Global Register Variables): Discourage use of type
6385 (Local Register Variables): Likewise.
6387 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
6389 PR tree-optimization/86506
6390 * hwint.h (ceil_log2): Resync with hwint.c implementation.
6392 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
6395 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
6396 hard_regno, make sure no insn between `from` and `to` clobbers it.
6398 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
6399 Tom de Vries <tdevries@suse.de>
6401 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
6402 (PTX_DEFAULT_RUNTIME_DIM): ... this.
6403 (nvptx_goacc_validate_dims): Set default worker and gang dims to
6404 PTX_DEFAULT_RUNTIME_DIM.
6405 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
6407 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
6409 * config/pa/pa.c (pa_output_addr_vec): Align address table.
6410 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
6411 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
6413 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
6415 * config/rs6000/constraints.md (wG constraint): Delete, no longer
6417 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
6418 predicate to reflect toc fusion has been deleted.
6419 (toc_fusion_mem_raw): Delete, no longer used.
6420 (toc_fusion_mem_wrapped): Likewise.
6421 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
6423 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
6424 Delete, no longer used.
6425 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
6426 meant to be used for toc fusion.
6427 (rs6000_debug_print_mode): Delete toc fusion debugging.
6428 (rs6000_debug_reg_global): Likewise.
6429 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
6430 fusion and secondary reload support that were never used.
6431 (rs6000_option_override_internal): Delete TOC fusion, that was only
6432 partially defined, and it did not work unless you also used the
6434 (rs6000_legitimate_address_p): Delete TOC fusion support.
6435 (rs6000_opt_masks): Likewise.
6436 (fusion_wrap_memory_address): Delete function, no longer used.
6437 (fusion_split_address); Delete TOC fusion support.
6438 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
6439 longer used with toc fusion being deleted.
6440 (TARGET_TOC_FUSION_FP): Likewise.
6441 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
6443 (toc fusion spliter): Delete TOC fusion support.
6444 (toc_fusionload_<mode>): Likewise.
6445 (toc_fusionload_di): Likewise.
6446 (fusion_gpr_load_<mode>): Delete generator function, this insn no
6447 longer needs to be named. Rename predicate to delete TOC fusion.
6448 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6449 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6450 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6451 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6452 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
6454 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
6456 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
6457 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
6458 __int128 in built-in function prototypes.
6459 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
6460 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
6462 2018-07-27 Martin Sebor <msebor@redhat.com>
6464 PR tree-optimization/86696
6465 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
6466 types, including enums.
6467 (handle_char_store): Be prepared for the above function to fail.
6469 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
6471 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
6472 when optimization level is lower than 2 or optimize for size.
6474 2018-07-26 Martin Sebor <msebor@redhat.com>
6476 PR tree-optimization/86043
6477 PR tree-optimization/86042
6478 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
6479 (get_string_cst_length): Rename...
6480 (get_min_string_length): ...to this. Add argument.
6481 (handle_char_store): Extend to handle multi-character stores by
6483 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
6484 * tree.h (initializer_zerop): Add argument.
6486 2018-07-26 Jakub Jelinek <jakub@redhat.com>
6489 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
6490 declare target to variables if they have always,{to,from,tofrom} map
6493 2018-07-26 Martin Liska <mliska@suse.cz>
6496 * lto-wrapper.c: Add linker_output as prefix
6497 for ltrans_output_file.
6499 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
6501 PR rtl-optimization/85805
6502 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
6503 value for hard registers if that was written in the same mode.
6505 2018-07-26 Martin Liska <mliska@suse.cz>
6507 PR gcov-profile/86536
6508 * gcov.c (format_gcov): Use printf format %.*f directly
6509 and do not handle special values.
6511 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6513 * common/config/arc/arc-common.c (arc_option_optimization_table):
6514 Update default optimizations for size.
6516 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6518 * config/arc/arc.md (movsf_insn): Add short instruction selection.
6519 * config/arc/constraints.md (CfZ): New constraint.
6520 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
6521 (subsf3_fpu): Likewise.
6522 (cmpsf_fpu): Likewise.
6523 (cmpsf_fpu_uneq): Likewise.
6525 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6527 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
6529 (arc_is_uncached_mem_p): uncached applies to both the variable and
6532 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
6534 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
6537 2018-07-25 David Malcolm <dmalcolm@redhat.com>
6539 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
6540 field "m_scopes" from vec to auto_vec.
6542 2018-07-25 Martin Liska <mliska@suse.cz>
6544 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
6547 2018-07-25 Richard Biener <rguenther@suse.de>
6550 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
6551 special wrt context_die late.
6552 (gen_subprogram_die): Re-use DIEs in local scope.
6554 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
6556 PR tree-optimization/86644
6557 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
6559 2018-07-25 Martin Liska <mliska@suse.cz>
6562 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
6563 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
6565 2018-07-25 Martin Liska <mliska@suse.cz>
6568 * params.def: Explain ASan abbreviation and provide
6569 a documentation link.
6571 2018-07-24 Martin Sebor <msebor@redhat.com>
6573 PR tree-optimization/86622
6574 PR tree-optimization/86532
6575 * builtins.h (string_length): Declare.
6576 * builtins.c (c_strlen): Correct handling of non-constant offsets.
6577 (check_access): Be prepared for non-constant length ranges.
6578 (string_length): Make extern.
6579 * expr.c (string_constant): Only handle the minor non-constant
6580 array index. Use string_constant to compute the length of
6581 a generic string constant.
6583 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
6585 PR tree-optimization/86618
6586 * tree-vect-stmts.c (vectorizable_call): Don't take the address
6587 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
6589 2018-07-24 David Malcolm <dmalcolm@redhat.com>
6591 PR tree-optimization/86636
6592 * json.cc (json::object::set): Fix comment. Add assertions.
6593 (json::array::append): Move here from json.h. Add comment and an
6595 (json::string::string): Likewise.
6596 * json.h (json::array::append): Move to json.cc.
6597 (json::string::string): Likewise.
6598 * optinfo-emit-json.cc
6599 (optrecord_json_writer::impl_location_to_json): Assert that we
6600 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
6601 wrapper around it. Expand the location once, rather than three
6603 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
6604 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
6606 (optrecord_json_writer::optinfo_to_json): Likewise, in four
6607 places. Fix some overlong lines.
6609 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
6611 * config/aarch64/aarch64-simd.md
6612 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
6613 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
6614 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
6615 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
6616 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
6617 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
6618 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
6619 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
6620 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
6622 2018-07-24 Jakub Jelinek <jakub@redhat.com>
6625 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
6626 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
6627 and abs_d == d, do the power of two handling if profitable.
6629 2018-07-24 Richard Biener <rguenther@suse.de>
6631 * match.pd: Add BIT_FIELD_REF canonicalizations.
6633 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6636 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
6638 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6640 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
6641 terminated STRING_CST object.
6643 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
6645 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
6647 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6649 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
6650 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
6651 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
6652 the elements into a register.
6653 (rs6000_split_v4si_init_di_reg): Delete.
6654 (rs6000_split_v4si_init): Delete.
6655 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
6656 (vsx_init_v4si): Rewrite as a define_expand.
6658 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
6660 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
6661 zero_extend argument from memory): New.
6663 2018-07-22 Martin Sebor <msebor@redhat.com>
6666 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
6667 diagnosing calls with unknown arguments unless -Walloca-larger-than
6668 is restricted to less than PTRDIFF_MAX bytes.
6670 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
6672 * doc/gcov.texi (Invoking Gcov): Editorial changes.
6674 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6676 * pretty-print.c (text_info::set_location): Remove redundant
6677 "line_table" parameter from call to rich_location::set_range.
6679 2018-07-20 Martin Sebor <msebor@redhat.com>
6682 * builtins.c (expand_builtin_alloca): Adjust.
6683 * calls.c (alloc_max_size): Simplify.
6684 * cgraphunit.c (cgraph_node::expand): Adjust.
6685 * common.opt (larger_than_size, warn_frame_larger_than): Remove
6687 (frame_larger_than_size): Same.
6688 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
6689 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
6691 * doc/invoke.texi (GCC Command Options): Document option arguments.
6692 Explain byte-size arguments and suffixes.
6693 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
6694 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
6695 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
6696 * doc/options.texi (UInteger): Expand.
6697 (Host_Wide_Int, ByteSize): Document new properties.
6698 * final.c (final_start_function_1): Include sizes in an error message.
6699 * function.c (frame_offset_overflow): Same.
6700 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
6701 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
6702 Diagnose unbounded alloca calls only for limits of less than
6704 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
6705 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
6707 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
6708 only for limits of less than PTRDIFF_MAX.
6709 * langhooks-def.h (lhd_handle_option): Change function argument
6711 * langhooks.c (lhd_handle_option): Same.
6712 * langhooks.h (handle_option): Same.
6713 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
6715 (var_type, var_type_struct): Same.
6716 (var_set): Handle ByteSize flag.
6717 * optc-gen.awk: Add comments to output to ease debugging. Make
6718 use of HOST_WIDE_INT where appropriate.
6719 * opts-gen-save.awk: Use %lx to format unsigned long.
6720 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
6721 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
6722 arguments. Parse bytes-size suffixes.
6723 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
6724 (enum_value_to_arg): Same.
6725 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
6726 (handle_option): Adjust.
6727 (generate_option): Change function argument to HOST_WIDE_INT.
6728 (cmdline_handle_error): Adjust.
6729 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
6730 (set_option): Change function argument to HOST_WIDE_INT.
6731 (option_enabled): Handle cl_host_wide_int.
6732 (get_option_state): Handle CLVC_SIZE.
6733 (control_warning_option): Same.
6734 * opts.c (common_handle_option): Change function argument to
6735 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
6736 OPT_Wvla_larger_than_.
6737 * opts.h (enum cl_var_type): Add an enumerator.
6738 * stor-layout.c (layout_decl): Print a more meaningful warning.
6739 * toplev.c (output_stack_usage): Adjust.
6741 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
6743 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
6744 call to inline_expand_builtin_string_cmp.
6745 (expand_builtin_strcmp): Likewise.
6746 (expand_builtin_strncmp): Likewise.
6747 (inline_string_cmp): Delete the last parameter, change char_type_node
6748 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
6750 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
6751 the inlining expansion on target where the type of the call has same or
6752 narrower precision than unsigned char.
6754 2018-07-20 David Malcolm <dmalcolm@redhat.com>
6756 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
6757 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
6758 * common.opt (fsave-optimization-record): New option.
6759 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
6760 * doc/invoke.texi (-fsave-optimization-record): New option.
6761 * dumpfile.c: Include "optinfo-emit-json.h".
6762 (struct kv_pair): Move to coretypes.h.
6763 (optgroup_options): Make non-static.
6764 (dump_context::end_scope): Call
6765 optimization_records_maybe_pop_dump_scope.
6766 * dumpfile.h (optgroup_options): New decl.
6767 * json.cc: New file.
6769 * optinfo-emit-json.cc: New file.
6770 * optinfo-emit-json.h: New file.
6771 * optinfo.cc: Include "optinfo-emit-json.h".
6772 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
6773 (optinfo_enabled_p): Check optimization_records_enabled_p.
6774 (optinfo_wants_inlining_info_p): Likewise.
6775 * optinfo.h: Update comment.
6776 * profile-count.c (profile_quality_as_string): New function.
6777 * profile-count.h (profile_quality_as_string): New decl.
6778 (profile_count::quality): New accessor.
6779 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
6780 and optinfo_emit_json_cc_tests.
6781 * selftest.h (selftest::json_cc_tests): New decl.
6782 (selftest::optinfo_emit_json_cc_tests): New decl.
6783 * toplev.c: Include "optinfo-emit-json.h".
6784 (compile_file): Call optimization_records_finish.
6785 (do_compile): Call optimization_records_start.
6786 * tree-ssa-live.c: Include optinfo.h.
6787 (remove_unused_scope_block_p): Retain inlining information if
6788 optinfo_wants_inlining_info_p returns true.
6790 2018-07-20 Richard Biener <rguenther@suse.de>
6793 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
6794 to cover -flto-partition=none.
6796 2018-07-20 Martin Liska <mliska@suse.cz>
6798 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
6799 (get_decl_source_range): Remove unused function.
6801 2018-07-20 Richard Biener <rguenther@suse.de>
6803 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
6804 (struct vn_phi_s): Likewise.
6805 (struct vn_reference_s): Likewise.
6806 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
6807 for searching the slot of an entry known to be in the hash itself.
6808 (vn_phi_hasher::equal): Likewise.
6809 (vn_reference_hasher::equal): Likewise.
6810 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
6812 (optimistic_info, current_info): Remove, keeping only valid_info.
6813 (vn_reference_lookup_1): Remove fallback lookup.
6814 (vn_reference_lookup_2): Likewise.
6815 (vn_nary_op_lookup_1): Likewise.
6816 (vn_phi_lookup): Likewise.
6817 (vn_nary_build_or_lookup_1): Make sure to not chain the built
6819 (vn_reference_insert): Adjust, chain the inserted hash element
6820 at last_inserted_ref.
6821 (vn_reference_insert_pieces): Likewise.
6822 (visit_reference_op_call): Likewise.
6823 (vn_nary_op_insert_into): Chain the inserted hash element at
6825 (vn_nary_op_insert_pieces): Adjust.
6826 (vn_nary_op_insert): Likewise.
6827 (vn_nary_op_insert_stmt): Likewise.
6828 (vn_phi_insert): Adjust, chain the inserted hash element at
6830 (process_scc): Remove clearing and copying the optimistic
6831 table. Instead remove elements inserted during an optimistic
6832 iteration from the single table we maintain.
6833 (init_scc_vn): Adjust.
6834 (free_scc_vn): Likewise.
6835 (sccvn_dom_walker::record_cond): Likewise.
6836 (sccvn_dom_walker::after_dom_children): Likewise.
6838 2018-07-19 Martin Sebor <msebor@redhat.com>
6840 PR tree-optimization/84047
6841 PR tree-optimization/83776
6842 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
6843 (check_array_bounds): Call it.
6845 2018-07-19 Martin Sebor <msebor@redhat.com>
6847 * align.h (align_flags): Use member initialization.
6849 2018-07-19 David Malcolm <dmalcolm@redhat.com>
6851 * Makefile.in (OBJS): Add optinfo.o.
6852 * coretypes.h (class symtab_node): New forward decl.
6853 (struct cgraph_node): New forward decl.
6854 (class varpool_node): New forward decl.
6855 * dump-context.h: New file.
6856 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
6858 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
6859 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
6860 (set_alt_dump_file): Likewise.
6861 (dump_context::~dump_context): New dtor.
6862 (dump_gimple_stmt): Move implementation to...
6863 (dump_context::dump_gimple_stmt): ...this new member function.
6864 Add the stmt to any pending optinfo, creating one if need be.
6865 (dump_gimple_stmt_loc): Move implementation to...
6866 (dump_context::dump_gimple_stmt_loc): ...this new member function.
6867 Start a new optinfo and add the stmt to it.
6868 (dump_gimple_expr): Move implementation to...
6869 (dump_context::dump_gimple_expr): ...this new member function.
6870 Add the stmt to any pending optinfo, creating one if need be.
6871 (dump_gimple_expr_loc): Move implementation to...
6872 (dump_context::dump_gimple_expr_loc): ...this new member function.
6873 Start a new optinfo and add the stmt to it.
6874 (dump_generic_expr): Move implementation to...
6875 (dump_context::dump_generic_expr): ...this new member function.
6876 Add the tree to any pending optinfo, creating one if need be.
6877 (dump_generic_expr_loc): Move implementation to...
6878 (dump_context::dump_generic_expr_loc): ...this new member
6879 function. Add the tree to any pending optinfo, creating one if
6881 (dump_printf): Move implementation to...
6882 (dump_context::dump_printf_va): ...this new member function. Add
6883 the text to any pending optinfo, creating one if need be.
6884 (dump_printf_loc): Move implementation to...
6885 (dump_context::dump_printf_loc_va): ...this new member function.
6886 Start a new optinfo and add the stmt to it.
6887 (dump_dec): Move implementation to...
6888 (dump_context::dump_dec): ...this new member function. Add the
6889 value to any pending optinfo, creating one if need be.
6890 (dump_context::dump_symtab_node): New member function.
6891 (dump_context::get_scope_depth): New member function.
6892 (dump_context::begin_scope): New member function.
6893 (dump_context::end_scope): New member function.
6894 (dump_context::ensure_pending_optinfo): New member function.
6895 (dump_context::begin_next_optinfo): New member function.
6896 (dump_context::end_any_optinfo): New member function.
6897 (dump_context::s_current): New global.
6898 (dump_context::s_default): New global.
6899 (dump_scope_depth): Delete global.
6900 (dumpfile_ensure_any_optinfo_are_flushed): New function.
6901 (dump_symtab_node): New function.
6902 (get_dump_scope_depth): Reimplement in terms of dump_context.
6903 (dump_begin_scope): Likewise.
6904 (dump_end_scope): Likewise.
6905 (selftest::temp_dump_context::temp_dump_context): New ctor.
6906 (selftest::temp_dump_context::~temp_dump_context): New dtor.
6907 (selftest::verify_item): New function.
6908 (ASSERT_IS_TEXT): New macro.
6909 (ASSERT_IS_TREE): New macro.
6910 (ASSERT_IS_GIMPLE): New macro.
6911 (selftest::test_capture_of_dump_calls): New test.
6912 (selftest::dumpfile_c_tests): Call it.
6913 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
6914 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
6915 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
6916 descriptive comment.
6917 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
6918 (dump_node, dump_bb): Move these unrelated decls.
6919 (class dump_manager): Add leading comment.
6920 * optinfo.cc: New file.
6921 * optinfo.h: New file.
6923 2018-07-19 Michael Collison <michael.collison@arm.com>
6924 Richard Henderson <rth@redhat.com>
6926 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
6927 (subti): Handle op1 zero.
6928 (subvti4, usub4ti4): New.
6929 (*sub<GPI>3_compare1_imm): New.
6930 (sub<GPI>3_carryinCV): New.
6931 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
6932 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
6934 2018-07-19 Michael Collison <michael.collison@arm.com>
6935 Richard Henderson <rth@redhat.com>
6937 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
6938 (addti3): Create simpler code if low part is already known to be 0.
6939 (addvti4, uaddvti4): New.
6940 (*add<GPI>3_compareC_cconly_imm): New.
6941 (*add<GPI>3_compareC_cconly): New.
6942 (*add<GPI>3_compareC_imm): New.
6943 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
6944 handle constants within this pattern..
6945 (*add<GPI>3_compareV_cconly_imm): New.
6946 (*add<GPI>3_compareV_cconly): New.
6947 (*add<GPI>3_compareV_imm): New.
6948 (add<GPI>3_compareV): New.
6949 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
6950 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
6951 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
6952 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
6953 with 'comparison' operator.
6954 (*add<GPI>3_compareV_cconly_imm): Ditto.
6955 (*add<GPI>3_compareV_cconly): Ditto.
6956 (*add<GPI>3_compareV_imm): Ditto.
6957 (add<GPI>3_compareV): Ditto.
6958 (add<mode>3_carryinC): Ditto.
6959 (*add<mode>3_carryinC_zero): Ditto.
6960 (*add<mode>3_carryinC): Ditto.
6961 (add<mode>3_carryinV): Ditto.
6962 (*add<mode>3_carryinV_zero): Ditto.
6963 (*add<mode>3_carryinV): Ditto.
6965 2018-07-19 Michael Collison <michael.collison@arm.com>
6966 Richard Henderson <rth@redhat.com>
6968 * config/aarch64/aarch64-modes.def (CC_V): New.
6969 * config/aarch64/aarch64-protos.h
6970 (aarch64_addti_scratch_regs): Declare
6971 (aarch64_subvti_scratch_regs): Declare.
6972 (aarch64_expand_subvti): Declare.
6973 (aarch64_gen_unlikely_cbranch): Declare
6974 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
6975 for signed overflow using CC_Vmode.
6976 (aarch64_get_condition_code_1): Handle CC_Vmode.
6977 (aarch64_gen_unlikely_cbranch): New function.
6978 (aarch64_addti_scratch_regs): New function.
6979 (aarch64_subvti_scratch_regs): New function.
6980 (aarch64_expand_subvti): New function.
6982 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6984 * config/aarch64/aarch64-option-extensions.def: New entry for profile
6986 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
6987 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
6990 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6993 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6994 address check not strict.
6996 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
6998 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
7000 (store_pair_lanes<mode>): Likewise.
7001 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
7002 enum value 'ADDR_QUERY_LDP_STP_N'.
7003 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
7004 (aarch64_print_address_internal): Add declaration.
7005 (aarch64_print_ldpstp_address): Remove.
7006 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
7007 (aarch64_print_operand): Change printing of 'y'.
7008 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
7009 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
7010 'true' rather than '1'.
7011 * config/aarch64/constraints.md (Uml): Likewise.
7012 (Uml): Rename to Umn.
7015 2018-07-19 Richard Biener <rguenther@suse.de>
7017 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
7019 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
7020 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
7021 (vn_reference_hasher): Likewise.
7022 (struct vn_tables_s): Remove obstack and alloc-pool members.
7023 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
7024 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
7025 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
7026 (vn_reference_insert_pieces): Likewise.
7027 (alloc_vn_nary_op_noinit): Adjust.
7028 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
7029 (vn_phi_eq): Adjust.
7030 (shared_lookup_phiargs): Remove.
7031 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
7032 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
7033 (visit_reference_op_call): Likewise.
7034 (copy_nary, copy_phi, copy_reference): Remove.
7035 (process_scc): Rewind the obstack when iterating. Do not
7036 copy the elements to valid_info but just move them from one
7037 hashtable to the other.
7038 (allocate_vn_table): Adjust.
7039 (free_vn_table): Likewise.
7040 (init_scc_vn): Likewise.
7041 (free_scc_vn): Likewise.
7043 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
7046 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
7047 indirect_return as function type attribute.
7048 (ix86_attribute_table): Change indirect_return to function
7050 * doc/extend.texi: Update indirect_return attribute.
7052 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
7054 * wide-int.h (widest2_int): New.
7055 * gimple-fold.c (arith_overflowed_p): Use it.
7056 * tree.h (widest2_int_cst): New.
7057 * tree-vrp.c (wide_int_binop_overflow): Rename from
7058 vrp_int_const_binop.
7059 Rewrite to work on trees.
7060 (extract_range_from_multiplicative_op_1): Abstract code to...
7061 (wide_int_range_min_max): ...here.
7062 (wide_int_range_cross_product): ...and here.
7063 (extract_range_from_binary_expr_1): Abstract overflow code to...
7064 (wide_int_range_mult_wrapping): ...here.
7065 * tree-vrp.h (wide_int_range_cross_product): New.
7066 (wide_int_range_mult_wrapping): New.
7068 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
7069 Julia Koval <julia.koval@intel.com>
7071 * config/i386/x86-tune-costs.h (skylake_memcpy,
7072 skylake_memset): Replace rep_prefix with unrolling for size 512.
7074 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
7077 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
7078 comparision with EQ_EXPR in last stmt.
7080 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
7082 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
7083 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
7084 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
7085 previously known as "PowerPC AltiVec Built-in Functions". Move
7086 some material to new subsubsections "PowerPC AltiVec Built-in
7087 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
7089 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
7090 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
7091 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
7092 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
7094 2018-07-18 Richard Biener <rguenther@suse.de>
7096 PR tree-optimization/86557
7097 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7100 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
7102 * config/s390/s390.c (s390_function_profiler): Generate CFI.
7104 2018-07-17 Jeff Law <law@redhat.com>
7106 * config/arm/arm.c (get_label_padding): Update for recent
7107 changes to label_to_alignment.
7109 PR tree-optimization/86010
7110 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7112 * config/mips/mips.c (vr4130_align_insns): Update for recent
7113 changes to label_to_alignment.
7115 * config/frv/frv.c (frv_label_align): Update for recent changes
7116 to label_to_alignment.
7118 * config/nios2/nios2.c (nios2_label_align): Update for recent
7119 changes which dropped ALIGN_LABELS_LOG.
7121 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
7123 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7124 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7126 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
7128 * config/arc/arc.c (arc_label_align): Use align_labels instead of
7129 deprecated align_labels_log.
7131 2018-07-17 Richard Biener <rguenther@suse.de>
7134 * dwarf2out.c (init_sections_and_labels): Always generate
7135 a debug_line_str_section for early LTO debug.
7136 (dwarf2out_finish): Reset debug_line_str_hash output early.
7137 Bump counter for extra dwarf5 .debug_loc labels to not conflict
7139 (dwarf2out_early_finish): Output debug_line_str.
7141 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7143 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7144 index register on z196 or later.
7146 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
7148 * config/s390/s390.c (s390_default_align): Set default function
7150 (s390_override_options_after_change): Call s390_default align.
7151 (s390_option_override_internal): Call s390_default_align.
7152 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7154 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7157 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7158 to _looptemp_ clauses, other than the first two.
7160 2018-07-17 Martin Liska <mliska@suse.cz>
7162 * opts.c: Do not enable OPT_falign_* for -Os.
7164 2018-07-17 Martin Liska <mliska@suse.cz>
7166 * align.h (MAX_CODE_ALIGN): New.
7167 (MAX_CODE_ALIGN_VALUE): New.
7168 * common/config/i386/i386-common.c (ix86_handle_option):
7169 (MAX_CODE_ALIGN): Moved to align.h.
7170 * final.c (MAX_CODE_ALIGN): Likewise.
7171 * opts.c (parse_and_check_align_values):
7172 (MAX_CODE_ALIGN): Likewise.
7173 (MAX_CODE_ALIGN_VALUE): Likewise.
7175 2018-07-17 Martin Liska <mliska@suse.cz>
7177 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7178 in order to fulfil coding style.
7179 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7180 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7181 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7182 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7183 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7184 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7185 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7186 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7187 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7189 2018-07-17 Martin Liska <mliska@suse.cz>
7191 * align.h: New file.
7192 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7194 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7195 align_flags of label_to_alignment.
7196 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7198 * config/m68k/m68k.c: Do not use removed align_labels_value and
7200 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7201 (LOOP_ALIGN): Likewise.
7202 (LABEL_ALIGN): Likewise.
7203 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7204 Remove not used macro.
7205 (rs6000_loop_align): Change return type to align_flags.
7206 (rs6000_loop_align_max_skip): Remove.
7207 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7208 Change return type to align_flags.
7209 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7210 Remove not used macro.
7211 (rs6000_loop_align): Change return type to align_flags.
7212 (rs6000_loop_align_max_skip): Remove.
7213 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7214 * config/rx/rx-protos.h (rx_align_for_label): Make it
7216 * config/rx/rx.c (rx_align_for_label): Change return type
7218 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7220 into align_flags class.
7221 (LABEL_ALIGN): Likewise.
7222 (LOOP_ALIGN): Likewise.
7223 * config/s390/s390.c (s390_label_align): Use align_flags
7225 (s390_asm_output_function_label): Likewise.
7226 * config/sh/sh.c (sh_override_options_after_change):
7227 Use align_flags class directly without macros.
7228 (find_barrier): Likewise.
7229 (barrier_align): Likewise.
7230 (sh_loop_align): Likewise.
7231 * config/spu/spu.c (spu_option_override):
7232 Use align_flags_tuple::get_value instead of removed macros.
7233 (spu_sched_init): Likewise.
7234 * config/spu/spu.h (GTY): Likewise.
7235 * config/visium/visium.c (visium_option_override):
7236 Set "8" as default secondary alignment.
7237 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
7238 in order to guarantee secondary alignment of 8.
7239 * coretypes.h: Include align.h header file.
7240 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
7241 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
7242 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
7243 * doc/tm.texi.in: Likewise.
7244 * final.c (struct label_alignment): Remove not used structure.
7245 (LABEL_ALIGN): Change type to align_flags.
7246 (LOOP_ALIGN): Likewise.
7247 (JUMP_ALIGN): Likewise.
7248 (default_loop_align_max_skip): Remove.
7249 (default_label_align_max_skip): Likewise.
7250 (default_jump_align_max_skip): Likewise.
7251 (default_label_align_after_barrier_max_skip):
7252 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
7253 (LABEL_TO_MAX_SKIP): Remove.
7254 (label_to_alignment): Return align_flags type instead of integer.
7255 (label_to_max_skip): Remove.
7256 (align_fuzz): Use align_flags type.
7257 (compute_alignments): Use align_flags type and use align_flags::max
7258 to combine multiple alignments.
7259 (grow_label_align): Grow vec instead of C array.
7260 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
7261 (shorten_branches): Use align_flags type and use align_flags::max
7262 to combine multiple alignments.
7263 (final_scan_insn_1): Remove usage of secondary alignment that comes
7264 from label alignment, but instead use proper secondary alignment
7265 which is computed in grow_label_align.
7266 * flags.h (struct align_flags_tuple): Move to align.h.
7267 (struct align_flags): Likewise.
7268 (state_align_loops): Rename to align_loops.
7269 (state_align_jumps): Rename to align_jumps.
7270 (state_align_labels): Rename to align_labels.
7271 (state_align_functions): Rename to align_functions.
7272 (align_loops_log): Remove.
7273 (align_jumps_log): Remove.
7274 (align_labels_log): Remove.
7275 (align_functions_log): Remove.
7276 (align_loops_max_skip): Remove.
7277 (align_jumps_max_skip): Remove.
7278 (align_labels_max_skip): Remove.
7279 (align_functions_max_skip): Remove.
7280 (align_loops_value): Remove.
7281 (align_jumps_value): Remove.
7282 (align_labels_value): Remove.
7283 (align_functions_value): Remove.
7284 * output.h (label_to_alignment): Change return type to align_flags.
7285 (label_to_max_skip): Remove.
7286 * target.def: Remove loop_align_max_skip, label_align_max_skip,
7287 jump_align_max_skip macros.
7288 * targhooks.h (default_loop_align_max_skip): Remove.
7289 (default_label_align_max_skip): Likewise.
7290 (default_jump_align_max_skip): Likewise.
7291 (default_label_align_after_barrier_max_skip): Remove.
7292 * toplev.c (read_log_maxskip): Use ::normalize function.
7293 (parse_N_M): Remove not used argument and also call ::normalize.
7294 (parse_alignment_opts): Do not pass unused arguments.
7295 * varasm.c (assemble_start_function): Use directly align_functions
7296 instead of removed macros.
7297 * system.h: Do not poison removed macros.
7299 2018-07-17 Jakub Jelinek <jakub@redhat.com>
7302 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
7303 and cond temporaries don't have reference type if iterator has
7304 pointer type. For init use &for_pre_body instead of pre_p if
7305 for_pre_body is non-empty.
7307 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7309 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
7310 double-double modes to SFmode directly directly.
7311 (trunc<mode>sf2_fprs): Delete.
7313 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7315 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
7316 for conversions between IFmode and the decimal floating point modes.
7317 (init_float128_ieee): Use the correct names for conversions between
7318 KFmode and the decimal floating point modes.
7320 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
7322 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
7323 for the conversions between TDmode and IFmode.
7324 (init_float128_ieee): Use more correct names for the conversions
7325 between TDmode and KFmode.
7327 2018-07-16 Jakub Jelinek <jakub@redhat.com>
7329 PR tree-optimization/86526
7330 * builtins.c (expand_builtin_memcmp): Formatting fixes.
7331 (inline_expand_builtin_string_cmp): Likewise.
7332 (inline_string_cmp): Likewise. Use c_readstr instead of
7333 builtin_memcpy_read_str. Add unit_mode temporary.
7335 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
7338 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
7339 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
7341 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
7343 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
7344 Alphabetize prototypes of built-in functions, separating out
7345 built-in functions that are listed in this section but should be
7346 described elsewhere.
7348 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
7351 * expmed.c (emit_store_flag): Do not emit setcc followed by a
7352 conditional move when trapping comparison was split to a
7353 non-trapping one (and vice versa).
7355 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7357 * config/s390/s390.c (s390_function_profiler): Generate nops
7358 instead of profiler call sequences.
7359 * config/s390/s390.opt: Add the new option.
7361 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7363 * config/s390/s390.c (s390_function_profiler): Generate
7364 __mcount_loc section.
7365 * config/s390/s390.opt: Add the new option.
7367 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
7369 * common.opt: Add the new warning.
7370 * config/s390/s390.c (s390_function_profiler): Emit "brasl
7371 %r0,__fentry__" when -mfentry is specified.
7372 (s390_option_override_internal): Disallow -mfentry for 31-bit
7374 * config/s390/s390.opt: Add the new option.
7376 2018-07-16 Richard Biener <rguenther@suse.de>
7379 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
7380 for function-local FUNCTION_DECL and RESULT_DECL immediately.
7382 2018-07-16 Martin Liska <mliska@suse.cz>
7385 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
7388 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
7390 * config/arc/arcHS.md: Update ARCHS scheduling rules.
7392 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
7394 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
7396 * config/arc/arc-cpus.def (hs4x): New cpu.
7398 * config/arc/arc-tables.opt: Regenerate.
7399 * config/arc/arc.c (arc_sched_issue_rate): New function.
7400 (TARGET_SCHED_ISSUE_RATE): Define.
7401 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
7402 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
7404 (attr tune): Add ARCHS4x tune values.
7405 (attr tune_dspmpy): Define.
7406 (*tst): Correct instruction type.
7407 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
7408 * config/arc/arcHS4x.md: New file.
7409 * config/arc/fpu.md: Update instruction type attributes.
7410 * config/arc/t-multilib: Regenerate.
7412 2018-07-16 Tom de Vries <tdevries@suse.de>
7415 * var-tracking.c (vt_initialize): Fix pre_dec handling.
7417 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7419 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
7422 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
7424 PR tree-optimization/86514
7425 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
7426 conversion to a boolean type from a type with greater precision.
7428 2018-07-16 Tom de Vries <tdevries@suse.de>
7430 * var-tracking.c (vt_initialize): Print adjusted insn slim if
7431 dump_flags request TDF_SLIM.
7433 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
7435 * fold-const.c (int_const_binop_1): Abstract...
7436 (wide_int_binop): ...wide int code here.
7437 (poly_int_binop): ...poly int code here.
7438 Abstract the rest of int_const_binop_1 into int_const_binop.
7439 * fold-const.h (wide_int_binop): New.
7440 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
7441 Remove useless PLUS/MINUS_EXPR case.
7442 (zero_nonzero_bits_from_vr): Move wide int code...
7443 (zero_nonzero_bits_from_bounds): ...here.
7444 (extract_range_from_binary_expr_1): Move mask optimization code...
7445 (range_easy_mask_min_max): ...here.
7446 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
7447 (range_easy_mask_min_max): New.
7449 2018-07-15 Jeff Law <law@redhat.com>
7452 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
7455 2018-07-14 Jim Wilson <jimw@sifive.com>
7457 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
7459 2018-07-14 Paul Koning <ni1d@arrl.net>
7461 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
7463 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
7465 * lto-streamer-out.c (copy_function_or_variable): Dump info about
7468 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
7469 Steve Munroe <munroesj52@gmail.com>
7471 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
7472 (_mm_andnot_si128): Likewise.
7473 (_mm_or_si128): Likewise.
7474 (_mm_xor_si128): Likewise.
7476 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
7479 * builtins.c (expand_builtin_memcmp): Inline the calls first
7480 when result_eq is false.
7481 (expand_builtin_strcmp): Inline the calls first.
7482 (expand_builtin_strncmp): Likewise.
7483 (inline_string_cmp): New routine. Expand a string compare
7484 call by using a sequence of char comparison.
7485 (inline_expand_builtin_string_cmp): New routine. Inline expansion
7486 a call to str(n)cmp/memcmp.
7487 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
7489 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
7491 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
7493 * config/arm/driver-arm.c: Include arm-native.h.
7494 (host_detect_local_cpu): Use auto-generated data tables.
7495 (vendors, arm_cpu_table): Delete. Move part information to ...
7496 * config/arm/arm-cpus.in: ... here.
7497 * config/arm/parsecpu.awk (gen_native): New function.
7498 (vendor, part): New CPU fields.
7499 (END): Add support for building the native CPU detection tables.
7500 * config/arm/t-arm (arm-native.h): Add build rule.
7501 (driver-arm.o): Add dependency on arm-native.h.
7503 2018-07-13 Richard Biener <rguenther@suse.de>
7506 * match.pd (addr1 - addr2): Allow either of the operand to
7509 2018-07-13 Tom de Vries <tdevries@suse.de>
7511 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
7514 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
7516 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
7517 arrays instead of numbered variables.
7519 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
7521 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
7522 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
7524 2018-07-13 Richard Biener <rguenther@suse.de>
7527 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
7528 instead of get_context_die.
7530 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
7531 Richard Biener <rguenther@suse.de>
7534 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
7535 that the loop latch destination where phi is defined.
7537 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
7539 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
7540 (riscv_expand_epilogue): Add assertion to check interrupt mode.
7541 (riscv_set_current_function): Extract getting interrupt type to new
7543 (riscv_get_interrupt_type): New function.
7544 (riscv_merge_decl_attributes): New function, checking interrupt type is
7546 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
7548 2018-07-12 Paul Koning <ni1d@arrl.net>
7550 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
7553 2018-07-12 Paul Koning <ni1d@arrl.net>
7555 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
7556 zero reference, add doloop_end instead.
7557 * doc/md.texi (decrement_and_branch_until_zero): Remove.
7558 (Looping patterns): Remove decrement_and_branch_until_zero. Add
7559 detail for doloop_end.
7561 2018-07-12 Martin Sebor <msebor@redhat.com>
7564 * attribs.c (decl_attributes): Reject conflicting attributes before
7565 calling attribute handlers.
7567 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7569 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
7571 (gcc::dump_manager::get_dump_file_name): likewise.
7572 (dump_begin): Likewise.
7573 * dumpfile.h (dump_begin): Update prototype.
7574 (gcc::dump_manager::get_dump_file_name,
7575 gcc::dump_manager::get_dump_file_name): Update prototype.
7577 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7579 * internal-fn.h (vectorizable_internal_fn_p): New function.
7580 * tree-vect-slp.c (compatible_calls_p): Likewise.
7581 (vect_build_slp_tree_1): Remove nops argument. Handle calls
7582 to internal functions.
7583 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
7585 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7587 * fold-const.h (inverse_conditions_p): Declare.
7588 * fold-const.c (inverse_conditions_p): New function.
7589 * match.pd: Use inverse_conditions_p. Add folds of view_converts
7590 that test the inverse condition of a conditional internal function.
7591 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
7592 * internal-fn.c (internal_fn_mask_index): Handle conditional
7594 (vectorized_internal_fn_supported_p): New function.
7595 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
7596 (any_pred_load_store): Replace with...
7597 (need_to_predicate): ...this new variable.
7598 (redundant_ssa_names): New variable.
7599 (ifcvt_can_use_mask_load_store): Move initial checks to...
7600 (ifcvt_can_predicate): ...this new function. Handle tree codes
7601 for which a conditional internal function exists.
7602 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
7603 instead of ifcvt_can_use_mask_load_store. Update after variable
7605 (predicate_load_or_store): New function, split out from
7606 predicate_mem_writes.
7607 (check_redundant_cond_expr): New function.
7608 (value_available_p): Likewise.
7609 (predicate_rhs_code): Likewise.
7610 (predicate_mem_writes): Rename to...
7611 (predicate_statements): ...this. Use predicate_load_or_store
7612 and predicate_rhs_code.
7613 (combine_blocks, tree_if_conversion): Update after above name changes.
7614 (ifcvt_local_dce): Handle redundant_ssa_names.
7615 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
7616 general conditional functions.
7617 * tree-vect-stmts.c (vectorizable_call): Likewise.
7619 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7620 Alan Hayward <alan.hayward@arm.com>
7621 David Sherwood <david.sherwood@arm.com>
7623 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
7624 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
7625 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
7626 plus and minus and convert them into IFN_COND_FMA-based sequences.
7627 (convert_mult_to_fma): Handle conditional plus and minus.
7629 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7631 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
7632 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
7633 (cond_fnms_optab): New optabs.
7634 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
7636 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
7637 * internal-fn.h (get_conditional_internal_fn): Declare.
7638 (get_unconditional_internal_fn): Likewise.
7639 * internal-fn.c (cond_ternary_direct): New macro.
7640 (expand_cond_ternary_optab_fn): Likewise.
7641 (direct_cond_ternary_optab_supported_p): Likewise.
7642 (FOR_EACH_COND_FN_PAIR): Likewise.
7643 (get_conditional_internal_fn): New function.
7644 (get_unconditional_internal_fn): Likewise.
7645 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
7646 (gimple_match_op::gimple_match_op): Add a new overload for 5
7648 (gimple_match_op::set_op): Likewise.
7649 (gimple_resimplify5): Declare.
7650 * genmatch.c (decision_tree::gen): Generate simplifications for
7652 * gimple-match-head.c (gimple_simplify): Define an overload for
7653 5 operands. Handle calls with 5 arguments in the top-level overload.
7654 (convert_conditional_op): Handle conversions from unconditional
7655 internal functions to conditional ones.
7656 (gimple_resimplify5): New function.
7657 (build_call_internal): Pass a fifth operand.
7658 (maybe_push_res_to_seq): Likewise.
7659 (try_conditional_simplification): Try converting conditional
7660 internal functions to unconditional internal functions.
7661 Handle 3-operand unconditional forms.
7662 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
7663 Define ternary equivalents of the current rules for binary conditional
7665 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
7667 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
7668 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
7669 (optab): Handle them.
7670 (SVE_COND_FP_TERNARY): New int iterator.
7671 (sve_fmla_op, sve_fmad_op): New int attributes.
7672 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
7673 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
7674 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
7676 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
7678 * target.def (preferred_else_value): New target hook.
7679 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
7680 * doc/tm.texi: Regenerate.
7681 * targhooks.h (default_preferred_else_value): Declare.
7682 * targhooks.c (default_preferred_else_value): New function.
7683 * internal-fn.h (conditional_internal_fn_code): Declare.
7684 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
7685 (get_conditional_internal_fn): Use it.
7686 (conditional_internal_fn_code): New function.
7687 * gimple-match.h (gimple_match_cond): New struct.
7688 (gimple_match_op): Add a cond member function.
7689 (gimple_match_op::gimple_match_op): Update all forms to take a
7691 * genmatch.c (expr::gen_transform): Use the same condition as res_op
7692 for the suboperation, but don't specify a particular else_value.
7693 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
7694 (visit_nary_op, visit_reference_op_load): Pass
7695 gimple_match_cond::UNCOND to the gimple_match_op constructor.
7696 * gimple-match-head.c: Include tree-eh.h
7697 (convert_conditional_op): New function.
7698 (maybe_resimplify_conditional_op): Likewise.
7699 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
7700 (gimple_resimplify2): Likewise.
7701 (gimple_resimplify3): Likewise.
7702 (gimple_resimplify4): Likewise.
7703 (maybe_push_res_to_seq): Return null for conditional operations.
7704 (try_conditional_simplification): New function.
7705 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
7707 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
7709 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
7711 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
7713 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
7715 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7717 (hash_tree): Do not hash DECL_FCONTEXT
7718 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
7719 Do not stream DECL_FCONTEXT.
7720 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
7721 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
7723 2018-07-12 Richard Biener <rguenther@suse.de>
7726 * dwarf2out.c (gen_block_die): Only output blocks when they have
7727 at least one !DECL_IGNORED_P variable.
7729 2018-07-12 Richard Biener <rguenther@suse.de>
7732 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
7733 Remove -mieee-fp handling.
7735 2018-07-12 Richard Biener <rguenther@suse.de>
7737 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
7738 left-over from last patch.
7740 2018-07-12 Jakub Jelinek <jakub@redhat.com>
7742 PR tree-optimization/86492
7743 * gimple-ssa-store-merging.c
7744 (imm_store_chain_info::coalesce_immediate_stores): Call
7745 check_no_overlap even for the merge_overlapping case. Formatting fix.
7747 2018-07-12 Richard Biener <rguenther@suse.de>
7750 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
7751 move possibly trapping operations into the conditional.
7753 2018-07-12 Richard Biener <rguenther@suse.de>
7755 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
7756 (vn_lookup_simplify_result): Remove recursion limit applied
7758 (vn_nary_build_or_lookup_1): Adjust.
7759 (try_to_simplify): Likewise.
7760 * gimple-match-head.c (gimple_resimplify1): Instead apply one
7762 (gimple_resimplify2): Likewise.
7763 (gimple_resimplify3): Likewise.
7764 (gimple_resimplify4): Likewise.
7766 2018-07-11 Jakub Jelinek <jakub@redhat.com>
7768 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
7769 Use __mmask64 type instead of __mmask8 for __M argument.
7770 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
7771 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
7773 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
7774 __mmask16 for __M argument.
7775 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
7776 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
7777 to __mmask16 instead of __mmask8.
7778 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
7779 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
7780 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
7781 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
7782 instead of __mmask16 for __U argument.
7783 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
7784 __mmask16 instead of __mmask8 for __U argument.
7785 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
7787 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
7789 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
7791 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
7793 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
7795 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
7797 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
7799 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
7801 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
7803 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
7805 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
7806 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
7807 return type as well as __M argument type and all casts from __mmask8
7809 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
7810 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
7811 return type as well as __M argument type and all casts from __mmask8
7813 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
7814 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
7815 return type as well as __M argument type and all casts from __mmask8
7817 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
7818 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
7819 return type as well as __M argument type and all casts from __mmask8
7821 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
7822 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
7825 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
7827 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
7828 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
7831 2018-07-11 Paul Koning <ni1d@arrl.net>
7833 * doc/md.texi (define_subst): Document how multiple occurrences of
7834 the same argument in the replacement pattern are handled.
7836 2018-07-11 Paul Koning <ni1d@arrl.net>
7838 * doc/extend.texi (Common Variable Attributes): Move "mode" into
7840 (Common Type Attributes): Add "mode" attribute.
7842 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
7844 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
7845 stream DECL_ORIGINAL_TYPE.
7846 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
7847 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
7848 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
7849 Do not walk original type.
7850 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
7852 (write_ts_decl_non_common_tree_pointers): Do not stream
7854 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
7855 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
7857 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7859 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
7860 thread twice from the same starting edge.
7862 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
7864 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
7865 * gimple.c (gimple_call_nonnull_result_p): ...here...
7866 (gimple_call_nonnull_arg): ...and here.
7867 * gimple.h (gimple_call_nonnull_result_p): New.
7868 (gimple_call_nonnull_arg): New.
7870 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
7872 * config/arm/arm-cpus.in: Move information from fpu field of each
7873 cpu definition to the isa field.
7874 * config/arm/parsecpu.awk (fpu): Delete match rule.
7875 (gen_comm_data): Don't add bits from the CPU's FPU entry.
7877 2018-07-11 Richard Biener <rguenther@suse.de>
7880 * dwarf2out.c (init_sections_and_labels): Use
7881 output_asm_line_debug_info consistently.
7882 (dwarf2out_early_finish): Likewise.
7883 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
7886 2018-07-11 Richard Biener <rguenther@suse.de>
7888 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
7889 Rework father_bb setting in a way to avoid propagating constants
7890 multiple times on a loop body.
7892 2018-07-10 Mark Wielaard <mark@klomp.org>
7895 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
7897 2018-07-10 Richard Biener <rguenther@suse.de>
7899 * hash-map.h (hash_map::iterator::operator*): Return
7900 references to key and value.
7902 2018-07-10 Jakub Jelinek <jakub@redhat.com>
7905 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
7906 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
7907 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
7908 (gimplify_omp_for): For composite loops, move outer
7909 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
7910 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
7911 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
7912 TREE_LIST for both the original class iterator and the "last" helper
7913 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
7914 loop, remember has_decl_expr from outer composite loops for the
7915 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
7917 2018-07-09 Martin Sebor <msebor@redhat.com>
7921 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
7922 accessing implicitly initialized array elements.
7923 * expr.c (string_constant): Handle string initializers of
7924 character arrays within aggregates.
7925 * gimple-fold.c (fold_array_ctor_reference): Add argument.
7926 Store element offset. As a special case, handle zero size.
7927 (fold_nonarray_ctor_reference): Same.
7928 (fold_ctor_reference): Add argument. Store subobject offset.
7929 * gimple-fold.h (fold_ctor_reference): Add argument.
7931 2018-07-09 Paul Koning <ni1d@arrl.net>
7933 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
7934 (pdp11_insn_cost): New function.
7935 (pdp11_md_asm_adjust): New function.
7936 (TARGET_INVALID_WITHIN_DOLOOP): Define.
7937 (pdp11_rtx_costs): Update to match machine better.
7938 (output_addr_const_pdp11): Correct format mismatch warnings.
7939 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
7940 * config/pdp11/pdp11.md: General change to add base_cost and/or
7941 length attributes for use by new pdp11_insn_cost function.
7942 (MIN_BRANCH): Correct definition.
7944 (doloop_end): Use standard pattern name for looping pattern.
7945 (doloop_end_nocc): New.
7946 (movsf): Add another constraint alternative.
7947 (zero_extendqihi2): Add constraint alternatives for not in place
7949 (zero_extendhisi2): Remove.
7950 (shift patterns): Add CC handling variants.
7954 (define_peephole2): New peephole to recognize mov that sets CC for
7957 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7959 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
7960 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
7963 2018-07-09 Jakub Jelinek <jakub@redhat.com>
7966 * real.c (real_nextafter): Return true if result is denormal.
7968 2018-07-09 Martin Liska <mliska@suse.cz>
7970 * common.opt: Add back wrongly removed attribute.
7972 2018-07-09 Richard Biener <rguenther@suse.de>
7975 * dwarf2out.c (gen_block_die): For an early generated DIE
7976 always output high/low PC attributes.
7978 2018-07-09 Tom de Vries <tdevries@suse.de>
7980 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
7981 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
7982 onto VAR_DECL with abstract origin.
7984 2018-07-07 Jim Wilson <jimw@sifive.com>
7986 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
7988 2018-07-07 Tom de Vries <tdevries@suse.de>
7990 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
7992 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
7994 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
7996 (combine_bound): Use wide-int overflow calculation instead of
7998 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
8000 * fold-const.c (int_const_binop_2): Same.
8001 (extract_muldiv_1): Same.
8002 (fold_div_compare): Same.
8003 (fold_abs_const): Same.
8005 * poly-int.h (add): Same.
8009 * predict.c (predict_iv_comparison): Same.
8010 * profile-count.c (slow_safe_scale_64bit): Same.
8011 * simplify-rtx.c (simplify_const_binary_operation): Same.
8012 * tree-chrec.c (tree_fold_binomial): Same.
8013 * tree-data-ref.c (split_constant_offset_1): Same.
8014 * tree-if-conv.c (idx_within_array_bound): Same.
8015 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
8016 * tree-ssa-phiopt.c (minmax_replacement): Same.
8017 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
8018 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
8019 * vr-values.c (vr_values::adjust_range_with_scev): Same.
8020 * wide-int.cc (wi::add_large): Same.
8021 (wi::mul_internal): Same.
8022 (wi::sub_large): Same.
8023 (wi::divmod_internal): Same.
8024 * wide-int.h: Change overflow type to overflow_type for neg, add,
8025 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
8026 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
8027 mul_internal, divmod_internal.
8028 (overflow_type): New enum.
8029 (accumulate_overflow): New.
8031 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
8033 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
8034 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
8036 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8038 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
8039 argument is checked for zero before entering loop, avoid checking again.
8041 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
8043 * gimplify.h (generic_expr_could_trap_p): Set as global function.
8044 * gimplify.h (generic_expr_could_trap_p): Likwise.
8045 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
8047 2018-07-06 Jakub Jelinek <jakub@redhat.com>
8049 PR tree-optimization/86401
8050 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
8051 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
8052 (fold_bit_and_mask): ... here. New helper function for match.pd.
8053 * fold-const.h (fold_bit_and_mask): Declare.
8054 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
8056 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
8059 * target.def (translate_mode_attribute): New hook.
8060 * targhooks.h (default_translate_mode_attribute): Declare.
8061 * targhooks.c (default_translate_mode_attribute): New function.
8062 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
8063 * doc/tm.texi: Regenerate.
8064 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
8065 (rs6000_translate_mode_attribute): New function.
8067 2018-07-06 Paul Koning <ni1d@arrl.net>
8069 * doc/md.texi (define_split): Document DONE and FAIL.
8070 (define_peephole2): Ditto.
8072 2018-07-05 Jeff Law <law@redhat.com>
8074 PR tree-optimization/86010
8075 * tree-ssa-dse.c (compute_trims): More aggressively trim at
8076 both the head and tail of mem* and str* calls.
8078 2018-07-05 Jim Wilson <jimw@sifive.com>
8080 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
8082 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
8084 * config/aarch64/aarch64-simd.md: correct flags text for
8085 MIN_EXPR replacement.
8087 2018-07-05 James Clarke <jrtc27@jrtc27.com>
8089 * configure: Regenerated.
8091 2018-07-05 Carl Love <cel@us.ibm.com>
8093 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
8094 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
8095 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
8096 VSX_BUILTIN_DOUBLEL_V4SF.
8098 2018-07-05 Martin Sebor <msebor@redhat.com>
8101 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8102 than its domain to compute its the upper bound of a char array.
8104 2018-07-05 Nathan Sidwell <nathan@acm.org>
8106 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8107 * doc/cpp.texi: Update comment.
8108 * doc/tm.texi: Rebuilt.
8109 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8110 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8111 * doc/extend.texi (Backwards Compatibility): Clarify it is system
8112 headers affected by extern "C".
8113 * system.h: Poison NO_IMPLICIT_EXTERN_C.
8114 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8115 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8116 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8117 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8118 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8119 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8120 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8121 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8122 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8123 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8124 config/sparc/sp64-elf.h, config/spu/spu.h,
8125 config/stormy16/stormy16.h, config/v850/v850.h,
8126 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8127 define NO_IMPLICIT_EXTERN_C.
8128 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8130 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8133 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8134 instead of GET_MODE_SIZE when comparing Units.
8136 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8139 * rtlanal.c (set_noop_p): Constrain on mode change,
8140 include hard-reg-set.h
8142 2018-07-05 Tamar Christina <tamar.christina@arm.com>
8144 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8146 2018-07-05 Jakub Jelinek <jakub@redhat.com>
8149 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8152 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8154 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8156 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
8159 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8161 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8163 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
8166 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8168 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
8170 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8171 neg_*_op* variables.
8173 2018-07-04 Martin Liska <mliska@suse.cz>
8175 * tree-switch-conversion.c: Define
8176 max_ratio_for_speed and max_ratio_for_size constants.
8178 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
8179 Martin Liska <mliska@suse.cz>
8184 * common.opt: Rename align options with 'str_' prefix.
8185 * common/config/i386/i386-common.c (set_malign_value): New
8187 (ix86_handle_option): Use it to set -falign-* options/
8188 * config/aarch64/aarch64-protos.h (struct tune_params): Change
8189 type from int to string.
8190 * config/aarch64/aarch64.c: Update default values from int
8192 * config/alpha/alpha.c (alpha_override_options_after_change):
8194 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8195 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8196 max skip conditionally.
8197 * config/i386/freebsd.h (SUBALIGN_LOG): New.
8198 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8199 max skip conditionally.
8200 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8201 max skip conditionally.
8202 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8203 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8204 max skip conditionally.
8205 * config/i386/i386.c (struct ptt): Change type from int to
8207 (ix86_default_align): Set default values.
8208 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8209 max skip conditionally.
8210 * config/i386/iamcu.h (SUBALIGN_LOG): New.
8211 (ASM_OUTPUT_MAX_SKIP_ALIGN):
8212 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8213 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8214 max skip conditionally.
8215 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8216 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8217 * config/i386/x86-64.h (SUBALIGN_LOG): New.
8218 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8219 max skip conditionally.
8220 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8221 * config/ia64/ia64.c (ia64_option_override): Set default values
8222 for alignment options.
8223 * config/m68k/m68k.c: Handle new str_align_* options.
8224 * config/mips/mips.c (mips_set_compression_mode): Change
8226 (mips_option_override): Set default values for options.
8227 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8229 * config/rs6000/rs6000.c (rs6000_option_override_internal):
8231 * config/rx/rx.c (rx_option_override): Likewise.
8232 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8233 (LABEL_ALIGN): Use align_labels_log.
8234 (LOOP_ALIGN): Use align_loops_align.
8235 * config/s390/s390.c (s390_asm_output_function_label): Use new
8237 * config/sh/sh.c (sh_override_options_after_change):
8238 Change type of constants.
8239 * config/spu/spu.c (spu_sched_init): Likewise.
8240 * config/sparc/sparc.c (sparc_option_override): Set default
8242 * config/visium/visium.c (visium_option_override): Likewise.
8243 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
8244 emit p2align format with last argument if it's not needed.
8245 * doc/invoke.texi: Document extended format of -falign-*.
8246 * final.c: Use align_labels alignment.
8247 * flags.h (struct target_flag_state): Change type to use
8249 (struct align_flags_tuple): New.
8250 (struct align_flags): Likewise.
8251 (align_loops_log): Redefine macro to use new types.
8252 (align_loops_max_skip): Redefine macro to use new types.
8253 (align_jumps_log): Redefine macro to use new types.
8254 (align_jumps_max_skip): Redefine macro to use new types.
8255 (align_labels_log): Redefine macro to use new types.
8256 (align_labels_max_skip): Redefine macro to use new types.
8257 (align_functions_log): Redefine macro to use new types.
8258 (align_loops): Redefine macro to use new types.
8259 (align_jumps): Redefine macro to use new types.
8260 (align_labels): Redefine macro to use new types.
8261 (align_functions): Redefine macro to use new types.
8262 (align_functions_max_skip): Redefine macro to use new types.
8263 (align_loops_value): New macro.
8264 (align_jumps_value): New macro.
8265 (align_labels_value): New macro.
8266 (align_functions_value): New macro.
8267 * function.c (invoke_set_current_function_hook): Propagate
8268 alignment values from flags to global variables default in
8270 * ipa-icf.c (sem_function::equals_wpa): Use
8271 cl_optimization_option_eq instead of memcmp.
8272 * lto-streamer.h (cl_optimization_stream_out): Support streaming
8274 (cl_optimization_stream_in): Likewise.
8275 * optc-save-gen.awk: Support strings in cl_optimization.
8276 * opth-gen.awk: Likewise.
8277 * opts.c (finish_options): Remove error checking of invalid
8279 (MAX_CODE_ALIGN): Remove.
8280 (MAX_CODE_ALIGN_VALUE): Likewise.
8281 (parse_and_check_align_values): New function.
8282 (check_alignment_argument): Likewise.
8283 (common_handle_option): Use check_alignment_argument.
8284 * opts.h (parse_and_check_align_values): Declare.
8285 * toplev.c (init_alignments): Remove.
8286 (read_log_maxskip): New.
8287 (parse_N_M): Likewise.
8288 (parse_alignment_opts): Likewise.
8289 (backend_init_target): Remove usage of init_alignments.
8290 * toplev.h (parse_alignment_opts): Declare.
8291 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
8293 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
8294 * tree.c (cl_option_hasher::equal): New.
8295 * varasm.c: Use new global macros.
8297 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
8299 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
8300 Use a simpler align directive also if MAXSKIP = ALIGN-1.
8301 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8302 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8303 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8304 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8305 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
8306 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
8307 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
8308 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8309 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8310 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8311 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8313 2018-07-04 Martin Liska <mliska@suse.cz>
8314 Jonathan Wakely <jwakely@redhat.com>
8316 * coverage.c: Use correct type.
8317 * doc/invoke.texi: Language correction.
8319 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
8322 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
8323 ENDBRANCH for non-tail call which may return via indirect branch.
8324 * doc/extend.texi: Document indirect_return attribute.
8326 2018-07-03 Martin Sebor <msebor@redhat.com>
8328 PR tree-optimization/86274
8329 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
8331 (format_floating): Correct handling of infinities and NaNs.
8333 2018-07-03 Martin Sebor <msebor@redhat.com>
8335 * print-tree.c (print_real_cst): New function.
8336 (print_node_brief): Call it.
8337 (print_node): Ditto.
8339 2018-07-03 Jeff Law <law@redhat.com>
8341 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
8342 into a single pattern.
8344 * config/h8300/h8300.md (ors code_iterator): New.
8345 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
8346 a single pattern and single splitter.
8347 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
8348 (iorqi3_1, xorqi3_1): Likewise.
8349 (iorqi3, xorqi3 expanders): Similarly.
8351 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
8352 (movmd_internal) into a single pattern using the P mode iterator.
8353 (movmd splitters): Similarly.
8354 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
8355 (movsd splitters): Similarly.
8357 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
8358 ADDB, ADDW and ADDL into a single ADD attribute which selects the
8359 right table based on the size of the operand.
8360 * config/h8300/h8300.md (length_table): Corresponding changes. All
8361 references to "addb", "addw" and "addl" changed to "add".
8362 (btst patterns): Merge two variants into a single pattern.
8363 (tstqi, tsthi): Likewise.
8364 (addhi3_incdec, addsi3_incdec): Likewise.
8365 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
8366 (mulhi3, mulsi3): Likewise.
8367 (udivhi3, udivsi3): Likewise.
8368 (divhi3, divsi3): Likewise.
8369 (andorqi3, andorhi3, andorsi3): Likewise.
8371 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
8374 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
8375 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
8377 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8379 PR tree-optimization/85694
8380 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
8381 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
8383 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
8384 (<u>avg<mode>3_ceil): New patterns.
8386 2018-07-03 David Malcolm <dmalcolm@redhat.com>
8388 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
8389 scan-tree-dump directive.
8390 * gcc.dg/vect/slp-perm-2.c: Likewise.
8391 * gcc.dg/vect/slp-perm-3.c: Likewise.
8392 * gcc.dg/vect/slp-perm-5.c: Likewise.
8393 * gcc.dg/vect/slp-perm-6.c: Likewise.
8394 * gcc.dg/vect/slp-perm-7.c: Likewise.
8395 * gcc.dg/vect/slp-perm-8.c: Likewise.
8397 2018-07-03 Marek Polacek <polacek@redhat.com>
8400 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
8402 2018-07-03 Richard Biener <rguenther@suse.de>
8405 * tree-ssa-structalias.c (find_func_clobbers): Properly
8406 handle indirect calls.
8408 2018-07-03 Jeff Law <law@redhat.com>
8410 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
8411 (shifts): New code iterator.
8412 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
8413 expander. Fix HImode handling on H8/SX.
8414 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
8415 (subqi3, subhi3, subsi3 expanders): Likewise.
8416 (andqi3, andhi3, andsi3 expanders): Likewise.
8417 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
8418 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
8419 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
8420 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
8421 (zero_extendqihi2, zero_extendqisi2): Likewise.
8422 (extendqihi2, extendqisi2): Likewise.
8423 (rotlqi3, rotlhi3, rotlsi3): Likewise.
8424 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
8425 (rotlqi3_1, rotlhi3_1): Likewise.
8426 (logicalhi3_sn, logicalsi3_sn): Likewise.
8427 (logicalhi3, logicalsi3): Likewise.
8429 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8431 * tree-vect-patterns.c (vect_recog_rotate_pattern)
8432 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8433 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
8434 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
8435 type to append_pattern_def_seq instead of creating a stmt_vec_info
8437 (build_mask_conversion): Likewise. Remove vinfo argument.
8438 (vect_add_conversion_to_patterm): Likewise, renaming to...
8439 (vect_add_conversion_to_pattern): ...this.
8440 (vect_recog_mask_conversion_pattern): Update call to
8441 build_mask_conversion. Pass the vector type to
8442 append_pattern_def_seq here too.
8443 (vect_recog_gather_scatter_pattern): Update call to
8444 vect_add_conversion_to_pattern.
8446 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8448 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
8449 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8450 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
8451 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
8452 STMT_VINFO_PATTERN_DEF_SEQ to null here.
8453 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
8454 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
8455 append_pattern_def_seq instead of new_pattern_def_seq.
8456 (vect_recog_divmod_pattern): Do both of the above.
8457 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
8460 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8462 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
8463 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
8464 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
8465 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
8466 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
8467 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8468 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
8469 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8470 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8471 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
8472 parameter with a single stmt_vec_info.
8473 (vect_recog_func_ptr): Likewise.
8474 (vect_recog_gather_scatter_pattern): Likewise, folding in...
8475 (vect_try_gather_scatter_pattern): ...this.
8476 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
8477 the stmt_vec_info of the statement to be matched. Don't clear
8478 STMT_VINFO_RELATED_STMT.
8479 (vect_pattern_recog): Update call accordingly.
8481 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8483 PR tree-optimization/85694
8484 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
8485 (uavgM3_ceil): Document new optabs.
8486 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
8487 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
8489 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
8490 (savg_ceil_optab): New optabs.
8491 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
8492 (vect_vect_recog_func_ptrs): Add it.
8493 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
8494 constant directly from the associated lhs.
8496 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8498 * tree-vect-patterns.c (vect_split_statement): New function.
8499 (vect_convert_input): Use it to try to split an existing cast.
8501 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8503 * poly-int.h (print_hex): New function.
8504 * dumpfile.h (dump_dec, dump_hex): Declare.
8505 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
8506 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
8507 min_input_precision, operation_precision and operation_sign.
8508 * tree-vect-patterns.c (vect_get_range_info): New function.
8509 (vect_same_loop_or_bb_p, vect_single_imm_use)
8510 (vect_operation_fits_smaller_type): Delete.
8511 (vect_look_through_possible_promotion): Add an optional
8512 single_use_p parameter.
8513 (vect_recog_over_widening_pattern): Rewrite to use new
8514 stmt_vec_info infomration. Handle one operation at a time.
8515 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
8516 (vect_truncatable_operation_p, vect_set_operation_type)
8517 (vect_set_min_input_precision): New functions.
8518 (vect_determine_min_output_precision_1): Likewise.
8519 (vect_determine_min_output_precision): Likewise.
8520 (vect_determine_precisions_from_range): Likewise.
8521 (vect_determine_precisions_from_users): Likewise.
8522 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
8523 (vect_vect_recog_func_ptrs): Put over_widening first.
8525 (vect_pattern_recog): Call vect_determine_precisions.
8527 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
8529 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
8530 statements that have been replaced by further pattern statements.
8531 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
8533 2018-07-03 Richard Biener <rguenther@suse.de>
8535 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
8536 always set *dt. Dump vectype in vectype overload.
8537 * dumpfile.h (dump_gimple_expr): New function.
8538 (dump_gimple_expr_loc): Likewise.
8539 * dumpfile.c (dump_gimple_expr): New function.
8540 (dump_gimple_expr_loc): Likewise.
8542 2018-07-02 Jeff Law <law@redhat.com>
8544 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
8545 the H8/300, H8/300H and H8/S variants into a single pattern.
8546 (movhi_h8300, movqi_h8300hs): Similarly.
8547 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
8548 (QHI mode iterator): New.
8550 * config/h8300/h8300.md: Remove trailing whitespace.
8552 2018-07-02 Jim Wilson <jimw@sifive.com>
8554 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
8555 instead of emit_insn for interrupt returns.
8556 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
8557 (riscv_sret, riscv_uret): Likewise.
8559 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8561 * pretty-print.c (selftest::test_pp_format): Move save and restore
8562 of quotes to class auto_fix_quotes, and add an instance.
8563 * selftest.c: Include "intl.h".
8564 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
8565 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
8566 * selftest.h (selftest::auto_fix_quotes): New class.
8568 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
8570 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
8571 (aarch64_sve_prepare_conditional_op): Remove.
8572 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
8573 Allow aarch64_simd_reg_or_zero as select operand; remove
8574 the aarch64_sve_prepare_conditional_op call.
8575 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
8576 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8577 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
8578 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
8579 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
8580 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
8581 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
8582 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
8583 and a splitters to match all of the *_any patterns.
8584 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
8586 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
8587 (SVE_COND_FP_BINARY_REV): Remove.
8588 (sve_int_op_rev, sve_fp_op_rev): New.
8589 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
8590 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
8591 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
8592 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
8593 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
8594 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
8595 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
8596 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
8597 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
8599 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
8600 Remove match_dup 1 from the inner unspec.
8601 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8603 * config/aarch64/aarch64.md (movprfx): New attr.
8604 (length): Default movprfx to 8.
8605 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
8606 (*madd<SVE_I>, *msub<SVE_I): Likewise.
8607 (*<su>mul<SVE_I>3_highpart): Likewise.
8608 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
8609 (*v<ASHIFT><SVE_I>3): Likewise.
8610 (*<su><MAXMIN><SVE_I>3): Likewise.
8611 (*<su><MAXMIN><SVE_F>3): Likewise.
8612 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
8613 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
8614 (*div<SVE_F>4): Likewise.
8616 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
8618 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
8621 2018-07-02 Richard Biener <rguenther@suse.de>
8623 PR tree-optimization/86363
8624 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
8625 memset argument refers to a non-variable address.
8627 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8629 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
8630 {PLUS,MINUS}_EXPR code to...
8631 (adjust_symbolic_bound): ...here,
8632 (combine_bound): ...here,
8633 (set_value_range_with_overflow): ...and here.
8635 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
8637 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
8639 (extract_range_from_abs_expr): ...here.
8641 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
8643 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
8644 -fno-omit-frame-pointer when not optimizing.
8646 2018-07-02 Martin Liska <mliska@suse.cz>
8649 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
8650 (propagate_nothrow): Likewise.
8652 2018-07-02 Martin Liska <mliska@suse.cz>
8655 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
8657 2018-07-02 David Malcolm <dmalcolm@redhat.com>
8659 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
8660 function in r262149, changing "loc" param from source_location to
8661 const dump_location_t &.
8662 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
8663 declaration, as above.
8665 2018-07-01 Paul Koning <ni1d@arrl.net>
8667 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8668 -munit-asm, -mgnu-asm, -mdec-asm.
8669 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
8670 (pdp11_output_labelref): New.
8671 (pdp11_output_def): New.
8672 (pdp11_output_addr_vec_elt): New.
8673 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
8674 %# and %@ format codes.
8675 (pdp11_option_override): New.
8676 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
8677 (pdp11_output_ident): New.
8678 (pdp11_asm_named_section): New.
8679 (pdp11_asm_init_sections): New.
8680 (pdp11_file_start): New.
8681 (pdp11_file_end): New.
8682 (output_ascii): Use .ascii/.asciz for -mdec-asm.
8683 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
8684 %o, like %c but octal.
8685 (pdp11_option_override): New.
8686 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
8688 (DATA_SECTION_ASM_OP): Ditto.
8689 (READONLY_DATA_SECTION_ASM_OP): New.
8690 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
8691 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
8692 (ASM_OUTPUT_LABELREF): Ditto.
8693 (ASM_OUTPUT_DEF): Ditto.
8694 (ASM_OUTPUT_EXTERNAL): New.
8695 (ASM_OUTPUT_SOURCE_FILENAME): New.
8696 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
8697 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
8698 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
8699 %# and %@ format codes.
8700 * config/pdp11/pdp11.opt (mgnu-asm): New.
8701 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
8702 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
8703 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
8705 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
8707 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
8708 dereferencing path[] beyond its length.
8710 (debug_all_paths): New.
8711 (rewire_first_differing_edge): New.
8712 (adjust_paths_after_duplication): New.
8713 (duplicate_thread_path): Call adjust_paths_after_duplication.
8715 (thread_through_all_blocks): Add new argument to
8716 duplicate_thread_path.
8718 2018-06-30 Jim Wilson <jimw@sifive.com>
8720 * config/riscv/predicates.md (p2m1_shift_operand): New.
8721 (high_mask_shift_operand): New.
8722 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
8723 pattern using p2m1_shift_operand.
8724 (lshsi3_zero_extend_3+2): New combiner pattern using
8725 high_mask_shift_operand.
8727 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8729 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
8731 (vect_recog_rotate_pattern): ...here.
8732 (vect_convert_input): Try to insert casts of invariants in the
8734 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
8735 preheader to be empty.
8737 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8739 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
8740 vector type. If given, install it in the new statement's
8742 (vect_element_precision): New function.
8743 (vect_unpromoted_value): New struct.
8744 (vect_unpromoted_value::vect_unpromoted_value): New function.
8745 (vect_unpromoted_value::set_op): Likewise.
8746 (vect_look_through_possible_promotion): Likewise.
8747 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
8748 (vect_widened_op_tree, vect_convert_input): Likewise.
8749 (vect_convert_inputs, vect_convert_output): Likewise.
8750 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
8751 to handle the optional cast of the multiplication result and
8752 vect_widened_op_tree to detect the widened multiplication itself.
8753 Do not require the input and output of promotion casts to have
8754 the same sign, but base the signedness of the operation on the
8755 input rather than the result. If the pattern includes two
8756 promotions, check that those promotions have the same sign.
8757 Do not restrict the MULT_EXPR handling to a double-width result;
8758 handle quadruple-width results and wider. Use vect_convert_inputs
8759 to convert the inputs to the common type.
8760 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
8761 to handle the optional cast of the ABS result. Also allow a sign
8762 change or a sign extension between the ABS and MINUS.
8763 Use vect_widened_op_tree to detect the widened subtraction and use
8764 vect_convert_inputs to convert the inputs to the common type.
8765 (vect_handle_widen_op_by_const): Delete.
8766 (vect_recog_widen_op_pattern): New function.
8767 (vect_recog_widen_mult_pattern): Use it.
8768 (vect_recog_widen_shift_pattern): Likewise.
8769 (vect_recog_widen_sum_pattern): Use
8770 vect_look_through_possible_promotion to handle the promoted
8773 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8775 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
8776 the containing gimple_seq *.
8777 * gimple-iterator.h (gsi_for_stmt): Declare it.
8778 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
8779 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
8780 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8781 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8782 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
8784 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
8786 (vect_mark_pattern_stmts): ...here. Handle cases in which the
8787 statement being replaced is part of an existing pattern
8788 definition sequence, inserting the new pattern statements before
8790 (vect_pattern_recog_1): Don't return a bool. If the statement
8791 is already part of a pattern, instead apply pattern matching
8792 to the pattern definition statements. Don't clear the
8793 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
8794 (vect_pattern_recog): Don't break after the first match;
8795 continue processing the pattern definition statements instead.
8796 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
8798 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8800 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8801 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8802 (vect_recog_widen_sum_pattern): Use it.
8804 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8806 * tree-vect-loop.c (vectorizable_reduction): Assert that the
8807 phi is not a pattern statement and has not been replaced by
8808 a pattern statement.
8809 * tree-vect-patterns.c (type_conversion_p): Don't check
8810 STMT_VINFO_IN_PATTERN_P.
8811 (vect_recog_vector_vector_shift_pattern): Likewise.
8812 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
8813 the pattern statement rather than the original statement; check
8814 directly for a WIDEN_MULT_EXPR here.
8815 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
8816 vect_is_simple_use to return the pattern statement rather
8817 than the original statement; use is_pattern_stmt_p to check
8818 for such a pattern statement.
8819 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
8820 to return the pattern statement rather than the original statement;
8821 don't do the same transformation here.
8822 (vect_is_simple_use): If the defining statement has been replaced
8823 by a pattern statement, return the pattern statement instead.
8824 Remove the corresponding (local) transformation from the vectype
8827 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
8829 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
8830 end and default to null.
8831 * tree-vect-loop.c (vect_create_epilog_for_reduction)
8832 (vectorizable_reduction): Update calls accordingly, dropping the
8833 gimple ** argument if the passed-back statement isn't needed.
8834 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
8835 (vect_recog_rotate_pattern): Likewise.
8836 (vect_recog_mask_conversion_pattern): Likewise.
8837 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8838 (vect_mask_constant_operand_p): Likewise.
8839 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
8840 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
8841 (get_group_load_store_type, get_load_store_type): Likewise.
8842 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
8843 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
8844 (vectorizable_conversion, vectorizable_assignment): Likewise.
8845 (vectorizable_shift, vectorizable_operation): Likewise.
8846 (vectorizable_store, vect_is_simple_cond): Likewise.
8847 (vectorizable_condition, vectorizable_comparison): Likewise.
8848 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
8849 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
8850 and move it to the end. Cope with null def_stmt_outs.
8852 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
8854 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
8856 2018-06-29 Jeff Law <law@redhat.com>
8858 * config/v850/v850.c (v850_legitimate_address_p): Handle large
8859 displacements for TARGET_V850E2V3 and newer.
8860 (TARGET_LRA_P): Remove. Defaults to LRA now.
8861 * config/v850/v850.md (sign23byte_load): Remove.
8862 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
8863 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
8865 2018-06-29 Martin Liska <mliska@suse.cz>
8868 * coverage.c (coverage_init): Mangle full path name.
8869 * doc/invoke.texi: Document the change.
8870 * gcov-io.c (mangle_path): New.
8871 * gcov-io.h (mangle_path): Likewise.
8872 * gcov.c (mangle_name): Use mangle_path for path mangling.
8874 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8876 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
8877 if starting source register is not even.
8879 2018-06-29 Martin Liska <mliska@suse.cz>
8881 PR tree-optimization/86263
8882 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
8883 Make edge redirection.
8885 2018-06-29 David Malcolm <dmalcolm@redhat.com>
8887 * dumpfile.c (dump_loc): Add indentation based on scope depth.
8888 (dump_scope_depth): New variable.
8889 (get_dump_scope_depth): New function.
8890 (dump_begin_scope): New function.
8891 (dump_end_scope): New function.
8892 * dumpfile.h (get_dump_scope_depth): New declaration.
8893 (dump_begin_scope): New declaration.
8894 (dump_end_scope): New declaration.
8895 (class auto_dump_scope): New class.
8896 (AUTO_DUMP_SCOPE): New macro.
8897 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
8900 2018-06-29 Richard Biener <rguenther@suse.de>
8902 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
8903 compute_all_dependences succeeds.
8904 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
8905 exceed --param loop-max-datarefs-for-datadeps.
8907 2018-06-29 Jakub Jelinek <jakub@redhat.com>
8909 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
8911 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
8914 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
8915 alternative 0 in preferred_for_speed attribute.
8917 2018-06-28 Paul Koning <ni1d@arrl.net>
8919 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
8920 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
8921 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
8922 * config/pdp11/pdp11.md: Correct "length" attribute calculation
8923 for shift insn patterns.
8925 2018-06-28 David Malcolm <dmalcolm@redhat.com>
8927 * cgraph.c (cgraph_node::get_body): Replace assignments to
8928 "dump_file" with calls to set_dump_file.
8929 * dumpfile.c (alt_dump_file): Make static, and group with...
8930 (alt_flags): ...this definition.
8931 (dumps_are_enabled): New variable.
8932 (refresh_dumps_are_enabled): New function.
8933 (set_dump_file): New function.
8934 (set_alt_dump_file): New function.
8935 (gcc::dump_manager::dump_start): Replace assignments to
8936 "dump_file" and "alt_dump_file" with calls to set_dump_file and
8938 (gcc::dump_manager::dump_finish): Likewise.
8939 * dumpfile.h (alt_dump_file): Delete decl.
8940 (dumps_are_enabled): New variable decl.
8941 (set_dump_file): New function decl.
8942 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
8944 * tree-nested.c (lower_nested_functions): Replace assignments to
8945 "dump_file" with calls to set_dump_file.
8947 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
8949 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
8950 goto_locus of each outgoing edge of each basic block.
8952 2018-06-28 Richard Biener <rguenther@suse.de>
8954 * dwarf2out.c (decl_scope_table): Remove.
8955 (push_decl_scope): Likewise.
8956 (pop_decl_scope): Likewise.
8957 (gen_type_die_for_member): Do not call push/pop_decl_scope.
8958 (gen_struct_or_union_type_die): Likewise.
8959 (gen_tagged_type_die): Likewise.
8960 (dwarf2out_init): Do not initialize decl_scope_table.
8961 (dwarf2out_c_finalize): Do not free it.
8963 2018-06-28 Richard Biener <rguenther@suse.de>
8965 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
8966 deciding whether to not re-use a DIE.
8968 2018-06-28 Richard Biener <rguenther@suse.de>
8970 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
8971 DW_AT_abstract_origin attribute.
8973 2018-06-28 Martin Liska <mliska@suse.cz>
8975 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8976 Use newly introduced constants.
8977 * tree-switch-conversion.h (struct jump_table_cluster):
8978 Define max_ratio_for_size and max_ratio_for_speed.
8980 2018-06-28 Martin Liska <mliska@suse.cz>
8982 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8983 Add new checking assert to catch invalid state.
8984 (jump_table_cluster::can_be_handled): Handle single case
8986 (jump_table_cluster::is_beneficial): Bail out for such case.
8987 (bit_test_cluster::find_bit_tests):
8988 Add new checking assert to catch invalid state.
8989 (bit_test_cluster::can_be_handled): Handle single case
8991 (bit_test_cluster::is_beneficial): Bail out for such case.
8992 (switch_decision_tree::analyze_switch_statement):
8995 2018-06-28 Martin Liska <mliska@suse.cz>
8997 * common.opt: Introduce -completion option.
8998 * gcc.c (driver_handle_option): Handle it.
8999 (driver::main): Print completions if completion
9001 * opt-suggestions.c (option_proposer::get_completions):
9003 (option_proposer::suggest_completion): Likewise.
9004 (option_proposer::find_param_completions): Likewise.
9005 (verify_autocompletions): Likewise.
9006 (test_completion_valid_options): Likewise.
9007 (test_completion_valid_params): Likewise.
9008 (in_completion_p): Likewise.
9009 (empty_completion_p): Likewise.
9010 (test_completion_partial_match): Likewise.
9011 (test_completion_garbage): Likewise.
9012 (opt_proposer_c_tests): Likewise.
9013 * opt-suggestions.h: Declare new functions.
9014 * opts.c (common_handle_option): Handle OPT__completion_.
9015 * selftest-run-tests.c (selftest::run_tests): Add
9016 opt_proposer_c_tests.
9017 * selftest.c (assert_str_startswith): New.
9018 * selftest.h (assert_str_startswith): Likewise.
9019 (opt_proposer_c_tests): New.
9020 (ASSERT_STR_STARTSWITH): Likewise.
9022 2018-06-28 Martin Liska <mliska@suse.cz>
9024 * Makefile.in: Add opt-suggestions.o.
9025 * gcc-main.c: Include opt-suggestions.h.
9026 * gcc.c (driver::driver): Likewise.
9027 (driver::~driver): Remove m_option_suggestions.
9028 (driver::build_option_suggestions): Moved to option_proposer.
9029 (driver::suggest_option): Likewise.
9030 (driver::handle_unrecognized_options): Use option_proposer.
9031 * gcc.h (class driver): Add new memver m_option_proposer.
9032 * opt-suggestions.c: New file.
9033 * opt-suggestions.h: New file.
9035 2018-06-28 Martin Liska <mliska@suse.cz>
9037 * vec.h (class auto_string_vec): New (moved from auto_argvec).
9038 (auto_string_vec::~auto_string_vec): Likewise.
9040 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
9042 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
9043 prevent_decl_creation_for_types fields up and add reset_location field.
9044 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
9045 statement if id->reset_location is true.
9046 (copy_edges_for_bb): Do not set goto_locus on the new edges if
9047 id->reset_location is true.
9048 (copy_phis_for_bb): Force input_location on the arguments if
9049 id->reset_location is true.
9050 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
9051 is set on the function to be inlined.
9053 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
9055 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
9057 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
9059 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
9060 registers for Pmode.
9061 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
9062 hard registers for the clobbered pseudo.
9064 2018-06-27 Paul Koning <ni1d@arrl.net>
9066 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
9067 mutually exclusive options.
9068 * config/pdp11/constraints.md (h): New constraint.
9069 (O): Update definition to match shift code generation.
9070 (D): New constraint.
9071 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
9073 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
9075 (output_jump): Change arguments.
9076 (pdp11_fixed_cc_regs): New function.
9077 (pdp11_cc_mode): Ditto.
9078 (pdp11_expand_shift): Ditto.
9079 (pdp11_assemble_shift): Ditto.
9080 (pdp11_small_shift): Ditto.
9081 (pdp11_branch_cost): Remove.
9082 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
9084 (pdp11_register_move_cost): Update for CC registers.
9085 (pdp11_rtx_costs): Add case for LSHIFTRT.
9086 (pdp11_output_jump): Add CCNZ mode conditional branches.
9087 (notice_update_cc_on_set): Remove.
9088 (pdp11_cc_mode): New function.
9089 (simple_memory_operand): Correct pre/post decrement case.
9090 (no_side_effect_operand): New function.
9091 (pdp11_regno_reg_class): Add CC_REGS class.
9092 (pdp11_fixed_cc_regs): New function.
9093 (pdp11_small_shift): New function.
9094 (pdp11_expand_shift): New function to expand shift insns.
9095 (pdp11_assemble_shift): New function to output shifts.
9096 (pdp11_branch_cost): Remove.
9097 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9098 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9099 (WCHAR_TYPE): Ditto.
9100 (PTRDIFF_TYPE): Ditto.
9101 (ADJUST_INSN_LENGTH): New macro.
9102 (FIXED_REGISTERS): Add CC registers.
9103 (CALL_USED_REGISTERS): Ditto.
9105 (REG_CLASS_NAMES): Ditto.
9106 (REG_CLASS_CONTENTS): Ditto.
9107 (SELECT_CC_MODE): Use new function.
9108 (TARGET_FLAGS_REGNUM): New macro.
9109 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9110 (cc0_reg_rtx): Remove.
9111 (CC_STATUS_MDEP): Remove.
9112 (CC_STATUS_MDEFP_INIT): Remove.
9113 (CC_IN_FPU): Remove.
9114 (NOTICE_UPDATE_CC): Remove.
9115 (REGISTER_NAMES): Add CC registers.
9116 (BRANCH_COST): Change to constant 1.
9117 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9119 * config/pdp11/pdp11.opt (mbcopy): Remove.
9120 (mbcopy-builtin): Remove.
9121 (mbranch-cheap): Remove.
9122 (mbranch-expensive): Remove.
9123 * config/pdp11/predicates.md (expand_shift_operand): Update to
9124 match shift code generation.
9125 (ccnz_operator): New predicate.
9126 * doc/invoke.texi (PDP-11 Options): Remove deleted options
9127 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9128 Remove non-existent option -mabshi, -mno-abshi. Document mutually
9130 * doc/md.texi (PDP-11): Document new D and h constraints. Update
9131 description of O constraint.
9133 2018-06-27 Jeff Law <law@redhat.com>
9134 Austin Law <austinklaw@gmail.com>
9136 * config/v850/v850.md (addsi3_set_flags): New pattern.
9137 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9138 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9139 (zero_extendhisi2_v850_set_flags): Likewise.
9140 (zero_extendqisi2_v850_set_flags): Likewise.
9141 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9142 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9143 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9145 * config/v850/v850-protos.h (notice_update_cc): Remove.
9146 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9147 (v850_print_operand): Handle 'D' and "d".
9148 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9149 Add handling of arithmetic/logical operations compared against zero.
9150 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9151 Do not look at v850_compare_op, instead get mode from last argument.
9152 (v850_gen_compare): Remove
9153 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9154 after reload for prologue insns.
9155 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9157 (construct_save_jarl): Likewise.
9158 (TARGET_FLAGS_REGNUM): Define.
9159 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9160 (NOTICE_UPDATE_CC): Remove.
9161 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9162 than cc0. Conditionalize on reload_completed.
9163 (cmpsi_insn, setfcc_insn): Likewise.
9164 (tst1 splitter): Turn into define_and_split which sets the flags
9166 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9167 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9168 (cstoresf4, cstoredf4): Clobber the flags.
9169 (cmpsi, cmpsf, cmpdf): Remove expanders.
9170 (setf_insn): Remove pattern.
9171 (addsi3): Turn into define_and_split which clobbers the flags after
9172 reload and a suitable pattern (addsi3_clobber_flags) for use after
9174 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9175 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9176 (ashrsi3, ashrsi3_v850e2): Likewise.
9177 (bins): Clobber the flags.
9178 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9179 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9180 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9181 (call_value_internal_short, call_value_internal_long): Likewise.
9182 (callt_save_interrupt, callt_return_interrupt): Likewise.
9183 (save_interrupt, return_interrupt): Likewise.
9184 (callt_save_all_interrupt, save_all_interrupt): Likewise.
9185 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9186 (restore_all_interrupt, _restore_all_interrupt): Likewise.
9187 (All FP comparisons): Only allow after reload has completed.
9189 (divh, divhu): Tweak output template.
9190 (branch_z_normal, branch_z_invert): Remove
9191 (branch_nz_normal, branch_nz_invert): Likewise.
9192 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9194 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9195 * config/v850/v850.c (notice_update_cc): Remove.
9196 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
9197 (CC_NO_CARRY): Likewise.
9198 (NOTICE_UPDATE_CC): Define to nothing.
9199 * config/v850/v850.md: Remove block comment on cc0 handling
9200 Remove "cc" attribute from all patterns. Remove cc_status handling
9201 from all patterns. Minor formatting fixes.
9203 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9205 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9206 (cortex-a76.cortex-a55): Likewise.
9207 * config/aarch64/aarch64-tune.md: Regenerate.
9208 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9209 cortex-a76.cortex-a55.
9211 2018-06-27 Jeff Law <law@redhat.com>
9213 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9214 (MULTILIB_DIRNAMES): Similarly.
9216 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
9218 * gimple.h (gimple_return_retbnd): Delete.
9219 (gimple_return_set_retbnd): Likewise.
9220 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9221 gimple_return_set_retbnd.
9222 * gimple-pretty-print.c (dump_gimple_return): Remove call to
9223 gimple_return_retbnd and adjust.
9224 * tree-inline.h (struct copy_body_data): Remove retbnd field.
9225 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9226 Explicitly return NULL in a couple more cases. Move assertion
9227 on debug statements and remove unreachable code.
9228 (reset_debug_binding): Do not test id->retbnd.
9229 (expand_call_inline): Do not set it.
9231 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9233 * configure.ac: Add --disable-gcov option.
9234 * configure: Regenerate.
9235 * Makefile.in: Honour @enable_gcov@.
9236 * doc/install.texi: Document --disable-gcov.
9238 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9240 * config/arm/arm-cpus.in (cortex-a76): New entry.
9241 (cortex-a76.cortex-a55): Likewise.
9242 * config/arm/arm-tables.opt: Regenerate.
9243 * config/arm/arm-tune.md: Likewise.
9244 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
9245 * doc/invoke.texi (ARM Options): Document cortex-a76 and
9246 cortex-a76.cortex-a55.
9248 2018-06-27 Tamar Christina <tamar.christina@arm.com>
9251 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
9253 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
9255 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
9257 (EPILOGUE_USES): Likewise.
9259 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
9261 * tree-inline.c (remap_location): New function extracted from...
9262 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
9263 (copy_phis_for_bb): ...here. Call remap_location.
9264 (copy_cfg_body): Adjust call to copy_edges_for_bb.
9266 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
9268 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
9269 unaligned vsx for 16B memset.
9271 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
9274 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
9275 ieee128_float_type_node to long_double_type_node unless
9276 TARGET_LONG_DOUBLE_128 is set.
9278 2018-06-26 David Malcolm <dmalcolm@redhat.com>
9280 * cfgloop.c (get_loop_location): Convert return type from
9281 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
9282 by implicit construction from rtx_insn *, and using
9283 dump_user_location_t::from_function_decl for the fallback case.
9284 * cfgloop.h (get_loop_location): Convert return type from
9285 location_t to dump_user_location_t.
9286 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
9287 dump_printf_loc to pass in a dump_location_t rather than a
9288 location_t, via the gimple stmt.
9289 * coverage.c (get_coverage_counts): Update calls to
9290 dump_printf_loc to pass in dump_location_t rather than a
9292 * doc/optinfo.texi (Dump types): Convert example of
9293 dump_printf_loc from taking "locus" to taking "insn". Update
9294 description of the "_loc" calls to cover dump_location_t.
9295 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
9297 (dump_user_location_t::dump_user_location_t): New constructors,
9298 from gimple *stmt and rtx_insn *.
9299 (dump_user_location_t::from_function_decl): New function.
9300 (dump_loc): Make static.
9301 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
9302 const dump_location_t &.
9303 (dump_generic_expr_loc): Delete.
9304 (dump_printf_loc): Convert param "loc" from location_t to
9305 const dump_location_t &.
9306 (selftest::test_impl_location): New function.
9307 (selftest::dumpfile_c_tests): New function.
9308 * dumpfile.h: Include "profile-count.h".
9309 (class dump_user_location_t): New class.
9310 (struct dump_impl_location_t): New struct.
9311 (class dump_location_t): New class.
9312 (dump_printf_loc): Convert 2nd param from source_location to
9313 const dump_location_t &.
9314 (dump_generic_expr_loc): Delete.
9315 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
9316 const dump_location_t &.
9317 * gimple-fold.c (fold_gimple_assign): Update call to
9318 dump_printf_loc to pass in a dump_location_t rather than a
9319 location_t, via the gimple stmt.
9320 (gimple_fold_call): Likewise.
9321 * gimple-loop-interchange.cc
9322 (loop_cand::analyze_iloop_reduction_var): Update for change to
9323 check_reduction_path.
9324 (tree_loop_interchange::interchange): Update for change to
9326 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
9327 change in return-type of find_loop_location.
9328 (graphite_regenerate_ast_isl): Likewise.
9329 * graphite-optimize-isl.c (optimize_isl): Likewise.
9330 * graphite.c (graphite_transform_loops): Likewise.
9331 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
9332 pass in a dump_location_t rather than a location_t, via the
9334 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9335 * ipa.c (walk_polymorphic_call_targets): Likewise.
9336 * loop-unroll.c (report_unroll): Convert "locus" param from
9337 location_t to dump_location_t.
9338 (decide_unrolling): Update for change to get_loop_location's
9340 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
9341 location_t to dump_user_location_t.
9342 (grid_find_single_omp_among_assignments_1): Updates calls to
9343 dump_printf_loc to pass in a dump_location_t rather than a
9344 location_t, via the gimple stmt.
9345 (grid_parallel_clauses_gridifiable): Convert "tloc" from
9346 location_t to dump_location_t. Updates calls to dump_printf_loc
9347 to pass in a dump_location_t rather than a location_t, via the
9349 (grid_inner_loop_gridifiable_p): Likewise.
9350 (grid_dist_follows_simple_pattern): Likewise.
9351 (grid_gfor_follows_tiling_pattern): Likewise.
9352 (grid_target_follows_gridifiable_pattern): Likewise.
9353 (grid_attempt_target_gridification): Convert initialization
9354 of local "grid" from memset to zero-initialization; FIXME: does
9355 this require C++11? Update call to dump_printf_loc to pass in a
9356 optinfo_location rather than a location_t, via the gimple stmt.
9357 * profile.c (read_profile_edge_counts): Updates call to
9358 dump_printf_loc to pass in a dump_location_t rather than a
9360 (compute_branch_probabilities): Likewise.
9361 * selftest-run-tests.c (selftest::run_tests): Call
9363 * selftest.h (dumpfile_c_tests): New decl.
9364 * tree-loop-distribution.c (pass_loop_distribution::execute):
9365 Update for change in return type of find_loop_location.
9366 * tree-parloops.c (parallelize_loops): Likewise.
9367 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
9368 "locus" from location_t to dump_user_location_t.
9369 (canonicalize_loop_induction_variables): Likewise.
9370 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
9371 for change in return type of find_loop_location.
9372 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
9373 to dump_printf_loc to pass in a dump_location_t rather than a
9374 location_t, via the stmt.
9375 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
9377 * tree-vect-loop-manip.c (find_loop_location): Convert return
9378 type from source_location to dump_user_location_t.
9379 (vect_do_peeling): Update for above change.
9380 (vect_loop_versioning): Update for change in type of
9382 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
9383 from location_t to dump_user_location_t.
9384 (vect_estimate_min_profitable_iters): Update for change in type
9386 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
9387 location_t to dump_location_t.
9388 (vect_slp_bb): Update for change in type of vect_location.
9389 * tree-vectorizer.c (vect_location): Convert from source_location
9390 to dump_user_location_t.
9391 (try_vectorize_loop_1): Update for change in vect_location's type.
9392 (vectorize_loops): Likewise.
9393 (increase_alignment): Likewise.
9394 * tree-vectorizer.h (vect_location): Convert from source_location
9395 to dump_user_location_t.
9396 (find_loop_location): Convert return type from source_location to
9397 dump_user_location_t.
9398 (check_reduction_path): Convert 1st param from location_t to
9399 dump_user_location_t.
9400 * value-prof.c (check_counter): Update call to dump_printf_loc to
9401 pass in a dump_user_location_t rather than a location_t; update
9402 call to error_at for change in type of "locus".
9403 (check_ic_target): Update call to dump_printf_loc to
9404 pass in a dump_user_location_t rather than a location_t, via the
9407 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
9409 * config/s390/s390.h (enum processor_flags): Do not use
9410 default tune parameter when -march was specified.
9412 2018-06-26 Jakub Jelinek <jakub@redhat.com>
9415 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
9416 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
9419 2018-06-26 Richard Biener <rguenther@suse.de>
9421 PR tree-optimization/86287
9423 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
9424 (vect_analyze_loop): Initialize n_stmts.
9426 2018-06-26 Richard Biener <rguenther@suse.de>
9429 * fold-const.c (fold_convertible_p): Pointer extension
9432 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
9435 * dwarf2out.c (loc_list_has_views): Adjust comments.
9436 (dw_loc_list): Split single cross-partition range with
9439 2018-06-25 Jeff Law <law@redhat.com>
9441 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
9442 on -mbig-switch by default.
9444 * config/v850/predicates.md (const_float_1_operand): Fix match_code
9446 (const_float_0_operand): Remove unused predicate.
9447 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
9448 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
9449 (recipsf2): New expander. Original pattern now called
9451 (recipdf2, recipdf2_insn): Similarly.
9452 (rsqrtsf2, rsqrtsf2_insn): Similarly
9453 (rsqrtdf2, rsqrtdf2_insn): Similarly
9455 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
9457 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
9458 Simplify logic for FreeBSD (twice).
9460 2018-06-25 Martin Sebor <msebor@redhat.com>
9462 PR tree-optimization/86204
9463 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
9464 a strnlen result if it's less than the length of the string.
9466 2018-06-25 Martin Sebor <msebor@redhat.com>
9468 PR tree-optimization/85700
9469 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
9470 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
9471 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
9473 2018-06-25 Martin Sebor <msebor@redhat.com>
9475 * doc/extend.texi (Zero-length arrays): Update and clarify.
9477 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
9479 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
9480 added IEEE/IBM long double multilib support on PowerPC little
9481 endian Linux systems.
9482 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
9483 (MULTILIB_DEFAULTS): Likewise.
9484 * config/rs6000/rs6000.c (rs6000_option_override_internal):
9486 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
9487 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
9488 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
9490 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
9493 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
9494 (REORDER_45): Likewise.
9496 2018-06-25 Jeff Law <law@redhat.com>
9498 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
9499 dividend to 32 bits. Adjust length.
9500 (udivmodhi4): Cleanup output template. Fix length.
9502 2018-06-25 Carl Love <cel@us.ibm.com>
9504 * config/rs6000/vsx.md: Change word selector to prefered location.
9506 2018-06-25 Richard Biener <rguenther@suse.de>
9508 PR tree-optimization/86304
9509 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
9510 epilogue-if-converted loops as well.
9512 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
9514 * lto-section-out.c (lto_begin_section): Do not print section
9515 name for noaddr and unnumbered dumps.
9517 2018-06-25 Richard Biener <rguenther@suse.de>
9519 * tree-vectorizer.h (struct vec_info_shared): New structure
9520 with parts split out from struct vec_info and loop_nest from
9521 struct _loop_vec_info.
9522 (struct vec_info): Adjust accordingly.
9523 (struct _loop_vec_info): Likewise.
9524 (LOOP_VINFO_LOOP_NEST): Adjust.
9525 (LOOP_VINFO_DATAREFS): Likewise.
9526 (LOOP_VINFO_DDRS): Likewise.
9527 (struct _bb_vec_info): Likewise.
9528 (BB_VINFO_DATAREFS): Likewise.
9529 (BB_VINFO_DDRS): Likewise.
9530 (struct _stmt_vec_info): Add dr_aux member.
9531 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
9532 (DR_MISALIGNMENT_UNINITIALIZED): New.
9533 (set_dr_misalignment): Adjust.
9534 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
9535 (vect_analyze_loop): Adjust prototype.
9536 (vect_analyze_loop_form): Likewise.
9537 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9538 Compute dependences lazily.
9539 (vect_record_base_alignments): Use shared datarefs/ddrs.
9540 (vect_verify_datarefs_alignment): Likewise.
9541 (vect_analyze_data_refs_alignment): Likewise.
9542 (vect_analyze_data_ref_accesses): Likewise.
9543 (vect_analyze_data_refs): Likewise.
9544 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
9545 constructor parameter for shared part.
9546 (vect_analyze_loop_form): Pass in shared part and adjust.
9547 (vect_analyze_loop_2): Pass in storage for the number of
9548 stmts. Move loop nest finding to the caller. Compute
9550 (vect_analyze_loop): Pass in shared part.
9551 (vect_transform_loop): Verify shared datarefs are unchanged.
9552 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
9553 constructor parameter for shared part.
9554 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
9555 (vect_slp_bb): Verify shared datarefs are unchanged before
9557 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
9559 (new_stmt_vec_info): Initialize DR_AUX misalignment to
9560 DR_MISALIGNMENT_UNINITIALIZED.
9561 * tree-vectorizer.c (vec_info::vec_info): Add constructor
9562 parameter for shared part.
9563 (vec_info::~vec_info): Adjust.
9564 (vec_info_shared::vec_info_shared): New.
9565 (vec_info_shared::~vec_info_shared): Likewise.
9566 (vec_info_shared::save_datarefs): Likewise.
9567 (vec_info_shared::check_datarefs): Likewise.
9568 (try_vectorize_loop_1): Construct shared part live for analyses
9569 of a single loop for multiple vector sizes.
9570 * tree-parloops.c (gather_scalar_reductions): Adjust.
9572 2018-06-25 Richard Biener <rguenther@suse.de>
9574 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
9575 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
9576 (vect_analyze_data_refs): Remove similar code from here and
9577 simplify accordingly.
9579 2018-06-25 Richard Biener <rguenther@suse.de>
9581 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
9582 for reverse storage order accesses rather than asserting
9583 they cannot happen here.
9585 2018-06-25 Tom de Vries <tdevries@suse.de>
9588 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
9589 Use data16 instead of .byte for insn prefix.
9591 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
9594 * parser.c (make_char_string_pack): Pass this literal chars
9595 through cpp_interpret_string.
9596 (cp_parser_userdef_numeric_literal): Check the result of
9597 make_char_string_pack.
9599 2018-06-24 Maya Rashish <coypu@sdf.org>
9601 * ginclude/stddef.h: Simplify conditions around avoiding
9602 re-definition of __size_t.
9604 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9606 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
9607 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
9609 2018-06-22 Maya Rashish <coypu@sdf.org>
9611 * doc/invoke.texi (mno-fancy-math-387): Update for changes
9612 made to OpenBSD and NetBSD through the years.
9614 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
9616 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9617 behavior of vec_pack (vector double, vector double) to match
9618 behavior of vec_float2 (vector double, vector double).
9620 2018-06-22 Olivier Hainque <hainque@adacore.com>
9622 * gimplify.c (gimplify_function_tree): Prevent creation
9623 of a trampoline for the address of the current function
9624 passed to entry/exit instrumentation hooks.
9626 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
9629 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
9632 2018-06-22 Martin Liska <mliska@suse.cz>
9634 PR tree-optimization/86263
9635 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9636 Bail out if is_enabled is false.
9637 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
9639 (jump_table_cluster::is_enabled): New function.
9641 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9643 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9644 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
9645 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
9646 (lto_input_ts_binfo_tree_pointers): Likewise.
9647 * tree-streamer-out.c (streamer_write_tree_bitfields,
9648 write_ts_binfo_tree_pointers): Likewise.
9649 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
9651 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
9653 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
9655 2018-06-22 Martin Liska <mliska@suse.cz>
9657 * symbol-summary.h (get): Make it pure and inline move
9658 functionality from ::get function.
9659 (get): Remove and inline into ::get and ::get_create.
9660 (get_create): Move code from ::get function.
9662 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9665 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
9666 -x assembler-with-cpp.
9668 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9670 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
9671 _FILE_OFFSET_BITS=64 for C++.
9673 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
9675 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
9676 conversion insn that shows up when pr85657-3.c is compiled using
9677 IEEE 128-bit long double.
9678 (neg<mode>2_internal): Use the correct mode to check whether the
9679 mode is IBM extended.
9680 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
9681 multiply and divide external functions from being created more
9684 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9686 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
9688 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
9689 the edge can be forwarded.
9690 (cfg_layout_merge_blocks): Likewise.
9692 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9694 * except.c (finish_eh_generation): Commit edge insertions only after
9695 the EH edges have been redirected from post-landing to landing pads.
9697 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9699 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
9700 create_tmp_var_for to create the FRAME decl.
9701 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
9703 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
9705 * tree-inline.c (copy_edges_for_bb): Minor tweak.
9706 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
9707 debug statement when resetting its value.
9708 (expand_call_inline): Copy the locus of the call onto the assignment
9709 of the return value, if any. Use local variable in more cases.
9711 2018-06-21 Martin Liska <mliska@suse.cz>
9713 * ipa-pure-const.c (propagate_nothrow): Use
9714 funct_state_summaries->get.
9715 (dump_malloc_lattice): Likewise.
9716 (propagate_malloc): Likewise.
9718 2018-06-21 Richard Biener <rguenther@suse.de>
9720 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
9721 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
9722 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
9724 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
9725 BLOCK_ABSTRACT_ORIGIN unconditionally.
9727 2018-06-21 David Malcolm <dmalcolm@redhat.com>
9729 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
9731 * ipa-reference.c (ipa_reference_c_finalize): Delete
9732 ipa_ref_opt_sum_summaries and set it to NULL.
9734 2018-06-21 Tom de Vries <tdevries@suse.de>
9736 PR tree-optimization/85859
9737 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
9738 test with comment from bb_no_side_effects_p.
9740 2018-06-21 Richard Biener <rguenther@suse.de>
9742 PR tree-optimization/86232
9743 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
9744 max for constant niter.
9746 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9748 * config/aarch64/aarch64-simd.md
9749 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
9751 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
9753 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
9754 Make opernads of the unspec commutative.
9756 2018-06-21 Richard Biener <rguenther@suse.de>
9758 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
9759 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
9760 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
9761 (vect_analyze_data_ref_dependence): Re-order checks to deal with
9763 (vect_record_base_alignments): Do not record base alignment
9764 for gathers or scatters.
9765 (vect_compute_data_ref_alignment): Drop return value that is always
9766 true. Bail out early for gathers or scatters.
9767 (vect_enhance_data_refs_alignment): Bail out early for gathers
9769 (vect_find_same_alignment_drs): Likewise.
9770 (vect_analyze_data_refs_alignment): Remove dead code.
9771 (vect_slp_analyze_and_verify_node_alignment): Likewise.
9772 (vect_analyze_data_refs): For possible gathers or scatters do
9773 not create an alternate DR, just check their possible validity
9774 and mark them. Adjust DECL_NONALIASED handling to not rely
9776 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
9777 update inits of gathers or scatters.
9778 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
9779 Also copy gather/scatter flag to pattern vinfo.
9781 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
9783 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9784 behavior of vec_packsu (vector unsigned long long, vector unsigned
9785 long long) to match behavior of vec_packs with same signature.
9787 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
9788 Thomas Schwinge <thomas@codesourcery.com>
9789 Cesar Philippidis <cesar@codesourcery.com>
9791 * gimplify.c (gimplify_scan_omp_clauses): Add support for
9792 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9793 (gimplify_adjust_omp_clauses): Likewise.
9794 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
9795 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
9796 (gimplify_omp_target_update): Update handling of acc update and
9798 * omp-low.c (install_var_field): Remove unused parameter
9799 base_pointers_restrict.
9800 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
9801 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
9803 (omp_target_base_pointers_restrict_p): Delete.
9804 (scan_omp_target): Update call to scan_sharing_clauses.
9805 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
9807 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
9808 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9809 (convert_local_omp_clauses): Likewise.
9810 * tree-pretty-print.c (dump_omp_clause): Likewise.
9811 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
9813 (omp_clause_code_name): Likewise.
9815 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9818 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
9821 PR tree-optimization/86231
9822 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
9823 anti-range don't overwrite *vr0min before using it to compute *vr0max.
9825 2018-06-20 Tom de Vries <tdevries@suse.de>
9827 PR tree-optimization/86097
9828 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
9829 iv type if signedness of iv type is not the same as that of *nit.
9831 2018-06-20 Jakub Jelinek <jakub@redhat.com>
9833 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
9834 EDGE_EH edges, verify they are all EDGE_EH.
9836 2018-06-20 Maya Rashish <coypu@sdf.org>
9838 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
9840 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
9842 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
9843 * config/aarch64/aarch64.c (xgene1_tunings): Add
9844 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
9845 (aarch64_mode_valid_for_sched_fusion_p):
9846 Allow 16-byte modes.
9847 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
9848 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
9850 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
9852 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
9853 * config/aarch64/iterators.md (VQ2): New mode iterator.
9855 2018-06-20 Martin Liska <mliska@suse.cz>
9857 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9858 Change default ratio from 10 to 8.
9860 2018-06-20 Martin Liska <mliska@suse.cz>
9862 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9864 (bit_test_cluster::find_bit_tests): Likewise.
9865 (switch_decision_tree::analyze_switch_statement): Find clusters.
9866 * tree-switch-conversion.h (struct jump_table_cluster): Document
9869 2018-06-20 Martin Liska <mliska@suse.cz>
9871 * tree-switch-conversion.c (switch_conversion::collect):
9872 Record m_uniq property.
9873 (switch_conversion::expand): Bail out for special conditions.
9874 (group_cluster::~group_cluster): New.
9875 (group_cluster::group_cluster): Likewise.
9876 (group_cluster::dump): Likewise.
9877 (jump_table_cluster::emit): New.
9878 (switch_decision_tree::fix_phi_operands_for_edges): New.
9879 (struct case_node): Remove struct.
9880 (jump_table_cluster::can_be_handled): New.
9881 (case_values_threshold): Moved to header.
9882 (reset_out_edges_aux): Likewise.
9883 (jump_table_cluster::is_beneficial): New.
9884 (bit_test_cluster::can_be_handled): Likewise.
9885 (add_case_node): Remove.
9886 (bit_test_cluster::is_beneficial): New.
9887 (case_bit_test::cmp): New.
9888 (bit_test_cluster::emit): New.
9889 (expand_switch_as_decision_tree_p): Remove.
9890 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
9891 (fix_phi_operands_for_edge): Likewise.
9892 (switch_decision_tree::analyze_switch_statement): New.
9893 (compute_cases_per_edge): Move ...
9894 (switch_decision_tree::compute_cases_per_edge): ... here.
9895 (try_switch_expansion): Likewise.
9896 (switch_decision_tree::try_switch_expansion): Likewise.
9897 (record_phi_operand_mapping): Likewise.
9898 (switch_decision_tree::record_phi_operand_mapping): Likewise.
9899 (emit_case_decision_tree): Likewise.
9900 (switch_decision_tree::emit): Likewise.
9901 (balance_case_nodes): Likewise.
9902 (switch_decision_tree::balance_case_nodes): Likewise.
9903 (dump_case_nodes): Likewise.
9904 (switch_decision_tree::dump_case_nodes): Likewise.
9905 (emit_jump): Likewise.
9906 (switch_decision_tree::emit_jump): Likewise.
9907 (emit_cmp_and_jump_insns): Likewise.
9908 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
9909 (emit_case_nodes): Likewise.
9910 (switch_decision_tree::emit_case_nodes): Likewise.
9911 (conditional_probability): Remove.
9912 * tree-switch-conversion.h (enum cluster_type): New.
9914 (struct cluster): Likewise.
9915 (cluster::cluster): Likewise.
9916 (struct simple_cluster): Likewise.
9917 (simple_cluster::simple_cluster): Likewise.
9918 (struct group_cluster): Likewise.
9919 (struct jump_table_cluster): Likewise.
9920 (struct bit_test_cluster): Likewise.
9921 (struct min_cluster_item): Likewise.
9922 (struct case_tree_node): Likewise.
9923 (case_tree_node::case_tree_node): Likewise.
9924 (jump_table_cluster::case_values_threshold): Likewise.
9925 (struct case_bit_test): Likewise.
9926 (struct switch_decision_tree): Likewise.
9927 (struct switch_conversion): Likewise.
9928 (switch_decision_tree::reset_out_edges_aux): Likewise.
9930 2018-06-20 Martin Liska <mliska@suse.cz>
9932 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
9933 (hoist_edge_and_branch_if_true): Likewise.
9934 (expand_switch_using_bit_tests_p): Likewise.
9935 (struct case_bit_test): Likewise.
9936 (case_bit_test_cmp): Likewise.
9937 (emit_case_bit_tests): Likewise.
9938 (switch_conversion::switch_conversion): New class.
9939 (struct switch_conv_info): Remove old struct.
9940 (collect_switch_conv_info): More to ...
9941 (switch_conversion::collect): ... this.
9942 (check_range): Likewise.
9943 (switch_conversion::check_range): Likewise.
9944 (check_all_empty_except_final): Likewise.
9945 (switch_conversion::check_all_empty_except_final): Likewise.
9946 (check_final_bb): Likewise.
9947 (switch_conversion::check_final_bb): Likewise.
9948 (create_temp_arrays): Likewise.
9949 (switch_conversion::create_temp_arrays): Likewise.
9950 (free_temp_arrays): Likewise.
9951 (gather_default_values): Likewise.
9952 (switch_conversion::gather_default_values): Likewise.
9953 (build_constructors): Likewise.
9954 (switch_conversion::build_constructors): Likewise.
9955 (constructor_contains_same_values_p): Likewise.
9956 (switch_conversion::contains_same_values_p): Likewise.
9957 (array_value_type): Likewise.
9958 (switch_conversion::array_value_type): Likewise.
9959 (build_one_array): Likewise.
9960 (switch_conversion::build_one_array): Likewise.
9961 (build_arrays): Likewise.
9962 (switch_conversion::build_arrays): Likewise.
9963 (gen_def_assigns): Likewise.
9964 (switch_conversion::gen_def_assigns): Likewise.
9965 (prune_bbs): Likewise.
9966 (switch_conversion::prune_bbs): Likewise.
9967 (fix_phi_nodes): Likewise.
9968 (switch_conversion::fix_phi_nodes): Likewise.
9969 (gen_inbound_check): Likewise.
9970 (switch_conversion::gen_inbound_check): Likewise.
9971 (process_switch): Use the newly created class.
9972 (switch_conversion::expand): New.
9973 (switch_conversion::~switch_conversion): New.
9974 * tree-switch-conversion.h: New file.
9976 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
9978 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
9979 tree-vect-patterns.c.
9980 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
9981 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
9982 (vect_recog_sad_pattern): Likewise.
9983 (vect_recog_widen_sum_pattern): Likewise.
9984 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
9985 (vect_recog_widen_shift_pattern): Remove the type_in argument.
9986 (vect_recog_rotate_pattern): Likewise.
9987 (vect_recog_mult_pattern): Likewise.
9988 (vect_recog_vector_vector_shift_pattern): Likewise.
9989 (vect_recog_divmod_pattern): Likewise.
9990 (vect_recog_mixed_size_cond_pattern): Likewise.
9991 (vect_recog_bool_pattern): Likewise.
9992 (vect_recog_mask_conversion_pattern): Likewise.
9993 (vect_try_gather_scatter_pattern): Likewise.
9994 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
9995 (vect_recog_over_widening_pattern): Likewise.
9996 (vect_recog_gather_scatter_pattern): Likewise.
9997 (vect_recog_func_ptr): Move from tree-vectorizer.h
9998 (vect_vect_recog_func_ptrs): Move further down the file.
9999 (vect_recog_func): Likewise. Remove the third argument.
10000 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
10001 (vect_pattern_recog_1): Expect the pattern function to do any
10002 necessary target tests. Also expect it to provide a vector type.
10003 Remove the type_in handling.
10005 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10007 * tree-vect-patterns.c (vect_pattern_detected): New function.
10008 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
10009 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
10010 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
10011 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
10012 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
10013 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
10014 (vect_recog_mask_conversion_pattern)
10015 (vect_try_gather_scatter_pattern): Likewise.
10017 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10019 * tree-vect-patterns.c (vect_get_internal_def): New function.
10020 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
10021 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
10022 (search_type_for_mask_1): Use it.
10024 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10026 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10027 redundant WIDEN_SUM_EXPR handling.
10028 (vect_recog_sad_pattern): Likewise.
10030 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10032 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
10033 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
10034 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
10035 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
10036 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
10038 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10040 * tree-vect-stmts.c (vectorizable_call): Make sure that we
10041 use the stmt_vec_info of the original bb statement for the
10042 new zero assignment, even if the call is part of a pattern.
10044 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10046 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
10047 that the sequence is attached to the original statement rather
10048 than the pattern statement.
10049 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
10050 PATTERN_DEF_SEQ from the original statement rather than
10051 the main pattern statement.
10052 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
10053 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
10054 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
10056 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
10058 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
10059 definition statements before the early exit for statements that aren't
10061 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
10063 (vect_transform_loop): ...here. Process pattern definition
10064 statements without first checking whether the main pattern
10065 statement is live or relevant.
10067 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10069 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
10070 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
10072 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
10074 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
10075 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
10076 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
10077 (expand_block_compare): Change select_block_compare_mode call.
10078 (expand_strncmp_align_check): Use new functions, fix comment.
10079 (emit_final_str_compare_gpr): New function.
10080 (expand_strn_compare): Refactor and clean up code.
10081 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
10083 2018-06-19 Tony Reix <tony.reix@atos.com>
10084 Damien Bergamini <damien.bergamini@atos.com>
10085 David Edelsohn <dje.gcc@gmail.com>
10087 * collect2.c (static_obj): New variable.
10088 (static_libs): New variable.
10089 (is_in_list): Uncomment declaration.
10090 (main): Track AIX libraries linked statically.
10091 (is_in_list): Uncomment definition.
10092 (scan_prog_file): Don't add AIX shared libraries initializer
10093 to constructor list if linking statically.
10095 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
10097 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10099 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10101 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
10103 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10106 2018-06-19 Martin Liska <mliska@suse.cz>
10108 * config/i386/i386.c (ix86_can_inline_p): Do not use
10109 ipa_fn_summaries::get_create.
10110 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10112 (devirtualization_time_bonus): Likewise.
10113 (ipcp_propagate_stage): Likewise.
10114 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10115 (edge_set_predicate): Likewise.
10116 (evaluate_conditions_for_known_args): Likewise.
10117 (evaluate_properties_for_edge): Likewise.
10118 (ipa_call_summary::reset): Tranform to ...
10119 (ipa_call_summary::~ipa_call_summary): ... this.
10120 (ipa_fn_summary::reset): Transform to ...
10121 (ipa_fn_summary::~ipa_fn_summary): ... this.
10122 (ipa_fn_summary_t::remove): Rename to ...
10123 (ipa_fn_summary_t::remove_callees): ... this.
10124 (ipa_fn_summary_t::duplicate): Use placement new
10125 instead of memory copy.
10126 (ipa_call_summary_t::duplicate): Likewise.
10127 (ipa_call_summary_t::remove): Remove.
10128 (dump_ipa_call_summary): Change get_create to get.
10129 (ipa_dump_fn_summary): Dump only when summary exists.
10130 (analyze_function_body): Use symbol_summary::get instead
10132 (compute_fn_summary): Likewise.
10133 (estimate_edge_devirt_benefit): Likewise.
10134 (estimate_edge_size_and_time): Likewise.
10135 (inline_update_callee_summaries): Likewise.
10136 (remap_edge_change_prob): Likewise.
10137 (remap_edge_summaries): Likewise.
10138 (ipa_merge_fn_summary_after_inlining): Likewise.
10139 (write_ipa_call_summary): Likewise.
10140 (ipa_fn_summary_write): Likewise.
10141 (ipa_free_fn_summary): Likewise.
10142 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10143 (struct ipa_call_summary): Likewise.
10144 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10146 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10147 (estimate_size_after_inlining): Likewise.
10148 (estimate_growth): Likewise.
10149 (growth_likely_positive): Likewise.
10150 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10151 (inline_call): Likewise.
10152 * ipa-inline.c (caller_growth_limits): Likewise.
10153 (can_inline_edge_p): Likewise.
10154 (can_inline_edge_by_limits_p): Likewise.
10155 (compute_uninlined_call_time): Likewise.
10156 (compute_inlined_call_time): Likewise.
10157 (want_inline_small_function_p): Likewise.
10158 (edge_badness): Likewise.
10159 (update_caller_keys): Likewise.
10160 (update_callee_keys): Likewise.
10161 (inline_small_functions): Likewise.
10162 (inline_to_all_callers_1): Likewise.
10163 (dump_overall_stats): Likewise.
10164 (early_inline_small_functions): Likewise.
10165 (early_inliner): Likewise.
10166 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10167 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10168 * ipa-pure-const.c (malloc_candidate_p): Likewise.
10169 * ipa-split.c (execute_split_functions): Likewise.
10170 * symbol-summary.h: Likewise.
10171 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10173 2018-06-19 Richard Biener <rguenther@suse.de>
10175 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10176 (vectorize_loops): ... here. Fix dbgcnt handling.
10177 (try_vectorize_loop): Wrap try_vectorize_loop_1.
10179 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
10182 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10183 ieee128 argument takes up only one (vector) register, not two (floating
10186 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
10188 * gimplify.c (gimplify_init_constructor): Really never clear for an
10189 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10191 2018-06-19 Richard Biener <rguenther@suse.de>
10193 PR tree-optimization/86179
10194 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10195 after failed recognition.
10197 2018-06-18 Martin Sebor <msebor@redhat.com>
10199 PR middle-end/85602
10200 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10201 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10202 Handle integer subtraction.
10203 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10204 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10206 2018-06-18 David Malcolm <dmalcolm@redhat.com>
10208 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10209 param from rtx to rtx_insn *.
10210 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10212 (frv_ifcvt_modify_insn): Likwise.
10213 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10214 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
10215 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10216 as_a <rtx_insn *> cast to local "unprotected_region" once
10217 it's been established that it's not NULL or pc_rtx.
10218 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10219 param "sethi" from rtx to rtx_insn *.
10220 (nds32_group_float_insns): Likewise for param "insn".
10221 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10223 (vax_output_int_subtract): Likewise.
10224 * config/vax/vax.c (vax_output_int_add): Likewise for param
10226 (vax_output_int_subtract): Likewise.
10227 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10228 (emit_pattern_after): Likewise for param "after".
10229 (emit_insn_after): Likewise.
10230 (emit_jump_insn_after): Likewise.
10231 (emit_call_insn_after): Likewise.
10232 (emit_debug_insn_after): Likewise.
10233 (emit_pattern_before): Likewise for param "before".
10234 (emit_insn_before): Likewise.
10235 (emit_jump_insn_before): Likewise.
10236 * final.c (get_insn_template): Likewise for param "insn", removing
10238 * output.h (get_insn_template): Likewise for 2nd param.
10239 * rtl.h (emit_insn_before): Likewise.
10240 (emit_jump_insn_before): Likewise.
10241 (emit_debug_insn_before_noloc): Likewise.
10242 (emit_insn_after): Likewise.
10243 (emit_jump_insn_after): Likewise.
10244 (emit_call_insn_after): Likewise.
10245 (emit_debug_insn_after): Likewise.
10246 (set_insn_deleted): Likewise for param.
10248 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
10251 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
10252 floating point modes, so that IFmode is numerically greater than
10253 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
10254 to declare the ordering. This prevents IFmode from being
10255 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
10256 machine. Include rs6000-modes.h to share the fractional values
10257 between genmodes* and the rest of the compiler.
10258 (IFmode): Likewise.
10259 (KFmode): Likewise.
10260 (TFmode): Likewise.
10261 * config/rs6000/rs6000-modes.h: New file.
10262 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
10263 meaning of rs6000_long_double_size so that 126..128 selects an
10264 appropriate 128-bit floating point type.
10265 (rs6000_option_override_internal): Likewise.
10266 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
10267 (TARGET_LONG_DOUBLE_128): Change the meaning of
10268 rs6000_long_double_size so that 126..128 selects an appropriate
10269 128-bit floating point type.
10270 (LONG_DOUBLE_TYPE_SIZE): Update comment.
10271 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
10272 source and destination to match the standard usage.
10273 (truncifkf2): Likewise.
10274 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
10275 ISA 2.07 to use an explicit clobber, instead of passing in a
10277 (copysign<mode>3_soft): Likewise.
10279 2018-06-18 David Malcolm <dmalcolm@redhat.com>
10281 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10282 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
10283 (vect_slp_analyze_instance_dependence): Likewise.
10284 (vect_enhance_data_refs_alignment): Likewise.
10285 (vect_analyze_data_refs_alignment): Likewise.
10286 (vect_slp_analyze_and_verify_instance_alignment
10287 (vect_analyze_data_ref_accesses): Likewise.
10288 (vect_prune_runtime_alias_test_list): Likewise.
10289 (vect_analyze_data_refs): Likewise.
10290 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
10291 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10292 (vect_analyze_scalar_cycles_1): Likewise.
10293 (vect_get_loop_niters): Likewise.
10294 (vect_analyze_loop_form_1): Likewise.
10295 (vect_update_vf_for_slp): Likewise.
10296 (vect_analyze_loop_operations): Likewise.
10297 (vect_analyze_loop): Likewise.
10298 (vectorizable_induction): Likewise.
10299 (vect_transform_loop): Likewise.
10300 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
10301 * tree-vect-slp.c (vect_analyze_slp): Likewise.
10302 (vect_make_slp_decision): Likewise.
10303 (vect_detect_hybrid_slp): Likewise.
10304 (vect_slp_analyze_operations): Likewise.
10305 (vect_slp_bb): Likewise.
10306 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
10307 (vectorizable_bswap): Likewise.
10308 (vectorizable_call): Likewise.
10309 (vectorizable_simd_clone_call): Likewise.
10310 (vectorizable_conversion): Likewise.
10311 (vectorizable_assignment): Likewise.
10312 (vectorizable_shift): Likewise.
10313 (vectorizable_operation): Likewise.
10314 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
10316 2018-06-18 Martin Sebor <msebor@redhat.com>
10318 PR tree-optimization/81384
10319 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
10320 * builtins.c (expand_builtin_strnlen): New function.
10321 (expand_builtin): Call it.
10322 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
10323 * builtins.def (BUILT_IN_STRNLEN): New.
10324 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
10325 Warn for bounds in excess of maximum object size.
10326 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
10327 single-value ranges. Handle strnlen.
10328 (handle_builtin_strlen): Handle strnlen.
10329 (strlen_check_and_optimize_stmt): Same.
10330 * doc/extend.texi (Other Builtins): Document strnlen.
10332 2018-06-18 Maya Rashish <coypu@sdf.org>
10334 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
10335 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
10336 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
10338 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
10340 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
10342 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
10344 * tree.c (escaped_string::escape): Replace cast to char * by
10345 const_cast<char *> (unescaped).
10347 2018-06-18 Nick Clifton <nickc@redhat.com>
10350 * tree.c (escaped_string): New class. Converts an unescaped
10351 string into its escaped equivalent.
10352 (warn_deprecated_use): Use the new class to convert the
10353 deprecation message, if present.
10354 (test_escaped_strings): New self test.
10355 (test_c_tests): Add test_escaped_strings.
10356 * doc/extend.texi (deprecated): Add a note that the
10357 deprecation message is affected by the -fmessage-length
10358 option, and that control characters will be escaped.
10359 (#pragma GCC error): Document this pragma.
10360 (#pragma GCC warning): Likewise.
10361 * doc/invoke.texi (-fmessage-length): Document this option's
10362 effect on the #warning and #error preprocessor directives and
10363 the deprecated attribute.
10365 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
10367 * tree.c (decl_value_expr_lookup): Revert latest change.
10368 (decl_value_expr_insert): Likewise.
10370 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
10372 * gimplify.c (nonlocal_vlas): Delete.
10373 (nonlocal_vla_vars): Likewise.
10374 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
10376 (gimplify_body): Do not create and destroy nonlocal_vlas.
10377 * tree-nested.c: Include diagnostic.h.
10378 (use_pointer_in_frame): Tweak.
10379 (lookup_field_for_decl): Add assertion and declare the transformation.
10380 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
10381 internal error when the reference is in a wrong context. Do not
10382 create a debug decl by default.
10383 (note_nonlocal_block_vlas): Delete.
10384 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
10385 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
10386 create a debug decl by default.
10387 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
10388 call is in a wrong context.
10389 (fixup_vla_decls): New function.
10390 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
10391 debug variables were created.
10392 * tree.c (decl_value_expr_lookup): Add checking assertion.
10393 (decl_value_expr_insert): Likewise.
10395 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
10397 PR middle-end/82479
10398 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
10399 * tree-scalar-evolution.c (interpret_expr): Likewise.
10400 (expression_expensive_p): Likewise.
10401 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
10402 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
10403 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
10404 (ssa_defined_by_minus_one_stmt_p): New.
10406 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
10408 PR middle-end/64946
10409 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
10410 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
10411 * dojump.c (do_jump): Likewise.
10412 * expr.c (expand_expr_real_2): Check operand type's sign.
10413 * fold-const.c (const_unop): Handle ABSU_EXPR.
10414 (fold_abs_const): Likewise.
10415 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
10416 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
10417 (strip_sign_op_1): Likesise.
10418 * match.pd: Add new pattern to generate ABSU_EXPR.
10419 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
10420 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10421 * tree-eh.c (operation_could_trap_helper_p): Likewise.
10422 * tree-inline.c (estimate_operator_cost): Likewise.
10423 * tree-pretty-print.c (dump_generic_node): Likewise.
10424 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
10425 * tree.def (ABSU_EXPR): New.
10427 2018-06-16 Jakub Jelinek <jakub@redhat.com>
10429 PR middle-end/86095
10430 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
10431 documented as preserved for backward compatibility only.
10432 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
10434 PR rtl-optimization/86108
10435 * bb-reorder.c (create_forwarder_block): Renamed to ...
10436 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
10437 jump from new landing pad to the second part.
10438 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
10441 2018-06-15 Jakub Jelinek <jakub@redhat.com>
10443 PR middle-end/85878
10444 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
10445 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
10446 Only call store_expr for halves if the mode is the same.
10448 PR middle-end/86123
10449 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
10450 Fix up comment formatting.
10452 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
10454 * typed-splay-tree.h (typed_splay_tree::remove): New function.
10455 (typed_splay_tree::closure,
10456 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
10457 (typed_splay_tree::typed_splay_tree,
10458 typed_splay_tree::operator =): Declared private.
10459 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
10460 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
10461 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
10462 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
10463 typed_splay_tree::splay_tree_splay,
10464 typed_splay_tree::splay_tree_foreach_helper,
10465 typed_splay_tree::splay_tree_insert,
10466 typed_splay_tree::splay_tree_remove,
10467 typed_splay_tree::splay_tree_lookup,
10468 typed_splay_tree::splay_tree_predecessor,
10469 typed_splay_tree::splay_tree_successor,
10470 typed_splay_tree::splay_tree_min,
10471 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
10472 (typed_splay_tree::root, typed_splay_tree::comp,
10473 typed_splay_tree::delete_key,
10474 typed_splay_tree::delete_value): New data members.
10475 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
10476 typed_splay_tree::remove.
10478 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
10480 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
10481 -mginv and -mno-ginv to the assembler.
10482 * config/mips/mips.opt (-mcrc): New option.
10483 (-mginv): Likewise.
10484 * doc/invoke.text (-mcrc): Document.
10485 (-mginv): Likewise.
10487 2018-06-15 Nick Clifton <nickc@redhat.com>
10490 * tree.c (escaped_string): New class. Converts an unescaped
10491 string into its escaped equivalent.
10492 (warn_deprecated_use): Use the new class to convert the
10493 deprecation message, if present.
10494 (test_escaped_strings): New self test.
10495 (test_c_tests): Add test_escaped_strings.
10496 * doc/extend.texi (deprecated): Add a note that the
10497 deprecation message is affected by the -fmessage-length
10498 option, and that control characters will be escaped.
10499 (#pragma GCC error): Document this pragma.
10500 (#pragma GCC warning): Likewise.
10501 * doc/invoke.texi (-fmessage-length): Document this option's
10502 effect on the #warning and #error preprocessor directives and
10503 the deprecated attribute.
10505 2018-06-15 Richard Biener <rguenther@suse.de>
10507 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
10508 here, also noting vector size used.
10509 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
10510 size used in MSG_OPTIMIZED_LOCATIONS dump.
10511 (pass_slp_vectorize::execute): Adjust.
10513 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
10516 * config/arc/arc.c (arc_return_address_register): Fix
10519 2018-06-15 Richard Biener <rguenther@suse.de>
10521 PR middle-end/86159
10522 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
10523 leave useless conversion stripping to force_gimple_operand_gsi.
10524 (gimplify_build2): Likewise.
10525 (gimplify_build1): Likewise.
10527 2018-06-15 Richard Biener <rguenther@suse.de>
10529 PR middle-end/86076
10530 * tree-cfg.c (move_stmt_op): unshare invariant addresses
10531 before adjusting their block.
10533 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
10535 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
10536 multilibs for *-*-rtems*.
10537 * config/riscv/t-rtems: New file.
10539 2018-06-14 Jakub Jelinek <jakub@redhat.com>
10541 PR middle-end/86122
10542 * match.pd ((A +- CST1) +- CST2): Punt if last resort
10543 unsigned_type_for returns NULL.
10546 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
10547 subregs of multi-word pseudos unless the float mode has word size.
10549 2018-06-14 Richard Biener <rguenther@suse.de>
10551 PR middle-end/86139
10552 * tree-vect-generic.c (build_word_mode_vector_type): Remove
10553 duplicate and harmful type_hash_canon.
10554 * tree.c (type_hash_canon): Assert we didn't find ourselves.
10556 2018-06-14 Richard Biener <rguenther@suse.de>
10559 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
10562 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
10564 * config/rtems.h (STDINT_LONG32): Define.
10566 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
10567 Prachi Godbole <prachi.godbole@imgtec.com>
10569 * config/mips/mips-cpus.def: Define P6600.
10570 * config/mips/mips-tables.opt: Regenerate.
10571 * config/mips/mips.c (mips_ucbranch_type): New enum.
10572 (mips_rtx_cost_data): Add support for P6600.
10573 (mips_issue_rate): Likewise.
10574 (mips_multipass_dfa_lookahead): Likewise.
10575 (mips_avoid_hazard): Likewise.
10576 (mips_reorg_process_insns): Likewise.
10577 (mips_classify_branch_p6600): New function.
10578 * config/mips/mips.h (TUNE_P6600): New define.
10579 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
10580 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
10581 * config/mips/mips.md: Include p6600.md.
10582 (processor): Add p6600.
10583 * config/mips/p6600.md: New file.
10584 * doc/invoke.texi: Add p6600 to supported architectures.
10586 2018-06-13 Martin Sebor <msebor@redhat.com>
10588 PR tree-optimization/86114
10589 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
10591 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
10593 2018-06-13 Richard Biener <rguenther@suse.de>
10595 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
10596 Properly set vector type of the intermediate stmt.
10597 * tree-vect-stmts.c (vectorizable_operation): The destination
10598 var always has vectype_out type.
10600 2018-06-13 Jeff Law <law@redhat.com>
10602 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
10603 integer 0 for argument to print_rtl_with_bb.
10604 (rl78_reorg): Likewise.
10606 2018-06-13 David Malcolm <dmalcolm@redhat.com>
10608 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
10609 from rtx to rtx_insn *.
10610 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
10612 (add_sched_insns_for_speculation): Likewise for local "target",
10613 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10614 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
10615 from rtx_insn ** to rtx_code_label **.
10616 (reorg_emit_nops): Likewise.
10617 (c6x_reorg): Likewise for local "call_labels".
10618 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
10620 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
10621 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
10622 the loops over LABEL_REFS.
10623 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
10625 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10626 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
10627 (split_branches): Strengthen local "olabel" from rtx to
10628 rtx_insn *, adding a safe_as_a cast.
10629 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
10631 (add_insn_after): Likewise for first two params.
10632 (add_insn_before): Likewise.
10633 (remove_insn): Likewise for param.
10634 (emit_pattern_before_noloc): Likewise for second and third params.
10635 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
10636 (emit_call_insn_before_noloc): Likewise.
10637 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
10639 (emit_barrier_before): Likewise.
10640 (emit_label_before): Strengthen "label" param from "rtx" to
10641 "rtx_code_label *". Strengthen "before" param from "rtx" to
10643 (emit_insn_after_1): Strengthen "after" param from "rtx" to
10645 (emit_pattern_after_noloc): Likewise.
10646 (emit_insn_after_noloc): Likewise.
10647 (emit_jump_insn_after_noloc): Likewise.
10648 (emit_call_insn_after_noloc): Likewise.
10649 (emit_debug_insn_after_noloc): Likewise.
10650 (emit_barrier_after): Likewise.
10651 (emit_label_after): Likewise for both params.
10652 (emit_pattern_after_setloc): Likewise for "after" param. Convert
10653 "loc" param from "int" to "location_t".
10654 (emit_insn_after_setloc): Likewise.
10655 (emit_jump_insn_after_setloc): Likewise.
10656 (emit_call_insn_after_setloc): Likewise.
10657 (emit_debug_insn_after_setloc): Likewise.
10658 (emit_pattern_before_setloc): Likewise for "before" param. Convert
10659 "loc" param from "int" to "location_t".
10660 (emit_pattern_before): Convert NULL_RTX to NULL.
10661 (emit_insn_before_setloc): Convert "loc" param from "int" to
10663 (emit_jump_insn_before_setloc): Likewise.
10664 (emit_call_insn_before_setloc): Likewise.
10665 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
10666 rtx_insn *. Convert "loc" param from "int" to "location_t".
10667 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
10668 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
10669 Convert 3rd param from "int" to "location_t".
10670 (emit_barrier_before, emit_barrier_after, next_real_insn):
10671 Strengthen param from rtx to rtx_insn *.
10672 (emit_label_before): Strengthen 1st param from "rtx" to
10673 "rtx_code_label *". Strengthen 2nd param from "rtx" to
10675 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
10676 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
10677 Strengthen 2nd param from "rtx" to "rtx_insn *".
10678 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
10679 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
10680 Likewise. Convert 3rd param from "int" to "location_t".
10681 (emit_label_after): Strengthen 1st param from "rtx" to
10682 "rtx_code_label *".
10683 (next_real_insn, remove_insn): Strengthen param from "rtx" to
10685 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
10686 from "rtx" to "rtx_insn *".
10688 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
10690 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
10691 bodies streamed in with -Q.
10692 * dumpfile.c (dump_files): Add lto-stream-out dump file.
10693 * dumpfile.h (tree_dump_index): Add lto_stream_out.
10694 * gimple-streamer-out.c: Include gimple-pretty-print.h
10695 (output_bb): Dump stmts streamed.
10696 * lto-section-out.c: Include print-tree.h
10697 (lto_begin_section): Dump sections created.
10698 (lto_output_decl_index): Dump decl encoded.
10699 * lto-streamer-out.c: Include print-tree.h
10700 (create_output_block): Dump output block created.
10701 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
10702 (output_function): Dump function output.
10703 (output_constructor): Dump constructor streamed.
10704 (write_global_stream): Output indexes encoded.
10705 (produce_asm_for_decls): Dump streams encoded.
10706 * lto-streamer.c (streamer_dump_file): New global var.
10707 * lto-streamer.h (streamer_dump_file): Declare.
10708 * passes.c (ipa_write_summaries): Initialize streamer dump.
10709 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
10712 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
10715 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
10716 offsets for register save directives. Emit a second batch of save
10717 directives, if need be, when the function accesses prior frames.
10719 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10721 * config/arc/fpu.md (fmasf4): Force operand to register.
10722 (fnmasf4): Likewise.
10724 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10726 * config/arc/arc-protos.h (arc_pad_return): Remove.
10727 * config/arc/arc.c (machine_function): Remove force_short_suffix
10729 (arc_print_operand): Adjust printing of '&'.
10730 (arc_verify_short): Remove conditional printing of short suffix.
10731 (arc_final_prescan_insn): Remove reference to size_reason.
10732 (pad_return): New function.
10733 (arc_reorg): Call pad_return.
10734 (arc_pad_return): Remove.
10735 (arc_init_machine_status): Remove reference to force_short_suffix.
10736 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
10737 (attr length): When attribute iscompact is true force to 2
10738 regardless; in the case of maybe check if we want to force the
10739 instruction to have 4 bytes length.
10740 (nopv): Change it to generate 4 byte long nop as well.
10741 (blockage): New pattern.
10742 (simple_return): Remove call to arc_pad_return.
10743 (p_return_i): Likewise.
10745 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10747 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
10749 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10751 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
10754 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10756 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
10757 for ARC700 and ARCv2.
10759 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
10762 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
10763 operands[2] instead of operands[1].
10766 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10768 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
10769 case, check whether the outer register overlaps an unallocatable
10770 register, not just whether it fits the required class.
10772 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
10774 * poly-int.h (can_div_trunc_p): Add new overload in which all values
10776 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
10777 (memrefs_conflict_p): Likewise.
10778 (init_alias_analysis): Likewise.
10779 * cfgexpand.c (expand_debug_expr): Likewise.
10780 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
10781 * cse.c (fold_rtx): Likewise.
10782 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
10783 * expr.c (emit_block_move_hints): Likewise.
10784 (clear_storage_hints, push_block, emit_push_insn): Likewise.
10785 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
10786 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
10787 (emit_group_store): Likewise.
10788 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
10789 to read the PRE/POST_MODIFY increment.
10790 * calls.c (store_one_arg): Use strip_offset.
10791 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
10793 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
10795 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
10796 (simplify_binary_operation_1): Extend CONST_INT handling to
10798 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
10799 than a HOST_WIDE_INT.
10800 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
10802 (adjust_mems, add_stores): Update accodingly.
10803 (vt_canonicalize_addr): Track polynomial offsets.
10804 (emit_note_insn_var_location): Likewise.
10805 (vt_add_function_parameter): Likewise.
10806 (vt_initialize): Likewise.
10808 2018-06-12 Jeff Law <law@redhat.com>
10810 * config.gcc (alpha*-*-freebsd*): Remove.
10811 * config/alpha/freebsd.h: Remove.
10813 2018-06-12 David Malcolm <dmalcolm@redhat.com>
10816 * spellcheck-tree.c (levenshtein_distance): Rename to...
10817 (get_edit_distance): ...this, and update for underlying renaming.
10818 * spellcheck-tree.h (levenshtein_distance): Rename to...
10819 (get_edit_distance): ...this.
10820 * spellcheck.c (levenshtein_distance): Rename to...
10821 (get_edit_distance): ...this. Convert from Levenshtein distance
10822 to Damerau-Levenshtein distance by supporting transpositions of
10823 adjacent characters. Rename "v1" to "v_next" and "v0" to
10825 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
10826 (selftest::test_edit_distance_unit_test_oneway): ...this, and
10827 update for underlying renaming.
10828 (selftest::levenshtein_distance_unit_test): Rename to...
10829 (selftest::test_get_edit_distance_unit): ...this, and update for
10830 underlying renaming.
10831 (selftest::test_find_closest_string): Add example from PR 69968
10832 where transposition helps
10833 (selftest::test_metric_conditions): Update for renaming.
10834 (selftest::test_metric_conditions): Likewise.
10835 (selftest::spellcheck_c_tests): Likewise.
10836 * spellcheck.h (levenshtein_distance): Rename both overloads to...
10837 (get_edit_distance): ...this.
10838 (best_match::consider): Update for renaming.
10840 2018-06-12 Martin Sebor <msebor@redhat.com>
10842 PR tree-optimization/85259
10843 * builtins.c (compute_objsize): Handle constant offsets.
10844 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
10845 true iff a warning has been issued.
10846 * gimple.h (gimple_nonartificial_location): New function.
10847 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
10848 gimple_nonartificial_location and handle -Wno-system-headers.
10849 (handle_builtin_stxncpy): Same.
10851 2018-06-12 Martin Sebor <msebor@redhat.com>
10854 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
10856 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
10858 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
10859 BUILTIN_VEC_XST entries for pointer to double and long long.
10861 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
10864 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10866 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10869 2018-06-12 Martin Liska <mliska@suse.cz>
10871 * doc/options.texi: Document IntegerRange.
10873 2018-06-12 Martin Liska <mliska@suse.cz>
10875 * config/i386/i386.opt: Make MPX-related options as Deprecated.
10876 * opt-functions.awk: Handle Deprecated flag.
10877 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
10879 (read_cmdline_option): Report warning for a deprecated option.
10880 * opts.h (struct cl_option): Add new field cl_deprecated.
10881 (CL_ERR_DEPRECATED): New.
10883 2018-06-12 Martin Liska <mliska@suse.cz>
10885 * doc/options.texi: Document Deprecated option flag.
10887 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10889 * config/arc/arc-arch.h (arc_extras): New enum.
10890 (arc_cpu_t):Add field extra.
10891 (arc_cpu_types): Consider the extras.
10892 * config/arc/arc-cpus.def: Add extras info.
10893 * config/arc/arc-opts.h (processor_type): Consider extra field.
10894 * config/arc/arc.c (arc_override_options): Handle extra field.
10896 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
10898 * config/arc/arc-arch.h: Update ARC_OPTX macro.
10899 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
10901 * config/arc/arc.c (arc_init): Update pic warning.
10902 (irq_range): Update irq range parsing warnings.
10903 (arc_override_options): Update various warning messages.
10904 (arc_handle_aux_attribute): Likewise.
10906 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
10908 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
10910 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10912 * doc/sourcebuild.texi: Document usage of line number 0 in verify
10913 compiler messages directives.
10915 2018-06-12 Matthew Fortune <mfortune@gmail.com>
10917 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
10918 * config/mips/mips-tables.opt: Regenerate.
10919 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
10921 * doc/invoke.texi: Document -march=i6500.
10923 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
10925 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
10926 (i6400_gpmul): Add cpu_unit.
10927 (i6400_gpdiv): Likewise.
10928 (i6400_msa_add_d): Update reservations.
10929 (i6400_msa_int_add) Likewise.
10930 (i6400_msa_short_logic3) Likewise.
10931 (i6400_msa_short_logic2) Likewise.
10932 (i6400_msa_short_logic) Likewise.
10933 (i6400_msa_move) Likewise.
10934 (i6400_msa_cmp) Likewise.
10935 (i6400_msa_short_float2) Likewise.
10936 (i6400_msa_div_d) Likewise.
10937 (i6400_msa_long_logic1) Likewise.
10938 (i6400_msa_long_logic2) Likewise.
10939 (i6400_msa_mult) Likewise.
10940 (i6400_msa_long_float2) Likewise.
10941 (i6400_msa_long_float4) Likewise.
10942 (i6400_msa_long_float5) Likewise.
10943 (i6400_msa_long_float8) Likewise.
10944 (i6400_fpu_fadd): Include frint type.
10945 (i6400_fpu_store): New define_insn_reservation.
10946 (i6400_fpu_load): Likewise.
10947 (i6400_fpu_move): Likewise.
10948 (i6400_fpu_fcmp): Likewise.
10949 (i6400_fpu_fmadd): Likewise.
10950 (i6400_int_mult): Include imul3nc type and update reservation.
10951 (i6400_int_div): Include idiv3 type and update reservation.
10952 (i6400_int_load): Update to check type not move_type.
10953 (i6400_int_store): Likewise.
10954 (i6400_int_prefetch): Set zero latency.
10956 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
10958 * gcc.c: Document new %@{...} sequence.
10959 (LINK_COMMAND_SPEC): Use it for the -L switches.
10960 (cpp_unique_options): Use it for the -I switches.
10961 (at_file_argbuf): New global variable.
10962 (in_at_file): Likewise.
10963 (alloc_args): Create at_file_argbuf.
10964 (clear_args): Truncate at_file_argbuf.
10965 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
10966 (open_at_file): New function.
10967 (close_at_file): Likewise.
10968 (create_at_file): Delete.
10969 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
10972 (validate_switches_from_spec): Deal with %@{...} sequence.
10973 (validate_switches): Likewise.
10974 (driver::finalize): Call clear_args.
10976 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
10978 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
10980 2018-06-11 Martin Sebor <msebor@redhat.com>
10982 * doc/invoke.texi (-Wall): List -Wc++17-compat.
10983 (Wno-class-memaccess): Add @opindex.
10984 (Wno-templates, Wno-multiple-inheritance): Same.
10985 (Wno-virtual-inheritance, Wno-namespaces): Same.
10986 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
10987 (Wno-format-overflow, Wno-format-truncation): Same.
10988 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
10989 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
10990 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
10991 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
10992 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
10993 (Wno-misspelled-isr): Same.
10995 2018-06-11 Martin Sebor <msebor@redhat.com>
10997 * PR tree-optimization/86083
10998 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
11000 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
11002 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
11004 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
11007 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
11008 on the correct operand.
11009 (*movdi_internal64): Ditto.
11011 2018-06-11 Martin Liska <mliska@suse.cz>
11013 PR tree-optimization/86089
11014 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
11016 2018-06-11 Julia Koval <julia.koval@intel.com>
11018 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
11019 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
11020 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
11022 2018-06-11 Olivier Hainque <hainque@adacore.com>
11024 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
11025 for Ada with strict dwarf2.
11027 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
11030 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
11033 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
11035 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
11037 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
11039 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
11042 2018-06-08 Martin Liska <mliska@suse.cz>
11044 * tree-cfg.h (debug_function): Fix argument type to match
11047 2018-06-08 Martin Liska <mliska@suse.cz>
11049 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
11050 Remove usage of MPX-related (and removed) fields.
11051 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
11053 2018-06-08 David Malcolm <dmalcolm@redhat.com>
11055 * cfg.c (debug): Use TDF_NONE rather than 0.
11056 * cfghooks.c (debug): Likewise.
11057 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
11058 (struct dump_option_value_info): Convert to...
11059 (struct kv_pair): ...this template type.
11060 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
11062 (optinfo_verbosity_options): Likewise.
11063 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
11065 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
11066 than int for "optgroup_flags" param.
11067 (dump_generic_expr_loc): Use dump_flags_t rather than int for
11069 (dump_dec): Likewise.
11070 (dump_finish): Use TDF_NONE rather than 0.
11071 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
11072 rather than int for "optgroup_flags" param. Use TDF_NONE rather
11073 than 0. Update for change to option_ptr.
11074 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
11075 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
11076 0. Update for changes to optinfo_verbosity_options and
11078 (opt_info_switch_p): Convert optgroup_flags from int to
11080 (dump_basic_block): Use dump_flags_t rather than int
11081 for "dump_kind" param.
11082 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
11083 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
11084 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
11085 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
11086 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
11087 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
11088 TDF_NONE): Convert from macros to...
11089 (enum dump_flag): ...this new enum.
11090 (dump_flags_t): Update to use enum.
11091 (operator|, operator&, operator~, operator|=, operator&=):
11092 Implement for dump_flags_t.
11093 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
11094 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
11095 Convert from macros to...
11096 (enum optgroup_flag): ...this new enum.
11097 (optgroup_flags_t): New typedef.
11098 (operator|, operator|=): Implement for optgroup_flags_t.
11099 (struct dump_file_info): Convert field "alt_flags" to
11100 dump_flags_t. Convert field "optgroup_flags" to
11102 (dump_basic_block): Use dump_flags_t rather than int for param.
11103 (dump_generic_expr_loc): Likewise.
11104 (dump_dec): Likewise.
11105 (dump_register): Convert param "optgroup_flags" to
11107 (opt_info_enable_passes): Likewise.
11108 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11110 * gimple-pretty-print.c (debug): Likewise.
11111 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11112 (merged_store_group::apply_stores): Likewise.
11113 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11114 * gimple.c (verify_gimple_pp): Likewise.
11115 * graphite-poly.c (print_pbb_body): Likewise.
11116 * passes.c (pass_manager::register_one_dump_file): Convert
11117 local "optgroup_flags" to optgroup_flags_t.
11118 * print-tree.c (print_node): Use TDF_NONE rather than 0.
11120 (debug_body): Likewise.
11121 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11122 to optgroup_flags_t.
11123 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11125 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11126 (convert_mult_to_fma): Likewise.
11127 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11128 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11129 * tree-vect-data-refs.c (dump_lower_bound): Convert param
11130 "dump_kind" to dump_flags_t.
11132 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11134 * config/rs6000/rs6000.c (min, max): Delete.
11136 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
11138 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11139 -mabi=spe and -mabi=no-spe.
11141 2018-06-08 Martin Liska <mliska@suse.cz>
11143 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11144 where we expect an existing summary.
11146 2018-06-08 Martin Liska <mliska@suse.cz>
11148 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11149 * ipa-inline.h (estimate_edge_growth): Likewise.
11151 2018-06-08 Martin Liska <mliska@suse.cz>
11153 * cgraph.c (function_version_hasher::hash): Use
11154 cgraph_node::get_uid ().
11155 (function_version_hasher::equal):
11156 * cgraph.h (cgraph_node::get_uid): New method.
11157 * ipa-inline.c (update_caller_keys): Use
11158 cgraph_node::get_uid ().
11159 (update_callee_keys): Likewise.
11160 * ipa-utils.c (searchc): Likewise.
11161 (ipa_reduced_postorder): Likewise.
11162 * lto-cgraph.c (input_node): Likewise.
11163 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11164 * symbol-summary.h (symtab_insertion): Likewise.
11165 (symtab_removal): Likewise.
11166 (symtab_duplication): Likewise.
11167 * tree-pretty-print.c (dump_function_header): Likewise.
11168 * tree-sra.c (convert_callers_for_node): Likewise.
11170 2018-06-08 Martin Liska <mliska@suse.cz>
11172 * cgraph.c (symbol_table::create_edge): Always assign a new
11174 (symbol_table::free_edge): Do not recycle numbers.
11175 * cgraph.h (cgraph_edge::get): New method.
11176 * symbol-summary.h (symtab_removal): Use it.
11177 (symtab_duplication): Likewise.
11178 (call_summary::hashable_uid): Remove.
11180 2018-06-08 Martin Liska <mliska@suse.cz>
11182 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11183 (initialize_growth_caches): Remove.
11184 (free_growth_caches): Likewise.
11185 (do_estimate_edge_time): Use edge_growth_cache.
11186 (do_estimate_edge_size): Likewise.
11187 (do_estimate_edge_hints): Likewise.
11188 * ipa-inline.c (reset_edge_caches): Likewise.
11189 (recursive_inlining): Likewise.
11190 (inline_small_functions): Likewise.
11191 * ipa-inline.h (initialize_growth_caches): Remove.
11192 (estimate_edge_size): Likewise.
11193 (estimate_edge_time): Likewise.
11194 (estimate_edge_hints): Likewise.
11195 (reset_edge_growth_cache): Likewise.
11196 * symbol-summary.h (call_summary::remove): New method.
11198 2018-06-08 Martin Liska <mliska@suse.cz>
11200 * ipa-cp.c (class edge_clone_summary): New summary.
11201 (grow_edge_clone_vectors): Remove.
11202 (ipcp_edge_duplication_hook): Remove.
11203 (class edge_clone_summary_t): New call_summary class.
11204 (ipcp_edge_removal_hook): Remove.
11205 (edge_clone_summary_t::duplicate): New function.
11206 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11207 (create_specialized_node): Likewise.
11208 (ipcp_driver): Initialize edge_clone_summaries and do not
11211 2018-06-08 Martin Liska <mliska@suse.cz>
11213 * symbol-summary.h (get): New function.
11214 (call_summary::m_initialize_when_cloning): New class member.
11216 2018-06-08 Martin Liska <mliska@suse.cz>
11218 * cgraph.c (cgraph_node::remove): Do not recycle uid.
11219 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11220 (symbol_table::allocate_cgraph_symbol): Do not set uid.
11221 * passes.c (uid_hash_t): Record removed_nodes by their uids.
11222 (remove_cgraph_node_from_order): Use the removed_nodes set.
11223 (do_per_function_toporder): Likwise.
11224 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11225 instead of summary_uid.
11226 (symtab_removal): Likewise.
11227 (symtab_duplication): Likewise.
11229 2018-06-08 Martin Liska <mliska@suse.cz>
11231 * ipa-cp.c (ipcp_store_bits_results): Use
11232 ipcp_transformation_sum.
11233 (ipcp_store_vr_results): Likewise.
11234 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
11236 (ipcp_transformation_initialize): ... this.
11237 (ipa_set_node_agg_value_chain):
11238 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
11239 (write_ipcp_transformation_info): Likewise.
11240 (read_ipcp_transformation_info): Likewise.
11241 (ipcp_update_bits): Likewise.
11242 (ipcp_update_vr): Likewise.
11243 (ipcp_transform_function): Likewise.
11244 * ipa-prop.h: Rename ipcp_transformation_summary to
11245 ipcp_transformation.
11246 (class ipcp_transformation_t): New function summary.
11247 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
11248 (ipa_get_agg_replacements_for_node): Likewise.
11250 2018-06-08 Martin Liska <mliska@suse.cz>
11252 * ipa-pure-const.c (struct funct_state_d): Do it class instead
11254 (class funct_state_summary_t): New function_summary class.
11255 (has_function_state): Remove.
11256 (get_function_state): Likewise.
11257 (set_function_state): Likewise.
11258 (add_new_function): Likewise.
11259 (funct_state_summary_t::insert): New function.
11260 (duplicate_node_data): Remove.
11261 (remove_node_data): Remove.
11262 (funct_state_summary_t::duplicate): New function.
11263 (register_hooks): Create new funct_state_summaries.
11264 (pure_const_generate_summary): Use it.
11265 (pure_const_write_summary): Likewise.
11266 (pure_const_read_summary): Likewise.
11267 (propagate_pure_const): Likewise.
11268 (propagate_nothrow): Likewise.
11269 (dump_malloc_lattice): Likewise.
11270 (propagate_malloc): Likewise.
11271 (execute): Do not register hooks, just remove summary
11273 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
11276 2018-06-08 Martin Liska <mliska@suse.cz>
11278 * ipa-reference.c (remove_node_data): Remove.
11279 (duplicate_node_data): Likewise.
11280 (class ipa_ref_var_info_summary_t): New class.
11281 (class ipa_ref_opt_summary_t): Likewise.
11282 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
11283 (get_reference_optimization_summary): Use
11284 ipa_ref_opt_sum_summaries.
11285 (set_reference_vars_info): Remove.
11286 (set_reference_optimization_summary): Likewise.
11287 (ipa_init): Create summaries.
11288 (init_function_info): Use function summary.
11289 (ipa_ref_opt_summary_t::duplicate): New function.
11290 (ipa_ref_opt_summary_t::remove): New function.
11291 (get_read_write_all_from_node): Fix GNU coding style.
11292 (propagate): Use function summary.
11293 (write_node_summary_p): Fix GNU coding style.
11294 (stream_out_bitmap): Likewise.
11295 (ipa_reference_read_optimization_summary): Use function summary.
11296 (ipa_reference_c_finalize): Do not release hooks.
11298 2018-06-08 Martin Liska <mliska@suse.cz>
11300 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
11301 (analyze_function_body): Extract multiple calls of get_create.
11302 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11303 * ipa-inline.c (recursive_inlining): Use ::get method.
11304 * ipa-inline.h (estimate_edge_growth): Likewise.
11306 2018-06-08 Martin Liska <mliska@suse.cz>
11308 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
11310 (hsa_function_summary::hsa_function_summary): Use the new enum
11312 (hsa_gpu_implementation_p): Use hsa_summaries::get.
11313 * hsa-gen.c (hsa_get_host_function): Likewise.
11314 (get_brig_function_name): Likewise.
11315 * ipa-hsa.c (process_hsa_functions): Likewise.
11316 (ipa_hsa_write_summary): Likewise.
11317 * symbol-summary.h (symtab_duplication): Use ::get function/
11318 (get): New function.
11320 2018-06-08 Martin Liska <mliska@suse.cz>
11322 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
11324 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
11325 (hsa_register_kernel): Likewise.
11326 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
11327 * hsa-gen.c (hsa_get_host_function): Likewise.
11328 (get_brig_function_name): Likewise.
11329 (generate_hsa): Likewise.
11330 (pass_gen_hsail::execute): Likewise.
11331 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
11332 (devirtualization_time_bonus): Likewise.
11333 (ipcp_propagate_stage): Likewise.
11334 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11335 (edge_set_predicate): Likewise.
11336 (evaluate_conditions_for_known_args): Likewise.
11337 (evaluate_properties_for_edge): Likewise.
11338 (ipa_fn_summary::reset): Likewise.
11339 (ipa_fn_summary_t::duplicate): Likewise.
11340 (dump_ipa_call_summary): Likewise.
11341 (ipa_dump_fn_summary): Likewise.
11342 (analyze_function_body): Likewise.
11343 (compute_fn_summary): Likewise.
11344 (estimate_edge_devirt_benefit): Likewise.
11345 (estimate_edge_size_and_time): Likewise.
11346 (estimate_calls_size_and_time): Likewise.
11347 (estimate_node_size_and_time): Likewise.
11348 (inline_update_callee_summaries): Likewise.
11349 (remap_edge_change_prob): Likewise.
11350 (remap_edge_summaries): Likewise.
11351 (ipa_merge_fn_summary_after_inlining): Likewise.
11352 (ipa_update_overall_fn_summary): Likewise.
11353 (read_ipa_call_summary): Likewise.
11354 (inline_read_section): Likewise.
11355 (write_ipa_call_summary): Likewise.
11356 (ipa_fn_summary_write): Likewise.
11357 (ipa_free_fn_summary): Likewise.
11358 * ipa-hsa.c (process_hsa_functions): Likewise.
11359 (ipa_hsa_write_summary): Likewise.
11360 (ipa_hsa_read_section): Likewise.
11361 * ipa-icf.c (sem_function::merge): Likewise.
11362 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11363 (do_estimate_edge_time): Likewise.
11364 (estimate_size_after_inlining): Likewise.
11365 (estimate_growth): Likewise.
11366 (growth_likely_positive): Likewise.
11367 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11368 (inline_call): Likewise.
11369 * ipa-inline.c (caller_growth_limits): Likewise.
11370 (can_inline_edge_p): Likewise.
11371 (can_inline_edge_by_limits_p): Likewise.
11372 (compute_uninlined_call_time): Likewise.
11373 (compute_inlined_call_time): Likewise.
11374 (want_inline_small_function_p): Likewise.
11375 (edge_badness): Likewise.
11376 (update_caller_keys): Likewise.
11377 (update_callee_keys): Likewise.
11378 (recursive_inlining): Likewise.
11379 (inline_small_functions): Likewise.
11380 (inline_to_all_callers_1): Likewise.
11381 (dump_overall_stats): Likewise.
11382 (early_inline_small_functions): Likewise.
11383 (early_inliner): Likewise.
11384 * ipa-inline.h (estimate_edge_growth): Likewise.
11385 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11386 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11387 * ipa-prop.h (IPA_NODE_REF): Likewise.
11388 (IPA_EDGE_REF): Likewise.
11389 * ipa-pure-const.c (malloc_candidate_p): Likewise.
11390 (propagate_malloc): Likewise.
11391 * ipa-split.c (execute_split_functions): Likewise.
11392 * symbol-summary.h: Rename get to get_create.
11394 (get_create): Likewise.
11395 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11397 2018-06-08 Martin Liska <mliska@suse.cz>
11399 * symbol-summary.h (release): Move definition out of class
11401 (symtab_removal): Likewise.
11402 (symtab_duplication): Likewise.
11404 2018-06-08 Martin Liska <mliska@suse.cz>
11406 * symbol-summary.h (function_summary): Move constructor
11407 implementation out of class declaration.
11408 (release): Likewise.
11409 (symtab_insertion): Likewise.
11410 (symtab_removal): Likewise.
11411 (symtab_duplication): Likewise.
11414 2018-06-08 Martin Liska <mliska@suse.cz>
11416 * Makefile.in: Remove support for MPX (macros, related functions,
11417 fields in cgraph_node, ...).
11418 * builtin-types.def (BT_BND): Likewise.
11419 (BT_FN_BND_CONST_PTR): Likewise.
11420 (BT_FN_CONST_PTR_BND): Likewise.
11421 (BT_FN_VOID_PTR_BND): Likewise.
11422 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
11423 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
11424 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
11425 (expand_builtin_mempcpy_with_bounds): Likewise.
11426 (expand_builtin_memset_with_bounds): Likewise.
11427 (expand_builtin_memset_args): Likewise.
11428 (std_expand_builtin_va_start): Likewise.
11429 (expand_builtin): Likewise.
11430 (expand_builtin_with_bounds): Likewise.
11431 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
11432 (DEF_LIB_BUILTIN_CHKP): Likewise.
11433 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
11434 (DEF_CHKP_BUILTIN): Likewise.
11435 (BUILT_IN_MEMCPY): Likewise.
11436 (BUILT_IN_MEMMOVE): Likewise.
11437 (BUILT_IN_MEMPCPY): Likewise.
11438 (BUILT_IN_MEMSET): Likewise.
11439 (BUILT_IN_STPCPY): Likewise.
11440 (BUILT_IN_STRCAT): Likewise.
11441 (BUILT_IN_STRCHR): Likewise.
11442 (BUILT_IN_STRCPY): Likewise.
11443 (BUILT_IN_STRLEN): Likewise.
11444 (BUILT_IN_MEMCPY_CHK): Likewise.
11445 (BUILT_IN_MEMMOVE_CHK): Likewise.
11446 (BUILT_IN_MEMPCPY_CHK): Likewise.
11447 (BUILT_IN_MEMSET_CHK): Likewise.
11448 (BUILT_IN_STPCPY_CHK): Likewise.
11449 (BUILT_IN_STRCAT_CHK): Likewise.
11450 (BUILT_IN_STRCPY_CHK): Likewise.
11451 * calls.c (store_bounds): Likewise.
11452 (emit_call_1): Likewise.
11453 (special_function_p): Likewise.
11454 (maybe_warn_nonstring_arg): Likewise.
11455 (initialize_argument_information): Likewise.
11456 (finalize_must_preallocate): Likewise.
11457 (compute_argument_addresses): Likewise.
11458 (expand_call): Likewise.
11459 * cfgexpand.c (expand_call_stmt): Likewise.
11460 (expand_return): Likewise.
11461 (expand_gimple_stmt_1): Likewise.
11462 (pass_expand::execute): Likewise.
11463 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
11464 (cgraph_node::remove): Likewise.
11465 (cgraph_node::dump): Likewise.
11466 (cgraph_node::verify_node): Likewise.
11467 * cgraph.h (chkp_function_instrumented_p): Likewise.
11468 (symtab_node::get_alias_target): Likewise.
11469 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
11470 (cgraph_local_p): Likewise.
11471 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
11472 (cgraph_edge::rebuild_references): Likewise.
11473 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
11474 (walk_polymorphic_call_targets): Likewise.
11475 (cgraph_node::expand_thunk): Likewise.
11476 (symbol_table::output_weakrefs): Likewise.
11477 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
11478 (ix86_handle_option): Likewise.
11479 * config/i386/constraints.md: Likewise.
11480 * config/i386/i386-builtin-types.def (BND): Likewise.
11484 * config/i386/i386-builtin.def (BDESC_END): Likewise.
11485 (BDESC_FIRST): Likewise.
11487 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
11488 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
11489 * config/i386/i386.c (enum reg_class): Likewise.
11490 (ix86_target_string): Likewise.
11491 (ix86_option_override_internal): Likewise.
11492 (ix86_conditional_register_usage): Likewise.
11493 (ix86_valid_target_attribute_inner_p): Likewise.
11494 (ix86_set_indirect_branch_type): Likewise.
11495 (ix86_set_current_function): Likewise.
11496 (ix86_function_arg_regno_p): Likewise.
11497 (init_cumulative_args): Likewise.
11498 (ix86_function_arg_advance): Likewise.
11499 (ix86_function_arg): Likewise.
11500 (ix86_pass_by_reference): Likewise.
11501 (ix86_function_value_regno_p): Likewise.
11502 (ix86_function_value_1): Likewise.
11503 (ix86_function_value_bounds): Likewise.
11504 (ix86_return_in_memory): Likewise.
11505 (ix86_setup_incoming_vararg_bounds): Likewise.
11506 (ix86_va_start): Likewise.
11507 (indirect_thunk_need_prefix): Likewise.
11508 (print_reg): Likewise.
11509 (ix86_print_operand): Likewise.
11510 (ix86_expand_call): Likewise.
11511 (ix86_output_function_return): Likewise.
11512 (reg_encoded_number): Likewise.
11513 (BDESC_VERIFYS): Likewise.
11514 (ix86_init_mpx_builtins): Likewise.
11515 (ix86_init_builtins): Likewise.
11516 (ix86_emit_cmove): Likewise.
11517 (ix86_emit_move_max): Likewise.
11518 (ix86_expand_builtin): Likewise.
11519 (ix86_builtin_mpx_function): Likewise.
11520 (ix86_get_arg_address_for_bt): Likewise.
11521 (ix86_load_bounds): Likewise.
11522 (ix86_store_bounds): Likewise.
11523 (ix86_load_returned_bounds): Likewise.
11524 (ix86_store_returned_bounds): Likewise.
11525 (ix86_class_likely_spilled_p): Likewise.
11526 (ix86_hard_regno_mode_ok): Likewise.
11527 (x86_order_regs_for_local_alloc): Likewise.
11528 (ix86_mitigate_rop): Likewise.
11529 (ix86_bnd_prefixed_insn_p): Likewise.
11530 (ix86_mpx_bound_mode): Likewise.
11531 (ix86_make_bounds_constant): Likewise.
11532 (ix86_initialize_bounds): Likewise.
11533 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
11534 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
11535 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
11536 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
11537 (TARGET_CHKP_BOUND_MODE): Likewise.
11538 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
11539 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
11540 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
11541 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
11542 * config/i386/i386.h (TARGET_MPX): Likewise.
11543 (TARGET_MPX_P): Likewise.
11544 (VALID_BND_REG_MODE): Likewise.
11545 (FIRST_BND_REG): Likewise.
11546 (LAST_BND_REG): Likewise.
11547 (enum reg_class): Likewise.
11548 (BND_REG_P): Likewise.
11549 (BND_REGNO_P): Likewise.
11550 (BNDmode): Likewise.
11551 (ADJUST_INSN_LENGTH): Likewise.
11552 * config/i386/i386.md: Likewise.
11553 * config/i386/i386.opt: Likewise.
11554 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
11555 (defined): Likewise.
11556 (LINK_MPX): Likewise.
11557 (MPX_SPEC): Likewise.
11558 (LIBMPX_SPEC): Likewise.
11559 (LIBMPXWRAPPERS_SPEC): Likewise.
11560 (CHKP_SPEC): Likewise.
11561 * config/i386/predicates.md: Likewise.
11562 * dbxout.c (dbxout_type): Likewise.
11563 * doc/extend.texi: Likewise.
11564 * doc/invoke.texi: Likewise.
11565 * doc/md.texi: Likewise.
11566 * doc/tm.texi: Likewise.
11567 * doc/tm.texi.in: Likewise.
11568 * dwarf2out.c (is_base_type): Likewise.
11569 (gen_formal_types_die): Likewise.
11570 (gen_subprogram_die): Likewise.
11571 (gen_type_die_with_usage): Likewise.
11572 (gen_decl_die): Likewise.
11573 (dwarf2out_late_global_decl): Likewise.
11574 * expr.c (expand_assignment): Likewise.
11575 (emit_storent_insn): Likewise.
11576 (store_expr_with_bounds): Likewise.
11577 (store_expr): Likewise.
11578 (expand_expr_real_1): Likewise.
11579 * expr.h (store_expr_with_bounds): Likewise.
11580 * function.c (use_register_for_decl): Likewise.
11581 (struct bounds_parm_data): Likewise.
11582 (assign_parms_augmented_arg_list): Likewise.
11583 (assign_parm_find_entry_rtl): Likewise.
11584 (assign_parm_is_stack_parm): Likewise.
11585 (assign_parm_load_bounds): Likewise.
11586 (assign_bounds): Likewise.
11587 (assign_parms): Likewise.
11588 (expand_function_start): Likewise.
11589 * gcc.c (CHKP_SPEC): Likewise.
11590 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
11591 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
11592 (wrestrict_dom_walker::check_call): Likewise.
11593 * gimple.c (gimple_build_call_from_tree): Likewise.
11594 * gimple.h (enum gf_mask): Likewise.
11595 (gimple_call_with_bounds_p): Likewise.
11596 (gimple_call_set_with_bounds): Likewise.
11597 * gimplify.c (gimplify_init_constructor): Likewise.
11598 * ipa-cp.c (initialize_node_lattices): Likewise.
11599 (propagate_constants_across_call): Likewise.
11600 (find_more_scalar_values_for_callers_subset): Likewise.
11601 * ipa-hsa.c (process_hsa_functions): Likewise.
11602 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
11603 * ipa-icf.c (sem_function::merge): Likewise.
11604 * ipa-inline.c (early_inliner): Likewise.
11605 * ipa-pure-const.c (warn_function_noreturn): Likewise.
11606 (warn_function_cold): Likewise.
11607 (propagate_pure_const): Likewise.
11608 * ipa-ref.h (enum GTY): Likewise.
11609 * ipa-split.c (find_retbnd): Likewise.
11610 (consider_split): Likewise.
11611 (split_function): Likewise.
11612 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11613 * ipa.c (walk_polymorphic_call_targets): Likewise.
11614 (symbol_table::remove_unreachable_nodes): Likewise.
11615 (process_references): Likewise.
11616 (cgraph_build_static_cdtor_1): Likewise.
11617 * lto-cgraph.c (lto_output_node): Likewise.
11618 (output_refs): Likewise.
11619 (compute_ltrans_boundary): Likewise.
11620 (input_overwrite_node): Likewise.
11621 (input_node): Likewise.
11622 (input_cgraph_1): Likewise.
11623 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
11624 * passes.c (pass_manager::execute_early_local_passes): Likewise.
11625 (class pass_chkp_instrumentation_passes): Likewise.
11626 (make_pass_chkp_instrumentation_passes): Likewise.
11627 * passes.def: Likewise.
11628 * rtl.h (struct GTY): Likewise.
11629 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
11630 * stor-layout.c (layout_type): Likewise.
11631 * symtab.c: Likewise.
11632 * target.def: Likewise.
11633 * targhooks.c (default_chkp_bound_type): Likewise.
11634 (default_chkp_bound_mode): Likewise.
11635 (default_builtin_chkp_function): Likewise.
11636 (default_chkp_function_value_bounds): Likewise.
11637 (default_chkp_make_bounds_constant): Likewise.
11638 (default_chkp_initialize_bounds): Likewise.
11639 * targhooks.h (default_chkp_bound_type): Likewise.
11640 (default_chkp_bound_mode): Likewise.
11641 (default_builtin_chkp_function): Likewise.
11642 (default_chkp_function_value_bounds): Likewise.
11643 (default_chkp_make_bounds_constant): Likewise.
11644 (default_chkp_initialize_bounds): Likewise.
11645 * toplev.c (compile_file): Likewise.
11646 (process_options): Likewise.
11647 * tree-core.h (DEF_BUILTIN): Likewise.
11648 (DEF_BUILTIN_CHKP): Likewise.
11649 * tree-inline.c (declare_return_variable): Likewise.
11650 (remap_gimple_stmt): Likewise.
11651 (copy_bb): Likewise.
11652 (initialize_inlined_parameters): Likewise.
11653 (expand_call_inline): Likewise.
11654 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
11655 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
11656 (make_pass_ipa_chkp_produce_thunks): Likewise.
11657 (make_pass_chkp): Likewise.
11658 (make_pass_chkp_opt): Likewise.
11659 (make_pass_chkp_instrumentation_passes): Likewise.
11660 * tree-pretty-print.c (dump_generic_node): Likewise.
11661 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
11662 * tree-ssa-dce.c (propagate_necessity): Likewise.
11663 (eliminate_unnecessary_stmts): Likewise.
11664 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
11665 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
11666 * tree-ssa-sccvn.h: Likewise.
11667 * tree-ssa-strlen.c (get_string_length): Likewise.
11668 (valid_builtin_call): Likewise.
11669 (adjust_last_stmt): Likewise.
11670 (handle_builtin_strchr): Likewise.
11671 (handle_builtin_strcpy): Likewise.
11672 (handle_builtin_stxncpy): Likewise.
11673 (handle_builtin_memcpy): Likewise.
11674 (handle_builtin_strcat): Likewise.
11675 (strlen_check_and_optimize_stmt): Likewise.
11676 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
11677 * tree-streamer-in.c: Likewise.
11678 * tree-streamer.c (record_common_node): Likewise.
11679 * tree.c (tree_code_size): Likewise.
11680 (wide_int_to_tree_1): Likewise.
11681 (type_contains_placeholder_1): Likewise.
11682 (build_common_tree_nodes): Likewise.
11683 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
11684 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
11685 (POINTER_BOUNDS_P): Likewise.
11686 (BOUNDED_TYPE_P): Likewise.
11687 (BOUNDED_P): Likewise.
11688 (CALL_WITH_BOUNDS_P): Likewise.
11689 (pointer_bounds_type_node): Likewise.
11690 * value-prof.c (gimple_ic): Likewise.
11691 * var-tracking.c (vt_add_function_parameters): Likewise.
11692 * varasm.c (make_decl_rtl): Likewise.
11693 (assemble_start_function): Likewise.
11694 (output_constant): Likewise.
11695 (maybe_assemble_visibility): Likewise.
11696 * varpool.c (ctor_for_folding): Likewise.
11697 * chkp-builtins.def: Remove.
11698 * ipa-chkp.c: Remove.
11699 * ipa-chkp.h: Remove.
11700 * rtl-chkp.c: Remove.
11701 * rtl-chkp.h: Remove.
11702 * tree-chkp-opt.c: Remove.
11703 * tree-chkp.c: Remove.
11704 * tree-chkp.h: Remove.
11706 2018-06-07 Carl Love <cel@us.ibm.com>
11708 * config/rs6000/vsx.md (vextract_fp_from_shorth,
11709 vextract_fp_from_shortl): Add BE support.
11711 2018-06-07 Paul Koning <ni1d@arrl.net>
11713 * compare-elim.c (try_merge_compare): Don't merge compare if
11714 address contains a side effect.
11715 (try_eliminate_compare): Likewise.
11717 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
11719 * config.gcc: Support "tremont".
11720 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
11721 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11723 * config/i386/i386.c (m_TREMONT): Define.
11724 (processor_target_table): Add "tremont".
11725 (PTA_TREMONT): Define.
11726 (ix86_lea_outperforms): Add TARGET_TREMONT.
11727 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
11728 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
11729 and M_INTEL_GOLDMONT_PLUS.
11730 (fold_builtin_cpu): Add "tremont".
11731 (ix86_add_stmt_cost): Add TARGET_TREMONT.
11732 (ix86_option_override_internal): Add "tremont".
11733 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
11734 (processor_type): Add PROCESSOR_TREMONT.
11735 * config/i386/x86-tune.def: Add m_TREMONT.
11736 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
11738 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11740 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
11741 symbol defined for msp430i* devices to be lower case.
11743 2018-06-07 Richard Biener <rguenther@suse.de>
11745 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
11746 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
11747 Properly wrap signed arithmetic if overflow wraps.
11749 2018-06-07 Jakub Jelinek <jakub@redhat.com>
11751 PR tree-optimization/69615
11752 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
11753 of a cast from a same precision integral SSA_NAME in a bb dominated
11754 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
11755 cast to utype if rhs2 has already a compatible type.
11757 2018-06-07 Richard Biener <rguenther@suse.de>
11759 PR tree-optimization/85935
11760 * graphite-scop-detection.c (find_params_in_bb): Analyze
11761 condition operands with respect to the correct loop. Assert
11762 the analysis doesn't fail.
11764 2018-06-04 Carl Love <cel@us.ibm.com>
11766 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
11767 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
11768 as it is slightly cheaper.
11769 (first_match_or_eos_index_<mode>):
11770 Calculate index using natural element order.
11771 (first_match_index_<mode>):
11772 Calculate index using natural element order.
11773 (first_match_or_eos_index_<mode>):
11774 Calculate index using natural order.
11775 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
11776 for BE and LE modes.
11777 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
11778 P9V_BUILTIN_VCLZLSBB_V16QI.
11779 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
11782 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11784 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
11785 indentation and line wrap for many prototypes. Add missing
11786 @smallexample directives around block of prototypes for vec_xl and
11789 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
11791 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
11792 track if we pass or return IEEE 128-bit floating point.
11793 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
11794 C++ mangling that is compatible with GCC 8.1.
11795 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
11796 (init_cumulative_args): Note if we pass or return IEEE 128-bit
11797 floating point types.
11798 (rs6000_function_arg_advance_1): Likewise.
11799 (rs6000_mangle_type): Optionally generate mangled names that match
11800 what GCC 8.1 generated for IEEE 128-bit floating point types.
11801 (rs6000_globalize_decl_name): If we have an external function that
11802 passes or returns IEEE 128-bit types, generate a weak reference
11803 from the mangled name used in GCC 8.1 to the current mangled
11805 (rs6000_init_builtins): Make __ibm128 use the long double type if
11806 long double is IBM extended double. Make __float128 use the long
11807 double type if long double is IEEE 128-bit.
11810 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
11811 macro for __ibm128 built-in functions.
11812 (PACK_IF): Add __ibm128 pack/unpack functions.
11813 (UNPACK_IF): Likewise.
11814 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
11815 enable long double built-in functions if long double is IEEE
11816 128-bit floating point.
11817 (rs6000_invalid_builtin): Update long double built-in function
11819 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
11820 functions, adjust the built-in function to use the long double
11821 built-in function if __ibm128 and long double are the same type.
11822 * doc/extend.texi (PowerPC builtins): Update documention for
11823 __builtin_{,un}pack_longdouble. Add documentation for
11824 __builtin_{,un}pack_ibm128.
11826 2018-06-06 Jim Wilson <jimw@sifive.com>
11828 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
11829 (struct machine_function): New field interrupt_mode.
11830 (riscv_handle_type_attribute): New function. Add forward declaration.
11831 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
11832 (riscv_expand_epilogue): Check interrupt_mode field.
11833 (riscv_set_current_function): Check interrupt attribute args and
11834 set interrupt_mode field.
11835 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
11836 (riscv_sret, riscv_uret): New.
11837 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
11838 new arguments to interrupt attribute.
11840 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
11843 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
11844 Don't handle -mcpu=power8 if -mpower9-vector is also used.
11846 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
11848 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
11849 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
11850 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11851 several redundant entries.
11853 2018-06-06 David Malcolm <dmalcolm@redhat.com>
11855 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
11856 type from "rtx" to "rtx_insn *".
11857 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
11858 for local "call_insn", removing cast.
11859 (ix86_expand_call): Likewise, introducing a "call_insn" local.
11861 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
11863 PR tree-optimization/86066
11864 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
11865 for BIT_INSERT_EXPR stores.
11867 2018-06-06 Richard Biener <rguenther@suse.de>
11869 PR tree-optimization/86062
11870 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11871 component refs ontop
11872 of to be offsetted base.
11874 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
11876 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
11877 to be static and remove check on interrupt attribute name.
11879 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
11881 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11882 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
11884 2018-06-05 Steve Ellcey <sellcey@cavium.com>
11887 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
11889 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
11890 Remove second argument, change how error is called.
11891 (aarch64_layout_arg): Remove second argument from
11892 aarch64_err_no_fpadvsimd call.
11893 (aarch64_init_cumulative_args): Ditto.
11894 (aarch64_gimplify_va_arg_expr): Ditto.
11895 * config/aarch64/aarch64.md (mov<mode>): Ditto.
11897 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
11899 * config/i386/i386.md (simple_return_indirect_internal): New expander.
11900 (*simple_return_indirect_internal<mode>): Rename from
11901 simple_return_indirect_internal. Use W mode iterator.
11902 (rstorssp): New expander.
11903 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
11904 (clrssbsy): New expander.
11905 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
11907 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
11909 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
11911 (cmse_check_pointed_object): Likewise.
11913 2018-06-05 Martin Liska <mliska@suse.cz>
11915 PR gcov-profile/47618
11916 * doc/invoke.texi: Document how -fprofile-dir format
11919 2018-06-05 Richard Biener <rguenther@suse.de>
11921 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
11922 removal pretend DOM info isn't available so we do not update
11923 it and only remove edges, not dominated blocks. Actually free
11924 DOM info in case we removed something. Remove unreachable blocks.
11925 (mfb_keep_latches): Work with either DOM info or marked backedges.
11926 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
11927 first. Mark backedges if DOM info isn't available.
11928 (Re-)compute DOM info after cleanup_control_flow_pre.
11930 2018-06-05 Richard Biener <rguenther@suse.de>
11932 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
11933 (locus_discrim_hasher::hash): Adjust.
11934 (locus_discrim_hasher::equal): Likewise.
11935 (next_discriminator_for_locus): Work on line directly.
11936 (same_line_p): Pass in expanded locus1 as well.
11937 (assign_discriminators): Avoid redundant location expansions.
11939 2018-06-05 Richard Biener <rguenther@suse.de>
11941 PR tree-optimization/86046
11942 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
11943 if required after clearing TREE_ADDRESSABLE.
11945 2018-06-05 Richard Biener <rguenther@suse.de>
11947 PR tree-optimization/86047
11948 * tree-ssa-loop.c (for_each_index): Glob handling of all
11949 decls and constants and really handle all of them.
11951 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11954 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
11955 qualifier_void_pointer and qualifier_const_void_pointer.
11956 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
11957 (arm_init_builtins): Handle the above.
11958 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
11959 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
11960 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
11963 2018-06-05 Martin Liska <mliska@suse.cz>
11965 * auto-profile.c (read_autofdo_file): Do not use
11966 gcov_ctr_summary struct.
11967 (afdo_callsite_hot_enough_for_early_inline): Likewise.
11968 * coverage.c (struct counts_entry): Likewise.
11969 (read_counts_file): Read just single summary entry.
11970 (get_coverage_counts): Use gcov_summary struct.
11971 * coverage.h (get_coverage_counts): Likewise.
11972 * gcov-dump.c (dump_working_sets): Likewise.
11973 (tag_summary): Dump just single summary.
11974 * gcov-io.c (gcov_write_summary): Write just histogram
11976 (gcov_read_summary): Read just single summary.
11977 (compute_working_sets): Use gcov_summary struct.
11978 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
11979 of GCOV_COUNTERS_SUMMABLE.
11980 (GCOV_COUNTERS_SUMMABLE): Remove.
11981 (GCOV_FIRST_VALUE_COUNTER): Replace with
11982 GCOV_COUNTER_V_INTERVAL.
11983 (struct gcov_ctr_summary): Remove.
11984 (struct gcov_summary): Directly use fields of former
11986 (compute_working_sets): Use gcov_summary struct.
11987 * gcov.c (read_count_file): Do not use ctrs fields.
11988 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
11990 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
11992 * profile.c: Likewise.
11993 * profile.h: Likewise.
11995 2018-06-05 Martin Liska <mliska@suse.cz>
11997 PR gcov-profile/84846
11998 * gcov.c (output_lines): Print working directory only
11999 in intermediate format.
12001 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
12003 * config/s390/s390-builtin-types.def: Add void function type.
12004 * config/s390/s390-builtins.def: Use the function type for the
12007 2018-06-04 Jim Wilson <jimw@sifive.com>
12009 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
12011 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
12012 and maybe_eh_return. Change regno to unsigned int. Use new args to
12013 handle EH_RETURN_DATA_REGNO registers properly.
12014 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
12015 (riscv_expand_epilogue): Update comment. Change argument name and
12016 type. Update code to use new name and type. Pass new args to
12017 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
12019 * config/riscv/riscv.md (NORMAL_RETURN): New.
12020 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
12021 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
12022 (eh_return): Call gen_eh_return_internal and emit barrier.
12023 (eh_return_internal): Call riscv_expand_epilogue.
12025 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
12027 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
12028 bit_insertion field and declare can_be_merged_into method.
12029 (merged_store_group::can_be_merged_into): New method.
12030 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
12031 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
12032 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
12034 2018-06-04 Richard Biener <rguenther@suse.de>
12036 PR tree-optimization/85955
12037 * builtins.c (fold_builtin_sincos): Convert pointers to
12038 destination to appropriate type before dereferencing.
12040 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
12042 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
12044 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
12046 * expr.c (expand_expr_real_1): Force the operand into memory if
12047 its TYPE_MODE is BLKmode and if there is no integer mode for
12048 the number of bits being extracted.
12050 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12054 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12055 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
12057 2018-06-04 Richard Biener <rguenther@suse.de>
12059 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
12060 (cleanup_tree_cfg_noloop): ... single caller. Do
12061 start_recording_case_labels later.
12063 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
12065 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
12066 to _IMMINTRIN_H_INCLUDED.
12067 * config/i386/pconfigintrin.h: Ditto.
12068 * config/i386/waitpkgintrin.h: Ditto.
12069 * config/i386/immintrin.h: Add includes for sgxintrin.h,
12070 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
12071 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
12072 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
12073 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
12074 waitpkgintrin.h and cldemoteintrin.h.
12076 2018-06-04 Richard Biener <rguenther@suse.de>
12078 PR tree-optimization/86038
12079 * tracer.c (find_best_successor): Check probability for
12080 being initialized, bail out if not.
12082 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
12085 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
12086 of bits to ignore when comparing architectures.
12088 2018-06-04 Jakub Jelinek <jakub@redhat.com>
12090 PR tree-optimization/69615
12091 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
12092 maximum or minimum of the type, try to merge it also as if
12093 range1 is + [-, x - 1] or + [x + 1, -].
12096 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12098 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
12100 PR tree-optimization/86034
12101 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12102 the unsigned bitfield type in a bit insertion sequence if it does not
12103 have a larger precision than the bitfield size.
12104 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12106 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
12108 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12110 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12112 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12113 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12114 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12115 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12117 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
12119 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12120 Disable -fdelete-null-pointer-checks for ELF toolchain.
12122 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12123 Kito Cheng <kito.cheng@gmail.com>
12125 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12126 (nds32le-*-*, nds32be-*-*): Integrate checking process.
12127 (nds32*-*-*): Add glibc and uclibc conditions.
12128 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12129 (TARGET_EXCEPT_UNWIND_INFO): Define.
12130 * config/nds32/elf.h: New file.
12131 * config/nds32/linux.h: New file.
12132 * config/nds32/nds32-elf.opt: New file.
12133 * config/nds32/nds32-linux.opt: New file.
12134 * config/nds32/nds32-fp-as-gp.c
12135 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12136 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12138 (nds32_asm_file_end): Ditto.
12139 (nds32_print_operand): Ditto.
12140 (nds32_insert_attributes): Ditto.
12141 (nds32_init_libfuncs): New function.
12142 (TARGET_HAVE_TLS): Define.
12143 (TARGET_INIT_LIBFUNCS): Define.
12144 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12146 (TARGET_ELF): Apply different mcmodel setting.
12147 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12148 been migrated into elf.h and linux.h files.
12149 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12150 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12151 (mcmodel): The content has been migrated into nds32-elf.opt and
12152 nds32-linux.opt files.
12153 * config/nds32/t-elf: New file.
12154 * config/nds32/t-linux: New file.
12156 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
12157 Shiva Chen <shiva0217@gmail.com>
12159 * config/nds32/constants.md (unspec_volatile_element): Add
12160 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12161 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12163 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12164 (make_pass_nds32_fp_as_gp): Declare.
12165 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12167 (nds32_asm_function_end_prologue): Remove unused asm output.
12168 (nds32_asm_function_begin_epilogue): Remove unused asm output.
12169 (nds32_asm_file_start): Output necessary fp_as_gp information.
12170 (nds32_option_override): Adjust register usage.
12171 (nds32_expand_prologue): Consider fp_as_gp situation.
12172 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12173 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12176 (simple_return): Ditto.
12177 (omit_fp_begin): Output special directive for fp_as_gp.
12178 (omit_fp_end): Output special directive for fp_as_gp.
12179 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12180 mforbid-fp-as-gp): New options.
12182 2018-06-01 Mark Wielaard <mark@klomp.org>
12184 * dwarf2out.c (dwarf2out_finish): Remove generation of
12185 DW_AT_loclists_base.
12187 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
12189 * gimple-ssa-store-merging.c: Include gimple-fold.h.
12190 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12191 (struct merged_store_group): Add bit_insertion field.
12192 (dump_char_array): Use standard hexadecimal format.
12193 (merged_store_group::merged_store_group): Set bit_insertion to false.
12194 (merged_store_group::apply_stores): Use optimal buffer size. Deal
12195 with BIT_INSERT_EXPR stores. Move up code updating the mask and
12196 also print the mask in the dump file.
12197 (pass_store_merging::gate): Minor tweak.
12198 (imm_store_chain_info::coalesce_immediate): Fix wrong association
12199 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
12200 stores with INTEGER_CST stores.
12201 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12202 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12203 and use it throughout. Generate bit insertion sequences if need be.
12204 (pass_store_merging::process_store): Remove redundant condition.
12205 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12207 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
12209 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12210 the 128-bit floating point types. Fix function comment.
12212 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12214 * config/aarch64/aarch64-simd.md
12215 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12217 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12220 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
12222 PR tree-optimization/85989
12223 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12225 (backprop::intersect_uses): Check it when deciding whether this
12226 is a backedge reference.
12227 (backprop::process_block): Add each phi to m_visited_phis
12228 after visiting it, then clear it at the end.
12230 2018-06-01 Richard Biener <rguenther@suse.de>
12232 * tree-vectorizer.h (vect_dr_stmt): New function.
12233 (vect_get_load_cost): Adjust.
12234 (vect_get_store_cost): Likewise.
12235 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
12236 Use vect_dr_stmt instead of DR_SMTT.
12237 (vect_record_base_alignments): Likewise.
12238 (vect_calculate_target_alignment): Likewise.
12239 (vect_compute_data_ref_alignment): Likewise and make static.
12240 (vect_update_misalignment_for_peel): Likewise.
12241 (vect_verify_datarefs_alignment): Likewise.
12242 (vector_alignment_reachable_p): Likewise.
12243 (vect_get_data_access_cost): Likewise. Pass down
12244 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
12245 (vect_get_peeling_costs_all_drs): Likewise.
12246 (vect_peeling_hash_get_lowest_cost): Likewise.
12247 (vect_enhance_data_refs_alignment): Likewise.
12248 (vect_find_same_alignment_drs): Likewise.
12249 (vect_analyze_data_refs_alignment): Likewise.
12250 (vect_analyze_group_access_1): Likewise.
12251 (vect_analyze_group_access): Likewise.
12252 (vect_analyze_data_ref_access): Likewise.
12253 (vect_analyze_data_ref_accesses): Likewise.
12254 (vect_vfa_segment_size): Likewise.
12255 (vect_small_gap_p): Likewise.
12256 (vectorizable_with_step_bound_p): Likewise.
12257 (vect_prune_runtime_alias_test_list): Likewise.
12258 (vect_analyze_data_refs): Likewise.
12259 (vect_supportable_dr_alignment): Likewise.
12260 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12261 (vect_gen_prolog_loop_niters): Likewise.
12262 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12263 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
12265 (vect_recog_mask_conversion_pattern): Likewise.
12266 (vect_try_gather_scatter_pattern): Likewise.
12267 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
12268 to vect_get_store_cost.
12269 (vect_get_store_cost): Get stmt_info instead of DR.
12270 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
12271 (vect_get_load_cost): Get stmt_info instead of DR.
12273 2018-06-01 Richard Biener <rguenther@suse.de>
12275 PR middle-end/86017
12276 * gimple-fold.c (var_decl_component_p): Also allow offsetted
12277 vars wrapped in MEM_REFs.
12279 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
12281 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12282 Fix subreg tests so that we only return a choice between
12283 GENERAL_REGS and FP_REGS if the original classes included both.
12285 2018-06-01 Richard Biener <rguenther@suse.de>
12288 * tree-ssa-structalias.c (get_function_part_constraint):
12289 Handle NULL fi->decl.
12290 (find_func_aliases_for_call): Properly handle indirect
12291 fi from direct call.
12292 (find_func_clobbers): Likewise.
12293 (ipa_pta_execute): Likewise.
12294 (create_variable_info_for): For functions that are ifunc_resolver
12295 resolve to a varinfo that contains the result of the resolver call.
12296 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
12299 2018-05-31 Michael Collison <michael.collison@arm.com>
12301 * config/aarch64/aarch64.md:
12302 (*fix_to_zero_extenddfdi2): New pattern.
12303 * gcc.target/aarch64/fix_extend1.c: New testcase.
12305 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
12307 PR middle-end/78809
12308 PR middle-end/83026
12309 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
12310 and BUILT_IN_STRNCMP_EQ.
12311 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
12312 BUILT_IN_STRNCMP_EQ.
12313 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
12314 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12315 (gimple_fold_builtin): Likewise.
12316 * tree-ssa-strlen.c (compute_string_length): New function.
12317 (determine_min_obsize): New function.
12318 (handle_builtin_string_cmp): New function to handle calls to
12319 string compare functions.
12320 (strlen_optimize_stmt): Add handling to builtin string compare
12322 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
12323 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12324 * tree.c (build_common_builtin_nodes): Add new defines of
12325 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
12327 2018-05-31 Jakub Jelinek <jakub@redhat.com>
12330 * bb-reorder.c (pass_partition_blocks::gate): Return false for
12331 functions with naked attribute.
12333 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
12335 * config/i386/sse.md (avx_vec_concat<mode>):
12336 Substitute concat_tg_mode mode attribute with xtg_mode.
12337 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
12338 (concat_tg_mode): Remove mode attribute.
12340 2018-05-31 Martin Sebor <msebor@redhat.com>
12343 * calls.c (alloc_max_size): Correct a logic error/typo.
12344 Treat excessive arguments as infinite. Warn for invalid arguments.
12345 * doc/invoke.texi (-Walloc-size-larger-than): Update.
12347 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
12350 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
12351 and movx for Haswell.
12353 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
12354 Cesar Philippidis <cesar@codesourcery.com>
12356 PR middle-end/85879
12357 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
12358 when emitting error on private/firstprivate reductions.
12359 * omp-low.c (lower_omp_target): Avoid reference-type processing
12360 on pointers for firstprivate clause.
12362 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12364 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
12367 * config/aarch64/aarch64-simd.md
12368 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
12369 (aarch64_ld1_x3_<mode>): Likewise
12370 (aarch64_st1x2<VALLDIF:mode>): Likewise
12371 (aarch64_st1_x2_<mode>): Likewise
12372 (aarch64_st1x3<VALLDIF:mode>): Likewise
12373 (aarch64_st1_x3_<mode>): Likewise
12374 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
12375 (vld1_s8_x3): Likewise.
12376 (vld1_u16_x3): Likewise.
12377 (vld1_s16_x3): Likewise.
12378 (vld1_u32_x3): Likewise.
12379 (vld1_s32_x3): Likewise.
12380 (vld1_u64_x3): Likewise.
12381 (vld1_s64_x3): Likewise.
12382 (vld1_f16_x3): Likewise.
12383 (vld1_f32_x3): Likewise.
12384 (vld1_f64_x3): Likewise.
12385 (vld1_p8_x3): Likewise.
12386 (vld1_p16_x3): Likewise.
12387 (vld1_p64_x3): Likewise.
12388 (vld1q_u8_x3): Likewise.
12389 (vld1q_s8_x3): Likewise.
12390 (vld1q_u16_x3): Likewise.
12391 (vld1q_s16_x3): Likewise.
12392 (vld1q_u32_x3): Likewise.
12393 (vld1q_s32_x3): Likewise.
12394 (vld1q_u64_x3): Likewise.
12395 (vld1q_s64_x3): Likewise.
12396 (vld1q_f16_x3): Likewise.
12397 (vld1q_f32_x3): Likewise.
12398 (vld1q_f64_x3): Likewise.
12399 (vld1q_p8_x3): Likewise.
12400 (vld1q_p16_x3): Likewise.
12401 (vld1q_p64_x3): Likewise.
12402 (vst1_s64_x2): Likewise.
12403 (vst1_u64_x2): Likewise.
12404 (vst1_f64_x2): Likewise.
12405 (vst1_s8_x2): Likewise.
12406 (vst1_p8_x2): Likewise.
12407 (vst1_s16_x2): Likewise.
12408 (vst1_p16_x2): Likewise.
12409 (vst1_s32_x2): Likewise.
12410 (vst1_u8_x2): Likewise.
12411 (vst1_u16_x2): Likewise.
12412 (vst1_u32_x2): Likewise.
12413 (vst1_f16_x2): Likewise.
12414 (vst1_f32_x2): Likewise.
12415 (vst1_p64_x2): Likewise.
12416 (vst1q_s8_x2): Likewise.
12417 (vst1q_p8_x2): Likewise.
12418 (vst1q_s16_x2): Likewise.
12419 (vst1q_p16_x2): Likewise.
12420 (vst1q_s32_x2): Likewise.
12421 (vst1q_s64_x2): Likewise.
12422 (vst1q_u8_x2): Likewise.
12423 (vst1q_u16_x2): Likewise.
12424 (vst1q_u32_x2): Likewise.
12425 (vst1q_u64_x2): Likewise.
12426 (vst1q_f16_x2): Likewise.
12427 (vst1q_f32_x2): Likewise.
12428 (vst1q_f64_x2): Likewise.
12429 (vst1q_p64_x2): Likewise.
12430 (vst1_s64_x3): Likewise.
12431 (vst1_u64_x3): Likewise.
12432 (vst1_f64_x3): Likewise.
12433 (vst1_s8_x3): Likewise.
12434 (vst1_p8_x3): Likewise.
12435 (vst1_s16_x3): Likewise.
12436 (vst1_p16_x3): Likewise.
12437 (vst1_s32_x3): Likewise.
12438 (vst1_u8_x3): Likewise.
12439 (vst1_u16_x3): Likewise.
12440 (vst1_u32_x3): Likewise.
12441 (vst1_f16_x3): Likewise.
12442 (vst1_f32_x3): Likewise.
12443 (vst1_p64_x3): Likewise.
12444 (vst1q_s8_x3): Likewise.
12445 (vst1q_p8_x3): Likewise.
12446 (vst1q_s16_x3): Likewise.
12447 (vst1q_p16_x3): Likewise.
12448 (vst1q_s32_x3): Likewise.
12449 (vst1q_s64_x3): Likewise.
12450 (vst1q_u8_x3): Likewise.
12451 (vst1q_u16_x3): Likewise.
12452 (vst1q_u32_x3): Likewise.
12453 (vst1q_u64_x3): Likewise.
12454 (vst1q_f16_x3): Likewise.
12455 (vst1q_f32_x3): Likewise.
12456 (vst1q_f64_x3): Likewise.
12457 (vst1q_p64_x3): Likewise.
12459 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
12461 * config/msp430/msp430.c (msp430_output_labelref): Prepend
12462 user_label_prefix to name.
12464 * tree-core.h: Update comment about the format of NAME string
12465 passed to handler in attribute_spec.
12467 * config/msp430/msp430.md: Remove erroneous subreg expression from
12468 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
12469 zero_extend{q,h}isi2.
12471 2018-05-30 Borislav Petkov <bp@suse.de>
12473 * doc/extend.texi: Document some architecture specific
12474 constraints and sort entries.
12476 2018-05-30 Martin Sebor <msebor@redhat.com>
12478 PR middle-end/85369
12479 * builtins.c (expand_builtin_stpcpy_1): New function.
12480 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
12481 only if the former succeeds.
12483 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
12485 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
12488 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12490 * doc/invoke.texi (-flinker-output): Document
12492 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12494 * passes.c (ipa_write_summaries): Only modify statements if body
12496 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
12497 incrementally linking.
12498 (ipa_passes): Likewise.
12499 * lto-cgraph.c (lto_output_node): When incrementally linking do not
12500 pass down resolution info.
12501 * common.opt (flag_incremental_link): Update info.
12502 * gcc.c (plugin specs): Turn flinker-output=* to
12503 -plugin-opt=-linker-output-known
12504 * toplev.c (compile_file): Also cut compilation when doing incremental
12506 * flag-types. (enum lto_partition_model): Add
12507 LTO_LINKER_OUTPUT_NOLTOREL.
12508 (invoke.texi): Add -flinker-output docs.
12509 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
12510 link same way as WPA; do not stream in dead initializers.
12512 * dwarf2out.c (dwarf2out_die_ref_for_decl,
12513 darf2out_register_external_decl): Support incremental link.
12515 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12517 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
12519 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
12521 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
12522 it down to simple_object_copy_lto_debug_sections.
12523 (run_gcc): Determine incremental LTO link time and configure
12524 lto1 into non-wpa mode, disable renaming of debug sections.
12526 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
12528 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12529 descriptions of various incorrectly documented functions.
12531 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
12534 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12535 address check not strict.
12537 2018-05-30 Richard Biener <rguenther@suse.de>
12539 PR tree-optimization/85964
12540 * tracer.c (better_p): Drop initialized count check, we only
12541 call the function with initialized counts now.
12542 (find_best_successor): Do find a best edge if one
12543 has uninitialized count.
12544 (find_best_predecessor): Likewise. Do BB frequency check only
12545 if count is initialized.
12547 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
12549 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
12550 (aarch64_ldrstr_offset_compare): New.
12551 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
12552 load/store orderings.
12553 (aarch64_gen_adjusted_ldpstp): Likewise.
12555 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
12557 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12558 Check for subset of GENERAL_REGS and FP_REGS.
12559 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
12562 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
12564 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
12565 and wi::to_poly_offset. Add the current offset and then check
12566 whether the sum fits, rather than using an unchecked addition of
12567 a checked term. Check for a shwi rather than a uhwi.
12568 * expr.c (get_bit_range): Use tree_to_poly_uint64.
12569 (store_constructor): Use poly_int_tree_p.
12570 (expand_expr_real_1): Likewise.
12571 * function.c (assign_temp): Likewise.
12572 * fold-const.c (const_binop): Use poly_int_tree_p and
12573 wi::to_poly_offset.
12574 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
12576 * ipa-icf-gimple.c (func_checker::compare_operand): Use
12577 to_poly_offset for MEM offsets.
12578 * ipa-icf.c (sem_variable::equals): Likewise.
12579 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
12580 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
12581 wi::to_poly_offset for BIT_FIELD_REF offsets.
12582 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
12583 wi::to_poly_offset.
12584 * var-tracking.c (emit_note_insn_var_location): Use
12585 tree_to_poly_uint64.
12587 2018-05-29 Jim Wilson <jimw@sifive.com>
12589 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
12591 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
12594 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
12595 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
12597 (sse4_1_round<mode>2): Use nonimmediate_operand
12598 for operand 1 predicate.
12600 2018-05-29 Martin Sebor <msebor@redhat.com>
12601 Richard Biener <rguenther@suse.de>
12604 * calls.c (get_size_range): Call determine_value_range instead
12605 of get_value_range..
12606 * tree-vrp.h (determine_value_range): Declared new function.
12607 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
12609 2018-05-29 Richard Biener <rguenther@suse.de>
12611 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
12612 sure to use non-pattern stmts for get_earlier_stmt arguments.
12613 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
12614 called on pattern stmts.
12615 (get_later_stmt): Likewise.
12617 2018-05-29 Martin Liska <mliska@suse.cz>
12619 PR gcov-profile/85759
12620 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
12623 2018-05-29 Jakub Jelinek <jakub@redhat.com>
12625 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
12627 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
12631 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12632 VEC_PACK_FLOAT_EXPR): New tree codes.
12633 * tree-pretty-print.c (op_code_prio): Handle
12634 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12635 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12636 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12637 * tree-inline.c (estimate_operator_cost): Likewise.
12638 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
12639 * fold-const.c (const_binop): Likewise.
12640 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
12641 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12642 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12643 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
12644 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12645 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12646 * expr.c (expand_expr_real_2): Likewise.
12647 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
12648 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
12649 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
12651 * optabs.c (expand_widen_pattern_expr): For
12652 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
12653 sign from result type rather than operand's type.
12654 (expand_binop_directly): For vec_packu_float_optab and
12655 vec_packs_float_optab allow result type to be different from operand's
12657 * optabs-tree.c (optab_for_tree_code): Handle
12658 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12659 VEC_PACK_FLOAT_EXPR. Formatting fixes.
12660 * tree-vect-generic.c (expand_vector_operations_1): Handle
12661 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12662 VEC_PACK_FLOAT_EXPR.
12663 * tree-vect-stmts.c (supportable_widening_operation): Handle
12665 (supportable_narrowing_operation): Handle FLOAT_EXPR.
12666 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
12667 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
12668 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
12669 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
12671 (vec_pack<floatprefix>_float_<mode>): New expander.
12672 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
12674 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
12675 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
12676 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
12677 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
12678 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
12680 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
12681 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
12682 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12683 VEC_PACK_FLOAT_EXPR): Document.
12685 2018-05-29 Richard Biener <rguenther@suse.de>
12687 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
12689 (stmt_vec_info_vec): Make pointer.
12690 (init_stmt_vec_info_vec): Remove.
12691 (free_stmt_vec_info_vec): Likewise.
12692 (set_stmt_vec_info_vec): New function.
12693 (free_stmt_vec_infos): Likewise.
12694 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
12695 (set_vinfo_for_stmt): Likewise.
12696 (get_earlier_stmt): Likewise.
12697 (get_later_stmt): Likewise.
12698 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
12699 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
12700 (vec_info::~vec_info): Free stmt_vec_infos.
12701 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
12702 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
12703 (pass_slp_vectorize::execute): Likewise.
12704 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
12705 (free_stmt_vec_info_vec): Likewise.
12706 (set_stmt_vec_info_vec): New function.
12707 (free_stmt_vec_infos): Likewise.
12708 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
12709 the global stmt_vec_info_vec.
12710 * tree-parloops.c (gather_scalar_reductions): Use
12711 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
12714 2018-05-29 Richard Biener <rguenther@suse.de>
12716 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
12718 2018-05-29 Martin Liska <mliska@suse.cz>
12719 David Malcolm <dmalcolm@redhat.com>
12721 * vec.c (test_reverse): New.
12722 (vec_c_tests): Add new test.
12723 * vec.h (vl_ptr>::reverse): New function.
12725 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
12727 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
12729 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
12732 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
12734 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
12736 2018-05-28 Richard Biener <rguenther@suse.de>
12738 PR tree-optimization/85933
12739 * tree-vect-data-refs.c (vect_record_base_alignments): Only
12740 look at stmts marked as vectorizable.
12742 2018-05-28 Richard Biener <rguenther@suse.de>
12744 PR tree-optimization/85934
12745 * tree-vect-generic.c (expand_vector_operations_1): Hoist
12746 vector boolean check before scalar optimization.
12748 2018-05-28 Jakub Jelinek <jakub@redhat.com>
12750 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
12753 2018-05-28 Mark Wielaard <mark@klomp.org>
12755 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
12756 if it is an expression containing a minus sign.
12758 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
12760 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
12762 2018-05-27 Paul Koning <ni1d@arrl.net>
12764 * config/pdp11/pdp11.md (truncsihi2): Remove.
12766 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
12767 Chung-Ju Wu <jasonwucj@gmail.com>
12769 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
12771 (unaligned_store_dw): Ditto.
12772 * config/nds32/nds32-memory-manipulation.c
12773 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
12774 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
12775 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
12776 (emit_setmem_word_loop): Rename to ...
12777 (emit_setmem_doubleword_loop): ... this.
12778 (nds32_gen_dup_4_byte_to_word_value): New function.
12779 (nds32_gen_dup_8_byte_to_double_word_value): New function.
12780 (nds32_expand_setmem_loop): Refine implementation.
12781 (nds32_expand_setmem_loop_v3m): Ditto.
12782 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
12785 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12787 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
12789 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
12791 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
12792 (nds32_init_machine_status): Initialize machine->attr_naked_p and
12793 machine->attr_no_prologue_p.
12794 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
12795 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
12796 (nds32_expand_epilogue): Consider attr_naked_p.
12797 (nds32_expand_epilogue_v3pop): Likewise.
12798 (nds32_can_use_return_insn): Likewise.
12799 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
12800 attr_no_prologue_p fields.
12801 * config/nds32/nds32.opt (mret-in-naked-func): New option.
12803 2018-05-27 Jakub Jelinek <jakub@redhat.com>
12806 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
12808 * config/i386/sse.md
12809 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
12811 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
12813 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
12815 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
12817 (*<floatsuffix>floatv2div2sf2): Rename to ...
12818 (*float<floatunssuffix>v2div2sf2): ... this.
12819 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
12820 (float<floatunssuffix>v2div2sf2_mask): ... this.
12821 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
12822 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
12823 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
12825 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
12827 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12829 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12831 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12833 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12835 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
12836 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
12837 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
12838 gen_ufix_truncv8dfv8si2.
12839 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
12840 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
12841 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
12842 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
12843 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
12844 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
12845 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
12846 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
12848 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
12852 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
12854 2018-05-25 Jim Wilson <jimw@sifive.com>
12856 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
12857 * config/riscv/riscv.c (struct machine_function): Add
12858 interrupt_handler_p and attribute_checked_p fields.
12859 (riscv_attribute_table): Add interrupt.
12860 (riscv_interrupt_type_p): New.
12861 (riscv_save_reg_p): Save extra regs for interrupt handler.
12862 (riscv_use_save_libcall): Return false for interrupt handler.
12863 (riscv_first_stack_step): Add forward declaration.
12864 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
12865 for interrupt handler with large frame. Use it for saved reg list.
12866 (riscv_expand_prologue): Move flag_stack_usage_info support to
12867 eliminate duplication.
12868 (riscv_expand_epilogue): Generate mret for interrupt handler.
12869 (riscv_epilogue_uses): New.
12870 (riscv_can_use_return_insn): Return false for interrupt handler.
12871 (riscv_function_ok_for_sibcall): Likewise.
12872 (riscv_set_current_function): Add interrupt handler support.
12873 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
12874 * config/riscv/riscv.md (UNSPECV_MRET): New.
12876 (riscv_frflags, riscv_fsflags): Use tab after opcode.
12878 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
12880 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
12882 PR tree-optimization/85712
12883 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
12884 this candidate has already been replaced in-situ by a copy.
12886 2018-05-25 Jason Merrill <jason@redhat.com>
12888 PR c++/80485 - inline function non-zero address.
12889 * symtab.c (nonzero_address): Check DECL_COMDAT.
12891 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
12894 * config/alpha/alpha.md (ashlsi3): New insn pattern.
12895 (*ashlsi_se): Rename from *ashldi_se. Define as sign
12896 extension of SImode operation. Use const123_operand predicate.
12897 (*saddsi_1): Remove.
12898 (*saddl_se_1): Ditto.
12899 (*ssubsi_1): Ditto.
12900 (*ssubl_se_1): Ditto.
12901 * config/alpha/predicates.md (const123_operand): New predicate.
12902 * config/alpha/constraints.md (P): Use IN_RANGE.
12904 2018-05-25 Richard Biener <rguenther@suse.de>
12906 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
12908 (ref_maybe_used_by_stmt_p): Likewise.
12909 (stmt_may_clobber_ref_p): Likewise.
12910 (stmt_may_clobber_ref_p_1): Likewise.
12911 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
12913 (ref_maybe_used_by_stmt_p): Likewise.
12914 (stmt_may_clobber_ref_p): Likewise.
12915 (stmt_may_clobber_ref_p_1): Likewise.
12916 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
12917 the alias oracle to disambiguate DRs with stmts DR analysis
12919 (vect_analyze_data_refs): Do not give up on not analyzable
12920 DRs for BB vectorization. Remove code truncating the dataref
12923 2018-05-25 Jakub Jelinek <jakub@redhat.com>
12926 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12927 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
12928 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
12930 2018-05-25 Richard Biener <rguenther@suse.de>
12932 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
12933 function, combining stmt data ref gathering and fatal analysis
12935 (vect_analyze_data_refs): Remove now redudnant code and simplify.
12936 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
12937 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
12938 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
12939 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
12941 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12943 PR tree-optimization/85720
12944 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
12945 SCC if all partitions are builtins.
12946 (version_loop_by_alias_check): New parameter. Generate cancelable
12947 runtime alias check if all partitions are builtins.
12948 (distribute_loop): Update call to above function.
12950 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12952 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
12953 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
12954 (parm_default_def_partition_arg): Ditto.
12955 (set_parm_default_def_partition): Ditto.
12956 (get_parm_default_def_partitions): Ditto and make it static.
12957 (get_undefined_value_partitions): Ditto and make it static.
12958 (remove_ssa_form): Refactor call to init_var_map here.
12959 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
12960 computation for loop region.
12961 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
12962 (register_default_def): Delete.
12963 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
12964 (parm_default_def_partition_arg): Ditto.
12965 (set_parm_default_def_partition): Ditto.
12966 (get_parm_default_def_partitions): Ditto and make it static.
12967 (get_undefined_value_partitions): Ditto and make it static.
12968 (coalesce_with_default, coalesce_with_default): Update comment.
12969 (create_coalesce_list_for_region): New func factored out from
12970 create_outofssa_var_map.
12971 (populate_coalesce_list_for_outofssa): New func factored out from
12972 create_outofssa_var_map and coalesce_ssa_name.
12973 (create_outofssa_var_map): Delete.
12974 (coalesce_ssa_name): Refactor to support live range computation.
12975 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
12976 (get_parm_default_def_partitions): Delete.
12977 (get_undefined_value_partitions): Ditto.
12978 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
12979 computation for loop region.
12980 (new_tree_live_info, loe_visit_block): Ditto.
12981 (live_worklist, set_var_live_on_entry): Ditto.
12982 (calculate_live_on_exit, verify_live_on_entry): Ditto.
12983 * tree-ssa-live.h (struct _var_map): New fields.
12984 (init_var_map): Change decl.
12985 (region_contains_p): New.
12987 2018-05-25 Bin Cheng <bin.cheng@arm.com>
12989 * tree-ssa-live.h (live_merge_and_clear): Delete.
12991 2018-05-25 Richard Biener <rguenther@suse.de>
12994 * tree-dump.c (dequeue_and_dump): Remove access to removed
12995 operand 2 of a SWITCH_EXPR.
12997 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
12999 * doc/sourcebuild.texi (vect_double_cond_arith): Include
13000 multiplication and division.
13001 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
13002 (cond_udiv@var{m}, cond_umod@var{m}): Document.
13003 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
13004 (cond_udiv_optab, cond_umod_optab): New optabs.
13005 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
13006 (IFN_COND_RDIV): New internal functions.
13007 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
13008 TRUNC_MOD_EXPR and RDIV_EXPR.
13009 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
13010 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
13012 (SVE_INT_BINARY): Include mult.
13013 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
13014 (optab, sve_int_op): Handle mult.
13015 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
13017 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
13018 for SVE_INT_BINARY_SD.
13020 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13022 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
13023 (optab, sve_int_op): Handle div and udiv.
13024 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
13025 for SVE_INT_BINARY_SD.
13026 (*<optab><mode>3): New insn for the same.
13028 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13030 * tree-vect-patterns.c: Include predict.h.
13031 (vect_recog_divmod_pattern): Restrict check for division support
13032 to when optimizing for size.
13034 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13036 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
13037 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
13038 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
13039 (gimple_match_op::set_op): Likewise.
13040 (gimple_resimplify4): Declare.
13041 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
13042 (expr::gen_transform): Likewise.
13043 (decision_tree::gen): Generate a simplification routine for 4 operands.
13044 * gimple-match-head.c (gimple_simplify): Add an overload for
13045 4 operands. In the top-level function, handle up to 4 call
13046 arguments and call gimple_resimplify4.
13047 (gimple_resimplify4): New function.
13048 (build_call_internal): Pass a fourth operand.
13049 (maybe_push_to_seq): Likewise.
13050 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
13051 Fold VEC_COND_EXPRs of an operation and a default value into
13052 an IFN_COND_* function if possible.
13053 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
13055 (SVE_COND_FP_BINARY): Include them.
13056 (optab, sve_fp_op): Handle them.
13057 (SVE_INT_BINARY_REV): New code iterator.
13058 (SVE_COND_FP_BINARY_REV): New int iterator.
13059 (commutative): New int attribute.
13060 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
13062 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
13064 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
13065 (*cond_<optab><mode>): New patterns for reversed operands.
13067 2018-05-25 Richard Biener <rguenther@suse.de>
13069 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
13070 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
13071 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
13072 (STMT_VINFO_GROUPED_ACCESS): Adjust.
13073 * tree-vect-data-refs.c (everywhere): Adjust users.
13074 * tree-vect-loop.c (everywhere): Likewise.
13075 * tree-vect-slp.c (everywhere): Likewise.
13076 * tree-vect-stmts.c (everywhere): Likewise.
13077 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
13079 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13081 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
13083 (gcc_cv_as_section_exclude): ... this.
13084 Try Solaris as #exclude syntax.
13085 * configure: Regenerate.
13086 * config.in: Regenerate.
13087 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
13089 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
13090 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
13092 * varasm.c (default_elf_asm_named_section): Don't check if
13093 HAVE_GAS_SECTION_EXCLUDE is defined.
13095 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13097 * doc/md.texi: Update the documentation of the cond_* optabs
13098 to mention the new final operand. Fix GET_MODE_NUNITS call.
13099 Describe the scalar case too.
13100 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13101 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13103 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13104 (get_conditional_internal_fn): Update comment.
13105 * tree-vect-loop.c (vectorizable_reduction): Pass the original
13106 accumulator value as a final argument to conditional functions.
13107 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13108 a define_expand and add an "else" operand. Assert for now that
13109 the else operand is equal to operand 2. Use SVE_INT_BINARY and
13110 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13111 (*cond_<optab><mode>): New patterns.
13112 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13113 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13114 (UNSPEC_COND_EOR): Delete.
13115 (optab): Remove associated mappings.
13116 (SVE_INT_BINARY): New code iterator.
13117 (sve_int_op): Remove int attribute and add "minus" to the code
13119 (SVE_COND_INT_OP): Delete.
13120 (SVE_COND_FP_OP): Rename to...
13121 (SVE_COND_FP_BINARY): ...this.
13123 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
13125 * optabs.c (can_reuse_operands_p): New function.
13126 (maybe_legitimize_operands): Try to reuse the results for
13129 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13131 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13132 Add {q} suffix to insn mnemonic.
13134 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
13136 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13137 (msp430_warn_func_return): New.
13139 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
13141 * fold-const.c (tree_nonzero_bits): New function.
13142 * fold-const.h (tree_nonzero_bits): Likewise.
13143 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13144 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13146 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
13150 * varasm.c (assemble_alias): Check ifunc_resolver only on
13153 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
13156 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13157 when memory input operand is handled.
13159 2018-05-24 Luis Machado <luis.machado@linaro.org>
13161 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13163 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13165 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13167 * match.pd: Delay FMA folds until after vectorization.
13169 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
13172 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13173 address check not strict.
13175 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
13177 * gimple-match.h (gimple_match_op): New class.
13178 (mprts_hook): Replace parameters with a gimple_match_op *.
13179 (maybe_build_generic_op): Likewise.
13180 (gimple_simplified_result_is_gimple_val): Replace parameters with
13181 a const gimple_match_op *.
13182 (gimple_simplify): Replace code_helper * and tree * parameters with
13183 a gimple_match_op * parameter.
13184 (gimple_resimplify1): Replace code_helper *, tree and tree *
13185 parameters with a gimple_match_op * parameter.
13186 (gimple_resimplify2): Likewise.
13187 (gimple_resimplify3): Likewise.
13188 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13189 parameters with a gimple_match_op * parameter.
13190 * gimple-match-head.c (gimple_simplify): Change prototypes of
13191 auto-generated functions to take a gimple_match_op * instead of
13192 separate code_helper * and tree * parameters. Make the same
13193 change in the top-level overload and update calls to the
13194 gimple_resimplify routines. Update calls to the auto-generated
13195 functions and to maybe_push_res_to_seq in the publicly-facing
13196 operation-specific gimple_simplify overloads.
13197 (gimple_match_op::MAX_NUM_OPS): Define.
13198 (gimple_resimplify1): Replace rcode and ops with a single res_op
13199 parameter. Update call to gimple_simplify.
13200 (gimple_resimplify2): Likewise.
13201 (gimple_resimplify3): Likewise.
13202 (mprts_hook): Replace parameters with a gimple_match_op *.
13203 (maybe_build_generic_op): Likewise.
13204 (build_call_internal): Replace type, nargs and ops with
13205 a gimple_match_op *.
13206 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13207 with a single gimple_match_op *. Update calls to mprts_hook,
13208 build_call_internal and gimple_simplified_result_is_gimple_val.
13209 Factor out code that is common to the tree_code and combined_fn cases.
13210 * genmatch.c (expr::gen_transform): Replace tem_code and
13211 tem_ops with a gimple_match_op called tem_op. Update calls
13212 to the gimple_resimplify functions and maybe_push_res_to_seq.
13213 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13214 res_ops. Update call to the gimple_resimplify functions.
13215 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13216 (decision_tree::gen): Make the functions take a gimple_match_op *
13217 called res_op instead of separate res_code and res_ops parameters.
13218 Update call accordingly.
13219 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13220 and ops with a single res_op parameter. Update calls to
13221 maybe_build_generic_op and maybe_push_res_to_seq.
13222 (fold_stmt_1): Update calls to gimple_simplify and
13223 replace_stmt_with_simplification.
13224 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13225 and gimple_simplified_result_is_gimple_val.
13226 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13228 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13229 with a gimple_match_op *.
13230 (vn_nary_build_or_lookup): Likewise. Update call to
13231 vn_nary_build_or_lookup_1.
13232 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13233 gimple_match_op *. Update calls to the gimple_resimplify routines
13234 and to gimple_simplified_result_is_gimple_val.
13235 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
13236 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
13237 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
13238 (visit_nary_op): Likewise.
13239 (visit_reference_op_load): Likewise.
13241 2018-05-23 Luis Machado <luis.machado@linaro.org>
13243 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
13244 modifier for printing the step amount.
13246 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
13249 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
13252 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
13254 * doc/sourcebuild.texi (Endianness): New subsubsection.
13256 2018-05-23 Luis Machado <luis.machado@linaro.org>
13258 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13259 <prefetch_dynamic_strides>: New const bool field.
13260 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13261 prefetch_dynamic_strides.
13262 (exynosm1_prefetch_tune): Likewise.
13263 (thunderxt88_prefetch_tune): Likewise.
13264 (thunderx_prefetch_tune): Likewise.
13265 (thunderx2t99_prefetch_tune): Likewise.
13266 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
13268 (aarch64_override_options_internal): Update to set
13269 PARAM_PREFETCH_DYNAMIC_STRIDES.
13270 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13271 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13272 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13273 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13274 prefetch-dynamic-strides setting.
13276 2018-05-23 Luis Machado <luis.machado@linaro.org>
13278 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13279 <minimum_stride>: New const int field.
13280 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13281 minimum_stride field defaulting to -1.
13282 (exynosm1_prefetch_tune): Likewise.
13283 (thunderxt88_prefetch_tune): Likewise.
13284 (thunderx_prefetch_tune): Likewise.
13285 (thunderx2t99_prefetch_tune): Likewise.
13286 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
13287 <default_opt_level>: Set to 3.
13288 (aarch64_override_options_internal): Update to set
13289 PARAM_PREFETCH_MINIMUM_STRIDE.
13290 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13291 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13292 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13293 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13294 stride is constant and is below the minimum stride threshold.
13296 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13298 * config/arm/arm-cpus.in (mode26): Delete.
13299 (armv4): Delete mode26 reference.
13300 * config/arm/arm.c (arm_configure_build_target): Delete use of
13303 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
13305 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
13307 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
13308 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
13310 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
13311 (floatunsdidf2): Ditto.
13313 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
13315 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
13316 (fixuns_trunc<mode>si2_avx512f): Ditto.
13317 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
13318 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
13319 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
13321 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
13323 PR rtl-optimization/79985
13324 * df-scan.c (df_insn_refs_collect): Remove special case for
13325 global registers and asm statements.
13327 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
13329 * extend.texi (Global Register Variables): Rewrite the bullet list.
13330 Note that the register is available for allocation. Note that access
13331 via inline asm must use constraints. Add note about async-signal
13332 handlers. Remove paragraph about automagic register selection.
13334 2018-05-23 Richard Biener <rguenther@suse.de>
13336 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
13337 of fixed offset from memset VN.
13339 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
13341 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
13342 first_interp field.
13343 (alloc_cand_and_find_basis): Initialize first_interp field.
13344 (slsr_process_mul): Modify first_interp field.
13345 (slsr_process_add): Likewise.
13346 (slsr_process_cast): Modify first_interp field for each new
13348 (slsr_process_copy): Likewise.
13349 (dump_candidate): Dump first_interp field.
13350 (replace_mult_candidate): Process all interpretations, not just
13352 (replace_rhs_if_not_dup): Likewise.
13353 (replace_one_candidate): Likewise.
13355 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
13357 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
13359 (aarch64_needs_frame_chain): New function.
13360 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
13362 2018-05-23 Sudakshina Das <sudi.das@arm.com>
13365 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
13366 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
13367 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
13368 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
13369 as true for strict-align.
13370 (aarch64_can_inline_p): Perform checks even when callee has no
13371 attributes to check for strict alignment.
13372 * doc/extend.texi (AArch64 Function Attributes): Document
13374 * doc/invoke.texi: (AArch64 Options): Likewise.
13376 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
13378 PR tree-optimization/85853
13379 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
13380 the handling of the root of the node to...
13381 (vect_slp_analyze_node_operations_1): ...this new function,
13382 and run the whole thing with the child nodes' def types
13383 set according to their SLP node's def type.
13385 2018-05-23 Richard Biener <rguenther@suse.de>
13387 PR middle-end/85874
13388 * tree-data-ref.c (create_runtime_alias_checks): Defer
13389 and ignore overflow warnings.
13391 2018-05-23 Yury Gribov <tetra2005@gmail.com>
13393 PR tree-optimization/85822
13394 * tree-vrp.c (is_masked_range_test): Fix handling of negative
13397 2018-05-23 Richard Biener <rguenther@suse.de>
13399 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13400 memset constants via native_interpret_expr.
13402 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
13405 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
13407 (cgraph_node::create_alias): Likewise.
13408 (cgraph_node::get_availability): Check ifunc_resolver instead
13409 of looking up ifunc attribute.
13410 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
13411 * varasm.c (do_assemble_alias): Likewise.
13412 (assemble_alias): Likewise.
13413 (default_binds_local_p_3): Likewise.
13414 * cgraph.h (cgraph_node): Add ifunc_resolver.
13415 (cgraph_node::only_called_directly_or_aliased_p): Return false
13416 for IFUNC resolver.
13417 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
13419 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
13420 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
13421 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
13422 instead of looking up ifunc attribute.
13424 2018-05-22 Luis Machado <luis.machado@linaro.org>
13426 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
13428 2018-05-22 Martin Sebor <msebor@redhat.com>
13430 PR middle-end/85359
13431 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
13432 only when expasion succeeds.
13433 (expand_builtin_strcmp): Same.
13434 (expand_builtin_strncmp): Same.
13436 2018-05-22 Martin Sebor <msebor@redhat.com>
13438 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
13440 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
13441 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13443 * config/aarch64/aarch64-ldpstp.md: Replace uses of
13444 aarch64_mem_pair_operand with memory_operand and delete operand swapping
13446 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13447 Add check for legitimate_address.
13448 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
13449 (aarch64_swap_ldrstr_operands): New.
13450 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
13453 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
13454 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13456 * config/aarch64/aarch64.md: New patterns to generate stp
13458 (store_pair_sw, store_pair_dw): New patterns to generate stp for
13459 single words and double words.
13460 (load_pair_sw, load_pair_dw): Likewise.
13461 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
13463 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
13465 * config/aarch64/aarch64-ldpstp.md: Modify peephole
13466 for different mode ldpstp and add peephole for merged zero stores.
13467 Likewise for loads.
13468 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13470 (aarch64_gen_store_pair): Rename calls to match new patterns.
13471 (aarch64_gen_load_pair): Rename calls to match new patterns.
13472 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
13473 (load_pair<DREG:mode><DREG2:mode>): ... This.
13474 (store_pair<mode>): Rename to...
13475 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
13476 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
13477 New mode iterators.
13478 (V_INT_EQUIV): Handle SImode.
13479 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
13482 2018-05-22 Martin Sebor <msebor@redhat.com>
13485 * calls.c (maybe_warn_nonstring_arg): Use string length to set
13486 or ajust the presumed bound on an operation to avoid unnecessary
13489 2018-05-22 Martin Sebor <msebor@redhat.com>
13491 PR tree-optimization/85826
13492 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
13493 assuming that a DECL necesarily has a constant size.
13495 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
13497 PR middle-end/85862
13498 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
13500 2018-05-22 Richard Biener <rguenther@suse.de>
13502 PR tree-optimization/85834
13503 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
13504 non-constant and non-zero memset arguments.
13506 2018-05-22 Martin Liska <mliska@suse.cz>
13509 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
13511 2018-05-22 Richard Biener <rguenther@suse.de>
13513 PR tree-optimization/85863
13514 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
13515 comparisons when vectype is specified.
13516 (vectorizable_condition): Do not specify vectype for
13517 vect_is_simple_cond when SLP vectorizing.
13519 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
13522 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
13523 define __ibm128 as long double.
13524 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
13525 as a distinct type when IEEE 128-bit support is enabled.
13526 (init_float128_ieee): Fix up conversions between IFmode and IEEE
13527 128-bit types to use the correct functions.
13528 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
13529 convert between 128-bit floating point types that have different
13530 modes but the same representation, instead of using gen_lowpart to
13532 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
13534 (IFKF_reg): New attributes to give the register constraints for
13536 (extend<mode>tf2_internal): New insns to mark an explicit
13537 conversion between 128-bit floating point types that have a
13538 different mode but share the same representation.
13540 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
13542 PR tree-optimization/85814
13543 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
13544 a null return from get_strinfo when unsharing the next
13545 strinfo in the chain.
13547 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
13550 * varasm.c (weak_finish): Clean up weak_decls.
13552 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13554 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
13555 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
13556 UNSPEC_UADALP values.
13557 * config/aarch64/iterators.md (ABAL): New int iterator.
13560 (sur): Add mappings for the above.
13561 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
13563 (aarch64_<sur>abal<mode>_4): Likewise.
13564 (aarch64_<sur>adalp<mode>_3): Likewise.
13565 (<sur>sadv16qi): New define_expand.
13567 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
13569 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
13570 (*movdf_internal): Ditto.
13571 (*rcpsf2_sse): Ditto.
13572 (*rsqrtsf2_sse): Ditto.
13573 (*sqrt<mode>2_sse): Ditto.
13575 2018-05-21 Tamar Christina <tamar.christina@arm.com>
13577 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
13579 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
13580 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
13581 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
13582 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
13584 * config/aarch64/arm_neon.h: Likewise.
13585 * config/aarch64/iterators.md (VQ_I): New.
13587 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
13589 * config.gcc: Add arc/t-multilib-linux to tmake_file for
13591 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
13594 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13596 * config/nds32/constraints.md (S): New constraint.
13597 * config/nds32/nds32.md (call_internal): Use constraint S.
13598 (call_value_internal): Likewise.
13599 (sibcall_internal): Likewise.
13600 (sibcall_value_internal): Likewise.
13602 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
13603 Chung-Ju Wu <jasonwucj@gmail.com>
13605 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
13606 into consideration.
13608 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
13609 Chung-Ju Wu <jasonwucj@gmail.com>
13611 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
13612 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
13613 (nds32_rtx_costs_impl): Simplify.
13614 (nds32_address_cost_impl): Simplify.
13615 (nds32_init_rtx_costs): New function.
13616 (nds32_rtx_costs_speed_prefer): Likewise.
13617 (nds32_rtx_costs_size_prefer): Likewise.
13618 (nds32_address_cost_speed_prefer): Likewise.
13619 (nds32_address_cost_speed_fwprop): Likewise.
13620 (nds32_address_cost_size_prefer): Likewise.
13621 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
13622 * config/nds32/nds32.c (nds32_option_override): Use
13623 nds32_init_rtx_costs function.
13625 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
13627 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
13628 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
13629 (TARGET_PIPELINE_N8): Likewise.
13630 (TARGET_PIPELINE_N10): Likewise.
13631 (TARGET_PIPELINE_N13): Likewise.
13632 (TARGET_PIPELINE_GRAYWOLF): Likewise.
13634 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
13636 * config/nds32/nds32-fpu.md: Update copyright year.
13638 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13640 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
13642 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13644 * config/nds32/nds32.c
13645 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
13646 * config/nds32/nds32.opt (minline-asm-r15): New option.
13648 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13650 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
13652 * config/nds32/nds32.md (abssi2): New pattern.
13654 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
13656 * config/i386/i386.md (rex64namesuffix): New mode attribute.
13657 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
13658 Merge insn pattern from sse_cvtsi2ss<round_name> and
13659 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
13660 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
13661 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
13662 using SWI48 mode iterator.
13663 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
13664 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
13665 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
13666 pattern from sse_cvttss2si<round_saeonly_name>
13667 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
13668 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
13669 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
13670 using SWI48 mode iterator.
13671 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
13672 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
13673 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
13674 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
13675 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
13676 using SWI48 mode iterator.
13677 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
13678 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
13679 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
13680 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
13681 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
13682 SWI48 mode iterator.
13683 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
13684 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
13685 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
13686 pattern from sse_cvttsd2si<round_saeonly_name>
13687 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
13689 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13691 * config/nds32/nds32-md-auxiliary.c
13692 (nds32_valid_smw_lwm_base_p): Refine.
13693 (nds32_output_smw_single_word): Refine.
13694 (nds32_output_smw_double_word): New.
13695 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
13697 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13699 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
13700 (nds32_output_stack_pop): Refine.
13701 (nds32_expand_unaligned_load): Refine.
13702 (nds32_expand_unaligned_store): Refine.
13704 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13705 Chung-Ju Wu <jasonwucj@gmail.com>
13707 * config/nds32/constants.md: Add TP_REGNUM constant.
13708 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
13709 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
13711 * config/nds32/nds32-doubleword.md: Consider flag_pic.
13712 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
13713 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
13714 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
13715 and PIC code generation.
13716 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
13718 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
13720 * config/nds32/nds32.md: Support TLS and PIC.
13721 * config/nds32/nds32.c: Support TLS and PIC.
13722 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
13723 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
13726 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
13728 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
13729 mode with E_ prefix.
13731 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
13732 Chung-Ju Wu <jasonwucj@gmail.com>
13734 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
13735 * config/nds32/nds32-md-auxiliary.c
13736 (symbolic_reference_mentioned_p): New.
13737 (nds32_legitimize_ict_address): New.
13738 (nds32_expand_ict_move): New.
13739 (nds32_indirect_call_referenced_p): New.
13740 (nds32_symbol_binds_local_p): Delete.
13741 (nds32_long_call_p): Modify.
13742 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
13743 * config/nds32/nds32-protos.h
13744 (symbolic_reference_mentioned_p): Declare.
13745 (nds32_legitimize_ict_address): Declare.
13746 (nds32_expand_ict_move): Declare.
13747 (nds32_indirect_call_referenced_p): Declare.
13748 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
13749 (nds32_relax_group): Use nds32_ict_const_p as condition.
13750 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
13751 (nds32_asm_file_start): Output ict_model directive in asm code.
13752 (nds32_legitimate_address_p): Consider indirect call.
13753 (nds32_print_operand): Consider indirect call.
13754 (nds32_print_operand_address): Consider indirect call.
13755 (nds32_insert_attributes): Handle "indirect_call" attribute.
13756 (TARGET_LEGITIMATE_ADDRESS_P): Define.
13757 (TARGET_LEGITIMATE_CONSTANT_P): Define.
13758 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
13759 (TARGET_DELEGITIMIZE_ADDRESS): Define.
13760 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
13761 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
13762 (TARGET_ICT_MODEL_SMALL): Define.
13763 (TARGET_ICT_MODEL_LARGE): Define.
13764 * config/nds32/nds32.md (movsi): Consider ict model.
13765 (call, call_value): Consider ict model.
13766 (sibcall, sibcall_value): Consider ict model.
13767 * config/nds32/nds32.opt (mict-model): New option.
13768 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
13771 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
13772 Monk Chiang <sh.chiang04@gmail.com>
13773 Jim Wilson <jimw@sifive.com>
13775 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
13776 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
13777 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
13778 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
13779 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
13780 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
13781 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
13782 compute save_libcall_adjustment properly.
13783 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
13784 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
13785 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
13786 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
13787 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
13788 (ABI_SPEC): Handle mabi=ilp32e.
13789 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
13790 (RVE): Add RVE mask.
13791 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
13792 <-march>: Add rv32e as an example.
13794 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
13797 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
13798 (intra_create_variable_infos): Handle C++ constructors.
13800 2018-05-18 Martin Liska <mliska@suse.cz>
13802 * passes.def: Remove a redundant pass.
13804 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
13807 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
13809 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13811 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
13813 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
13815 (armv2, armv2a, armv3, armv3m): Delete architectures.
13816 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
13817 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
13818 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
13820 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
13821 (*mulsidi3adddi): Likewise.
13822 (mulsidi3): Likewise.
13823 (*mulsidi3_nov6): Likewise.
13824 (umulsidi3): Likewise.
13825 (umulsidi3_nov6): Likewise.
13826 (umaddsidi4): Likewise.
13827 (*umulsidi3adddi): Likewise.
13828 (smulsi3_highpart): Likewise.
13829 (*smulsi3_highpart_nov6): Likewise.
13830 (umulsi3_highpart): Likewise.
13831 (*umulsi3_highpart_nov6): Likewise.
13832 * config/arm/arm.h (arm_arch3m): Delete.
13833 * config/arm/arm.c (arm_arch3m): Delete.
13834 (arm_option_override_internal): Update armv3-related comment.
13835 (arm_configure_build_target): Delete use of isa_bit_mode32.
13836 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
13837 (arm_rtx_costs_internal): Delete check of arm_arch3m.
13838 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
13839 (mulsa3): Likewise.
13840 (mulusa3): Likewise.
13841 * config/arm/arm-protos.h (arm_arch3m): Delete.
13842 * config/arm/arm-tables.opt: Regenerate.
13843 * config/arm/arm-tune.md: Likewise.
13844 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
13845 deleted architectures.
13847 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13849 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
13850 (armv5t, armv5te): New features.
13851 (ARMv5, ARMv5e): Delete fgroups.
13852 (ARMv5t, ARMv5te): Adjust for above changes.
13853 (ARMv6m): Likewise.
13854 (armv5, armv5e): Delete arches.
13855 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
13857 (*call_reg_arm): Likewise.
13858 (*call_value_reg_armv5): Likewise.
13859 (*call_value_reg_arm): Likewise.
13860 (*call_symbol): Likewise.
13861 (*call_value_symbol): Likewise.
13862 (*sibcall_insn): Likewise.
13863 (*sibcall_value_insn): Likewise.
13864 (clzsi2): Likewise.
13865 (prefetch): Likewise.
13866 (define_split and define_peephole2 dependent on arm_arch5):
13868 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
13870 (TARGET_ARM_QBIT): Likewise.
13871 (TARGET_DSP_MULTIPLY): Likewise.
13872 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
13873 (arm_arch5, arm_arch5e): Delete.
13874 (arm_arch5t, arm_arch5te): Declare.
13875 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
13876 (arm_arch5t): Declare.
13877 (arm_option_reconfigure_globals): Update for the above.
13878 (arm_options_perform_arch_sanity_checks): Update comment, replace
13879 use of arm_arch5 with arm_arch5t.
13880 (use_return_insn): Likewise.
13881 (arm_emit_call_insn): Likewise.
13882 (output_return_instruction): Likewise.
13883 (arm_final_prescan_insn): Likewise.
13884 (arm_coproc_builtin_available): Likewise.
13885 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
13886 arm_arch5e with arm_arch5t and arm_arch5te.
13887 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
13888 (arm_arch5t, arm_arch5te): Declare.
13889 * config/arm/arm-tables.opt: Regenerate.
13890 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
13891 * config/arm/t-multilib: Likewise.
13892 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
13893 instead of arm_arch5.
13894 (*call_reg_thumb1): Likewise.
13895 (*call_value_reg_thumb1_v5): Likewise.
13896 (*call_value_reg_thumb1): Likewise.
13897 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
13899 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
13901 2018-05-18 Martin Liska <mliska@suse.cz>
13903 PR gcov-profile/84846
13904 * doc/gcov.texi: Document -t option of gcov tool.
13906 2018-05-18 Martin Liska <mliska@suse.cz>
13908 PR gcov-profile/84846
13909 * gcov.c (print_usage): Add new -t option.
13910 (process_args): Handle the option.
13911 (generate_results): Use stdout as output when requested by
13914 2018-05-18 Martin Liska <mliska@suse.cz>
13916 PR gcov-profile/84846
13917 * coverage.c (coverage_init): Write PWD to .gcno file.
13918 * doc/gcov.texi: Document how working directory is printed.
13919 * gcov-dump.c (dump_gcov_file): Print PWD.
13920 * gcov.c (output_intermediate_file): Likewise.
13921 (read_graph_file): Read PWD string.
13922 (output_lines): Print PWD.
13924 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
13926 PR middle-end/85817
13927 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
13928 for retval and return false if all args to phi are zero.
13930 2018-05-18 Richard Biener <rguenther@suse.de>
13932 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
13934 (evrp_dom_walker::before_dom_children): Call it.
13936 2018-05-18 Richard Biener <rguenther@suse.de>
13938 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
13939 results when processing array refs with variable index.
13941 2018-05-18 Toon Moene <toon@moene.org>
13943 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
13944 directly after that of -floop-interchange. Indicate that both
13945 options are enabled by default when specifying -O3.
13947 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13949 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
13950 iterator. Delete separate integer-mode vec_set<mode> expander.
13951 (aarch64_simd_vec_setv2di): Delete.
13952 (vec_setv2di): Delete.
13953 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
13954 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
13955 the "w, r" alternative.
13957 2018-05-18 Martin Liska <mliska@suse.cz>
13959 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
13960 * tree-pass.h (make_pass_lower_switch_O0): New function.
13961 * tree-switch-conversion.c (node_has_low_bound): Remove.
13962 (node_has_high_bound): Likewise.
13963 (node_is_bounded): Likewise.
13964 (class pass_lower_switch): Make it a template type and create
13966 (pass_lower_switch::execute): Add template argument.
13967 (make_pass_lower_switch): New function.
13968 (make_pass_lower_switch_O0): New function.
13969 (do_jump_if_equal): Remove.
13970 (emit_case_nodes): Simplify to just handle all 3 cases and leave
13971 all the hard work to tree optimization passes.
13973 2018-05-18 Martin Liska <mliska@suse.cz>
13975 * dbgcnt.c (limit_low): Renamed from limit.
13976 (limit_high): New variable.
13977 (dbg_cnt_is_enabled): Check for upper limit.
13978 (dbg_cnt): Adjust dumping.
13979 (dbg_cnt_set_limit_by_index): Add new argument for high
13981 (dbg_cnt_set_limit_by_name): Likewise.
13982 (dbg_cnt_process_single_pair): Parse new format.
13983 (dbg_cnt_process_opt): Use strtok.
13984 (dbg_cnt_list_all_counters): Remove 'value' and add
13986 * doc/invoke.texi: Document changes.
13988 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
13990 * doc/sourcebuild.texi (scalar_all_fma): Document.
13991 * tree.def (FMA_EXPR): Delete.
13992 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
13993 * internal-fn.c (ternary_direct): New macro.
13994 (expand_ternary_optab_fn): Likewise.
13995 (direct_ternary_optab_supported_p): Likewise.
13996 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
13997 * builtins.c (fold_builtin_fma): Delete.
13998 (fold_builtin_3): Don't call it.
13999 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
14000 * expr.c (expand_expr_real_2): Likewise.
14001 * fold-const.c (operand_equal_p): Likewise.
14002 (fold_ternary_loc): Likewise.
14003 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
14004 * gimple.c (DEFTREECODE): Likewise.
14005 * gimplify.c (gimplify_expr): Likewise.
14006 * optabs-tree.c (optab_for_tree_code): Likewise.
14007 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
14008 * tree-eh.c (operation_could_trap_p): Likewise.
14009 (stmt_could_throw_1_p): Likewise.
14010 * tree-inline.c (estimate_operator_cost): Likewise.
14011 * tree-pretty-print.c (dump_generic_node): Likewise.
14012 (op_code_prio): Likewise.
14013 * tree-ssa-loop-im.c (stmt_cost): Likewise.
14014 * tree-ssa-operands.c (get_expr_operands): Likewise.
14015 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
14016 * fold-const-call.h (fold_fma): Delete.
14017 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
14018 CFN_FNMA and CFN_FNMS.
14019 (fold_fma): Delete.
14020 * genmatch.c (combined_fn): New enum.
14021 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
14022 (commutative_op): New function.
14023 (commutate): Use it. Handle more than 2 operands.
14024 (dt_operand::gen_gimple_expr): Use commutative_op.
14025 (parser::parse_expr): Allow :c to be used with non-binary
14026 operators if the commutative operand is known.
14027 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
14028 CFN_FMS, CFN_FNMA and CFN_FNMS.
14029 (backprop::process_assign_use): Remove FMA_EXPR handling.
14030 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
14031 (gen_hsa_fma): New function.
14032 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
14033 IFN_FNMA and IFN_FNMS.
14034 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
14035 * gimple-fold.h (follow_all_ssa_edges): Declare.
14036 * gimple-fold.c (follow_all_ssa_edges): New function.
14037 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
14038 gimple_build interface and use follow_all_ssa_edges to fold the result.
14039 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
14040 instead of checking for optabs directly.
14041 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
14042 rather than FMA_EXPRs.
14043 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
14044 call to IFN_FMA instead of an FMA_EXPR.
14046 2018-05-17 Jim Wilson <jimw@sifive.com>
14048 * expr.c (do_tablejump): When converting index to Pmode, if we have a
14049 sign extended promoted subreg, and the range does not have the sign bit
14050 set, then do a sign extend.
14052 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
14053 test, check for sign extended subreg and/or constant operands, and
14054 do a sign extend in that case.
14056 2018-05-17 Steve Ellcey <sellcey@cavium.com>
14058 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
14059 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
14061 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
14062 Change logics_shift_reg to logics_shift_imm.
14063 (thunderx2t99_fp_loadpair_basic): Delete.
14064 (thunderx2t99_fp_storepair_basic): Delete.
14065 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
14066 (thunderx2t99_asimd_polynomial): Delete.
14067 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
14068 and neon_fp_mul_d_scalar_q.
14069 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
14070 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
14071 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
14072 (thunderx2t99_asimd_lut): Add missing tbl types.
14073 (thunderx2t99_asimd_ext): Delete.
14074 (thunderx2t99_asimd_load1_1_mult): Delete.
14075 (thunderx2t99_asimd_load1_2_mult): Delete.
14076 (thunderx2t99_asimd_load1_ldp): New.
14077 (thunderx2t99_asimd_load1): New.
14078 (thunderx2t99_asimd_load2): Add missing *load2* types.
14079 (thunderx2t99_asimd_load3): New.
14080 (thunderx2t99_asimd_load4): New.
14081 (thunderx2t99_asimd_store1_1_mult): Delete.
14082 (thunderx2t99_asimd_store1_2_mult): Delete.
14083 (thunderx2t99_asimd_store2_mult): Delete.
14084 (thunderx2t99_asimd_store2_onelane): Delete.
14085 (thunderx2t99_asimd_store_stp): New.
14086 (thunderx2t99_asimd_store1): New.
14087 (thunderx2t99_asimd_store2): New.
14088 (thunderx2t99_asimd_store3): New.
14089 (thunderx2t99_asimd_store4): New.
14091 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
14093 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
14094 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
14096 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
14097 Segher Boessenkool <segher@kernel.crashing.org>
14100 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14103 2018-05-17 Richard Biener <rguenther@suse.de>
14105 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14106 for pruning loop and prune defs feeding only already visited PHIs.
14108 2018-05-17 Richard Biener <rguenther@suse.de>
14110 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14112 2018-05-17 Bin Cheng <bin.cheng@arm.com>
14113 Richard Biener <rguenther@suse.de>
14115 PR tree-optimization/85793
14116 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14117 for VMAT_ELEMENTWISE.
14119 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14121 * internal-fn.h (lookup_internal_fn): Declare
14122 * internal-fn.c (lookup_internal_fn): New function.
14123 * gimple.c (gimple_build_call_from_tree): Handle calls to
14124 internal functions.
14125 * gimple-pretty-print.c (dump_gimple_call): Print "." before
14126 internal function names.
14127 * tree-pretty-print.c (dump_generic_node): Likewise.
14128 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14130 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
14132 * gimple-fold.h (gimple_build): Make the function forms take
14133 combined_fn rather than built_in_function.
14134 (gimple_simplify): Likewise.
14135 * gimple-match-head.c (gimple_simplify): Likewise.
14136 * gimple-fold.c (gimple_build): Likewise.
14137 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14138 rather than gimple_build_call_internal.
14139 (get_initial_defs_for_reduction): Likewise.
14140 (vect_create_epilog_for_reduction): Likewise.
14141 (vectorizable_live_operation): Likewise.
14143 2018-05-17 Martin Liska <mliska@suse.cz>
14145 * gimple-ssa-sprintf.c (format_directive): Do not use
14146 space in between 'G_' and '('.
14148 2018-05-17 Jakub Jelinek <jakub@redhat.com>
14151 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14152 even if the mask is not all ones.
14155 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14157 (ix86_gimple_fold_builtin): Likewise.
14160 * config/i386/i386.c: Include tree-vector-builder.h.
14161 (ix86_vector_shift_count): New function.
14162 (ix86_fold_builtin): Fold shift builtins by scalar count.
14163 (ix86_gimple_fold_builtin): Likewise.
14165 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14166 _mm512_setzero): New intrinsics.
14168 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
14169 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14171 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14172 code generation for cases where splatting a value is not useful.
14173 * simplify-rtx.c (simplify_ternary_operation): Simplify
14174 vec_merge across a vec_duplicate and a paradoxical subreg forming
14175 a vector mode to a vec_concat.
14177 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
14179 * config.gcc: Support "goldmont-plus".
14180 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14182 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14183 PROCESSOR_GOLDMONT_PLUS.
14184 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14185 (processor_target_table): Add "goldmont-plus".
14186 (PTA_GOLDMONT_PLUS): Define.
14187 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14188 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14189 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14190 (fold_builtin_cpu): Add "goldmont-plus".
14191 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14192 (ix86_option_override_internal): Add "goldmont-plus".
14193 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14194 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14195 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14196 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14198 2018-05-17 Richard Biener <rguenther@suse.de>
14200 PR tree-optimization/85757
14201 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14202 remove defs that only feed that PHI from further processing.
14204 2018-05-16 Jim Wilson <jimw@sifive.com>
14206 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14208 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14210 2018-05-16 Mark Wielaard <mark@klomp.org>
14212 * dwarf2out.c (count_index_strings): New function.
14213 (output_indirect_strings): Call count_index_strings and generate
14214 header for dwarf_version >= 5.
14216 2018-05-16 Mark Wielaard <mark@klomp.org>
14218 * dwarf2out.c (dwarf_FORM): New function.
14219 (set_indirect_string): Use dwarf_FORM.
14220 (reset_indirect_string): Likewise.
14221 (size_of_die): Likewise.
14222 (value_format): Likewise.
14223 (output_die): Likewise.
14224 (add_skeleton_AT_string): Likewise.
14225 (output_macinfo_op): Likewise.
14226 (index_string): Likewise.
14227 (output_index_string_offset): Likewise.
14228 (output_index_string): Likewise.
14229 (count_index_strings): Likewise.
14231 2018-05-16 Carl Love <cel@us.ibm.com>
14233 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14234 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
14236 2018-05-16 Martin Jambor <mjambor@suse.cz>
14238 * ipa-prop.c (ipa_free_all_edge_args): Remove.
14239 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
14241 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
14243 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
14244 (fnma<mode>4): Likewise.
14245 (fms<mode>4): Likewise.
14246 (fnms<mode>4): Likewise.
14247 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
14248 (aarch64_fnma<mode>4): Likewise.
14249 (aarch64_fms<mode>4): Likewise.
14250 (aarch64_fnms<mode>4): Likewise.
14251 (aarch64_fnmadd<mode>4): Likewise.
14253 2018-05-16 Jason Merrill <jason@redhat.com>
14255 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
14257 2018-05-16 Richard Biener <rguenther@suse.de>
14259 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
14260 (dump_stmt_cost): Declare.
14261 (add_stmt_cost): Dump cost we add.
14262 (add_stmt_costs): New function.
14263 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
14264 No longer exported.
14265 (vect_analyze_stmt): Adjust prototype.
14266 (vectorizable_condition): Likewise.
14267 (vectorizable_live_operation): Likewise.
14268 (vectorizable_reduction): Likewise.
14269 (vectorizable_induction): Likewise.
14270 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
14271 cost vector to pass to vectorizable_ and record afterwards.
14272 (vect_model_reduction_cost): Take cost vector argument and adjust.
14273 (vect_model_induction_cost): Likewise.
14274 (vectorizable_reduction): Likewise.
14275 (vectorizable_induction): Likewise.
14276 (vectorizable_live_operation): Likewise.
14277 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
14278 SLP_TREE_NUMBER_OF_VEC_STMTS.
14279 (vect_analyze_slp_cost_1): Remove.
14280 (vect_analyze_slp_cost): Likewise.
14281 (vect_slp_analyze_node_operations): Take visited args and
14282 a target cost vector. Avoid processing already visited stmt sets.
14283 (vect_slp_analyze_operations): Use a local cost vector to gather
14284 costs and register those of non-discarded instances.
14285 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
14286 (vect_schedule_slp_instance): Remove copying of
14287 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
14289 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
14290 adding cost. Record cost entry location.
14291 (vect_prologue_cost_for_slp_op): Function to compute cost of
14292 a constant or invariant generated for SLP vect in the prologue,
14293 split out from vect_analyze_slp_cost_1.
14294 (vect_model_simple_cost): Make static. Adjust for SLP costing.
14295 (vect_model_promotion_demotion_cost): Likewise.
14296 (vect_model_store_cost): Likewise, make static.
14297 (vect_model_load_cost): Likewise.
14298 (vectorizable_bswap): Add cost vector arg and adjust.
14299 (vectorizable_call): Likewise.
14300 (vectorizable_simd_clone_call): Likewise.
14301 (vectorizable_conversion): Likewise.
14302 (vectorizable_assignment): Likewise.
14303 (vectorizable_shift): Likewise.
14304 (vectorizable_operation): Likewise.
14305 (vectorizable_store): Likewise.
14306 (vectorizable_load): Likewise.
14307 (vectorizable_condition): Likewise.
14308 (vectorizable_comparison): Likewise.
14309 (can_vectorize_live_stmts): Likewise.
14310 (vect_analyze_stmt): Likewise.
14311 (vect_transform_stmt): Adjust calls to vectorizable_*.
14312 * tree-vectorizer.c: Include gimple-pretty-print.h.
14313 (dump_stmt_cost): New function.
14315 2018-05-16 Richard Biener <rguenther@suse.de>
14317 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
14318 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
14319 * tree-ssa-dse.c: Include tree-ssa-loop.h.
14320 (check_name): New callback.
14321 (dse_classify_store): Track cycles via a visited bitmap of PHI
14322 defs and simplify handling of in-loop and across loop dead stores
14323 and properly fail for loop-variant refs. Handle byte-tracking with
14324 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
14327 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
14329 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
14330 (vect_get_mask_type_for_stmt): Likewise.
14331 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
14333 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
14334 to determine the statement's vector type and the vector type that
14335 should be used for calculating nunits. Deal with cases in which
14336 the type has to be deferred.
14337 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
14338 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
14339 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
14340 (vect_determine_vf_for_stmt): New functions, split out from...
14341 (vect_determine_vectorization_factor): ...here.
14342 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
14343 (vect_get_mask_type_for_stmt): New functions, split out from
14344 vect_determine_vectorization_factor.
14346 2018-05-16 Richard Biener <rguenther@suse.de>
14348 * tree-cfg.c (verify_gimple_assign_ternary): Properly
14349 verify the [VEC_]COND_EXPR embedded comparison.
14351 2018-05-15 Martin Sebor <msebor@redhat.com>
14353 PR tree-optimization/85753
14354 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
14355 RECORD_TYPE in addition to ARRAY_TYPE.
14357 2018-05-15 Martin Sebor <msebor@redhat.com>
14359 PR middle-end/85643
14360 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
14362 2018-05-15 Richard Biener <rguenther@suse.de>
14364 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
14365 add by_clobber_p one. Change algorithm to collect all defs
14366 representing uses we need to walk and try reducing them to
14367 a single one before failing.
14368 (dse_dom_walker::dse_optimize_stmt): Adjust.
14370 2018-05-13 Mark Wielaard <mark@klomp.org>
14372 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
14373 (size_of_loc_descr): Likewise.
14374 (output_loc_operands): Likewise.
14375 (output_loc_operands_raw): Likewise.
14376 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
14377 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
14378 (hash_loc_operands): Likewise.
14379 (compare_loc_operands): Likewise.
14381 2018-05-14 Mark Wielaard <mark@klomp.org>
14383 * dwarf2out.c (count_index_addrs): New function.
14384 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
14386 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14388 PR tree-optimization/83648
14389 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
14390 return value as malloc candidate.
14392 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14395 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
14396 param as true in call to suggest_attribute.
14398 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
14400 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
14401 -mreadonly-in-sdata.
14403 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14405 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
14407 (aarch64_crypto_aesd_fused): Likewise.
14409 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
14411 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
14412 (movsi_aarch64): Likewise.
14413 (load_pairsi): Likewise.
14414 (load_pairdi): Likewise.
14415 (store_pairsi): Likewise.
14416 (store_pairdi): Likewise.
14417 (load_pairsf): Likewise.
14418 (load_pairdf): Likewise.
14419 (store_pairsf): Likewise.
14420 (store_pairdf): Likewise.
14421 (zero_extend): Likewise.
14422 (trunc): Swap alternatives.
14423 (fcvt_target): Add '?' to prefer w over r.
14425 2018-05-14 Jakub Jelinek <jakub@redhat.com>
14428 * config/i386/i386.md: Disallow non-commutative arithmetics in
14429 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
14430 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
14431 in the peephole2 before it.
14433 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
14435 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
14436 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
14437 (ix86_handle_option): Handle -mcldemote.
14438 * config.gcc: New header.
14439 * config/i386/cldemoteintrin.h: New file.
14440 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
14441 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14443 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14444 OPTION_MASK_ISA_CLDEMOTE.
14445 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
14446 (ix86_valid_target_attribute_inner_p): Ditto.
14447 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
14448 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
14449 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
14450 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
14451 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
14453 * config/i386/i386.opt: Add -mcldemote.
14454 * config/i386/x86intrin.h: New header.
14455 * doc/invoke.texi: Add -mcldemote.
14457 2018-05-14 Richard Biener <rguenther@suse.de>
14459 * doc/match-and-simplify.texi: Adjust :s documentation.
14461 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
14463 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
14464 intended memcpy size.
14465 (REORDER_45): Likewise.
14467 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
14469 * sort.cc: New file.
14470 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
14471 * vec.c (qsort_chk): Use gcc_qsort.
14472 * Makefile.in (OBJS-libcommon): Add sort.o.
14473 (build/sort.o): New target. Use it...
14474 (BUILD_RTL): ... here, and...
14475 (build/gencfn-macros): ... here, and...
14476 (build/genmatch): ... here.
14478 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14479 Chung-Ju Wu <jasonwucj@gmail.com>
14481 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
14482 * config/nds32/nds32-graywolf.md: New file.
14483 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
14484 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
14486 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
14487 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
14488 * config/nds32/nds32.md (pipeline_model): Add graywolf.
14489 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
14490 * config/nds32/pipelines.md: Include n15 settings.
14492 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14493 Chung-Ju Wu <jasonwucj@gmail.com>
14495 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
14496 * config/nds32/nds32-n13.md: New file.
14497 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
14498 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
14500 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
14501 * config/nds32/nds32.md (pipeline_model): Add n13.
14502 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
14503 * config/nds32/pipelines.md: Include n13 settings.
14505 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
14506 Chung-Ju Wu <jasonwucj@gmail.com>
14508 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
14509 * config/nds32/nds32-n10.md: New file.
14510 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
14511 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
14513 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
14514 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
14515 * config/nds32/nds32.md (pipeline_model): Add n10.
14516 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
14517 * config/nds32/pipelines.md: Include n10 settings.
14519 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
14520 Kito Cheng <kito.cheng@gmail.com>
14521 Chung-Ju Wu <jasonwucj@gmail.com>
14523 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
14524 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14525 Add enum values for DSP extension instructions.
14526 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
14528 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
14529 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
14530 New code iterators.
14531 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
14532 * config/nds32/nds32-dspext.md: New file for DSP implementation.
14533 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
14534 * config/nds32/nds32-intrinsic.md: Likewise.
14535 * config/nds32/nds32_intrinsic.h: Likewise.
14536 * config/nds32/nds32-md-auxiliary.c: Likewise.
14537 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
14538 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
14539 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
14540 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
14541 * config/nds32/nds32-protos.h: New declarations for DSP extension.
14542 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
14543 TYPE_DMAC in switch statement.
14544 * config/nds32/nds32.c: New checking and implementation for DSP
14545 extension instructions.
14546 * config/nds32/nds32.h: Likewise.
14547 * config/nds32/nds32.md: Likewise.
14548 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
14549 * config/nds32/predicates.md: Implement new predicates for DSP
14552 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14554 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
14555 Reformat alternatives and attributes so it is easier to identify
14556 which constraints/attributes go with which instruction.
14557 (mov<mode>_hardfloat32, FMOVE64): Likewise.
14558 (mov<mode>_softfloat32, FMOVE64): Likewise.
14559 (mov<mode>_hardfloat64, FMOVE64): Likewise.
14560 (mov<mode>_softfloat64, FMOVE64): Likewise.
14562 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
14564 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14566 (Basic PowerPC Built-in Functions): The new name of the
14567 subsection previously known as "PowerPC Built-in Functions".
14568 (Basic PowerPC Built-in Functions Available on all Configurations):
14570 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
14571 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
14572 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
14573 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
14575 2018-05-11 Martin Jambor <mjambor@suse.cz>
14578 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
14581 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
14584 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
14586 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
14588 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
14589 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
14590 (ix86_handle_option): Handle -mwaitpkg.
14591 * config.gcc: New header.
14592 * config/i386/cpuid.h (bit_WAITPKG): New bit.
14593 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
14594 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
14596 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14597 OPTION_MASK_ISA_WAITPKG.
14598 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
14599 (ix86_option_override_internal): Add PTA_WAITPKG.
14600 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
14601 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
14602 IX86_BUILTIN_TPAUSE.
14603 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
14604 __builtin_ia32_umwait and __builtin_ia32_tpause.
14605 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
14606 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
14607 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
14608 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
14609 UNSPECV_TPAUSE): New.
14610 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
14611 * config/i386/i386.opt: Add -mwaitpkg.
14612 * config/i386/waitpkgintrin.h: New file.
14613 * config/i386/x86intrin.h: New header.
14614 * doc/invoke.texi: Add -mwaitpkg.
14616 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
14619 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
14621 (cortex-m0): Use armv6s-m isa.
14622 (cortex-m0plus): Likewise.
14623 (cortex-m1): Likewise.
14624 (cortex-m0.small-multiply): Likewise.
14625 (cortex-m0plus.small-multiply): Likewise.
14626 (cortex-m1.small-multiply): Likewise.
14628 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
14629 Jakub Jelinek <jakub@redhat.com>
14631 PR tree-optimization/85692
14632 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
14633 source permute as well.
14635 2018-05-11 Martin Liska <mliska@suse.cz>
14638 * doc/extend.texi: Document LLVM style format for no_sanitize
14641 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14643 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
14644 mode_supports_vsx_dform_quad to mode_supports_dq_form.
14645 (mode_supports_vsx_dform_quad): Likewise.
14646 (mode_supports_vmx_dform): Move these functions to be next to the
14647 other mode_supports functions.
14648 (mode_supports_dq_form): Likewise.
14649 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
14650 mode_supports_dq_form.
14651 (reg_offset_addressing_ok_p): Likewise.
14652 (offsettable_ok_by_alignment): Likewise.
14653 (rs6000_legitimate_offset_address_p): Likewise.
14654 (legitimate_lo_sum_address_p): Likewise.
14655 (rs6000_legitimize_address): Likewise.
14656 (rs6000_legitimize_reload_address): Likewise.
14657 (rs6000_secondary_reload_inner): Likewise.
14658 (rs6000_preferred_reload_class): Likewise.
14659 (rs6000_output_move_128bit): Likewise.
14661 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14663 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
14664 Generate SImode target register for null target.
14665 <case IX86_BUILTIN_XGETBV>: Ditto.
14666 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
14667 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
14669 2018-05-10 Carl Love <cel@us.ibm.com>
14671 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14672 dcbtt and dcbtstt if operands[2] is 0.
14674 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
14677 * config/i386/sse.md (usadv64qi): New expander.
14679 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
14681 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
14682 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
14683 -maltivec=be support.
14684 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
14685 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
14686 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
14687 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
14688 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
14689 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
14690 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
14691 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
14692 altivec_vsumsws): Adjust.
14693 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
14694 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
14695 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
14697 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
14698 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
14699 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
14700 (altivec_lve<VI_char>x): Delete expand.
14701 (*altivec_lve<VI_char>x_internal): Rename to...
14702 (altivec_lve<VI_char>x): ... this.
14703 (altivec_lvxl_<mode>): Delete expand.
14704 (*altivec_lvxl_<mode>_internal): Rename to ...
14705 (altivec_lvxl_<mode>): ... this.
14706 (altivec_stvxl_<mode>): Delete expand.
14707 (*altivec_stvxl_<mode>_internal): Rename to ...
14708 (altivec_stvxl_<mode>): ... this.
14709 (altivec_stve<VI_char>x): Delete expand.
14710 (*altivec_stve<VI_char>x_internal): Rename to ...
14711 (altivec_stve<VI_char>x): ... this.
14712 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
14713 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
14714 reduc_plus_scal_<mode>): Adjust.
14715 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
14717 (rs6000_cpu_cpp_builtins): Adjust.
14718 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
14719 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
14720 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
14721 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
14722 -maltivec=be support.
14723 (rs6000_split_vec_extract_var): Adjust.
14724 (rs6000_split_v4si_init): Adjust.
14725 (swap_selector_for_mode): Delete.
14726 (altivec_expand_lvx_be, altivec_expand_stvx_be,
14727 altivec_expand_stvex_be): Delete.
14728 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
14729 -maltivec=be support.
14730 (rs6000_gimple_fold_builtin): Ditto.
14731 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
14733 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
14734 (TARGET_DIRECT_MOVE_64BIT): Adjust.
14735 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
14736 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
14737 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
14738 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
14739 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
14740 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
14741 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
14742 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
14743 anonymous split): Adjust.
14744 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
14745 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
14747 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
14749 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
14750 when --with-gxx-include-dir is also specified.
14751 * configure: Regenerate.
14753 2018-05-09 Jim Wilson <jimw@sifive.com>
14756 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
14757 * config/riscv/t-withmultilib: New.
14758 * config/riscv/withmultilib.h: New.
14759 * doc/install.texi: Document RISC-V --with-multilib-list support.
14761 2018-05-09 Richard Biener <rguenther@suse.de>
14763 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
14765 (vect_bb_vectorization_profitable_p): Adjust. Compute
14766 actual scalar cost using the cost vector and the add_stmt_cost
14769 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14771 PR rtl-optimization/85645
14772 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
14773 in the REG_CFA_REGISTER note for LR, don't leave it empty.
14775 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14777 PR rtl-optimization/85645
14778 * shrink-wrap.c (spread_components): Return a boolean saying if
14779 anything was changed.
14780 (try_shrink_wrapping_separate): Iterate spread_components until
14781 nothing changes anymore.
14783 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14785 PR rtl-optimization/85645
14786 * regrename.c (build_def_use): Also kill the chains that include the
14787 destination of a REG_CFA_REGISTER note.
14789 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
14791 PR rtl-optimization/85645
14792 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
14793 insn that has a REG_CFA_REGISTER note.
14795 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
14797 * cfgexpand.c (expand_clobber): New function.
14798 (expand_gimple_stmt_1): Use it.
14799 * tree-vect-stmts.c (vect_clobber_variable): New function,
14801 (vectorizable_simd_clone_call): ...here.
14802 (vectorizable_store): Emit a clobber either side of an
14803 IFN_STORE_LANES sequence.
14804 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
14806 2018-05-09 Tom de Vries <tom@codesourcery.com>
14809 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
14810 (define_insn "trap_if_false"): Add exit after trap.
14812 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
14814 PR rtl-optimization/85638
14815 * bb-reorder.c: Include common/common-target.h.
14816 (create_forwarder_block): New function extracted from...
14817 (fix_up_crossing_landing_pad): ...here. Rename into...
14818 (dw2_fix_up_crossing_landing_pad): ...this.
14819 (sjlj_fix_up_crossing_landing_pad): New function.
14820 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
14821 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
14822 from both partitions and exit the loop after one iteration.
14824 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14827 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14829 (Basic PowerPC Built-in Functions): The new name of the
14830 subsection previously known as "PowerPC Built-in Functions".
14831 (Basic PowerPC Built-in Functions Available on all Configurations):
14833 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14835 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14837 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14839 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14842 2018-05-08 Jim Wilson <jimw@sifive.com>
14844 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
14845 (LD_EMUL_SUFFIX): New.
14846 (LINK_SPEC): Use it.
14848 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
14850 * doc/extend.texi (PowerPC Built-in Functions): Rename this
14852 (Basic PowerPC Built-in Functions): The new name of the
14853 subsection previously known as "PowerPC Built-in Functions".
14854 (Basic PowerPC Built-in Functions Available on all Configurations):
14856 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14858 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14860 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14862 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14865 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14868 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
14869 after cmpelim optimization.
14871 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
14873 * config.gcc: Support "goldmont".
14874 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
14875 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14876 PROCESSOR_GOLDMONT.
14877 * config/i386/i386.c (m_GOLDMONT): Define.
14878 (processor_target_table): Add "goldmont".
14879 (PTA_GOLDMONT): Define.
14880 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
14881 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
14882 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
14883 (fold_builtin_cpu): Add "goldmont".
14884 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
14885 (ix86_option_override_internal): Add "goldmont".
14886 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
14887 (processor_type): Add PROCESSOR_GOLDMONT.
14888 * config/i386/i386.md: Add CPU "glm".
14889 * config/i386/glm.md: New file.
14890 * config/i386/x86-tune.def: Add m_GOLDMONT.
14891 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
14893 2018-05-08 Jakub Jelinek <jakub@redhat.com>
14896 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
14898 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
14899 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
14900 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
14903 * config/i386/i386.c (ix86_fold_builtin): Handle
14904 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
14907 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
14908 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
14910 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
14913 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
14914 (check_arch): Likewise.
14915 (check_fpu): Return the result rather than printing it.
14916 (end arch): Fix operator precedence.
14917 (end cpu): Likewise.
14918 (END): Print the result from check_fpu.
14920 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
14921 Alan Hayward <alan.hayward@arm.com>
14922 David Sherwood <david.sherwood@arm.com>
14924 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
14925 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
14926 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
14927 (*fcmuo<mode>_and): New patterns.
14929 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
14931 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
14932 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
14933 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
14934 (cmp_op, sve_imm_con): New code attributes.
14935 (SVE_COND_INT_CMP, imm_con): Delete.
14936 (cmp_op): Remove above unspecs from int attribute.
14937 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
14939 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
14940 comparison-specific unspecs.
14941 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
14942 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
14943 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
14944 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
14945 (*vec_fcm<cmp_op><mode>): Rename to...
14946 (*fcm<cmp_op><mode>): ...this and adjust likewise.
14947 (*vec_fcmuo<mode>): Rename to...
14948 (*fcmuo<mode>): ...this and adjust likewise.
14949 (*pred_fcm<cmp_op><mode>): New pattern.
14950 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
14951 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
14953 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
14955 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
14956 (aarch64_emit_sve_predicated_cond): New function.
14957 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
14958 (aarch64_emit_unspec_cond_or): Replace with...
14959 (aarch64_emit_sve_or_conds): ...this new function. Use
14960 aarch64_emit_sve_ptrue_op for the individual comparisons and
14961 aarch64_emit_binop to OR them together.
14962 (aarch64_emit_inverted_unspec_cond): Replace with...
14963 (aarch64_emit_sve_inverted_cond): ...this new function. Use
14964 aarch64_emit_sve_ptrue_op for the comparison and
14965 aarch64_emit_unop to invert the result.
14966 (aarch64_expand_sve_vec_cmp_float): Update after the above
14967 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
14969 2018-05-07 Nathan Sidwell <nathan@acm.org>
14971 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
14972 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
14973 (Backwards Compatibility): Likewise.
14975 2018-05-07 Luis Machado <luis.machado@linaro.org>
14979 2018-05-07 Luis Machado <luis.machado@linaro.org>
14981 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14982 <prefetch_dynamic_strides>: New const bool field.
14983 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14984 prefetch_dynamic_strides.
14985 (exynosm1_prefetch_tune): Likewise.
14986 (thunderxt88_prefetch_tune): Likewise.
14987 (thunderx_prefetch_tune): Likewise.
14988 (thunderx2t99_prefetch_tune): Likewise.
14989 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14991 (aarch64_override_options_internal): Update to set
14992 PARAM_PREFETCH_DYNAMIC_STRIDES.
14993 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14994 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14995 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14996 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14997 prefetch-dynamic-strides setting.
14999 2018-05-07 Luis Machado <luis.machado@linaro.org>
15001 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15002 <minimum_stride>: New const int field.
15003 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15004 minimum_stride field.
15005 (exynosm1_prefetch_tune): Likewise.
15006 (thunderxt88_prefetch_tune): Likewise.
15007 (thunderx_prefetch_tune): Likewise.
15008 (thunderx2t99_prefetch_tune): Likewise.
15009 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15010 (aarch64_override_options_internal): Update to set
15011 PARAM_PREFETCH_MINIMUM_STRIDE.
15012 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15013 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15014 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15015 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15016 stride is constant and is below the minimum stride threshold.
15018 2018-05-07 Luis Machado <luis.machado@linaro.org>
15020 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
15023 2018-05-07 Luis Machado <luis.machado@linaro.org>
15025 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15026 <prefetch_dynamic_strides>: New const bool field.
15027 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15028 prefetch_dynamic_strides.
15029 (exynosm1_prefetch_tune): Likewise.
15030 (thunderxt88_prefetch_tune): Likewise.
15031 (thunderx_prefetch_tune): Likewise.
15032 (thunderx2t99_prefetch_tune): Likewise.
15033 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
15035 (aarch64_override_options_internal): Update to set
15036 PARAM_PREFETCH_DYNAMIC_STRIDES.
15037 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
15038 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
15039 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
15040 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
15041 prefetch-dynamic-strides setting.
15043 2018-05-07 Luis Machado <luis.machado@linaro.org>
15045 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
15046 <minimum_stride>: New const int field.
15047 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
15048 minimum_stride field.
15049 (exynosm1_prefetch_tune): Likewise.
15050 (thunderxt88_prefetch_tune): Likewise.
15051 (thunderx_prefetch_tune): Likewise.
15052 (thunderx2t99_prefetch_tune): Likewise.
15053 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
15054 (aarch64_override_options_internal): Update to set
15055 PARAM_PREFETCH_MINIMUM_STRIDE.
15056 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
15057 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
15058 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
15059 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
15060 stride is constant and is below the minimum stride threshold.
15062 2018-05-06 Jakub Jelinek <jakub@redhat.com>
15065 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
15066 the type is addressable. Don't force op into register if it has
15069 2018-05-05 Roland McGrath <mcgrathr@google.com>
15072 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
15073 any section for which we don't know a specific type it should have,
15074 regardless of name. Previously this was done only for the exact
15075 names ".init_array", ".fini_array", and ".preinit_array".
15076 (default_elf_asm_named_section): Add comment about
15077 relationship with default_section_type_flags and SECTION_NOTYPE.
15078 (get_section): Don't consider it a type conflict if one side has
15079 SECTION_NOTYPE and the other doesn't, as long as neither has the
15080 SECTION_BSS et al used in the default_section_type_flags logic.
15082 2018-05-05 Tom de Vries <tom@codesourcery.com>
15085 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
15086 (workaround_barsyncs): New function.
15087 (nvptx_reorg): Use workaround_barsyncs.
15088 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
15089 (define_expand "nvptx_membar_cta"): New define_expand.
15090 (define_insn "*nvptx_membar_cta"): New insn.
15092 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
15094 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
15095 To improve optimization opportunities.
15096 * builtin-types.def: The new needed builtin types for the above.
15098 2018-05-04 Richard Biener <rguenther@suse.de>
15100 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15101 * gimple-ssa-store-merging.c
15102 (imm_store_chain_info::output_merged_store): Remove redundant create,
15103 release split_store vector contents on failure.
15104 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15105 scalar stmt vector on cache hit.
15107 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
15109 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15111 * config.gcc (powerpc-xilinx-eabi*): Remove.
15112 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15114 (fusion_addis_mem_combo_load): Ditto.
15115 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15117 (rs6000_cpu_cpp_builtins): Ditto.
15118 * config/rs6000/rs6000-linux.c
15119 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15120 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15121 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15123 (rs6000_setup_reg_addr_masks): Ditto.
15124 (rs6000_init_hard_regno_mode_ok): Ditto.
15125 (rs6000_option_override_internal): Ditto.
15126 (legitimate_lo_sum_address_p): Ditto.
15127 (rs6000_legitimize_address): Ditto.
15128 (rs6000_legitimize_reload_address): Ditto.
15129 (rs6000_legitimate_address_p): Ditto.
15130 (abi_v4_pass_in_fpr): Ditto.
15131 (setup_incoming_varargs): Ditto.
15132 (rs6000_gimplify_va_arg): Ditto.
15133 (rs6000_split_multireg_move): Ditto.
15134 (rs6000_savres_strategy): Ditto.
15135 (rs6000_emit_prologue_components): Ditto.
15136 (rs6000_emit_epilogue_components): Ditto.
15137 (rs6000_emit_prologue): Ditto.
15138 (rs6000_emit_epilogue): Ditto.
15139 (rs6000_elf_file_end): Ditto.
15140 (rs6000_function_value): Ditto.
15141 (rs6000_libcall_value): Ditto.
15142 * config/rs6000/rs6000.h: Ditto.
15143 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15144 (TARGET_MINMAX): ... this. New.
15145 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15146 * config/rs6000/rs6000.md: Remove Xilinx FP support.
15147 (*movsi_internal1_single): Delete.
15148 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15149 mfpu=, mxilinx-fpu): Delete.
15150 * config/rs6000/singlefp.h: Delete.
15151 * config/rs6000/sysv4.h: Remove Xilinx FP support.
15152 * config/rs6000/t-rs6000: Ditto.
15153 * config/rs6000/t-xilinx: Delete.
15154 * config/rs6000/titan.md: Adjust for fp_type removal.
15155 * config/rs6000/vsx.md: Remove Xilinx FP support.
15156 (VStype_simple): Delete.
15157 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15158 * config/rs6000/xfpu.h: Delete.
15159 * config/rs6000/xfpu.md: Delete.
15160 * config/rs6000/xilinx.h: Delete.
15161 * config/rs6000/xilinx.opt: Delete.
15162 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15163 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15165 2018-05-04 Tom de Vries <tom@codesourcery.com>
15168 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15171 2018-05-04 Richard Biener <rguenther@suse.de>
15173 PR middle-end/85627
15174 * tree-complex.c (update_complex_assignment): We are always in SSA form.
15175 (expand_complex_div_wide): Likewise.
15176 (expand_complex_operations_1): Likewise.
15177 (expand_complex_libcall): Preserve EH info of the original stmt.
15178 (tree_lower_complex): Handle removed blocks.
15179 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15180 on complex multiplication and division libcall builtins.
15182 2018-05-04 Richard Biener <rguenther@suse.de>
15184 PR middle-end/85574
15185 * fold-const.c (negate_expr_p): Restrict negation of operand
15186 zero of a division to when we know that can happen without
15188 (fold_negate_expr_1): Likewise.
15190 2018-05-04 Jakub Jelinek <jakub@redhat.com>
15193 * real.h (real_nextafter): Declare.
15194 * real.c (real_nextafter): New function.
15195 * fold-const-call.c (fold_const_nextafter): New function.
15196 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15197 CASE_CFN_NEXTTOWARD.
15198 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15199 even when arg1_mode is different from arg0_mode.
15201 2018-05-03 Nathan Sidwell <nathan@acm.org>
15203 * doc/extend.texi (Deprecated Features): Remove
15204 -ffriend-injection.
15205 (Backwards Compatibility): Likewise.
15206 * doc/invoke.texi (C++ Language Options): Likewise.
15207 (C++ Dialect Options): Likewise.
15209 2018-05-03 Jakub Jelinek <jakub@redhat.com>
15212 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15213 _mm512_mask_mullox_epi64): New intrinsics.
15215 2018-05-03 Tom de Vries <tom@codesourcery.com>
15218 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15219 dump files): Add offload-tree.
15221 2018-05-03 Richard Biener <rguenther@suse.de>
15223 PR tree-optimization/85615
15224 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15225 to loops not nested in BBs loop father to avoid creating multi-entry
15228 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15230 PR tree-optimization/70291
15231 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15232 arguments. Change return type to tree. Emit libcall as a new
15233 statement rather than replacing existing one when inplace_p is true.
15234 (expand_complex_multiplication_components): New function.
15235 (expand_complex_multiplication): Expand floating-point complex
15236 multiplication using the above.
15237 (expand_complex_division): Rename inner_type parameter to type.
15238 Update expand_complex_libcall call-site.
15239 (expand_complex_operations_1): Update expand_complex_multiplication
15240 and expand_complex_division call-sites.
15242 2018-05-02 Jakub Jelinek <jakub@redhat.com>
15245 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15246 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15247 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
15248 the highest significant bit of the shift count mask is clear. In
15249 check whether and[sq]i3 is needed verify that all significant bits
15250 of the shift count other than the highest are set.
15252 2018-05-02 Tom de Vries <tom@codesourcery.com>
15255 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
15256 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
15257 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
15258 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
15259 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
15260 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
15261 __builtin_goacc_parlevel_size.
15263 2018-05-02 Richard Biener <rguenther@suse.de>
15265 PR tree-optimization/85597
15266 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
15267 do not use split vect_get_vec_defs call but call vect_get_slp_defs
15270 2018-05-02 Tom de Vries <tom@codesourcery.com>
15273 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15274 dump files): Add ltrans-tree.
15276 2018-05-02 Tom de Vries <tom@codesourcery.com>
15279 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15280 dump files): Add wpa-ipa.
15282 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
15284 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
15285 powerpc*-*-linux*paired* target.
15286 * config/rs6000/750cl.h: Delete.
15287 * config/rs6000/paired.h: Delete.
15288 * config/rs6000/paired.md: Delete.
15289 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
15291 * config/rs6000/rs6000-builtin.def: Remove paired float support.
15292 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
15293 comment. Remove paired float support.
15294 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
15295 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
15297 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
15298 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
15300 * config/rs6000/rs6000.c: Remove paired float support.
15301 (paired_expand_vector_init, paired_expand_vector_move,
15302 paired_emit_vector_compare, paired_emit_vector_cond_expr,
15303 (paired_expand_lv_builtin, paired_expand_stv_builtin,
15304 paired_expand_builtin, paired_expand_predicate_builtin,
15305 paired_init_builtins): Delete.
15306 * config/rs6000/rs6000.h: Remove paired float support.
15307 * config/rs6000/rs6000.md: Remove paired float support.
15308 (move_from_CR_ov_bit): Delete.
15309 * config/rs6000/rs6000.opt (mpaired): Delete.
15310 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
15311 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
15313 2018-05-02 Richard Biener <rguenther@suse.de>
15315 PR middle-end/85567
15316 * gimplify.c (gimplify_save_expr): When in SSA form allow
15317 SAVE_EXPRs to compute to SSA vars.
15319 2018-05-02 Jakub Jelinek <jakub@redhat.com>
15322 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15323 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15324 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
15325 clobber operands[2], instead use a new pseudo. Formatting fixes.
15327 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
15329 PR tree-optimization/85586
15330 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
15331 exit early for statements in the same group if the accesses are
15334 2018-05-02 Tom de Vries <tom@codesourcery.com>
15337 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
15340 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
15342 PR tree-optimization/85143
15343 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
15345 2018-05-01 Tom de Vries <tom@codesourcery.com>
15348 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
15349 not found" error message.
15351 2018-05-01 Tom de Vries <tom@codesourcery.com>
15354 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
15355 * vec.c (test_ordered_remove_if): New function.
15356 (vec_c_tests): Call test_ordered_remove_if.
15357 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
15358 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
15359 * tree-vect-patterns.c (vect_pattern_recog_1): Use
15360 VEC_ORDERED_REMOVE_IF.
15362 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15364 PR tree-optimization/82665
15365 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
15366 pointer subtraction where arguments come from a memchr call.
15368 2018-05-01 Jakub Jelinek <jakub@redhat.com>
15370 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
15371 --push-state --as-needed and --pop-state instead of --as-needed and
15372 --no-as-needed if ld supports it.
15373 * configure: Regenerated.
15376 * doc/install.texi2html: Replace _002d with - and _002a with * in
15377 generated html files using sed.
15379 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15382 * gcc-rich-location.c (blank_line_before_p): New function.
15383 (use_new_line): New function.
15384 (gcc_rich_location::add_fixit_insert_formatted): New function.
15385 * gcc-rich-location.h
15386 (gcc_rich_location::add_fixit_insert_formatted): New function.
15388 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15390 * selftest.c (assert_streq): Rename "expected" and "actual" to
15391 "val1" and "val2". Extend NULL-handling to cover both inputs
15392 symmetrically, while still requiring both to be non-NULL for a pass.
15393 * selftest.h (assert_streq): Rename "expected" and "actual" to
15395 (ASSERT_EQ): Likewise.
15396 (ASSERT_EQ_AT): Likewise.
15397 (ASSERT_KNOWN_EQ): Likewise.
15398 (ASSERT_KNOWN_EQ_AT): Likewise.
15399 (ASSERT_NE): Likewise.
15400 (ASSERT_MAYBE_NE): Likewise.
15401 (ASSERT_MAYBE_NE_AT): Likewise.
15402 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
15403 the assertion to pass.
15404 (ASSERT_STREQ_AT): Likewise.
15406 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
15408 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
15409 interaction with -pie.
15411 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15413 * selftest.h: Fix alphabetization of per-source-file selftest
15416 2018-04-30 Jason Merrill <jason@redhat.com>
15418 PR c++/61982 - dead stores to destroyed objects.
15419 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
15422 2018-04-30 Jason Merrill <jason@redhat.com>
15424 * tree.c (build_clobber): New.
15425 * tree.h: Declare it.
15426 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
15428 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15430 * diagnostic-show-locus.c (layout::layout): Update for
15431 location_get_source_line returning a char_span.
15432 (struct char_span): Move to input.h.
15433 (struct correction): Update for fields in char_span becoming
15435 (struct source_line): Update for location_get_source_line
15436 returning a char_span.
15437 (layout::print_line): Likewise.
15438 * edit-context.c (edited_file::print_content): Likewise.
15439 (edited_file::print_diff_hunk): Likewise.
15440 (edited_file::print_run_of_changed_lines): Likewise.
15441 (edited_file::get_num_lines): Likewise.
15442 (edited_line::edited_line): Likewise.
15443 * final.c (asm_show_source): Likewise.
15444 * input.c (location_get_source_line): Convert return type
15445 from const char * to char_span, losing the final "line_len"
15447 (dump_location_info): Update for the above.
15448 (get_substring_ranges_for_loc): Likewise. Use a char_span
15449 when handling the literal within the line.
15450 (test_reading_source_line): Update for location_get_source_line
15451 returning a char_span.
15452 * input.h (class char_span): Move here from
15453 diagnostic-show-locus.c, converting from a struct to a class.
15454 Make data members private.
15455 (char_span::operator bool): New.
15456 (char_span::length): New.
15457 (char_span::get_buffer): New.
15458 (char_span::operator[]): New.
15459 (char_span::subspan): Make const.
15460 (char_span::xstrdup): New.
15461 (location_get_source_line): Convert return type from const char *
15462 to char_span, losing the final "line_size" param.
15464 2018-04-30 Jan Hubicka <jh@suse.cz>
15466 * lto-wrapper.c (ltrans_priorities): New static var.
15467 (cmp_priority): New.
15468 (run_gcc): Read priorities and if doing parallel build order
15469 the Makefile by them.
15471 2018-04-30 David Malcolm <dmalcolm@redhat.com>
15473 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
15475 2018-04-30 Richard Biener <rguenther@suse.de>
15477 * tree-cfg.c (verify_address): Remove base argument, add
15478 flag whether to check TREE_ADDRESSABLE and do that.
15479 (verify_expr): Remove.
15480 (verify_types_in_gimple_reference): Add pieces from verify_expr.
15481 (verify_gimple_assign_single): Likewise.
15482 (verify_gimple_switch): Likewise.
15483 (verify_expr_location_1): Dereference tp once. Add (disabled)
15484 piece from verify_expr.
15485 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
15487 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15489 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
15491 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15493 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
15494 (small_data_pattern): Likewise.
15495 (arc_rewrite_small_data): Likewise.
15496 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
15497 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
15498 (get_symbol_alignment): New function.
15499 (legitimate_small_data_address_p): Likewise.
15500 (legitimate_scaled_address): Update, call
15501 legitimate_small_data_address_p.
15502 (output_sdata): New static variable.
15503 (arc_print_operand): Update how we handle small data operands.
15504 (arc_print_operand_address): Likewise.
15505 (arc_legitimate_address_p): Update, use
15506 legitimate_small_data_address_p.
15507 (arc_rewrite_small_data_p): Remove.
15508 (arc_rewrite_small_data_1): Likewise.
15509 (arc_rewrite_small_data): Likewise.
15510 (small_data_pattern): Likewise.
15511 (compact_sda_memory_operand): Update to use
15512 legitimate_small_data_address_p and get_symbol_alignment.
15513 (prepare_move_operands): Don't rewite sdata pattern.
15514 (prepare_extend_operands): Remove.
15515 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
15517 (zero_extendqisi2): Likewise.
15518 (zero_extendhisi2): Likewise.
15519 (extendqihi2): Likewise.
15520 (extendqisi2): Likewise.
15521 (extendhisi2): Likewise.
15522 (addsi3): Likewise.
15523 (subsi3): Likewise.
15524 (andsi3): Likewise.
15525 * config/arc/constraints.md (Usd): Change it to memory constraint.
15527 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
15529 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
15530 as source of std instructions.
15531 * config/arc/arc.md (movsi_insn): Update pattern predicate to
15532 allow 6-bit constants as source for store instructions.
15533 (movdi_insn): Update instruction pattern to allow 6-bit constants
15534 as source for store instructions.
15536 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
15538 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
15540 2018-04-30 Nathan Sidwell <nathan@acm.org>
15541 Sandra Loosemore <sandra@codesourcery.com>
15543 * dumpfile.c (dump_open): Allow '-' for stdout.
15544 * doc/invoke.texi (Developer Options): Document dump filename
15545 determination early. Document stdin/stdout selection.
15547 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
15549 Microblaze Target: PIC data text relative
15551 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
15552 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
15554 * config/microblaze/microblaze.h (microblaze_constant_address_p):
15555 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
15556 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
15557 New addressing mode for data-text relative position indepenedent code.
15558 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
15559 'ADDRESS_SYMBOLIC_TXT_REL'.
15560 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
15561 (microblaze_legitimate_pic_operand): Exclude function calls from
15562 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
15563 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
15565 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15566 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15567 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
15568 for 'address + offset'.
15569 (microblaze_expand_prologue): Add new function prologue call for
15571 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
15572 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
15573 table in case of TARGET_PIC_DATA_TEXT_REL.
15574 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
15575 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
15576 Add new macros 'UNSPEC_TEXT',
15577 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
15578 + exclude function calls from 'UNSPEC_PLT' in case of data text
15580 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15581 new target hook for generating address diff vector tables in case of
15583 * doc/tm.texi : Regenerate.
15584 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
15585 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
15586 of addr diff vector generation.
15587 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15588 target hook definition.
15589 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
15590 Add default function for generate_pic_addr_diff_vec -> flag_pic.
15591 * doc/invoke.texi (Add new pic option): Add new microblaze pic
15592 option for data text relative.
15594 2018-04-30 Richard Biener <rguenther@suse.de>
15596 * tree-chrec.h (evolution_function_is_constant_p): Remove
15598 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
15600 2018-04-30 Richard Biener <rguenther@suse.de>
15603 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
15605 2018-04-30 Richard Biener <rguenther@suse.de>
15607 PR tree-optimization/28364
15608 PR tree-optimization/85275
15609 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
15610 copying first exit test.
15612 2018-04-28 Mark Wielaard <mark@klomp.org>
15614 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
15615 dwarf_version >= 5.
15616 (dwarf_AT): Handle DW_AT_addr_base.
15617 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
15619 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
15622 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
15623 (*ashl<dwi>3_doubleword_mask_1): Ditto.
15624 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
15625 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
15627 2018-04-28 Richard Biener <rguenther@suse.de>
15629 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
15630 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
15631 to reflect use. Only add interesting stmts.
15633 2018-04-27 Martin Jambor <mjambor@suse.cz>
15636 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
15637 the jump function allows for passing through aggregate values.
15639 2018-04-27 David Malcolm <dmalcolm@redhat.com>
15641 * input.h (in_system_header_at): Convert from macro to inline
15643 (from_macro_expansion_at): Likewise.
15644 (from_macro_definition_at): Likewise.
15646 2018-04-27 Jeff Law <law@redhat.com>
15648 * config.gcc: Mark tile* targets as deprecated/obsolete.
15650 2018-04-27 Richard Biener <rguenther@suse.de>
15652 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
15655 2018-04-27 Richard Biener <rguenther@suse.de>
15657 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
15659 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
15661 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
15662 with Yd constraint. Set "preferred_for_speed" attribute from
15663 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
15664 with Yd constraint.
15665 (*movdi_internal): Ditto.
15666 (movti_interunit splitters): Remove
15667 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
15668 (movdi_interunit splitters): Ditto.
15669 * config/i386/constraints.md (Ye): Remove.
15670 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
15672 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15675 * config/aarch64/constraints.md (Usg): Limit to 31.
15676 (Usj): Limit to 63.
15678 2018-04-27 Jakub Jelinek <jakub@redhat.com>
15680 PR tree-optimization/85529
15681 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
15682 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
15683 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
15684 zero extension or masking of the MSB bit.
15685 (optimize_range_tests): Add FIRST_BB argument, pass it through
15686 to optimize_range_tests_var_bound.
15687 (maybe_optimize_range_tests, reassociate_bb): Adjust
15688 optimize_range_tests callers.
15690 2018-04-26 Richard Biener <rguenther@suse.de>
15691 Jakub Jelinek <jakub@redhat.com>
15693 * cgraph.h (symbol_table): Just declare debug method here.
15694 * symtab.c (symbol_table::debug): Define.
15696 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
15698 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
15700 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
15702 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
15703 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
15704 (*movdi_internal): Substitute Yi and Yj constraint with x
15705 and Ym and Yn constraint with y constraint. Update "isa"
15706 attribute and set "preferred_for_speed" attribute from
15707 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
15708 (*movsi_internal): Ditto.
15709 (*movdf_internal): Ditto.
15710 (*movsf_internal): Ditto.
15711 (*zero_extendsidi2): Ditto.
15712 * config/i386/sse.md (vec_set<mode>_0): Ditto.
15713 (sse2_loadld): Ditto.
15714 (*vec_extract<ssevecmodelower>_0): Ditto.
15715 (*vec_extractv4si_0_zext_sse4): Ditto.
15716 (vec_concatv2di): Ditto.
15717 (*vec_dup<mode>): Ditto.
15718 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
15719 * config/i386/constraints.md (Yi): Remove.
15724 2018-04-26 Nathan Sidwell <nathan@acm.org>
15726 * dumpfile.c (dump_open): New.
15727 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
15728 (dump_finish): Detect stdio/stderr by value not name.
15730 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
15732 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
15734 2018-04-26 Tom de Vries <tom@codesourcery.com>
15737 * config/nvptx/nvptx.c (verify_neutering_jumps)
15738 (verify_neutering_labels): New function
15739 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
15741 2018-04-26 Tom de Vries <tom@codesourcery.com>
15744 * config/nvptx/nvptx.c (needs_neutering_p): New function.
15745 (nvptx_single): Use needs_neutering_p to skip over insns that do not
15748 2018-04-26 Richard Biener <rguenther@suse.de>
15749 Tom de Vries <tom@codesourcery.com>
15752 * lto-streamer-out.c (output_function): Fixup loops if required to match
15753 discovery done in the reader.
15755 2018-04-26 Richard Biener <rguenther@suse.de>
15757 PR tree-optimization/85116
15758 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
15759 have a loop exit from the single latch predecessor. Remove
15760 case of header with just condition.
15761 (ch_base::copy_headers): Exclude infinite loops from any
15763 (pass_ch::execute): Record exits.
15765 2018-04-26 Richard Biener <rguenther@suse.de>
15767 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
15768 prologue cost vector and pass it to vect_get_load_cost.
15769 (vect_get_peeling_costs_all_drs): Likewise.
15770 (vect_peeling_hash_get_lowest_cost): Likewise.
15771 (vect_enhance_data_refs_alignment): Likewise.
15773 2018-04-26 Richard Biener <rguenther@suse.de>
15775 PR middle-end/85450
15776 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
15777 checking of integer<->pointer conversions.
15778 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
15779 sign-/zero-extending pointer types.
15780 (expand_omp_for_static_chunk): Likewise.
15782 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
15783 Jean Lee <xiaoyur347@gmail.com>
15785 * config/mips/mips.c (mips_asan_shadow_offset): New function.
15786 (TARGET_ASAN_SHADOW_OFFSET): Define.
15787 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
15788 true for -fsanitize=address.
15790 2018-04-25 Mark Wielaard <mark@klomp.org>
15792 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
15795 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15797 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
15798 than "alu", remove explicit "memory" and "imm_disp" attributes.
15799 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
15801 PR middle-end/85414
15802 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
15803 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
15804 gen_lowpart_no_emit.
15806 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
15809 * config/i386/i386.c (ix86_expand_builtin): Change memory
15810 operand to XI, extend p0 to Pmode.
15811 * config/i386/i386.md: Change unspec volatile and operand
15812 1 mode to XI, change operand 0 mode to P.
15814 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15816 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
15817 GET_MODE_MASK before any checking.
15818 (nds32_can_use_bset_p): Likewise.
15819 (nds32_can_use_btgl_p): Likewise.
15821 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15823 * config/nds32/nds32-doubleword.md: New define_split pattern for
15824 illegal register number.
15826 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15828 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
15830 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
15832 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
15834 2018-04-25 Richard Biener <rguenther@suse.de>
15836 * lto-streamer.h (LTO_major_version): Bump to 8.
15838 2018-04-25 Jakub Jelinek <jakub@redhat.com>
15840 * BASE-VER: Set to 9.0.0.
15842 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
15844 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
15845 in __abskf2 and __powikf2.
15847 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15850 * config/aarch64/constraints.md (Usg, Usj): New constraints.
15851 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
15852 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
15853 Use the above on operand 2. Reindent.
15854 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
15856 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15859 * common/config/i386/i386-common.c (ix86_handle_option): Don't
15861 * config/i386/i386.opt (mcet): Removed.
15862 * doc/install.texi: Remove -mcet documentation.
15863 * doc/invoke.texi: Likewise.
15865 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
15868 * doc/install.texi: Remove -mcet from bootstrap-cet.
15870 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15873 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
15874 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
15878 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
15879 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
15880 containing a CONST_VECTOR.
15882 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
15884 * doc/install.texi: Update newlib dependency for nvptx.
15886 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15889 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
15890 instead of INTVAL when shifting x left.
15892 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
15894 PR tree-optimization/85478
15895 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
15896 vect_grouped_store_supported for single element vectors.
15898 2018-04-24 Richard Biener <rguenther@suse.de>
15901 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
15902 load cost increase to the case of non-constant step.
15904 2018-04-24 Jakub Jelinek <jakub@redhat.com>
15907 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
15908 destination if any_malformed_asm.
15910 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
15912 PR middle-end/85496
15913 * expr.c (store_field): In the bitfield case, if the value comes from
15914 a function call and is returned in registers by means of a PARALLEL,
15915 do not change the mode of the temporary unless BLKmode and VOIDmode.
15917 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
15919 PR rtl-optimization/85423
15920 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
15921 dependencies to debug insns when the previous insn is non-debug.
15923 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
15925 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
15926 enums into a single definition.
15927 (fls): Fix predicates and printing.
15930 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
15932 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
15933 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
15934 and short u6 immediate.
15935 (check_if_valid_sleep_operand): Remove.
15936 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
15938 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15940 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
15941 flag_always_save_lp condition.
15942 * config/nds32/nds32.opt (malways-save-lp): New option.
15944 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15946 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
15947 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
15948 * config/nds32/nds32.h
15949 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
15950 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
15952 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15954 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
15955 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
15957 2018-04-22 Shiva Chen <shiva0217@gmail.com>
15958 Chung-Ju Wu <jasonwucj@gmail.com>
15960 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
15962 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
15963 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
15965 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15967 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
15969 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15971 * config/nds32/nds32-protos.h (nds32_data_alignment,
15972 nds32_local_alignment): Declare.
15973 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
15974 nds32_local_alignment): New functions.
15975 (TARGET_CONSTANT_ALIGNMENT): Define.
15976 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
15978 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15980 * config/nds32/nds32.c
15981 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
15982 (TARGET_MODES_TIEABLE_P): Likewise.
15984 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
15986 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
15987 level Ofast and Og.
15989 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
15990 Chung-Ju Wu <jasonwucj@gmail.com>
15992 * config/nds32/constants.md (unspec_volatile_element): Add enum values
15993 for unaligned access.
15994 * config/nds32/nds32-intrinsic.c: Implementation of expanding
15996 * config/nds32/nds32-intrinsic.md: Likewise.
15997 * config/nds32/nds32_intrinsic.h: Likewise.
15998 * config/nds32/nds32.h (nds32_builtins): Likewise.
15999 * config/nds32/nds32.opt (munaligned-access): New option.
16000 * config/nds32/nds32.c (nds32_asm_file_start): Display
16001 flag_unaligned_access status.
16003 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16005 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
16006 -mno-relax is present.
16007 * config/riscv/linux.h (LINK_SPEC): Ditto.
16009 2018-04-20 Martin Sebor <msebor@redhat.com>
16012 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
16014 (gimple_fold_builtin_stxcpy_chk): Same.
16015 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
16017 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
16020 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
16021 __powikf2 when long double is IEEE 128-bit.
16023 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
16025 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
16026 step to make sure stack always aligned.
16028 2018-04-20 Carl Love <cel@us.ibm.com>
16031 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
16032 size check for arg0.
16034 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
16035 Tom de Vries <tom@codesourcery.com>
16038 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
16039 Emit insns for calls too.
16040 (nvptx_find_par): Always look for worker-level predecessor insn.
16041 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
16043 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
16044 (nvptx_process_pars): Propagate frames for calls.
16046 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
16049 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
16051 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
16052 (ix86_handle_option): Don't handle OPT_mibt.
16053 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
16055 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
16057 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
16058 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
16059 (ix86_target_macros): Define __CET__ with flag_cf_protection
16060 for -fcf-protection.
16061 * config/i386/i386.c (isa2_opts): Remove -mibt.
16062 * config/i386/i386.h (TARGET_IBT): Removed.
16063 (TARGET_IBT_P): Likewise.
16064 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
16065 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
16066 * config/i386/i386.opt (mcet): Update help message.
16067 (mshstk): Likewise.
16069 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
16070 -mcet as an alias for -mshstk.
16072 2018-04-20 Richard Biener <rguenther@suse.de>
16074 PR middle-end/85475
16075 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
16076 complexity by forcing a single use of the multiply operand.
16078 2018-04-20 Martin Jambor <mjambor@suse.cz>
16081 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
16082 recursion dependency to only apply to non-clones.
16084 2018-04-20 Martin Jambor <mjambor@suse.cz>
16087 * ipa-cp.c (create_specialized_node): Check that clones of
16088 self-recursive edges exist during IPA-CP.
16090 2018-04-19 Toon Moene <toon@moene.org>
16092 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
16095 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16097 PR tree-optimization/85467
16098 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16099 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
16100 VECTOR_CST element to type.
16102 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16105 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16106 * config/i386/i386.md (builtin_setjmp_setup): Removed.
16107 (builtin_longjmp): Likewise.
16108 (save_stack_nonlocal): New pattern.
16109 (restore_stack_nonlocal): Likewise.
16111 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16114 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16115 Replace ASM_OUTPUT_LABEL with fprintf.
16117 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
16120 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16121 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16122 * config/i386/i386-c.c (ix86_target_macros_internal): Also
16123 define __IBT__ and __SHSTK__ for -fcf-protection.
16124 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16126 (ix86_trampoline_init): Likewise.
16127 (x86_output_mi_thunk): Likewise.
16128 (ix86_notrack_prefixed_insn_p): Likewise.
16129 (ix86_option_override_internal): Don't disallow -fcf-protection.
16130 * config/i386/i386.md (rdssp<mode>): Also enable for
16132 (incssp<mode>): Likewise.
16133 (nop_endbr): Likewise.
16134 * config/i386/i386.opt (mcet): Change help message to built-in
16137 (mshstk): Likewise.
16138 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16139 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
16140 enable CET built-in functions.
16142 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
16144 * common/config/i386/i386-common.c
16145 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16146 OPTION_MASK_ISA_MOVDIRI_UNSET,
16147 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16148 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16149 * config.gcc (movdirintrin.h): New header.
16150 * config/i386/cpuid.h (bit_MOVDIRI,
16151 bit_MOVDIR64B): New bits.
16152 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16154 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16155 (VOID, PVOID, PCVOID)): New function types.
16156 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16157 __builtin_ia32_directstoreu_u64,
16158 __builtin_ia32_movdir64b): New builtins.
16159 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16160 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16162 (ix86_valid_target_attribute_inner_p): Ditto.
16163 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16164 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16165 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16166 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16167 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16168 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16169 (movdiri<mode>, movdir64b_<mode>): New.
16170 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16171 * config/i386/immintrin.h: Include movdirintrin.h.
16172 * config/i386/movdirintrin.h: New file.
16173 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16175 2018-04-19 Richard Biener <rguenther@suse.de>
16177 PR middle-end/85455
16178 * cfg.c (clear_bb_flags): When loop state says we have
16179 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16181 2018-04-19 Richard Biener <rguenther@suse.de>
16183 PR tree-optimization/84737
16184 * tree-vect-data-refs.c (vect_copy_ref_info): New function
16185 copying restrict info.
16186 (vect_setup_realignment): Use it.
16187 * tree-vectorizer.h (vect_copy_ref_info): Declare.
16188 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16189 the first DR to all generated stores.
16190 (vectorizable_load): Likewise for loads.
16192 2018-04-19 Jakub Jelinek <jakub@redhat.com>
16194 PR tree-optimization/85446
16195 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16196 the integral and pointer types to have the same precision.
16198 * doc/install.texi: Document --disable-cet being the default and
16201 2018-04-18 Martin Liska <mliska@suse.cz>
16203 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16206 2018-04-18 Martin Liska <mliska@suse.cz>
16209 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
16212 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16213 arguments if they are comparable.
16215 2018-04-18 Martin Liska <mliska@suse.cz>
16218 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
16221 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16224 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
16227 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16228 ENDBR after calling __morestack.
16230 2018-04-18 David Malcolm <dmalcolm@redhat.com>
16233 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
16234 by using gcc_base_ver to generate a gcc_driver_version, and use
16235 it when generating GCC_DRIVER_NAME.
16236 * configure: Regenerate.
16238 2018-04-18 Jakub Jelinek <jakub@redhat.com>
16241 * config.gcc: Obsolete powerpc*-*-*spe*.
16243 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16246 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
16247 (stabstr_D): Change type of unum from unsigned int to
16248 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
16251 2018-04-17 Jim Wilson <jimw@sifive.com>
16254 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
16255 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
16256 Set arg_pointer_offset after using pretend_args_size.
16258 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16260 PR rtl-optimization/85431
16261 * dse.c (record_store): Ignore zero width stores.
16264 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
16265 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
16266 __builtin_stack_restore rather than after it.
16267 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
16268 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
16269 argument instead of virtual_dynamic_stack_rtx.
16271 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
16273 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
16275 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
16276 Add note to error message to explain internal mapping of overloaded
16277 built-in function name to non-overloaded built-in function name.
16278 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
16281 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
16284 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
16285 where the inputs overlap with the output.
16287 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16289 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
16290 (=v, v) alternative and explicit "memory" attribute.
16291 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
16292 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16294 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
16295 "sselog1" type instead of "sselog".
16296 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
16297 "sselog". Remove explicit "memory" attribute.
16298 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
16299 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16301 (vec_extract_hi_v32hi): Merge all alternatives into one, use
16302 "sselog1" type instead of "sselog". Remove explicit "memory"
16304 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
16305 use "sselog1" type instead of "sselog". Remove explicit "memory"
16307 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
16308 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16310 (vec_extract_hi_v64qi): Merge all alternatives into one, use
16311 "sselog1" type instead of "sselog". Remove explicit "memory"
16313 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
16314 use "sselog1" type instead of "sselog". Remove explicit "memory"
16318 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
16320 PR middle-end/85414
16321 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
16324 2018-04-17 Martin Jambor <mjambor@suse.cz>
16327 * ipa-cp.c (create_specialized_node): Call
16328 expand_all_artificial_thunks if necessary.
16330 2018-04-17 Martin Liska <mliska@suse.cz>
16333 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
16334 in message, remote space in between '_G' and '('.
16336 2018-04-17 Jakub Jelinek <jakub@redhat.com>
16339 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
16340 avx512f_vmcmp<mode>3<round_saeonly_name>,
16341 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
16342 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
16343 avx512f_rndscale<mode><round_saeonly_name>,
16344 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
16345 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
16346 Use %<iptr>2 instead of %2 for -masm=intel.
16347 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
16348 avx512f_vcvttss2usi<round_saeonly_name>,
16349 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
16351 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
16352 avx512f_vcvttsd2usi<round_saeonly_name>,
16353 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
16354 Use %q1 instead of %1 for -masm=intel.
16355 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
16356 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
16357 of %3 for -masm=intel.
16358 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
16360 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
16362 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
16364 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
16365 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
16367 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
16369 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
16370 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
16371 %g1 and one with %0 and %1.
16372 (avx512er_vmrcp28<mode><round_saeonly_name>,
16373 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
16374 %1 for -masm=intel.
16375 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
16376 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
16377 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
16378 of %0 and %{%4%} for -masm=intel.
16379 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
16380 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
16381 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
16382 order of %0 and %{%5%}%{z%} for -masm=intel.
16384 2018-04-17 Jan Hubicka <jh@suse.cz>
16387 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
16389 2018-04-17 Martin Liska <mliska@suse.cz>
16392 * multiple_target.c (create_dispatcher_calls): Set apostrophes
16393 for target_clone error message. Make default implementation
16394 clone to be a local declaration.
16395 (separate_attrs): Add new argument and check for an empty
16397 (expand_target_clones): Handle it.
16398 (ipa_target_clone): Make redirection just for target_clones
16401 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
16402 Tom de Vries <tom@codesourcery.com>
16404 PR middle-end/84955
16405 * omp-expand.c (expand_oacc_for): Add dummy false branch for
16406 tiled basic blocks without omp continue statements.
16408 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
16411 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
16412 vec_extract expression as having side effects to make sure it gets
16415 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
16418 * config/i386/i386.c (get_builtin_code_for_version): Check
16421 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
16424 * config.gcc: Support "skylake".
16425 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16427 * config/i386/i386.c (m_SKYLAKE): Define.
16428 (processor_target_table): Add "skylake".
16429 (ix86_option_override_internal): Add "skylake".
16430 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
16431 PROCESSOR_CANNONLAKE.
16432 (get_builtin_code_for_version): Fix priority for
16433 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
16434 PROCESSOR_SKYLAKE-AVX512.
16435 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
16436 (processor_type): Add PROCESSOR_SKYLAKE.
16438 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
16439 Jason Merrill <jason@redhat.com>
16442 * convert.c (convert_to_integer_1): Use direct recursion for
16443 enumeral types and types with a precision less than the number
16444 of bits in their mode.
16446 2018-04-16 Julia Koval <julia.koval@intel.com>
16449 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
16450 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
16452 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
16455 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
16456 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
16457 and -mno-direct-move.
16459 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
16462 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
16463 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
16464 Ensure negative shifts result in {0}.
16466 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
16468 PR rtl-optimization/79916
16469 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
16470 regs (if any) to define how to gnerate SD moves when LRA is in
16473 2018-04-13 Jakub Jelinek <jakub@redhat.com>
16475 PR rtl-optimization/85393
16476 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
16477 * except.c (expand_dw2_landing_pad_for_region): Make static.
16478 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
16479 a label and unconditional jump to old_bb, rather than
16480 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
16483 PR rtl-optimization/85376
16484 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
16485 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
16486 instead of a specific value.
16488 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
16489 Bin Cheng <bin.cheng@arm.com>
16491 PR tree-optimization/82965
16492 PR tree-optimization/83991
16493 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
16494 by_profile_only parameter.
16495 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
16496 information if the loop was predicted to iterate too many times.
16497 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
16499 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
16502 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
16505 2018-04-13 Martin Liska <mliska@suse.cz>
16506 Jakub Jelinek <jakub@redhat.com>
16508 PR middle-end/81657
16509 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
16510 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
16511 * builtins.c (expand_builtin_memory_copy_args): Use
16512 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
16513 handle dest_addr == pc_rtx.
16515 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
16518 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
16519 asked to not generate direct moves.
16520 (fix_trunc<mode>si2_stfiwx): Similar.
16521 (fix_trunc<mode>si2_internal): Similar.
16523 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16526 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
16527 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
16528 lookup if dest in some wider mode is known to be const0_rtx and
16529 if so, record permanent equivalence for it to be ZERO_EXTEND of
16530 the narrower mode destination.
16532 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
16534 * lto-streamer-out.c (output_function): Revert 259346.
16535 * omp-expand.c (expand_oacc_for): Likewise.
16537 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
16539 PR rtl-optimization/85354
16540 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
16541 * sel-sched.c (sel_global_init): ... here.
16543 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
16546 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
16547 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
16548 mode for PE-COFF targets.
16549 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
16550 (i386_pe_asm_lto_end): Likewise.
16551 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
16552 (TARGET_ASM_LTO_END): Likewise.
16553 * config/i386/winnt.c (saved_debug_info_level): New static variable.
16554 (i386_pe_asm_lto_start): New function.
16555 (i386_pe_asm_lto_end): Likewise.
16557 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
16558 Richard Biener <rguenther@suse.de>
16560 PR middle-end/84955
16561 * lto-streamer-out.c (output_function): Fix CFG loop state before
16563 * omp-expand.c (expand_oacc_for): Handle calls to internal
16564 functions like regular functions.
16566 2018-04-12 Richard Biener <rguenther@suse.de>
16569 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
16570 for the early LTO debug to properly generate references to it
16571 during DIE emission. Do not re-use that for the skeleton for
16573 (dwarf2out_early_finish): Likewise.
16575 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16578 * config/i386/sse.md
16579 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
16580 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
16581 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
16582 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
16583 and output is a reg, avoid creating invalid lowpart subreg, but
16584 instead split into a 512-bit move. Don't split if not AVX512VL,
16585 input is xmm16+ reg and output is a mem.
16586 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
16587 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
16588 xmm16+ reg and output is a mem.
16590 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
16592 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
16593 also for flag_dwarf2_cfi_asm.
16595 2018-04-12 Jakub Jelinek <jakub@redhat.com>
16597 PR rtl-optimization/85342
16598 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
16599 a bool scalar var inside of the loop instead. Don't try to update
16600 recog_data.operand after failed apply_change_group.
16602 2018-04-12 Tom de Vries <tom@codesourcery.com>
16605 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
16606 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
16607 array with flexible array member as array without given dimension.
16608 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
16609 argument for undefined param to true.
16611 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
16614 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16615 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
16616 from PowerPC section.
16617 * config/rs6000/sysv4.opt (mcall-): Improve help text.
16618 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
16619 help text that is too long.
16620 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
16621 help text that is too long.
16622 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
16623 help text that is too long.
16625 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
16627 * config/alpha/alpha.md (stack_probe_internal): Rename
16628 from "probe_stack". Update all callers.
16630 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16632 PR rtl-optimization/84566
16633 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
16634 sched_macro_fuse_insns.
16636 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16639 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
16640 (compute_block_dependences): ... from here.
16642 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16644 PR tree-optimization/85331
16645 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
16646 from int to HOST_WIDE_INT.
16648 2018-04-11 Martin Jambor <mjambor@suse.cz>
16651 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
16652 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
16653 not the same as the source val.
16654 (cgraph_edge_brings_value_p): New parameter.
16655 (gather_edges_for_value): Pass destination value to
16656 cgraph_edge_brings_value_p.
16657 (perhaps_add_new_callers): Likewise.
16658 (get_info_about_necessary_edges): Likewise and exclude values brought
16659 only by self-recursive edges.
16660 (create_specialized_node): Redirect only clones of self-calling edges.
16661 (+self_recursive_pass_through_p): New function.
16662 (find_more_scalar_values_for_callers_subset): Use it.
16663 (find_aggregate_values_for_callers_subset): Likewise.
16664 (known_aggs_to_agg_replacement_list): Removed.
16665 (decide_whether_version_node): Re-calculate known constants for all
16666 remaining context clones.
16668 2018-04-11 Richard Biener <rguenther@suse.de>
16671 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
16672 from early DWARF output.
16673 (dwarf2out_early_finish): Output line info unconditionally into
16674 early DWARF and add reference to it.
16676 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16679 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
16680 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
16681 other than V2DFmode using iptr mode attribute.
16682 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
16684 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
16686 PR rtl-optimization/84659
16687 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
16689 2018-04-11 Jakub Jelinek <jakub@redhat.com>
16692 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
16694 (output_loc_list): Pass address of a dummy size variable even in the
16695 locview handling loop.
16696 (index_location_lists): Add comment on why skip_loc_list_entry can't
16699 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
16702 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
16705 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
16708 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16709 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
16710 and -mstring-compare-inline-limit.
16712 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16715 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
16716 for stack clash protection in a register whenever we need it to be in
16719 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16721 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
16722 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
16724 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
16727 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
16729 (mlong-double-): Ditto.
16730 * config/rs6000/sysv4.opt (msdata=): Ditto.
16731 (mtls-size=): Ditto.
16733 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
16735 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16736 erroneous entries for
16737 "vector int vec_ldl (int, long int *)", and
16738 "vector unsigned int vec_ldl (int, unsigned long int *)".
16739 Add comments and entries for
16740 "vector bool char vec_ldl (int, bool char *)",
16741 "vector bool short vec_ldl (int, bool short *)",
16742 "vector bool int vec_ldl (int, bool int *)",
16743 "vector bool long long vec_ldl (int, bool long long *)",
16744 "vector pixel vec_ldl (int, pixel *)",
16745 "vector long long vec_ldl (int, long long *)",
16746 "vector unsigned long long vec_ldl (int, unsigned long long *)".
16747 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
16748 type tree bool_long_long_type_node and correct definition of
16749 bool_V2DI_type_node to make reference to this new type tree.
16750 (rs6000_mangle_type): Replace erroneous reference to
16751 bool_long_type_node with bool_long_long_type_node.
16752 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
16753 comments to emphasize sign distinctions for char and int types and
16754 replace RS6000_BTI_bool_long constant with
16755 RS6000_BTI_bool_long_long constant. Also add comment to restrict
16756 use of RS6000_BTI_pixel.
16757 (bool_long_type_node): Remove this macro definition.
16758 (bool_long_long_type_node): New macro definition
16760 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16762 PR rtl-optimization/85300
16763 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
16764 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
16765 simplify_unary_operation fails.
16767 2018-04-10 Martin Liska <mliska@suse.cz>
16769 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
16770 cgraph_edge and ipa_ref.
16772 2018-04-10 Jakub Jelinek <jakub@redhat.com>
16776 * config/i386/sse.md
16777 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
16778 computation of the VEC_MERGE selector from mask.
16779 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
16780 Fix decoding of the VEC_MERGE selector into mask.
16782 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16784 PR tree-optimization/85286
16785 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
16787 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
16789 * final.c (final_1): Set insn_last_address as well as
16790 insn_current_address.
16792 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
16795 * explow.c (emit_stack_probe): Call validize_mem on memory location
16796 before passing it to gen_probe_stack. Create address operand and
16797 legitimize it for the probe_stack_address case.
16799 2018-04-09 Jan Hubicka <jh@suse.cz>
16802 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
16803 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
16804 * tree.c (free_lang_data_in_type): Fix handling of binfos;
16806 (free_lang_data): Rebuild type inheritance graph.
16808 2018-04-09 Martin Sebor <msebor@redhat.com>
16810 * invoke.texi (-finline-small-functions): Mention other optimization
16812 (-findirect-inlining, -fpartial-inlining): Same.
16813 (-finline-functions-called-once): Same.
16814 (-freorder-blocks-and-partition): Same.
16816 2018-04-09 Jan Hubicka <jh@suse.cz>
16819 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
16820 jumps; choose last target that matches the criteria (i.e.
16821 no partition changes for non-crossing jumps).
16822 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
16823 support for redirecting crossing jumps to non-crossing.
16825 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
16827 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
16828 also for naked functions.
16830 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
16832 * config/arc/arc.md (add_shift): New pattern.
16833 (add_shift2): Likewise.
16834 (sub_shift): Likewise.
16835 (sub_shift_cmp0_noout): Likewise.
16836 (compare_si_ashiftsi): Likewise.
16837 (xbfu_cmp0_noout): New combine pattern.
16838 (xbfu_cmp0"): Likewise.
16839 (movsi_set_cc_insn): Place the predicable variant first.
16840 (commutative_binary_cmp0_noout): Remove clobber.
16841 (commutative_binary_cmp0): New pattern.
16842 (noncommutative_binary_cmp0): Likewise.
16843 (noncommutative_binary_cmp0_noout): Likewise.
16844 (noncommutative_binary_comparison_result_used): Removed.
16845 (rsub_cmp0): New pattern.
16846 (rsub_cmp0_noout): Likewise.
16847 (extzvsi): Changed, keep only meaningful variants.
16848 (SQH, SEZ): New iterators.
16849 (SQH_postfix): New mode attribute.
16850 (SEZ_prefix): New code attribute.
16851 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
16852 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
16853 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
16854 of numerical value.
16855 (noncommutative_operator): Check the availability of barrel
16858 2018-04-09 Richard Biener <rguenther@suse.de>
16860 PR tree-optimization/85284
16861 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
16862 Only use the niter constraining form of simple_iv when the exit
16863 is always executed.
16865 2018-04-09 Tom de Vries <tom@codesourcery.com>
16868 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16869 (define_expand "*memory_barrier"): New define_expand.
16870 (define_insn "memory_barrier"): New insn.
16872 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16874 PR rtl-optimization/80463
16875 PR rtl-optimization/83972
16876 PR rtl-optimization/83480
16878 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
16879 correct producer for the insn.
16880 (tidy_control_flow): Fixup seqnos in case of debug insns.
16882 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16884 PR rtl-optimization/83913
16886 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
16887 different sched-times when merging exprs.
16889 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16891 PR rtl-optimization/83962
16893 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
16894 tidy_fallthru_edge and tidy_control_flow.
16896 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
16898 PR rtl-optimization/83530
16900 * sel-sched.c (force_next_insn): New global variable.
16901 (remove_insn_for_debug): When force_next_insn is true, also leave only
16902 next insn in the ready list.
16903 (sel_sched_region): When the region wasn't scheduled, make another pass
16904 over it with force_next_insn set to 1.
16906 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
16908 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
16910 * config/nds32/constants.md (unspec_volatile_element): Add enum values
16911 for interrupt control.
16912 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16913 functions for interrupt control.
16914 * config/nds32/nds32-intrinsic.md: Likewise.
16915 * config/nds32/nds32_intrinsic.h: Likewise.
16916 * config/nds32/nds32.h (nds32_builtins): Likewise.
16918 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
16920 * config/nds32/nds32.c (nds32_init_machine_status,
16921 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
16922 strict_aligned_p field.
16923 (nds32_expand_to_rtl_hook): New function.
16924 (TARGET_EXPAND_TO_RTL_HOOK): Define.
16925 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
16927 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16928 Chung-Ju Wu <jasonwucj@gmail.com>
16930 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
16931 * config/nds32/nds32-n7.md: New file.
16932 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
16933 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
16935 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
16936 * config/nds32/nds32.md (pipeline_model): Add n7.
16937 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
16938 * config/nds32/pipelines.md: Include n7 settings.
16940 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16941 Chung-Ju Wu <jasonwucj@gmail.com>
16943 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
16944 * config/nds32/nds32-e8.md: New file.
16945 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
16946 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
16948 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
16949 * config/nds32/nds32.md (pipeline_model): Add e8.
16950 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
16951 * config/nds32/pipelines.md: Include e8 settings.
16953 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16954 Chung-Ju Wu <jasonwucj@gmail.com>
16956 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
16957 * config/nds32/nds32-n8.md: New file.
16958 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
16959 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
16961 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
16962 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
16963 * config/nds32/nds32.md (pipeline_model): Add n8.
16964 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
16965 * config/nds32/pipelines.md: Include n8 settings.
16967 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
16968 Chung-Ju Wu <jasonwucj@gmail.com>
16970 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
16971 * config/nds32/nds32-n9-2r1w.md: New file.
16972 * config/nds32/nds32-n9-3r2w.md: New file.
16973 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
16974 nds32_register_ports): New or modify for cpu n9.
16975 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
16977 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
16978 * config/nds32/nds32-utils.c: New file.
16979 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
16980 TARGET_MUL_SLOW): Define.
16981 * config/nds32/nds32.md (pipeline_model): New attribute.
16982 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
16983 New options that support cpu n9.
16984 * config/nds32/pipelines.md: Include n9 settings.
16985 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
16987 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
16989 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
16990 information if necessary.
16991 (output_cond_branch_compare_zero): Likewise.
16992 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
16993 (nds32_target_alignment): Refine for alignment.
16994 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
16995 (FUNCTION_BOUNDARY): Modify.
16996 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
16998 * config/nds32/nds32.opt (malways-align, malign-functions): New.
17000 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
17002 * config/nds32/constants.md (unspec_volatile_element): Add values for
17003 TLB operation and data prefetch.
17004 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
17005 functions for TLB operation and data prefetch.
17006 * config/nds32/nds32-intrinsic.md: Likewise.
17007 * config/nds32/nds32_intrinsic.h: Likewise.
17008 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
17009 (nds32_print_operand): Likewise.
17010 * config/nds32/nds32.h (nds32_builtins): Likewise.
17012 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
17013 Andrew Pinski <pinsika@gcc.gnu.org>
17015 PR middle-end/82976
17016 * match.pd: Use constant_boolean_node of correct type instead of
17017 boolean_true_node or boolean_false_node for simplifying
17018 pointer comparisons to zero.
17020 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17022 PR tree-optimization/80021
17023 * tree.c (verify_type_variant): Make error call in verify_variant_match
17024 translatable and remove final full stop.
17026 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17028 * config/nds32/constants.md (unspec_volatile_element): Add
17029 UNSPEC_VOLATILE_EH_RETURN.
17030 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
17031 nds32_output_stack_pop): Support dwarf exception handling process.
17032 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
17033 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
17034 exception handling process.
17035 (nds32_compute_stack_frame): Likewise.
17036 (nds32_return_addr_rtx): Likewise.
17037 (nds32_initial_elimination_offset): Likewise.
17038 (nds32_expand_prologue): Likewise.
17039 (nds32_expand_epilogue): Likewise.
17040 (nds32_dynamic_chain_address): New function.
17041 * config/nds32/nds32.h (machine_function): Add fields for dwarf
17042 exception handling.
17043 (DYNAMIC_CHAIN_ADDRESS): Define.
17044 (EH_RETURN_DATA_REGNO): Define.
17045 (EH_RETURN_STACKADJ_RTX): Define.
17046 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
17047 patterns for dwarf exception handling.
17049 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17051 * config/nds32/nds32.h: Clean up obsolete macros.
17053 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17055 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17056 Add enum values for particular instructions.
17057 * config/nds32/nds32-intrinsic.c: Implementation of expanding
17058 particular intrinsic functions.
17059 * config/nds32/nds32-intrinsic.md: Likewise.
17060 * config/nds32/nds32_intrinsic.h: Likewise.
17061 * config/nds32/nds32.h (nds32_builtins): Likewise.
17062 * config/nds32/nds32.md (type): Add pbsad and pbsada.
17063 (btst, ave): New patterns for particular instructions.
17065 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17067 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
17068 Add enum values for atomic load/store and memory sync.
17069 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
17071 * config/nds32/nds32-intrinsic.md: Likewise.
17072 * config/nds32/nds32_intrinsic.h: Likewise.
17073 * config/nds32/nds32.h (nds32_builtins): Likewise.
17075 2018-04-07 Jakub Jelinek <jakub@redhat.com>
17077 PR tree-optimization/85257
17078 * fold-const.c (native_encode_vector): If not all elts could fit
17079 and off is -1, return 0 rather than offset.
17080 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
17081 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
17082 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
17083 adjust buffer in native_interpret_expr call.
17085 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17087 * config/nds32/constants.md (unspec_volatile_element): Add cache
17088 control enum values.
17089 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
17090 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
17091 * config/nds32/nds32.c (nds32_cctl_names): New.
17092 (nds32_print_operand): Handle cache control register names.
17093 * config/nds32/nds32.h (nds32_builtins): New enum values.
17094 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
17096 * config/nds32/nds32.md (type): Add mmu.
17097 * config/nds32/pipelines.md (simple_insn): Add mmu.
17099 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17101 * config/nds32/nds32.md (type): Remove call.
17102 * config/nds32/pipelines.md (simple_insn): Likewise.
17104 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17106 * config/nds32/constants.md (unspec_volatile_element): Add
17107 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17108 UNSPEC_VOLATILE_FMFCFG.
17109 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17110 description for fmfcfg and fmfcsr.
17111 (bdesc_1arg): Add fmtcsr.
17112 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17113 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17114 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17115 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17116 unspec_fmfcfg): New patterns.
17117 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17118 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17119 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17120 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17121 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17122 __nds32__fmfcfg): Define.
17124 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
17126 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17127 intrinsic register names.
17128 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17129 intrinsic register enum values and macros.
17131 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
17133 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17134 for load/store addressing form.
17135 (nds32_print_operand_address): Likewise.
17137 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
17140 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17141 based on LABEL_REF. Remove useless assertion.
17142 (pic_address_needs_scratch): Fix formatting.
17143 (sparc_legitimize_pic_address): Minor tweaks.
17144 (sparc_delegitimize_address): Adjust assertion accordingly.
17145 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17146 into symbolic_operand.
17147 (movsi_high_pic_label_ref): Likewise.
17148 (movsi_lo_sum_pic_label_ref): Likewise.
17149 (movdi_pic_label_ref): Likewise.
17150 (movdi_high_pic_label_ref): Likewise.
17151 (movdi_lo_sum_pic_label_ref): Likewise.
17153 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
17155 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17156 custom LIB_SPEC setup.
17158 2018-04-06 Ruslan Bukin <br@bsdpad.com>
17159 Kito Cheng <kito.cheng@gmail.com>
17161 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17162 * config/riscv/freebsd.h: New.
17164 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17166 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17167 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17170 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17171 Kito Cheng <kito.cheng@gmail.com>
17173 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17174 nds32_output_call, nds32_symbol_binds_local_p): New functions.
17175 * config/nds32/nds32-protos.h (nds32_output_call,
17176 nds32_output_return): Declare.
17177 * config/nds32/nds32.md: Refine all the call and return patterns.
17179 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17182 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17183 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17185 PR rtl-optimization/84872
17186 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17187 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17188 EDGE_CROSSING edge.
17190 2018-04-06 Tamar Christina <tamar.christina@arm.com>
17192 * expr.c (copy_blkmode_to_reg): Revert 254862.
17193 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17195 2018-04-06 Richard Biener <rguenther@suse.de>
17197 PR middle-end/85244
17198 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17199 after seeing a component reference with an adjacent field. Treat
17200 refs to arrays at struct end of external decls similar to
17201 refs to unconstrained commons.
17203 2018-04-06 Jakub Jelinek <jakub@redhat.com>
17206 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17207 look through SAVE_EXPRs with non-side-effects argument. Adjust
17209 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17212 2018-04-06 Richard Biener <rguenther@suse.de>
17214 PR middle-end/85180
17215 * alias.c (find_base_term): New wrapper around find_base_term
17216 unwinding CSELIB_VAL_PTR changes.
17217 (find_base_term): Do not restore CSELIB_VAL_PTR during the
17220 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
17222 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17224 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17225 constant definitions.
17226 ("nop"): lr 0,0 -> nopr r0
17227 ("nop_lr0", "nop_lr1"): New insn definitions.
17229 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
17231 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17232 NDS32_V3PUSH_AVAILABLE_P macro.
17234 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
17235 Chung-Ju Wu <jasonwucj@gmail.com>
17237 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
17238 (nds32*-*-*): Add float and fpu_config into supported_defaults.
17239 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
17240 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
17241 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
17242 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
17243 * config/nds32/constraints.md: New constraints and checking for hard
17244 float configuration.
17245 * config/nds32/iterators.md: New mode iterator and attribute for hard
17246 float configuration.
17247 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
17249 * config/nds32/nds32-fpu.md: New file.
17250 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
17251 deal with hard float code generation.
17252 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
17254 (abi_type, float_reg_number): New enum type.
17255 * config/nds32/nds32-predicates.c: New predicates for hard float.
17256 * config/nds32/nds32-protos.h: Declare functions for hard float.
17257 * config/nds32/nds32.c: Implementation for hard float configuration.
17258 * config/nds32/nds32.h: Definitions for hard float configuration.
17259 * config/nds32/nds32.md: Include hard float machine description and
17260 modify patterns for hard float configuration.
17261 * config/nds32/nds32.opt: New options for hard float configuration.
17262 * config/nds32/predicates.md: New predicates for hard float
17265 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17267 * common/config/nds32/nds32-common.c
17268 (nds32_option_optimization_table): Enable -mreleax-hint by default.
17270 2018-04-05 Jakub Jelinek <jakub@redhat.com>
17272 PR middle-end/85195
17273 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
17274 CONSTRUCTOR_ELT (ctor, ...)->value.
17276 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
17279 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
17281 2018-04-05 Tom de Vries <tom@codesourcery.com>
17284 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
17287 2018-04-05 Shiva Chen <shiva0217@gmail.com>
17288 Kito Cheng <kito.cheng@gmail.com>
17290 * config/nds32/constraints.md (U33): Fine-tune checking condition.
17291 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
17292 * config/nds32/nds32.h (nds32_16bit_address_type): Add
17293 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
17295 2018-04-05 Shiva Chen <shiva0217@gmail.com>
17296 Kito Cheng <kito.cheng@gmail.com>
17298 * config/nds32/constraints.md (Ufe): New memory constraint.
17299 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
17300 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
17301 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
17303 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
17304 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
17306 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17308 * config/nds32/nds32.md: Use optimize_size in the condition for
17309 alu-shift instructions.
17311 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17313 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
17315 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17317 * config/nds32/nds32.md (negsi2): Refine pattern.
17319 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
17320 Chung-Ju Wu <jasonwucj@gmail.com>
17322 * config/nds32/iterators.md (shift_rotate): New code iterator.
17323 (shift): New code attribute.
17324 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
17325 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
17326 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
17327 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
17328 bit-wise operations.
17329 (andsi3, *andsi3): Ditto.
17330 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
17331 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
17332 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
17333 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
17334 nds32_ior_operand, nds32_xor_operand): New predicates.
17336 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17338 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
17339 (addsi3, subsi3): ... this.
17341 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17343 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
17345 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17347 * config/nds32/nds32.md: Adjust indention.
17349 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
17351 * config/nds32/nds32.md (feature): New attribute.
17353 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
17355 * config/nds32/nds32.md (subtype): New attribute.
17357 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
17360 * config/arm/arm-builtins.c (arm_expand_builtin): Change
17361 expansion to perform a bitwise AND of the argument followed by a
17362 boolean negation of the result.
17364 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
17366 PR rtl-optimization/84878
17367 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
17368 the basic block. Assert the use reference is not artificial and that
17369 it has an associated insn.
17371 2018-04-04 Michael Matz <matz@suse.de>
17373 * builtins.c (compute_objsize): Pass correct operand
17374 to array_at_struct_end_p.
17376 2018-04-04 Richard Biener <rguenther@suse.de>
17379 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
17380 from contexts for DINFO_LEVEL_TERSE and below.
17382 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17384 * config/nds32/nds32-doubleword.md (move_<mode>): Require
17385 resiter_operand condition.
17386 * config/nds32/nds32.md (*move<mode>): Ditto.
17388 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17389 Monk Chiang <sh.chiang04@gmail.com>
17391 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
17393 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17395 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
17397 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17398 Kito Cheng <kito.cheng@gmail.com>
17400 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
17401 nds32_cond_code_str, output_cond_branch,
17402 output_cond_branch_compare_zero, nds32_expand_cbranch,
17403 nds32_expand_cstore, nds32_expand_movcc,
17404 nds32_output_cbranchsi4_equality_zero,
17405 nds32_output_cbranchsi4_equality_reg,
17406 nds32_output_cbranchsi4_equality_reg_or_const_int,
17407 nds32_output_cbranchsi4_greater_less_zero: New functions.
17408 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
17409 nds32_expand_cstore, nds32_expand_movcc,
17410 nds32_output_cbranchsi4_equality_zero,
17411 nds32_output_cbranchsi4_equality_reg,
17412 nds32_output_cbranchsi4_equality_reg_or_const_int,
17413 nds32_output_cbranchsi4_greater_less_zero): Declare.
17414 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
17415 nds32_rimm11s_operand): New predicates.
17416 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
17417 * config/nds32/nds32.md: Rewrite all the branch and conditional move
17420 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
17422 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
17423 * config/nds32/nds32.md: Ditto.
17424 * config/nds32/pipelines.md: Ditto.
17426 2018-04-04 Richard Biener <rguenther@suse.de>
17428 PR tree-optimization/85168
17429 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
17430 propagating abnormals.
17432 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17434 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
17436 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
17437 Kito Cheng <kito.cheng@gmail.com>
17439 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
17440 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
17441 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
17442 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
17443 * config/nds32/nds32.md (sibcall_internal): New.
17444 (sibcall_register): Remove.
17445 (sibcall_immediate): Remove.
17446 (sibcall_value_internal): New.
17447 (sibcall_value_register): Remove.
17448 (sibcall_value_immediate): Remove.
17449 * config/nds32/predicates.md (nds32_general_register_operand): New.
17450 (nds32_call_address_operand): New.
17452 2018-04-03 Jakub Jelinek <jakub@redhat.com>
17454 PR rtl-optimization/85167
17455 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
17456 bb_defs if *split_p, instead preinitialize it to NULL.
17458 PR tree-optimization/85156
17459 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
17460 evaluating the argument multiple times.
17462 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
17464 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
17466 (_mm_cvtpd_ps): Likewise.
17467 (_mm_cvttpd_epi32): Likewise.
17468 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
17469 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
17470 vector, pixel, and bool following altivec.h include.
17472 2018-04-03 Martin Sebor <msebor@redhat.com>
17474 * doc/extend.texi (Common Function Attributes): Clarify.
17475 (const attribute): Likewise.
17476 (pure attribute): Likewise.
17478 2018-04-03 Jakub Jelinek <jakub@redhat.com>
17481 * config/i386/i386.c (ix86_expand_vector_set): Use
17482 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
17484 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
17486 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
17487 instructions when changing rounding bits to preserve precision bits
17488 in the x87 control word.
17490 2018-04-03 Martin Liska <mliska@suse.cz>
17492 PR tree-optimization/82491
17493 * rtl.h (strip_offset_and_add): Replace += suboffset with
17494 poly_uint64 () + suboffset.
17496 2018-03-29 Martin Liska <mliska@suse.cz>
17497 Martin Jambor <mjambor@suse.cz>
17500 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
17501 param_type is not an integral or pointer type.
17503 2018-04-03 Richard Biener <rguenther@suse.de>
17505 * sese.h (recompute_all_dominators): Remove.
17507 2018-04-02 Martin Sebor <msebor@redhat.com>
17509 * doc/invoke.texi (-Wrestrict): Fix typos.
17511 2018-04-02 Jim Wilson <jimw@sifive.com>
17513 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
17514 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
17515 (<optab>di3, <optab>si3_extend): Likewise.
17516 (<optab>si3_mask, <optab>si3_mask_1): New.
17517 (<optab>di3_mask, <optab>di3_mask_1): New.
17518 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
17519 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
17520 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
17522 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
17524 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
17527 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
17529 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
17530 (nds32_canonicalize_comparison): New function.
17532 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17533 Kito Cheng <kito.cheng@gmail.com>
17534 Kuan-Lin Chen <kuanlinchentw@gmail.com>
17536 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
17537 * config/nds32/constants.md (unspec_volatile_element): Add
17538 UNSPEC_VOLATILE_RELAX_GROUP.
17539 * config/nds32/nds32-relax-opt.c: New file.
17540 * config/nds32/nds32-predicates.c
17541 (nds32_symbol_load_store_p): New function.
17542 * config/nds32/nds32-protos.h
17543 (nds32_symbol_load_store_p): Declare function.
17544 (make_pass_nds32_relax_opt): Declare new rtl pass function.
17545 * config/nds32/nds32.c
17546 (nds32_register_pass): New function to register pass.
17547 (nds32_register_passes): New function to register passes.
17548 * config/nds32/nds32.md (relax_group): New pattern.
17549 * config/nds32/nds32.opt (mrelax-hint): New option.
17550 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
17552 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
17554 * config/nds32/t-nds32: Modify files dependency.
17556 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17558 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
17559 (PROFILE_HOOK): Define its implementation.
17561 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
17563 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
17564 type and 32-bit size.
17566 2018-04-01 Jakub Jelinek <jakub@redhat.com>
17568 PR middle-end/85090
17569 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
17570 (V_128_256): New mode iterator.
17571 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
17572 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
17573 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
17575 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
17576 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
17578 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
17581 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
17582 NaN inputs correctly.
17584 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
17587 * config/rs6000/vsx.md (??r): New mode attribute.
17588 (*vsx_mov<mode>_64bit): Use it.
17589 (*vsx_mov<mode>_32bit): Likewise.
17591 2018-03-30 Martin Sebor <msebor@redhat.com>
17593 PR tree-optimization/84818
17594 * builtins.c (check_access): Use warning_n.
17596 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
17599 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
17601 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
17604 2018-03-30 Julia Koval <julia.koval@intel.com>
17607 * x86-tune.def (movx, partial_reg_dependency): Enable for
17610 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
17612 PR inline-asm/84985
17613 * lra-constraints.c (process_alt_operands): Move setting
17614 this_alternative_matches below.
17616 2018-03-29 Martin Liska <mliska@suse.cz>
17619 * doc/invoke.texi: Document how LTO works with debug info.
17620 Describe auto-load support of binutils. Mention 'x86-64'
17621 as valid option value of -march option.
17623 2018-03-29 Jakub Jelinek <jakub@redhat.com>
17625 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
17628 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
17629 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
17630 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
17633 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17636 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
17637 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
17638 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
17639 for RS6000_BTM_POWERPC64.
17640 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
17641 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
17642 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
17645 (DIVDEU): Likewise.
17647 2018-03-28 Carl Love <cel@us.ibm.com>
17650 2017-09-27 Carl Love <cel@us.ibm.com>
17652 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
17653 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
17654 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
17657 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
17659 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
17660 instead of __vector bool.
17661 (_mm_max_pu8): Likewise.
17662 (_mm_min_pi16): Likewise.
17664 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
17667 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
17668 (DIVWEUO): Likewise.
17669 (DIVDEO): Likewise.
17670 (DIVDEUO): Likewise.
17671 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
17672 DIVWEUO and DIVDEUO.
17673 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
17674 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
17675 (div_extend): Likewise.
17676 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
17678 (__builtin_divweuo): Likewise.
17679 (__builtin_divdeo): Likewise.
17680 (__builtin_divdeuo): Likewise.
17682 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17685 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
17686 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
17688 PR tree-optimization/82004
17689 * gimple-match-head.c (optimize_pow_to_exp): New function.
17690 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
17691 Don't fold to exp if optimize_pow_to_exp is false.
17693 2018-03-28 Martin Liska <mliska@suse.cz>
17696 * calls.c (initialize_argument_information): Fix trailing space.
17697 * common.opt: Fix typo and provide better explanation for
17698 -fsanitize-coverage option.
17699 * config/i386/i386.opt: Fix typo.
17701 2018-03-28 Jakub Jelinek <jakub@redhat.com>
17702 Martin Liska <mliska@suse.cz>
17705 * gimplify.c (asan_poison_variable): Don't do the check for
17706 gimplify_omp_ctxp here.
17707 (gimplify_decl_expr): Do it here.
17708 (gimplify_target_expr): Likewise.
17710 2018-03-28 Martin Liska <mliska@suse.cz>
17713 * config/i386/i386.c (ix86_function_arg_advance): Do not call
17714 chkp_type_bounds_count if MPX is not enabled.
17716 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
17718 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
17720 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
17723 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
17724 function to create the function decl for complex long double
17725 multiply and divide for -mabi=ieeelongdouble.
17726 (init_float128_ieee): Call it.
17728 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
17731 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
17732 -fcf-protection=branch -mibt.
17733 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
17735 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17738 * config/arm/arm.c (arm_valid_symbolic_address): Handle
17739 arm_word_relocations.
17741 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
17744 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
17745 extern array declarations.
17747 2018-03-27 Richard Biener <rguenther@suse.de>
17749 PR middle-end/84067
17750 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
17751 explicit single_use checks.
17753 2018-03-27 Richard Biener <rguenther@suse.de>
17755 PR tree-optimization/85082
17756 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
17759 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
17761 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
17762 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
17763 Turn on fasynchronous-unwind-tables and funwind-tables.
17765 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
17768 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
17769 (*bmi_blsr_<mode>_ccz): Ditto.
17771 2018-03-26 Tom de Vries <tom@codesourcery.com>
17773 PR tree-optimization/85063
17774 * omp-general.c (offloading_function_p): New function. Factor out
17776 * omp-offload.c (pass_omp_target_link::gate): ... here.
17777 * omp-general.h (offloading_function_p): Declare.
17778 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
17779 with attribute omp declare target for offloading functions.
17781 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
17783 PR tree-optimization/84005
17784 * tree-data-ref.h (get_base_for_alignment): Declare.
17785 * tree-data-ref.c (get_base_for_alignment_1): New function.
17786 (get_base_for_alignment): Likewise.
17787 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
17788 get_base_for_alignment to find a suitable base object, instead
17789 of always using drb->base_address.
17791 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17793 PR inline-asm/85022
17794 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
17795 known size by default.
17797 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
17799 PR inline-asm/85030
17800 * lra-constraints.c (process_alt_operands): Don't match BLKmode
17801 and non BLKmode operands.
17803 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17806 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
17807 Clean up attributes.
17809 2018-03-23 Richard Biener <rguenther@suse.de>
17812 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
17813 we are going to emit early debug for LTO.
17815 2018-03-23 Jakub Jelinek <jakub@redhat.com>
17817 PR inline-asm/85034
17818 * function.c (match_asm_constraints_1): Don't optimize if input
17819 doesn't satisfy general_operand predicate for output's mode.
17821 PR inline-asm/85022
17822 * alias.c (write_dependence_p): Don't require for x_canonicalized
17823 non-VOIDmode if x has VOIDmode.
17826 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
17827 just don't try to optimize it rather than assert it never happens.
17829 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17831 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
17832 macro expansions for definition of ST_INTERNAL_<mode> and
17833 LD_INTERNAL_<mode> builtins.
17834 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
17836 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
17838 (altivec_expand_st_builtin): Likewise.
17839 (altivec_expand_builtin): Remove calls to deleted functions.
17840 (rs6000_address_for_altivec): Delete this function.
17841 * config/rs6000/vector.md: Remove expands for
17842 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
17844 2018-03-22 Sudakshina Das <sudi.das@arm.com>
17847 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
17848 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
17849 re-computing once computed.
17850 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
17851 (arm_init_machine_status): Initialize
17852 machine->static_chain_stack_bytes.
17854 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
17857 * doc/extend.texi: Add four new prototypes for vec_ld.
17858 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
17859 definitions for more logical presentation.
17860 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
17861 entries for V1TI variants of __builtin_altivec_ld builtin.
17862 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
17863 handling of V1TI variant of LVX icode pattern.
17864 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
17865 (rs6000_gimple_fold_builtin): Likewise.
17866 (altivec_init_builtins): Add code to define
17867 __builtin_altivec_lvx_v1ti function.
17869 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17871 PR inline-asm/84941
17872 * function.c (match_asm_constraints_1): Don't do the optimization
17873 if input isn't a REG, SUBREG, MEM or constant.
17875 2018-03-22 Tom de Vries <tom@codesourcery.com>
17877 PR tree-optimization/84956
17878 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
17879 bb_has_abnormal_pred.
17881 2018-03-22 Jakub Jelinek <jakub@redhat.com>
17884 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
17885 DECL_INITIAL (decl) to decl at the end.
17886 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
17887 adjust the comment.
17889 2018-03-21 Joseph Myers <joseph@codesourcery.com>
17891 * doc/extend.texi (__builtin_tgmath): Document when complex
17892 integer types are treated as _Complex _Float64.
17894 2018-03-21 Tom de Vries <tom@codesourcery.com>
17896 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
17898 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17900 PR tree-optimization/84960
17901 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
17902 if it is ENTRY block, move them into single succ of ENTRY in that case.
17904 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
17906 PR tree-optimization/84811
17907 * poly-int.h (poly_span_traits): Remove the T3 parameter and
17908 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
17909 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
17910 (known_subrange_p): Update accordingly. Cast each value involved
17911 in the size comparison, rather than casting the result of the
17914 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17916 PR tree-optimization/84982
17917 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
17918 by flipping the least significant bit rather than all bits from
17919 bitpos to bitpos + bitsize - 1.
17921 2018-03-21 Nathan Sidwell <nathan@acm.org>
17923 * doc/extend.texi (Deprecated Features): Remove mention of
17924 long-deleted deprecations.
17926 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
17929 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
17930 * configure: Regenerate.
17932 2018-03-21 Tom de Vries <tom@codesourcery.com>
17934 PR tree-optimization/83126
17935 * tree-parloops.c (num_phis): New function.
17936 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
17938 2018-03-21 Nathan Sidwell <nathan@acm.org>
17940 * doc/extend.texi (Deprecated Features): Update deprecated flags,
17941 mention anon-struct/union members and trailing attributes.
17943 2018-03-21 Bin Cheng <bin.cheng@arm.com>
17945 PR tree-optimization/84969
17946 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
17947 builtin memset partitions if they set different rhs values.
17949 2018-03-21 Jakub Jelinek <jakub@redhat.com>
17951 PR rtl-optimization/84989
17952 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
17953 VEC_DUPLICATE with scalar result mode.
17955 2018-03-21 Martin Liska <mliska@suse.cz>
17958 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
17959 not intended return statement.
17961 2018-03-21 Martin Liska <mliska@suse.cz>
17964 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
17965 (chkp_find_bound_slots_1): Limit number of iterations.
17967 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
17970 * Minor grammar fixes for x86 options.
17972 2018-03-20 Jakub Jelinek <jakub@redhat.com>
17975 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
17976 holding REG_CFA_RESTORE notes, instead turn them into a USE.
17978 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
17981 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
17982 (altivec_lvx_<mode>_1op): Likewise.
17983 (altivec_stvx_<mode>_2op): Likewise.
17984 (altivec_stvx_<mode>_1op): Likewise.
17985 (altivec_lvx_<VM2:mode>): New define_expand.
17986 (altivec_stvx_<VM2:mode>): Likewise.
17987 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
17988 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17989 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
17990 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17991 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
17992 (rs6000_gen_lvx): Likewise.
17993 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
17994 (altivec_expand_stv_builtin): Likewise.
17995 (altivec_expand_builtin): Likewise.
17996 * config/rs6000/vector.md: Likewise.
17998 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18001 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
18004 2018-03-20 Richard Biener <rguenther@suse.de>
18007 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
18008 sign-conversions as zero, fall back to standard scalar_stmt
18011 2018-03-20 Martin Liska <mliska@suse.cz>
18014 * predict.c (rebuild_frequencies): Handle case when we have
18015 PROFILE_ABSENT, but flag_guess_branch_prob is false.
18017 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18020 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
18021 flag_section_anchors.
18022 * varasm.c (use_blocks_for_decl_p): Remove hack for
18023 dw2_force_const_mem.
18026 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
18028 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
18029 be created, use lowpart_subreg of operands[0] rather than operands[0]
18031 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
18032 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
18033 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
18034 and n constraint instead of aarch64_shift_imm_di and Usd.
18035 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
18036 (*aarch64_<optab>_reg_minus<mode>3): ... this.
18038 2018-03-20 Sudakshina Das <sudi.das@arm.com>
18041 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
18042 to favor GPR over NEON registers.
18043 (<shift>di3_neon): Likewise.
18045 2018-03-20 Tom de Vries <tom@codesourcery.com>
18048 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
18049 (nvptx_process_pars): Emit bar.sync asap and alap.
18051 2018-03-20 Tom de Vries <tom@codesourcery.com>
18054 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
18055 seen_label if seen_label is already set.
18057 2018-03-20 Jakub Jelinek <jakub@redhat.com>
18060 * config/i386/i386.c (fold_builtin_cpu): For features above 31
18061 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
18062 Use 1U instead of 1. Formatting fixes.
18065 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
18066 instead of TREE_TYPE (s1) for the return value.
18068 2018-03-19 Jakub Jelinek <jakub@redhat.com>
18070 PR tree-optimization/84946
18071 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
18072 bitsize + bitsize in poly_uint64 rather than poly_int64.
18075 * dwarf2asm.c: Include fold-const.c.
18076 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
18077 of decl rather than decl itself.
18079 PR rtl-optimization/84643
18080 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
18082 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
18085 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
18086 calling assemble_variable.
18088 2018-03-19 Sudakshina Das <sudi.das@arm.com>
18091 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
18092 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
18094 2018-03-19 Jim Wilson <jimw@sifive.com>
18097 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18098 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18099 (riscv_first_stack_step): Likewise.
18100 (riscv_option_override): Use STACK_BOUNDARY instead of
18101 MIN_STACK_BOUNDARY.
18102 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18103 MIN_STACK_BOUNDARY.
18104 (BIGGEST_ALIGNMENT): Set to 128.
18105 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18106 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18109 2018-03-19 Richard Biener <rguenther@suse.de>
18111 PR tree-optimization/84933
18112 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18113 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18115 2018-03-19 Richard Biener <rguenther@suse.de>
18117 PR tree-optimization/84859
18118 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18119 (cond_if_else_store_replacement): Perform sinking operation on
18120 single-store BBs regardless of MAX_STORES_TO_SINK setting.
18121 Generalize what a BB with a single eligible store is.
18123 2018-03-19 Richard Biener <rguenther@suse.de>
18125 PR tree-optimization/84929
18126 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18127 chrec_is_positive against non-chrec arg.
18129 2018-03-19 Tamar Christina <tamar.christina@arm.com>
18132 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18134 2018-03-18 Martin Liska <mliska@suse.cz>
18136 PR rtl-optimization/84635
18137 * regrename.c (build_def_use): Use matches_mode only when
18140 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
18142 PR tree-optimization/84913
18143 * tree-vect-loop.c (vectorizable_reduction): Don't try to
18144 vectorize chains of COND_EXPRs.
18146 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18148 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18150 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18152 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18154 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
18156 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18158 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18159 Kito Cheng <kito.cheng@gmail.com>
18161 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18162 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18163 (nds32_adjust_reg_alloc_order): New function.
18164 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18166 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
18168 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18169 nds32_print_operand, nds32_print_operand_address): Use
18170 HOST_WIDE_INT_PRINT_DEC instead.
18172 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
18174 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18176 2018-03-17 Jakub Jelinek <jakub@redhat.com>
18179 * config/i386/i386.c (initial_ix86_tune_features,
18180 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18181 unsigned long long.
18182 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18183 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18184 rather than 1u << ix86_tune. Formatting fix.
18185 (ix86_option_override_internal): Change ix86_arch_mask from
18186 unsigned int to unsigned HOST_WIDE_INT, initialize to
18187 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18188 (ix86_function_specific_restore): Likewise.
18190 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18193 * postreload.c (reload_combine_recognize_pattern): Perform
18194 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18195 truncate_int_for_mode the result for the destination's mode.
18198 * hsa-gen.c (mem_type_for_type): Fix comment typo.
18199 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18201 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18204 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
18207 * lra-assigns.c (lra_split_hard_reg_for): Don't use
18208 regno_allocno_class_array and sorted_pseudos.
18209 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18210 insns where regno is used.
18212 2018-03-16 Martin Liska <mliska@suse.cz>
18215 * multiple_target.c (create_dispatcher_calls): Redirect
18216 reference in the symbol table.
18218 2018-03-16 Martin Liska <mliska@suse.cz>
18221 * multiple_target.c (create_dispatcher_calls): Redirect also
18224 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18228 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18229 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18230 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18232 2018-03-16 Julia Koval <julia.koval@intel.com>
18234 * doc/invoke.texi (Skylake Server): Add CLWB.
18235 Cannonlake): Remove CLWB.
18237 2018-03-16 Jakub Jelinek <jakub@redhat.com>
18239 PR tree-optimization/84841
18240 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
18242 (FLOAT_ONE_CONST_TYPE): Define.
18243 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
18244 (sort_by_operand_rank): Put entries with higher constant_type last
18245 rather than first to match comments.
18247 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
18249 * config/nios2/nios2.md (movsi_internal): Fix thinko in
18252 2018-03-15 Jakub Jelinek <jakub@redhat.com>
18255 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
18256 check and use address of target always.
18258 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
18261 * config/i386/i386.c (indirect_thunk_needed): Update comments.
18262 (indirect_thunk_bnd_needed): Likewise.
18263 (indirect_thunks_used): Likewise.
18264 (indirect_thunks_bnd_used): Likewise.
18265 (indirect_return_needed): New.
18266 (indirect_return_bnd_needed): Likewise.
18267 (output_indirect_thunk_function): Add a bool argument for
18269 (output_indirect_thunk_function): Don't generate alias for
18270 function return thunk.
18271 (ix86_code_end): Call output_indirect_thunk_function to generate
18272 function return thunks.
18273 (ix86_output_function_return): Set indirect_return_bnd_needed
18274 and indirect_return_needed instead of indirect_thunk_bnd_needed
18275 and indirect_thunk_needed.
18277 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
18279 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
18280 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
18281 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
18283 2018-03-15 David Malcolm <dmalcolm@redhat.com>
18284 Paul Hua <paul.hua.gm@gmail.com>
18287 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
18289 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
18291 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
18292 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
18293 resp. SFmode cases.
18295 2018-03-15 Tamar Christina <tamar.christina@arm.com>
18298 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
18299 instead of GET_MODE_SIZE when comparing Units.
18301 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
18304 * varasm.c (hash_section): Return an unchangeble hash value
18305 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
18306 Return !aarch64_can_use_per_function_literal_pools_p ().
18308 2018-03-15 Jakub Jelinek <jakub@redhat.com>
18311 * optabs.c (emit_conditional_move): Pass address of cmode's copy
18312 rather than address of cmode as last argument to prepare_cmp_insn.
18314 2018-03-15 Julia Koval <julia.koval@intel.com>
18316 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
18317 F_AVX512VNNI, F_AVX512BITALG): New.
18319 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
18322 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
18323 insn for floating-point loads and stores.
18325 2018-03-14 Carl Love <cel@us.ibm.com>
18327 * config/rs6000/rs6000-c.c: Add macro definitions for
18328 ALTIVEC_BUILTIN_VEC_PERMXOR.
18329 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
18330 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
18331 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
18332 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
18334 * config/doc/extend.texi: Add prototypes for vec_permxor.
18336 2018-03-14 David Malcolm <dmalcolm@redhat.com>
18339 * diagnostic-show-locus.c (class layout_point): Convert m_line
18340 from int to linenum_type.
18341 (line_span::comparator): Use linenum "compare" function when
18342 comparing line numbers.
18343 (test_line_span): New function.
18344 (layout_range::contains_point): Convert param "row" from int to
18346 (layout_range::intersects_line_p): Likewise.
18347 (layout::will_show_line_p): Likewise.
18348 (layout::print_source_line): Likewise.
18349 (layout::should_print_annotation_line_p): Likewise.
18350 (layout::print_annotation_line): Likewise.
18351 (layout::print_leading_fixits): Likewise.
18352 (layout::annotation_line_showed_range_p): Likewise.
18353 (struct line_corrections): Likewise for field m_row.
18354 (line_corrections::line_corrections): Likewise for param "row".
18355 (layout::print_trailing_fixits): Likewise.
18356 (layout::get_state_at_point): Likewise.
18357 (layout::get_x_bound_for_row): Likewise.
18358 (layout::print_line): Likewise.
18359 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
18360 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
18361 * input.c (selftest::test_linenum_comparisons): New function.
18362 (selftest::input_c_tests): Call it.
18363 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
18364 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
18365 * selftest.h (ASSERT_GT): New macro.
18366 (ASSERT_GT_AT): New macro.
18367 (ASSERT_LT): New macro.
18368 (ASSERT_LT_AT): New macro.
18370 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
18372 PR rtl-optimization/84780
18373 * combine.c (distribute_links): Don't make a link based on pc_rtx.
18375 2018-03-14 Martin Liska <mliska@suse.cz>
18377 * tree.c (record_node_allocation_statistics): Use
18378 get_stats_node_kind.
18379 (get_stats_node_kind): New function extracted from
18380 record_node_allocation_statistics.
18381 (free_node): Use get_stats_node_kind.
18383 2018-03-14 Richard Biener <rguenther@suse.de>
18385 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
18386 that the value-set of ANTIC_IN doesn't grow.
18389 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18391 (BB_VISITED_WITH_VISITED_SUCCS): New define.
18392 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18394 2018-03-14 Julia Koval <julia.koval@intel.com>
18396 * config.gcc (icelake-client, icelake-server): New.
18398 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
18399 (initial_ix86_arch_features): Ditto.
18400 (PTA_SKYLAKE): Add SGX.
18401 (PTA_ICELAKE): Remove.
18402 (PTA_ICELAKE_CLIENT): New.
18403 (PTA_ICELAKE_SERVER): New.
18404 (ix86_option_override_internal): Split up icelake on icelake client and
18406 (get_builtin_code_for_version): Ditto.
18407 (fold_builtin_cpu): Ditto.
18408 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
18409 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
18410 * config/i386/i386.h (processor_type): Ditto.
18411 * doc/invoke.texi: Ditto.
18413 2018-03-14 Jakub Jelinek <jakub@redhat.com>
18416 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
18417 INTEGER_CST offset, add it together with bitpos / 8 and
18418 sign extend based on POINTER_SIZE.
18422 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
18425 * config/i386/constraints.md (Yc): New register constraint.
18426 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
18427 Use Yc constraint for alternative 2 of operand 0. Remove
18428 preferred_for_speed attribute.
18430 2018-03-14 Richard Biener <rguenther@suse.de>
18432 PR tree-optimization/84830
18433 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
18434 with the old one to avoid oscillations.
18436 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
18439 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
18441 (assign_by_spills): Return a flag of reload assignment failure.
18442 Do not process the reload assignment failures. Do not spill other
18443 reload pseudos if they has the same reg class. Update n if
18445 (lra_assign): Add a return arg. Set up from the result of
18446 assign_by_spills call.
18447 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18448 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18449 usage_insns if it is not NULL.
18450 (spill_hard_reg_in_range): New function.
18451 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18452 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18453 function prototypes.
18454 (lra_assign): Change prototype.
18455 * lra.c (lra): Add code to deal with fails by splitting hard reg
18458 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
18460 * config/riscv/riscv.opt (mrelax): New option.
18461 * config/riscv/riscv.c (riscv_file_start): Emit ".option
18462 "norelax" when riscv_mrelax is disabled.
18463 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
18465 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
18468 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
18469 reassociation for int modes.
18471 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18473 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
18474 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
18476 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
18477 * config/aarch64/aarch64-sve.md
18478 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
18479 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
18480 (*extend<mode><Vwide>2): Rename to...
18481 (aarch64_sve_extend<mode><Vwide>2): ...this.
18482 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
18483 renaming the old pattern to...
18484 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
18486 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
18487 define_expand, renaming the old pattern to...
18488 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
18489 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
18490 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
18491 account when deciding which SVE instruction the optab should use.
18492 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
18494 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18496 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
18497 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
18498 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
18499 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
18500 (tlsdesc_small_<mode>): Turn a define_expand and use
18501 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
18502 (tlsdesc_small_advsimd_<mode>): ...this.
18503 (tlsdesc_small_sve_<mode>): New pattern.
18505 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
18507 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
18508 (UNSPEC_UMUL_HIGHPART): New constants.
18509 (MUL_HIGHPART): New int iteraor.
18510 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
18511 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
18513 (*<su>mul<mode>3_highpart): New define_insn.
18515 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
18518 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18521 2018-03-13 Martin Liska <mliska@suse.cz>
18524 * (sem_item_optimizer::sem_item_optimizer): Initialize new
18526 (sem_item_optimizer::~sem_item_optimizer): Release it.
18527 (sem_item_optimizer::merge_classes): Register variable aliases.
18528 (sem_item_optimizer::fixup_pt_set): New function.
18529 (sem_item_optimizer::fixup_points_to_sets): Likewise.
18530 * ipa-icf.h: Declare new variables and functions.
18532 2018-03-13 Jakub Jelinek <jakub@redhat.com>
18534 PR middle-end/84834
18535 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
18536 integer_pow2p@2 and test integer_pow2p in condition.
18537 (A < 0 ? C : 0): Similarly for @1.
18539 PR middle-end/84831
18540 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
18541 characters starting at p contain '\0' character, don't look beyond
18545 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
18546 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
18549 * reg-stack.c (change_stack): Change update_end var from int to
18550 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
18551 also call set_block_for_insn on the newly added insns and rescan.
18554 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
18555 on the last operand.
18558 * tree.c (stabilize_reference_1): Return save_expr (e) for
18559 STATEMENT_LIST even if it doesn't have side-effects.
18561 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
18563 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
18565 2018-03-12 Renlin Li <renlin.li@arm.com>
18567 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
18568 aarch64_output_scalar_simd_mov_immediate.
18570 2018-03-12 Martin Sebor <msebor@redhat.com>
18572 PR tree-optimization/83456
18573 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
18574 for perfectly overlapping calls to memcpy.
18575 (gimple_fold_builtin_memory_chk): Same.
18576 (gimple_fold_builtin_strcpy): Handle no-warning.
18577 (gimple_fold_builtin_stxcpy_chk): Same.
18578 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
18580 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
18582 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
18583 parameter. Use it for SFmode.
18584 (rs6000_function_arg_advance_1): Adjust.
18585 (rs6000_function_arg): Adjust.
18586 (rs6000_gimplify_va_arg): Pass false for that new parameter.
18588 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
18590 PR rtl-optimization/84169
18591 PR rtl-optimization/84780
18592 * combine.c (can_combine_p): Check for a 2-insn combination whether
18593 the destination register is used between the two insns, too.
18595 2018-03-12 Richard Biener <rguenther@suse.de>
18597 PR tree-optimization/84803
18598 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
18599 for refs DR analysis didn't process.
18601 2018-03-12 Richard Biener <rguenther@suse.de>
18603 PR tree-optimization/84777
18604 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
18605 force-vectorize loops ignore whether we are optimizing for size.
18607 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
18609 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
18610 (TARGET_MD_ASM_ADJUST): Define.
18612 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
18613 Kito Cheng <kito.cheng@gmail.com>
18614 Chung-Ju Wu <jasonwucj@gmail.com>
18616 * config/nds32/nds32.c (nds32_compute_stack_frame,
18617 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
18618 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
18619 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
18620 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
18621 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
18622 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
18623 * config/nds32/nds32.md (prologue, epilogue): Use macro
18624 NDS32_V3PUSH_AVAILABLE_P to do checking.
18626 2018-03-11 Jakub Jelinek <jakub@redhat.com>
18629 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
18630 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
18631 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
18632 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
18633 addition of most attributes on !orig_type_die or the attribute not
18634 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
18636 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18637 Chung-Ju Wu <jasonwucj@gmail.com>
18639 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
18640 __NDS32_VH__ macro.
18641 * config/nds32/nds32.opt (mvh): New option.
18643 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18644 Chung-Ju Wu <jasonwucj@gmail.com>
18646 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
18648 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
18649 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
18652 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
18653 Chung-Ju Wu <jasonwucj@gmail.com>
18655 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
18657 * config/nds32/nds32-multiple.md (strlensi): New pattern.
18658 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
18660 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
18661 Kito Cheng <kito.cheng@gmail.com>
18662 Chung-Ju Wu <jasonwucj@gmail.com>
18664 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
18665 UNSPEC_FFMISM and UNSPEC_FLMISM.
18666 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
18667 for ffb, ffmism and flmism.
18668 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
18669 (unspec_ffmism): Ditto.
18670 (unspec_flmism): Ditto.
18671 (nds32_expand_builtin_impl): Check if string extension is available.
18672 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
18673 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
18675 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
18678 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18681 * lra-assigns.c (assign_by_spills): Return a flag of reload
18682 assignment failure. Do not process the reload assignment
18683 failures. Do not spill other reload pseudos if they has the same
18685 (lra_assign): Add a return arg. Set up from the result of
18686 assign_by_spills call.
18687 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18688 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18689 usage_insns if it is not NULL.
18690 (spill_hard_reg_in_range): New function.
18691 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18692 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18693 function prototypes.
18694 (lra_assign): Change prototype.
18695 * lra.c (lra): Add code to deal with fails by splitting hard reg
18698 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
18701 * config/i386/i386.opt: Replace Enforcment with Enforcement.
18703 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
18706 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
18707 (dw_val_node): Add val_symbolic_view.
18708 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
18709 (symview_upper_bound): New.
18710 (new_line_info_table): Initialize symviews_since_reset.
18711 (dwarf2out_source_line): Count symviews_since_reset and set
18712 symview_upper_bound.
18713 (dw_val_equal_p): Handle symview.
18714 (add_AT_symview): New.
18715 (print_dw_val): Handle symview.
18716 (attr_checksum, attr_checksum_ordered): Likewise.
18717 (same_dw_val_p, size_of_die): Likewise.
18718 (value_format, output_die): Likewise.
18719 (add_high_low_attributes): Use add_AT_symview for entry_view.
18720 (dwarf2out_finish): Reset symview_upper_bound, clear
18723 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
18726 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
18727 Add strict argument and use it.
18728 (rs6000_split_multireg_move): Update for new strict argument.
18729 (mem_operand_gpr): Disallow all non-offsettable addresses.
18730 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
18732 2018-03-09 Jakub Jelinek <jakub@redhat.com>
18735 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
18736 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
18737 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
18740 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
18741 decl, use remap_type if we want to use the type.
18743 2018-03-09 Martin Sebor <msebor@redhat.com>
18745 PR tree-optimization/84526
18746 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18748 (builtin_access::generic_overlap): Be prepared to handle non-array
18751 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
18753 PR rtl-optimization/84682
18754 * lra-constraints.c (process_address_1): Check is_address flag
18755 for address constraints.
18756 (process_alt_operands): Likewise.
18757 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
18758 preprocess_constraints.
18759 * recog.h (preprocess_constraints): Add oploc parameter.
18761 * recog.c (preprocess_constraints): Test address_operand for
18762 CT_ADDRESS constraints.
18764 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
18767 * lra-assigns.c (assign_by_spills): Return a flag of reload
18768 assignment failure. Do not process the reload assignment
18769 failures. Do not spill other reload pseudos if they has the same
18771 (lra_assign): Add a return arg. Set up from the result of
18772 assign_by_spills call.
18773 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18774 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
18775 usage_insns if it is not NULL.
18776 (spill_hard_reg_in_range): New function.
18777 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18778 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18779 function prototypes.
18780 (lra_assign): Change prototype.
18781 * lra.c (lra): Add code to deal with fails by splitting hard reg
18784 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18787 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
18788 Accept complain bool parameter. Only emit errors if it is true.
18789 (arm_parse_cpu_option_name): Likewise.
18790 (arm_target_thumb_only): Adjust callers of the above.
18791 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
18792 prototype to take a default true bool parameter.
18793 (arm_parse_arch_option_name): Likewise.
18795 2018-03-09 David Malcolm <dmalcolm@redhat.com>
18796 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
18800 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
18801 * configure: Regenerate.
18802 * configure.ac ("linker --version-script option"): New.
18803 ("linker soname option"): New.
18805 2018-03-09 Richard Biener <rguenther@suse.de>
18807 PR tree-optimization/84775
18808 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
18809 immediate uses of predicate stmts and mark them modified.
18812 PR tree-optimization/84178
18813 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18815 (version_loop_for_if_conversion): Delay update_ssa call.
18816 (tree_if_conversion): Delay update_ssa until after predicate
18819 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
18822 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
18823 when the function accesses prior frames.
18825 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18828 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
18829 gen_llsym, otherwise call maybe_gen_llsym.
18831 PR inline-asm/84742
18832 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
18833 has ',' character inside of it.
18835 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18838 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
18839 as clobbering CC_REGNUM.
18841 2018-03-08 Richard Biener <rguenther@suse.de>
18843 PR middle-end/84552
18844 * tree-scalar-evolution.c: Include tree-into-ssa.h.
18845 (follow_copies_to_constant): Do not follow SSA names registered
18848 2018-03-08 Richard Biener <rguenther@suse.de>
18850 PR tree-optimization/84178
18851 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18853 (version_loop_for_if_conversion): Delay update_ssa call.
18854 (tree_if_conversion): Delay update_ssa until after predicate
18857 2018-03-08 David Malcolm <dmalcolm@redhat.com>
18859 PR tree-optimization/84178
18860 * tree-if-conv.c (release_bb_predicate): Remove the
18861 the assertion that the stmts have NULL use_ops.
18862 Discard the statements, asserting that they haven't
18863 yet been added to a BB.
18865 2018-03-08 Richard Biener <rguenther@suse.de>
18867 PR tree-optimization/84746
18868 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
18869 (phi_translate): Pass in destination ANTIC_OUT set.
18870 (phi_translate_1): Likewise. For a simplified result lookup
18871 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
18872 (phi_translate_set): Adjust.
18873 (do_pre_regular_insertion): Likewise.
18874 (do_pre_partial_partial_insertion): Likewise.
18876 2018-03-08 Martin Liska <mliska@suse.cz>
18878 PR gcov-profile/84735
18879 * doc/gcov.texi: Document usage of profile files.
18880 * gcov-io.h: Document changes in the format.
18882 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
18886 * dwarf2out.c (struct dw_line_info_table): Update comments for
18888 (FORCE_RESET_NEXT_VIEW): New.
18889 (FORCE_RESETTING_VIEW_P): New.
18890 (RESETTING_VIEW_P): Check for -1 too.
18891 (ZERO_VIEW_P): Likewise.
18892 (new_line_info_table): Force-reset next view.
18893 (dwarf2out_begin_function): Likewise.
18894 (dwarf2out_source_line): Simplify zero_view_p initialization.
18895 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
18896 view directly. Omit view when omitting .loc at line 0.
18898 2018-03-08 Jakub Jelinek <jakub@redhat.com>
18900 PR tree-optimization/84740
18901 * tree-switch-conversion.c (process_switch): Call build_constructors
18902 only if info.phi_count is non-zero.
18904 PR tree-optimization/84739
18905 * tree-tailcall.c (find_tail_calls): Check call arguments against
18906 DECL_ARGUMENTS (current_function_decl) rather than
18907 DECL_ARGUMENTS (func) when checking for tail recursion.
18909 2018-03-07 Jakub Jelinek <jakub@redhat.com>
18911 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
18912 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
18913 Volker Reichelt's entry and add entries for people that perform
18914 GCC fuzzy testing and report numerous bugs.
18916 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
18919 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
18920 readonly data in sdata, if that is disabled.
18921 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
18922 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
18923 -mreadonly-in-sdata option.
18925 2018-03-07 Martin Sebor <msebor@redhat.com>
18927 PR tree-optimization/84468
18928 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
18929 basic block when looking for nul assignment.
18931 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
18934 * except.h (output_function_exception_table): Adjust prototype.
18935 * except.c (output_function_exception_table): Remove FNNAME parameter
18936 and add SECTION parameter. Ouput one part of the table at a time.
18937 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
18938 the first part of the exception table and emit unwind directives.
18939 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
18940 (i386_pe_seh_cold_init): Likewise.
18941 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
18942 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
18943 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
18944 (ix86_output_call_insn): Emit a nop in one more case for SEH.
18945 * config/i386/winnt.c: Include except.h.
18946 (struct seh_frame_state): Add reg_offset, after_prologue and
18947 in_cold_section fields.
18948 (i386_pe_seh_end_prologue): Set seh->after_prologue.
18949 (i386_pe_seh_cold_init): New function.
18950 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
18951 to seh->in_cold_section.
18952 (seh_emit_push): Record the offset of the push.
18953 (seh_emit_save): Record the offet of the save.
18954 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
18955 Test seh->after_prologue to disregard the epilogue.
18956 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
18957 (i386_pe_end_cold_function): New function.
18959 2018-03-07 Jakub Jelinek <jakub@redhat.com>
18962 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
18963 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
18966 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
18968 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
18969 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
18971 PR middle-end/84723
18972 * multiple_target.c: Include tree-inline.h and intl.h.
18973 (expand_target_clones): Diagnose and fail if node->definition and
18974 !tree_versionable_function_p (node->decl).
18976 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
18978 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
18980 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
18981 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
18982 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
18984 2018-03-06 Jakub Jelinek <jakub@redhat.com>
18987 * combine.c (try_combine): Use reg_or_subregno instead of handling
18988 just paradoxical SUBREGs and REGs.
18990 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
18992 * config/arc/arc.c (arc_finalize_pic): Remove function.
18993 (arc_must_save_register): We use single base PIC register, remove
18994 checks to save/restore the PIC register.
18995 (arc_expand_prologue): Likewise.
18996 * config/arc/arc-protos.h (arc_set_default_type_attributes):
18998 (arc_verify_short): Likewise.
18999 (arc_attr_type): Likewise.
19000 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
19001 (walk_stores): Likewise.
19002 (arc_address_cost): Make it static.
19003 (arc_verify_short): Likewise.
19004 (branch_dest): Likewise.
19005 (arc_attr_type): Likewise.
19006 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
19007 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
19008 (arc_final_prescan_insn): Remove inserting the nops due to
19009 hardware hazards. It is done in reorg step.
19010 (insn_length_variant_t): Remove.
19011 (insn_length_parameters_t): Likewise.
19012 (arc_insn_length_parameters): Likewise.
19013 (arc_get_insn_variants): Likewise.
19014 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
19016 2018-03-06 Jakub Jelinek <jakub@redhat.com>
19018 PR inline-asm/84683
19019 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
19022 PR tree-optimization/84687
19023 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
19025 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
19027 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
19029 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
19030 Rename to ppc_speculation_barrier.
19031 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
19032 __builtin_ppc_speculation_barrier.
19034 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19037 * combine.c (combine_simplify_rtx): Don't try to simplify if
19038 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
19041 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
19043 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
19044 to 32 bytes when compiling for POWER9.
19046 2018-03-05 Jakub Jelinek <jakub@redhat.com>
19049 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
19050 regparm >= 3 with no arg reg available also for calls with
19051 flag_force_indirect_call. Pass decl to ix86_function_regparm.
19054 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
19056 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
19058 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
19061 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
19063 2018-03-05 Richard Biener <rguenther@suse.de>
19065 PR tree-optimization/84486
19066 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
19067 When inserting a __builtin_assume_aligned call set the LHS
19068 SSA name alignment info accordingly.
19070 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
19072 PR tree-optimization/84114
19073 * config/aarch64/aarch64.c (aarch64_reassociation_width)
19074 Avoid reassociation of FLOAT_MODE addition.
19076 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
19078 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
19079 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
19080 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
19081 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
19082 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
19083 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
19084 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
19086 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
19087 __builtin_ia32_wbinvd): New builtins.
19088 (SPECIAL_ARGS2): New.
19089 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
19090 (SPECIAL_ARGS2): New.
19091 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
19092 (ix86_valid_target_attribute_inner_p): Ditto.
19093 (ix86_init_mmx_sse_builtins): Add special_args2.
19094 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
19095 TARGET_WBNOINVD_P): New.
19096 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
19097 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19098 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19099 * config/i386/immintrin.h (_wbinvd): New intrinsic.
19100 * config/i386/pconfigintrin.h: New file.
19101 * config/i386/wbnoinvdintrin.h: Ditto.
19102 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19104 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19106 2018-03-05 Richard Biener <rguenther@suse.de>
19108 PR tree-optimization/84670
19109 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19111 (BB_VISITED_WITH_VISITED_SUCCS): New define.
19112 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19113 (compute_antic_aux): Only assert the number of values in ANTIC_IN
19114 doesn't grow if all successors (recursively) were visited at least
19117 2018-03-05 Richard Biener <rguenther@suse.de>
19119 PR tree-optimization/84650
19120 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19121 if executed in the loop pipeline.
19123 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
19125 * doc/configfiles.texi (Configuration Files): Move info about
19126 conditionalizing $target-protos.h to...
19127 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
19128 differs from $target-protos.h.
19130 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
19131 Chung-Ju Wu <jasonwucj@gmail.com>
19133 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19134 * config/nds32/nds32-multiple.md (setmemsi): Define.
19135 * config/nds32/nds32-memory-manipulation.c
19136 (nds32_gen_dup_4_byte_to_word_value): New.
19137 (emit_setmem_word_loop): New.
19138 (emit_setmem_byte_loop): New.
19139 (nds32_expand_setmem_loop): New.
19140 (nds32_expand_setmem_loop_v3m): New.
19141 (nds32_expand_setmem_unroll): New.
19142 (nds32_expand_setmem): New.
19144 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19145 Chung-Ju Wu <jasonwucj@gmail.com>
19147 * config/nds32/nds32-memory-manipulation.c
19148 (nds32_emit_load_store): New.
19149 (nds32_emit_post_inc_load_store): New.
19150 (nds32_emit_mem_move): New.
19151 (nds32_emit_mem_move_block): New.
19152 (nds32_expand_movmemsi_loop_unknown_size): New.
19153 (nds32_expand_movmemsi_loop_known_size): New.
19154 (nds32_expand_movmemsi_loop): New.
19155 (nds32_expand_movmemsi_unroll): New.
19156 (nds32_expand_movmemqi): Rename ...
19157 (nds32_expand_movmemsi): ... to this.
19158 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19159 (movmemsi): ... to this.
19160 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19161 (nds32_expand_movmemsi): ... to this.
19163 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19164 Monk Chiang <sh.chiang04@gmail.com>
19165 Chung-Ju Wu <jasonwucj@gmail.com>
19167 * config/nds32/nds32-protos.h
19168 (nds32_expand_load_multiple): New arguments.
19169 (nds32_expand_store_multiple): Ditto.
19170 (nds32_valid_multiple_load_store): Rename ...
19171 (nds32_valid_multiple_load_store_p): ... to this.
19172 * config/nds32/nds32-memory-manipulation.c
19173 (nds32_expand_load_multiple): Refine implementation.
19174 (nds32_expand_store_multiple): Ditto.
19175 * config/nds32/nds32-multiple.md
19176 (load_multiple): Update nds32_expand_load_multiple interface.
19177 (store_multiple): Update nds32_expand_store_multiple interface.
19178 * config/nds32/nds32-predicates.c
19179 (nds32_valid_multiple_load_store): Rename ...
19180 (nds32_valid_multiple_load_store_p): ... to this and refine
19182 * config/nds32/predicates.md
19183 (nds32_load_multiple_and_update_address_operation): New predicate.
19184 (nds32_store_multiple_and_update_address_operation): New predicate.
19186 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
19187 Chung-Ju Wu <jasonwucj@gmail.com>
19189 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19190 (combo): New attribute.
19191 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19193 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
19195 * config/nds32/nds32.opt: Change -mcmodel= default value.
19197 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
19198 Monk Chiang <sh.chiang04@gmail.com>
19199 Chung-Ju Wu <jasonwucj@gmail.com>
19201 * config/nds32/constants.md (unspec_element): New enum.
19202 * config/nds32/constraints.md (Umw): New constraint.
19203 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19204 * config/nds32/nds32-intrinsic.md: Likewise.
19205 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19206 (nds32_valid_smw_lwm_base_p): New.
19207 (nds32_output_smw_single_word): New.
19208 (nds32_output_lmw_single_word): New.
19209 (nds32_expand_unaligned_load): New.
19210 (nds32_expand_unaligned_store): New.
19211 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19212 (nds32_output_smw_single_word): Declare.
19213 (nds32_output_lmw_single_word): Declare.
19214 (nds32_expand_unaligned_load): Declare.
19215 (nds32_expand_unaligned_store): Declare.
19216 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19217 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19218 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19219 NDS32_BUILTIN_UASTORE_DW.
19220 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19223 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
19224 Kito Cheng <kito.cheng@gmail.com>
19225 Chung-Ju Wu <jasonwucj@gmail.com>
19227 * config/nds32/nds32-intrinsic.c
19228 (nds32_expand_builtin_null_ftype_reg): Delete.
19229 (nds32_expand_builtin_reg_ftype_imm): Ditto.
19230 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19231 (nds32_read_argument): New.
19232 (nds32_legitimize_target): Ditto.
19233 (nds32_legitimize_argument): Ditto.
19234 (nds32_check_constant_argument): Ditto.
19235 (nds32_expand_unop_builtin): Ditto.
19236 (nds32_expand_unopimm_builtin): Ditto.
19237 (nds32_expand_binop_builtin): Ditto.
19238 (nds32_builtin_decl_impl): Ditto.
19239 (builtin_description): Ditto.
19240 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
19241 (nds32_init_builtins_impl): Ditto.
19242 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
19243 (nds32_builtin_decl): New.
19244 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
19245 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
19247 2018-03-02 Jeff Law <law@redhat.com>
19249 * reorg.c (stop_search_p): Handle DEBUG_INSN.
19250 (redundant_insn, fill_simple_delay_slots): Likewise.
19251 (fill_slots_from_thread): Likewise.
19252 * resource.c (mark_referenced_resources): Likewise.
19253 (mark_set_resources, find_dead_or_set_registers): Likewise.
19255 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19257 * substring-locations.h (format_warning_va): Formatting fix for
19258 ATTRIBUTE_GCC_DIAG.
19259 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
19261 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
19262 * substring-locations.c: Include intl.h.
19263 (format_warning_va): Turned into small wrapper around
19264 format_warning_n_va, renamed to ...
19265 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
19266 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
19268 (format_warning_at_substring_n): New function.
19269 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
19270 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
19271 format_warning_at_substring with just a shorter name instead of
19272 const function pointer.
19273 (fmtwarn_n): New function.
19274 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
19275 appropriate, get rid of all the fmtstr temporaries, move conditionals
19276 with G_() wrapped string literals directly into fmtwarn arguments,
19277 cast dir.len to (int), formatting fixes.
19279 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
19281 * doc/invoke.texi: Remove "Cilk Plus" references.
19283 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19284 Richard Biener <rguenther@suse.de>
19287 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
19288 for error or warning attributes if CALL_FROM_THUNK_P is set.
19291 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19294 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
19295 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
19298 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
19299 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
19301 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
19302 instead of -1U in last predictors element's probability member.
19304 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
19307 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19308 arguments if they are comparable.
19310 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
19312 PR tree-optimization/84634
19313 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
19314 masks and masked_loop_p with a single loop_masks, making sure it's
19315 null for bb vectorization.
19317 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
19319 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
19320 (vect_analyze_data_ref_access): Use loop->safe_len rather than
19321 loop->force_vectorize to check whether there is no alias.
19323 2018-03-02 Jakub Jelinek <jakub@redhat.com>
19326 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
19328 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
19330 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
19331 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
19332 instead of a loop around prev_real_insn.
19333 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
19336 PR inline-asm/84625
19337 * config/i386/i386.c (ix86_print_operand): Use conditional
19338 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
19341 2018-03-02 Richard Biener <rguenther@suse.de>
19343 PR tree-optimization/84427
19344 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
19345 (bitmap_set_subtract_values): Rewrite to handle multiple
19348 (prune_clobbered_mems): Likewise.
19349 (phi_translate): Take edge instead of pred/phiblock.
19350 (phi_translate_1): Likewise.
19351 (phi_translate_set): Likewise. Insert all translated
19352 exprs for a value into the set, keeping possibly multiple
19353 expressions per value.
19354 (compute_antic_aux): Adjust for phi_translate changes.
19355 When intersecting union the expressions and prune those
19356 not in the final value set, keeping possibly multiple
19357 expressions per value. Do not use value-insertion
19358 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
19359 all expressions. Add verification that the value-sets
19360 only shrink during iteration.
19361 (compute_partial_antic_aux): Adjust for the phi_translate changes.
19362 (do_pre_regular_insertion): Likewise.
19363 (do_pre_partial_partial_insertion): Likewise.
19365 2018-03-02 Richard Biener <rguenther@suse.de>
19368 * config/darwin.c (saved_debug_info_level): New static global.
19369 (darwin_asm_lto_start): Disable debug info generation for LTO out.
19370 (darwin_asm_lto_end): Restore debug info generation settings.
19372 2018-03-01 Martin Liska <mliska@suse.cz>
19375 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
19376 volatile arguments.
19378 2018-03-01 Richard Biener <rguenther@suse.de>
19381 * dwarf2out.c (gen_variable_die): Properly handle late VLA
19382 type annotation with LTO when debug was disabled at compile-time.
19384 2018-03-01 Matthew Fortune <mfortune@gmail.com>
19386 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
19388 (mips_final_postscan_insn): Likewise.
19390 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
19392 PR rtl-optimization/84528
19393 * alias.c (init_alias_target): Add commentary.
19394 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
19395 a unique base value if the frame pointer is not eliminated
19396 to the stack pointer.
19398 2018-03-01 Tom de Vries <tom@codesourcery.com>
19400 PR rtl-optimization/83327
19401 * lra-int.h (hard_regs_spilled_into): Declare.
19402 * lra.c (hard_regs_spilled_into): Define.
19403 (init_reg_info): Init hard_regs_spilled_into.
19404 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
19405 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
19406 (process_bb_lives): Handle hard_regs_spilled_into.
19407 (lra_create_live_ranges_1): Before doing liveness propagation, clear
19408 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
19410 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
19412 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
19413 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
19414 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
19415 * config/rs6000/aix72.h: New file.
19417 2018-02-28 Jakub Jelinek <jakub@redhat.com>
19419 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
19420 instead of warning_at with conditional singular and plural messages
19424 * stor-layout.c (update_alignment_for_field): For
19425 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
19426 && !DECL_PACKED (field), do the alignment update, just use
19427 only desired_align instead of MAX (type_align, desired_align)
19429 (place_field): Don't do known_align < desired_align handling
19430 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
19431 is non-NULL, instead do it after rli->prev_field handling and
19432 only if not within a bitfield word. For DECL_PACKED (field)
19433 use type_align of BITS_PER_UNIT.
19435 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
19437 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
19438 superfluous parentheses and trailing spaces.
19440 2018-02-28 Richard Biener <rguenther@suse.de>
19442 PR tree-optimization/84584
19443 * graphite-scop-detection.c (scop_detection::add_scop): Discard
19444 SCoPs with fake exit edge.
19446 2018-02-28 Martin Liska <mliska@suse.cz>
19449 * timevar.c (timer::print): Fix format to properly print 100%
19452 2018-02-28 Richard Biener <rguenther@suse.de>
19454 PR middle-end/84607
19455 * genmatch.c (capture_info::walk_match): Do not mark
19456 captured expressions without operands as expr_p given
19457 they act more like predicates and should be subject to
19458 "lost tail" side-effect preserving.
19460 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
19462 PR rtl-optimization/81611
19463 * auto-inc-dec.c (attempt_change): Move dead note from
19464 mem_insn if it's the next use of regno
19465 (find_address): Take address use of reg holding
19466 non-incremented value. Add parm to limit search to the named
19468 (merge_in_block): Attempt to use a mem insn that is the next
19469 use of the original regno.
19471 2018-02-27 Martin Sebor <msebor@redhat.com>
19474 * doc/invoke.texi (-Wmissing-attributes): New option.
19475 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
19477 2018-02-27 Martin Sebor <msebor@redhat.com>
19479 PR translation/84207
19480 * diagnostic-core.h (warning_n, error_n, inform_n): Change
19481 n argument to unsigned HOST_WIDE_INT.
19482 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
19483 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
19484 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
19485 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
19487 2018-02-27 Richard Biener <rguenther@suse.de>
19489 PR tree-optimization/84512
19490 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19491 Do not use the estimate returned from record_stmt_cost for
19492 the scalar iteration cost but sum properly using add_stmt_cost.
19494 2018-02-27 Richard Biener <rguenther@suse.de>
19496 PR tree-optimization/84466
19497 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19498 Adjust last change to less strictly validate use operands.
19500 2018-02-27 Martin Liska <mliska@suse.cz>
19502 PR gcov-profile/84548
19503 * gcov.c (process_file): Allow partial overlap and consider it
19504 also as group functions.
19505 (output_lines): Properly calculate range of lines for a group.
19507 2018-02-27 Martin Liska <mliska@suse.cz>
19509 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
19510 'ggc' suffixes. Change first column width.
19511 (timer::print): Fix formatting of the column.
19513 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
19515 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
19516 preserve inline entry blocks for the sake of debug inline
19517 entry point markers alone.
19518 (remove_unused_locals): Suggest in comments a better place to
19519 force the preservation of inline entry blocks that are
19520 otherwise unused, but do not preserve them.
19522 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19524 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
19526 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19529 * config/i386/constraints.md (Bs): Replace
19530 ix86_indirect_branch_register with
19531 TARGET_INDIRECT_BRANCH_REGISTER.
19533 * config/i386/i386.md (indirect_jump): Likewise.
19534 (tablejump): Likewise.
19535 (*sibcall_memory): Likewise.
19536 (*sibcall_value_memory): Likewise.
19537 Peepholes of indirect call and jump via memory: Likewise.
19538 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
19539 (*sibcall_value_GOT_32): Likewise.
19540 * config/i386/predicates.md (indirect_branch_operand): Likewise.
19541 (GOT_memory_operand): Likewise.
19542 (call_insn_operand): Likewise.
19543 (sibcall_insn_operand): Likewise.
19544 (GOT32_symbol_operand): Likewise.
19545 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
19547 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
19549 PR rtl-optimization/83496
19550 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
19551 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
19552 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
19553 redundant insn, if any.
19554 (relax_delay_slots): Likewise.
19555 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
19557 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
19559 PR tree-optimization/83965
19560 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
19561 that grouped statements are part of a reduction chain. Return
19562 true if the statement is not marked as a reduction itself but
19563 is part of a group.
19564 (vect_recog_dot_prod_pattern): Don't check whether the statement
19565 is part of a group here.
19566 (vect_recog_sad_pattern): Likewise.
19567 (vect_recog_widen_sum_pattern): Likewise.
19569 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
19572 * final.c (rest_of_clean_state): Also look for calls inside sequences.
19574 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
19577 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
19579 (ix86_output_indirect_function_return): New prototype.
19580 (ix86_split_simple_return_pop_internal): Likewise.
19581 * config/i386/i386.c (indirect_return_via_cx): New.
19582 (indirect_return_via_cx_bnd): Likewise.
19583 (indirect_thunk_name): Handle return va CX_REG.
19584 (output_indirect_thunk_function): Create alias for
19585 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
19586 (ix86_output_indirect_jmp): Remove the bool argument.
19587 (ix86_output_indirect_function_return): New function.
19588 (ix86_split_simple_return_pop_internal): Likewise.
19589 * config/i386/i386.md (*indirect_jump): Don't pass false
19590 to ix86_output_indirect_jmp.
19591 (*tablejump_1): Likewise.
19592 (simple_return_pop_internal): Change it to define_insn_and_split.
19593 Call ix86_split_simple_return_pop_internal to split it for
19594 -mfunction-return=.
19595 (simple_return_indirect_internal): Call
19596 ix86_output_indirect_function_return instead of
19597 ix86_output_indirect_jmp.
19599 2018-02-26 Jakub Jelinek <jakub@redhat.com>
19602 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
19603 memset and value initialization afterwards.
19605 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
19607 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
19609 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
19612 * common/config/aarch64/aarch64-common.c
19613 (aarch_option_optimization_table[]): Switch
19614 off fomit-frame-pointer
19616 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19617 Chung-Ju Wu <jasonwucj@gmail.com>
19619 * config/nds32/nds32-multiple.md (load_multiple): Disallow
19621 (store_multiple): Ditto.
19623 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
19625 * config.gcc: Add --with-cpu support for nds32 target.
19626 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
19627 * config/nds32/nds32.opt: Add -mcpu= option.
19629 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
19631 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
19632 isel=yes): Warn for these deprecated options.
19634 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
19636 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
19637 ISA_2_5_MASKS_EMBEDDED.
19639 2018-02-23 Jakub Jelinek <jakub@redhat.com>
19641 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
19642 p->max as pointers rather than using iterative_hash_expr.
19644 2018-02-23 Carl Love <cel@us.ibm.com>
19646 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
19647 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
19649 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
19650 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
19651 P8V_BUILTIN_VEC_VUNSIGNED2.
19653 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
19656 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
19657 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
19659 * lra-constraints.c (curr_insn_transform): Set up
19660 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
19662 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
19663 * lra-eliminations.c (spill_pseudos): Ditto.
19664 (process_insn_for_elimination): Ditto.
19665 * lra-lives.c (reg_early_clobber_p): Use the new macros.
19666 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
19667 LRA_NON_CLOBBERED_ALT.
19669 2018-02-22 Martin Sebor <msebor@redhat.com>
19671 PR tree-optimization/84480
19672 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
19673 to maybe_diag_stxncpy_trunc. Call it.
19674 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
19675 from gimple_fold_builtin_strcpy. Print inlining stack.
19676 (handle_builtin_stxncpy): Print inlining stack.
19677 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
19679 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
19682 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
19683 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
19684 and -fcheck-pointer-bounds are used together.
19685 (indirect_thunk_prefix): New enum.
19686 (indirect_thunk_need_prefix): New function.
19687 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
19688 "_nt" instead of "_bnd" for NOTRACK prefix.
19689 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
19690 (output_indirect_thunk_function): Likewise.
19692 (ix86_code_end): Update output_indirect_thunk_function calls.
19693 (ix86_output_indirect_branch_via_reg): Replace
19694 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
19695 (ix86_output_indirect_branch_via_push): Likewise.
19696 (ix86_output_function_return): Likewise.
19697 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
19698 incompatible with -fcf-protection=branch and
19699 -fcheck-pointer-bounds.
19701 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19704 * config/aarch64/aarch64.c (aarch64_print_address_internal):
19705 Change gcc_assert call to output_operand_lossage.
19707 2018-02-22 Steve Ellcey <sellcey@cavium.com>
19709 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
19711 2018-02-22 DJ Delorie <dj@redhat.com>
19712 Sebastian Perta <sebastian.perta@renesas.com>
19713 Oleg Endo <olegendo@gcc.gnu.org>
19715 * config/rx/rx.c (rx_rtx_costs): New function.
19716 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
19718 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
19720 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
19722 2018-02-22 Martin Liska <mliska@suse.cz>
19725 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
19726 Add "native" as a possible value.
19728 2018-02-22 Martin Liska <mliska@suse.cz>
19731 * config/i386/i386.c (ix86_option_override_internal):
19732 Add "native" as a possible value for -march and -mtune.
19734 2018-02-22 Jakub Jelinek <jakub@redhat.com>
19737 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
19738 to all type variants.
19740 PR tree-optimization/84503
19741 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
19742 width as info->bitpos + info->bitsize - start.
19743 (merged_store_group::merge_overlapping): Simplify width computation.
19744 (check_no_overlap): New function.
19745 (imm_store_chain_info::try_coalesce_bswap): Compute expected
19746 start + width and last_order of the group, fail if check_no_overlap
19748 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
19749 to group if check_no_overlap fails.
19751 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19753 * config/rs6000/altivec.md: Delete contraint arguments to
19754 define_expand, define_split, and define_peephole2, and in
19755 define_insn_and_split if always unused.
19756 * config/rs6000/darwin.md: Ditto.
19757 * config/rs6000/dfp.md: Ditto.
19758 * config/rs6000/rs6000.md: Ditto.
19759 * config/rs6000/sync.md: Ditto.
19760 * config/rs6000/vector.md: Ditto.
19761 * config/rs6000/vsx.md: Ditto.
19763 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
19765 * config/rs6000/altivec.md: Write output control strings as braced
19766 blocks instead of double-quoted strings.
19767 * config/rs6000/darwin.md: Ditto.
19768 * config/rs6000/rs6000.md: Ditto.
19769 * config/rs6000/vector.md: Ditto.
19770 * config/rs6000/vsx.md: Ditto.
19772 2018-02-21 Jason Merrill <jason@redhat.com>
19774 PR c++/84314 - ICE with templates and fastcall attribute.
19775 * attribs.c (build_type_attribute_qual_variant): Remove assert.
19777 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19779 * ipa-cp.c (determine_versionability): Fix comment typos.
19781 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
19784 * ipa-cp.c (determine_versionability): Do not version functions caling
19787 2018-02-21 Martin Liska <mliska@suse.cz>
19790 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
19791 Add "native" as a possible value.
19792 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
19793 the macro when native cpu detection is available.
19795 2018-02-21 Martin Liska <mliska@suse.cz>
19798 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
19799 Add "native" as a possible value.
19800 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
19801 when native cpu detection is available.
19803 2018-02-21 Jakub Jelinek <jakub@redhat.com>
19804 Martin Sebor <msebor@redhat.com>
19806 PR tree-optimization/84478
19807 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
19809 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
19810 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
19811 support which is conservatively correct, for 2 only stay conservative
19812 for maxlen. Formatting and comment capitalization fixes. Add STRICT
19813 argument to the 2 argument get_range_strlen, adjust 6 arg
19814 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
19816 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
19817 (gimple_fold_builtin_strlen): Pass true as last argument to
19820 2018-02-20 Martin Sebor <msebor@redhat.com>
19822 PR middle-end/84095
19823 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
19824 (builtin_memref::set_base_and_offset): Same. Handle inner references.
19825 (builtin_memref::builtin_memref): Factor out parts into
19826 set_base_and_offset and call it.
19828 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19830 PR middle-end/84406
19831 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
19832 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
19833 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
19834 search at the associated MODE_INT.
19836 2018-02-20 Jeff Law <law@redhat.com>
19838 PR middle-end/82123
19839 PR tree-optimization/81592
19840 PR middle-end/79257
19841 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
19842 for range data rather than using global data.
19843 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
19844 range data rather than using global data.
19845 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
19846 pass it to children as needed.
19847 (struct directive::fmtresult): Similarly.
19848 (struct directive::set_width): Similarly.
19849 (struct directive::set_precision): Similarly.
19850 (format_integer, format_directive, parse_directive): Similarly.
19851 (format_none): Accept unnamed vr_values parameter.
19852 (format_percent, format_floating, format_character): Similarly.
19853 (format_string, format_plain): Similarly.
19854 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
19855 the EVRP range analyzer for range data rather than using global data.
19856 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
19857 gimple-ssa-evrp-analyze.h
19858 (class sprintf_dom_walker): Add after_dom_children member function.
19859 Add evrp_range_analyzer member.
19860 (sprintf_dom_walker::before_dom_children): Call into the EVRP
19861 range analyzer as needed.
19862 (sprintf_dom_walker::after_dom_children): New member function.
19863 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
19865 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
19866 (evrp_range_analyzer::pop_to_marker): Likewise.
19868 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
19870 PR tree-optimization/84419
19871 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
19872 with the required type if its current type is compatible but
19875 2018-02-20 Jakub Jelinek <jakub@redhat.com>
19877 PR middle-end/82004
19878 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
19879 after vectorization.
19881 2018-02-20 Martin Liska <mliska@suse.cz>
19884 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
19885 possible values if we don't have a hint.
19887 2018-02-20 Martin Liska <mliska@suse.cz>
19891 * final.c (shorten_branches): Build align_tab array with one
19893 * opts.c (finish_options): Add alignment option limit check.
19894 (MAX_CODE_ALIGN): Likewise.
19895 (MAX_CODE_ALIGN_VALUE): Likewise.
19896 * doc/invoke.texi: Document maximum allowed option value for
19897 all -falign-* options.
19899 2018-02-19 Jakub Jelinek <jakub@redhat.com>
19902 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
19903 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
19904 * var-tracking.c (emit_note_insn_var_location): Remove all references
19905 to NOTE_INSN_CALL_ARG_LOCATION.
19906 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
19907 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
19908 Use copy_rtx_if_shared.
19909 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
19910 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
19911 (dwarf2out_var_location): Remove handling of
19912 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
19914 * final.c (final_scan_insn): Remove all references to
19915 NOTE_INSN_CALL_ARG_LOCATION.
19916 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
19917 before dumping final insns.
19918 * except.c (emit_note_eh_region_end): Remove all references to
19919 NOTE_INSN_CALL_ARG_LOCATION.
19920 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
19921 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
19922 * config/arc/arc.c (hwloop_optimize): Likewise.
19923 * config/arm/arm.c (create_fix_barrier): Likewise.
19924 * config/s390/s390.c (s390_chunkify_start): Likewise.
19925 * config/sh/sh.c (find_barrier): Likewise.
19926 * config/i386/i386.c (rest_of_insert_endbranch,
19927 ix86_seh_fixup_eh_fallthru): Likewise.
19928 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
19929 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
19930 * config/frv/frv.c (frv_function_prologue): Likewise.
19931 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
19933 (note_outside_basic_block_p): Remove all references to
19934 NOTE_INSN_CALL_ARG_LOCATION.
19935 * gengtype.c (adjust_field_rtx_def): Likewise.
19936 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
19938 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
19939 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
19942 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
19945 PR tree-optimization/84452
19946 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
19947 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
19950 2018-02-19 Martin Liska <mliska@suse.cz>
19953 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
19955 2018-02-19 Martin Liska <mliska@suse.cz>
19956 Richard Sandiford <richard.sandiford@linaro.org>
19958 PR tree-optimization/82491
19959 * gimple-fold.c (get_base_constructor): Make earlier bail out
19962 2018-02-19 Carl Love <cel@us.ibm.com>
19964 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
19965 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
19967 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
19968 P8V_BUILTIN_VEC_NEG.
19970 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
19972 * config/rl78/rl78.md (movdf): New define expand.
19974 2018-02-19 Martin Liska <mliska@suse.cz>
19977 * doc/invoke.texi: Fix typo.
19978 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
19980 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
19982 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
19983 handle rs6000_single_float and rs6000_double_float specially for
19986 2018-02-16 Jeff Law <law@redhat.com>
19988 * config/rx/rx.c (add_pop_cfi_notes): New function.;
19989 (pop_regs): Use it.
19991 2018-02-16 Jakub Jelinek <jakub@redhat.com>
19994 * ipa-inline.c (inline_small_functions): Fix a typo.
19996 2018-02-16 Nathan Sidwell <nathan@acm.org>
19998 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
20000 2018-02-16 Carl Love <cel@us.ibm.com>
20002 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
20003 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
20004 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
20005 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
20006 expansion to P8V_BUILTIN_VEC_FLOAT2.
20008 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
20010 PR rtl-optimization/70023
20011 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
20012 src_regno into account.
20014 2018-02-16 Carl Love <cel@us.ibm.com>
20016 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
20017 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
20018 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
20019 * config/rs6000/rs6000.c: Remove case statements for
20020 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
20021 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
20022 and P9V_BUILTIN_VEC_VINSERT4B.
20023 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
20024 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
20025 * config/rs6000/vsx.md:
20026 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
20029 2018-02-16 Carl Love <cel@us.ibm.com>
20031 * config/rs6000/altivec.h: Add builtin names vec_extract4b
20033 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
20035 * config/rs6000/rs6000-c.c: Add the definitions for
20036 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
20037 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
20038 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
20039 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
20040 definition for insert4b and define insn *insert3b_internal.
20041 * doc/extend.texi: Add documentation for vec_extract4b.
20043 2018-02-16 Nathan Sidwell <nathan@acm.org>
20045 * doc/extend.texi (Backwards Compatibility): Mention friend
20046 injection. Note for-scope is deprecated.
20047 * doc/invoke.texi (-ffriend-injection): Deprecate.
20049 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
20051 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
20052 that moved to I2, also allow destinations that are a paradoxical
20053 subreg (instead of a normal reg).
20055 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
20058 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
20061 2018-02-16 Richard Biener <rguenther@suse.de>
20063 PR tree-optimization/84037
20064 PR tree-optimization/84016
20066 * config/i386/i386.c (ix86_builtin_vectorization_cost):
20067 Adjust vec_construct for the fact we need additional higher latency
20068 128bit inserts for AVX256 and AVX512 vector builds.
20069 (ix86_add_stmt_cost): Scale vector construction cost for
20072 2018-02-16 Richard Biener <rguenther@suse.de>
20074 PR tree-optimization/84417
20075 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
20076 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
20077 (non_rewritable_lvalue_p): Likewise, use poly-ints.
20079 2018-02-16 Martin Liska <mliska@suse.cz>
20082 * internal-fn.def (ASAN_CHECK): Set proper flags.
20083 (ASAN_MARK): Likewise.
20085 2018-02-16 Julia Koval <julia.koval@intel.com>
20087 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
20088 from PTA_CANNONLAKE.
20090 2018-02-16 Jakub Jelinek <jakub@redhat.com>
20093 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
20094 Use ++iter rather than iter++ for std::list iterators.
20095 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
20096 defer deleting them until all nodes in the forest are processed. Do
20097 free even leaf nodes. Change to_process into auto_vec.
20100 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20101 * vec.h (vec_default_construct): Use memset instead of placement new
20102 if BROKEN_VALUE_INITIALIZATION is defined.
20103 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20104 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20107 PR rtl-optimization/83723
20108 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20109 * lra.c (lra_substitute_pseudo): Likewise. If true, use
20110 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
20112 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20114 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20116 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
20118 PR rtl-optimization/81443
20119 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20120 from inner REGs to paradoxical SUBREGs.
20122 2018-02-16 Richard Biener <rguenther@suse.de>
20124 PR tree-optimization/84399
20125 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20126 For operands we can analyze at their definition make sure we can
20127 analyze them at each use as well.
20129 2018-02-16 Richard Biener <rguenther@suse.de>
20131 PR tree-optimization/84190
20132 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20133 volatile accesses if the decl isn't volatile.
20135 2018-02-15 Jason Merrill <jason@redhat.com>
20137 PR c++/84314 - ICE with templates and fastcall attribute.
20138 * attribs.c (build_type_attribute_qual_variant): Don't clobber
20139 TYPE_CANONICAL on an existing type.
20141 2018-02-15 Jakub Jelinek <jakub@redhat.com>
20143 PR tree-optimization/84383
20144 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20145 dstoff nor call operand_equal_p if dstbase is NULL.
20147 PR tree-optimization/84334
20148 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20149 also a CONSTANT_CLASS_P, punt.
20151 2018-02-14 Jim Wilson <jimw@sifive.com>
20153 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20154 first SMALL_OPERAND check. New local min_second_step. Move assert
20155 to where locals are set. Add TARGET_RVC support.
20156 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20158 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
20160 * doc/invoke.texi: Correct -Wformat-overflow code sample.
20162 2018-02-14 Martin Sebor <msebor@redhat.com>
20164 PR tree-optimization/83698
20165 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20166 arrays constrain the offset range to their bounds.
20167 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20168 (builtin_access::overlap): Avoid setting the size of overlap if it's
20170 (maybe_diag_overlap): Also consider arrays when deciding what values
20171 of offsets to include in diagnostics.
20173 2018-02-14 Martin Sebor <msebor@redhat.com>
20176 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20177 that correspond to the kind of a declaration.
20179 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
20182 * config/pa/pa.md: Load address of PIC label using the linkage table
20183 if the label is nonlocal.
20185 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
20187 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20188 warning message if user requests -maltivec=be.
20189 * doc/invoke.texi: Document deprecation of -maltivec=be.
20191 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
20194 * config/rs6000/rs6000-c.c: Update definitions for
20195 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20196 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20198 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
20201 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20202 add _get_ssp intrinsics. Remove argument from
20203 __builtin_ia32_rdssp[d|q].
20204 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20205 * config/i386/i386-builtin.def: Remove argument from
20206 __builtin_ia32_rdssp[d|q].
20207 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20208 ix86_expand_special_args_builtin for _rdssp[d|q].
20209 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20210 Clear register before usage.
20211 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20212 Add documentation for new _get_ssp and _inc_ssp intrinsics.
20214 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
20216 PR tree-optimization/84357
20217 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20218 operand 1 of an ARRAY_REF too.
20220 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
20223 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20224 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20226 (set_of_reg): New struct.
20227 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20228 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20229 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20231 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20232 Split into bitclr, bitset, bitinvert patterns if appropriate.
20233 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
20234 use rx_fuse_in_memory_bitop.
20235 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
20236 to named insn, correct maximum insn length.
20238 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
20241 * machmode.def: Define a complex mode for PARTIAL_INT.
20242 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
20244 * doc/rtl.texi: Document CSPImode.
20245 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
20247 (msp430_hard_regno_nregs_with_padding): Likewise.
20249 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
20252 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
20254 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
20256 PR rtl-optimization/84169
20257 * combine.c (try_combine): New variable split_i2i3. Set it to true if
20258 we generated a parallel as new i3 and we split that to new i2 and i3
20259 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
20260 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
20261 those to i2, not i1. Partially rewrite this scan code.
20263 2018-02-13 Jakub Jelinek <jakub@redhat.com>
20266 * stor-layout.c (place_field): For variable length fields, adjust
20267 offset_align afterwards not just based on the field's alignment,
20268 but also on the size.
20270 PR middle-end/84309
20271 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
20272 of exps and logs in the use_exp2 case.
20274 2018-02-13 Jeff Law <law@redhat.com>
20276 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
20277 entry for "vector".
20279 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
20282 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
20286 * common.opt (gas-loc-support, gas-locview-support): New.
20287 (ginline-points, ginternal-reset-location-views): New.
20288 * doc/invoke.texi: Document them. Use @itemx where intended.
20289 (gvariable-location-views): Adjust.
20290 * target.def (reset_location_view): New.
20291 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
20292 (TARGET_RESET_LOCATION_VIEW): New.
20293 * doc/tm.texi: Rebuilt.
20294 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
20295 (dwarf2out_default_as_locview_support): New.
20296 (output_asm_line_debug_info): Use option variables.
20297 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
20298 (output_loc_list): Likewise.
20299 (add_high_low_attributes): Check option variables.
20300 Don't output entry view attribute in strict mode.
20301 (gen_inlined_subroutine_die): Check option variables.
20302 (dwarf2out_inline_entry): Likewise.
20303 (init_sections_and_labels): Likewise.
20304 (dwarf2out_early_finish): Likewise.
20305 (maybe_reset_location_view): New, from...
20306 (dwarf2out_var_location): ... here. Call it.
20307 * debug.h (dwarf2out_default_as_loc_support): Declare.
20308 (dwarf2out_default_as_locview_support): Declare.
20309 * hooks.c (hook_int_rtx_insn_0): New.
20310 * hooks.h (hook_int_rtx_insn_0): Declare.
20311 * toplev.c (process_options): Take -gas-loc-support and
20312 -gas-locview-support from dwarf2out. Enable
20313 -gvariable-location-views by default only with locview
20314 assembler support. Enable -ginternal-reset-location-views by
20315 default only if the target defines the corresponding hook.
20316 Enable -ginline-points by default if location views are
20317 enabled; force it disabled if statement frontiers are
20319 * tree-inline.c (expand_call_inline): Check option variables.
20320 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
20322 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
20324 PR tree-optimization/84321
20325 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
20326 handling. Also check whether the anti-range contains any values
20327 that satisfy the mask; switch to a VR_RANGE if not.
20329 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
20332 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
20334 2018-02-13 Martin Jambor <mjambor@suse.cz>
20337 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
20338 of call statements, also set location of a load to a temporary.
20340 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
20342 * config/rl78/rl78.c (add_vector_labels): New function.
20343 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
20344 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
20345 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
20346 which checks that no arguments are passed.
20347 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
20348 * doc/extend.texi: Documentation for the new attribute.
20350 2018-02-13 Andreas Schwab <schwab@suse.de>
20352 * config/riscv/linux.h (CPP_SPEC): Define.
20354 2018-02-13 Jakub Jelinek <jakub@redhat.com>
20357 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
20358 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
20359 OPTION_MASK_ISA_AES as first argument to def_builtin_const
20360 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
20361 instead of OPTION_MASK_ISA_PCLMUL as first argument to
20362 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
20363 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
20364 temporarily for AES and PCLMUL builtins.
20366 PR tree-optimization/84339
20367 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
20368 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
20371 PR middle-end/84309
20372 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
20373 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
20374 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
20376 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
20378 * omp-simd-clone.h: New file.
20379 * omp-simd-clone.c: Include omp-simd-clone.h.
20380 (expand_simd_clones): No longer static.
20381 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
20382 cgraph.h and omp-simd-clone.h.
20383 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
20384 (vect_recog_widen_shift_pattern): Formatting fix.
20385 (vect_pattern_recog_1): Don't check optab for calls.
20388 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
20389 operands[2] into a REG before using gen_lowpart on it.
20391 2018-02-12 Jeff Law <law@redhat.com>
20394 * config/sh/sh.c (find_barrier): Consider a sibling call
20397 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
20398 successfully back substituting a reg.
20400 2018-02-12 Richard Biener <rguenther@suse.de>
20402 PR tree-optimization/84037
20403 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
20404 parameter, move visited init to caller.
20405 (vect_slp_analyze_operations): Separate cost from validity
20406 check, initialize visited once for all instances.
20407 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
20409 * tree-vect-stmts.c (vect_model_simple_cost): Make early
20411 (vect_model_promotion_demotion_cost): Likewise.
20412 (vectorizable_bswap): Guard cost modeling with !slp_node
20413 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
20415 (vectorizable_call): Likewise.
20416 (vectorizable_conversion): Likewise.
20417 (vectorizable_assignment): Likewise.
20418 (vectorizable_shift): Likewise.
20419 (vectorizable_operation): Likewise.
20420 (vectorizable_store): Likewise.
20421 (vectorizable_load): Likewise.
20422 (vectorizable_condition): Likewise.
20423 (vectorizable_comparison): Likewise.
20425 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
20428 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
20429 (ASAN_MARK): Fix fnspec to account for return value, change pointer
20430 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
20432 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20434 PR middle-end/83665
20435 * params.def (inline-min-speedup): Increase from 8 to 15.
20436 (max-inline-insns-auto): Decrease from 40 to 30.
20437 * ipa-split.c (consider_split): Add some buffer for function to
20438 be considered inlining candidate.
20439 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
20442 2018-02-12 Richard Biener <rguenther@suse.de>
20444 PR tree-optimization/84037
20445 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
20446 matched stmts if we cannot swap the non-matched ones.
20448 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
20450 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
20451 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
20452 _mm_maskz_scalef_round_ss): New intrinsics.
20453 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
20454 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
20455 __builtin_ia32_scalefss_round): Remove.
20456 (__builtin_ia32_scalefsd_mask_round,
20457 __builtin_ia32_scalefss_mask_round): New intrinsics.
20458 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
20459 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
20460 ((match_operand:VF_128 2 "<round_nimm_predicate>"
20461 "<round_constraint>")): Changed to ...
20462 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
20463 "<round_scalar_constraint>")): ... this.
20464 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
20465 %0, %1, %2<round_op3>}"): Changed to ...
20466 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
20467 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
20468 %2<round_scalar_mask_op3>}"): ... this.
20469 * config/i386/subst.md (round_scalar_nimm_predicate): New.
20471 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
20473 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
20474 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
20475 (_mm_maskz_sqrt_round_ss): New intrinsics.
20476 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
20477 (__builtin_ia32_sqrtsd_mask_round)
20478 (__builtin_ia32_sqrtss_mask_round): New builtins.
20479 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
20480 (__builtin_ia32_sqrtss_round): Remove.
20481 (__builtin_ia32_sqrtsd_mask_round)
20482 (__builtin_ia32_sqrtss_mask_round): New builtins.
20483 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
20484 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
20485 ((match_operand:VF_128 1 "vector_operand"
20486 "xBm,<round_constraint>")): Changed to ...
20487 ((match_operand:VF_128 1 "vector_operand"
20488 "xBm,<round_scalar_constraint>")): ... this.
20489 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
20490 %0, %2, %<iptr>1<round_op3>}): Changed to ...
20491 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
20492 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
20493 %<iptr>1<round_scalar_mask_op3>}): ... this.
20494 ((set_attr "prefix" "<round_prefix>")): Changed to ...
20495 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
20497 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
20500 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
20501 Cast vec_cmpeq result to correct type.
20502 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
20503 Cast vec_cmpgt result to correct type.
20505 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
20507 * final.c (final_scan_insn_1): Renamed from...
20508 (final_scan_insn): ... this. New wrapper, to recover
20509 seen from the outermost call in recursive ones.
20510 * config/sparc/sparc.c (output_return): Drop seen from call.
20511 (output_sibcall): Likewise.
20512 * config/visium/visium.c (output_branch): Likewise.
20514 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
20516 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
20519 2018-02-10 Alan Modra <amodra@gmail.com>
20522 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
20523 Specify LR as an input.
20525 2018-02-10 Jakub Jelinek <jakub@redhat.com>
20528 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
20529 remove_member_access_dummy_vars): New functions.
20530 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
20531 lower_omp_1, execute_lower_omp): Use them.
20533 PR rtl-optimization/84308
20534 * shrink-wrap.c (spread_components): Release todo vector.
20536 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
20538 PR rtl-optimization/57193
20539 * ira-color.c (struct allocno_color_data): Add member
20540 conflict_allocno_hard_prefs.
20541 (update_conflict_allocno_hard_prefs): New.
20542 (bucket_allocno_compare_func): Add a preference based on
20543 conflict_allocno_hard_prefs.
20544 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
20545 (color_allocnos): Remove a dead code. Initiate
20546 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
20548 2018-02-09 Jakub Jelinek <jakub@redhat.com>
20551 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
20552 constraint from =wa to wa. Avoid a subreg on the output operand,
20553 instead use a pseudo and subreg it in a move.
20554 (p9_xxbrd_<mode>): Changed to ...
20555 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
20556 (p9_xxbrd_v2df): New expander.
20557 (p9_xxbrw_<mode>): Changed to ...
20558 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
20559 (p9_xxbrw_v4sf): New expander.
20561 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
20563 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
20565 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
20568 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
20569 multiply in 32-bit mode.
20570 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
20571 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
20574 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
20576 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
20577 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
20578 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
20579 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
20581 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
20584 * dwarf2out.c (is_trivial_indirect_ref): New function.
20585 (dwarf2out_late_global_decl): Do not generate a location
20586 attribute for variables that have a non-trivial DECL_VALUE_EXPR
20587 and that are not defined in the current unit.
20589 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
20591 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
20592 instead of a libcall for UNORDERED.
20594 2018-02-09 Tamar Christina <tamar.christina@arm.com>
20597 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
20598 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
20600 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20603 * config/s390/s390.c (s390_set_current_function): Invoke
20604 s390_indirect_branch_settings also if fndecl didn't change.
20606 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20608 * config/rs6000/rs6000.md (blockage): Set length to zero.
20610 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
20612 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
20614 2018-02-09 Jakub Jelinek <jakub@redhat.com>
20617 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
20618 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
20622 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
20623 PARALLEL incoming that failed vt_get_decl_and_offset check.
20625 PR middle-end/84237
20626 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
20627 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
20629 (get_variable_section): For decls in named .bss* sections pass true as
20630 second argument to bss_initializer_p.
20632 2018-02-09 Marek Polacek <polacek@redhat.com>
20633 Jakub Jelinek <jakub@redhat.com>
20636 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
20637 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
20638 Sync some changes from cxx_fold_indirect_ref.
20640 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
20642 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
20644 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
20645 (BLOCK_INLINE_ENTRY_LABEL): New.
20646 (dwarf2out_var_location): Disregard inline entry markers.
20647 (inline_entry_data): New struct.
20648 (inline_entry_data_hasher): New hashtable type.
20649 (inline_entry_data_hasher::hash): New.
20650 (inline_entry_data_hasher::equal): New.
20651 (inline_entry_data_table): New variable.
20652 (add_high_low_attributes): Add DW_AT_entry_pc and
20653 DW_AT_GNU_entry_view attributes if a pending entry is found
20654 in inline_entry_data_table. Add old entry_pc attribute only
20655 if debug nonbinding markers are disabled.
20656 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
20657 markers are enabled.
20658 (block_within_block_p, dwarf2out_inline_entry): New.
20659 (dwarf2out_finish): Check that no entries remained in
20660 inline_entry_data_table.
20661 * final.c (reemit_insn_block_notes): Handle inline entry notes.
20662 (final_scan_insn, notice_source_line): Likewise.
20663 (rest_of_clean_state): Skip inline entry markers.
20664 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
20666 * gimple.c (gimple_build_debug_inline_entry): New.
20667 * gimple.h (enum gimple_debug_subcode): Add
20668 GIMPLE_DEBUG_INLINE_ENTRY.
20669 (gimple_build_debug_inline_entry): Declare.
20670 (gimple_debug_inline_entry_p): New.
20671 (gimple_debug_nonbind_marker_p): Adjust.
20672 * insn-notes.def (INLINE_ENTRY): New.
20673 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
20674 inline entry marker notes.
20675 (print_insn): Likewise.
20676 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
20677 (INSN_DEBUG_MARKER_KIND): Likewise.
20678 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
20679 * tree-inline.c (expand_call_inline): Build and insert
20680 debug_inline_entry stmt.
20681 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
20682 inline entry blocks early, if nonbind markers are enabled.
20683 (dump_scope_block): Dump fragment info.
20684 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
20685 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
20686 (gimple_build_debug_inline_entry): New.
20687 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
20688 Enable/disable inline entry points too.
20689 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
20690 (DEBUG_INSN): Describe inline entry markers.
20692 * common.opt (gvariable-location-views): New.
20693 (gvariable-location-views=incompat5): New.
20694 * config.in: Rebuilt.
20695 * configure: Rebuilt.
20696 * configure.ac: Test assembler for view support.
20697 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
20698 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
20699 * dwarf2out.c (var_loc_view): New typedef.
20700 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
20701 (dwarf2out_locviews_in_attribute): New.
20702 (dwarf2out_locviews_in_loclist): New.
20703 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
20704 (enum dw_line_info_opcode): Add LI_adv_address.
20705 (struct dw_line_info_table): Add view.
20706 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
20707 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
20708 (zero_view_p): New variable.
20709 (ZERO_VIEW_P): New macro.
20710 (output_asm_line_debug_info): New.
20711 (struct var_loc_node): Add view.
20712 (add_AT_view_list, AT_loc_list): New.
20713 (add_var_loc_to_decl): Add view param. Test it against last.
20714 (new_loc_list): Add view params. Record them.
20715 (AT_loc_list_ptr): Handle loc and view lists.
20716 (view_list_to_loc_list_val_node): New.
20717 (print_dw_val): Handle dw_val_class_view_list.
20718 (size_of_die): Likewise.
20719 (value_format): Likewise.
20720 (loc_list_has_views): New.
20721 (gen_llsym): Set vl_symbol too.
20722 (maybe_gen_llsym, skip_loc_list_entry): New.
20723 (dwarf2out_maybe_output_loclist_view_pair): New.
20724 (output_loc_list): Output view list or entries too.
20725 (output_view_list_offset): New.
20726 (output_die): Handle dw_val_class_view_list.
20727 (output_dwarf_version): New.
20728 (output_compilation_unit_header): Use it.
20729 (output_skeleton_debug_sections): Likewise.
20730 (output_rnglists, output_line_info): Likewise.
20731 (output_pubnames, output_aranges): Update version comments.
20732 (output_one_line_info_table): Output view numbers in asm comments.
20733 (dw_loc_list): Determine current endview, pass it to new_loc_list.
20734 Call maybe_gen_llsym.
20735 (loc_list_from_tree_1): Adjust.
20736 (add_AT_location_description): Create view list attribute if
20737 needed, check it's absent otherwise.
20738 (convert_cfa_to_fb_loc_list): Adjust.
20739 (maybe_emit_file): Call output_asm_line_debug_info for test.
20740 (dwarf2out_var_location): Reset views as needed. Precompute
20741 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
20742 attribute. Set view.
20743 (new_line_info_table): Reset next view.
20744 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
20745 (dwarf2out_source_line): Likewise. Output view resets and labels to
20746 the assembler, or select appropriate line info opcodes.
20747 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
20748 (optimize_string_length): Catch it. Adjust.
20749 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
20750 dw_val_class_view_list, and remove it if no longer needed.
20751 (hash_loc_list): Hash view numbers.
20752 (loc_list_hasher::equal): Compare them.
20753 (optimize_location_lists): Check whether a view list symbol is
20754 needed, and whether the locview attribute is present, and
20755 whether they match. Remove the locview attribute if no longer
20757 (index_location_lists): Call skip_loc_list_entry for test.
20758 (dwarf2out_finish): Call output_asm_line_debug_info for test.
20759 Use output_dwarf_version.
20760 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
20761 (struct dw_val_node): Add val_view_list.
20762 * final.c (SEEN_NEXT_VIEW): New.
20763 (set_next_view_needed): New.
20764 (clear_next_view_needed): New.
20765 (maybe_output_next_view): New.
20766 (final_start_function): Rename to...
20767 (final_start_function_1): ... this. Take pointer to FIRST,
20768 add SEEN parameter. Emit param bindings in the initial view.
20769 (final_start_function): Reintroduce SEEN-less interface.
20770 (final): Rename to...
20771 (final_1): ... this. Take SEEN parameter. Output final pending
20772 next view at the end.
20773 (final): Reintroduce seen-less interface.
20774 (final_scan_insn): Output pending next view before switching
20775 sections or ending a block. Mark the next view as needed when
20776 outputting variable locations. Notify debug backend of section
20777 changes, and of location view changes.
20778 (rest_of_handle_final): Adjust.
20779 * toplev.c (process_options): Autodetect value for debug variable
20780 location views option. Warn on incompat5 without -gdwarf-5.
20781 * doc/invoke.texi (gvariable-location-views): New.
20782 (gvariable-location-views=incompat5): New.
20783 (gno-variable-location-views): New.
20785 2018-02-08 David Malcolm <dmalcolm@redhat.com>
20787 PR tree-optimization/84136
20788 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
20789 that the result of find_edge is non-NULL.
20791 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20794 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
20795 storing integer register in SImode. Fix cost of 256 and 512
20796 byte aligned SSE register store.
20798 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
20800 * config/i386/i386.c (ix86_multiplication_cost): Fix
20801 multiplication cost for TARGET_AVX512DQ.
20803 2018-02-08 Marek Polacek <polacek@redhat.com>
20805 PR tree-optimization/84238
20806 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
20809 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20811 PR tree-optimization/84265
20812 * tree-vect-stmts.c (vectorizable_store): Don't treat
20813 VMAT_CONTIGUOUS accesses as grouped.
20814 (vectorizable_load): Likewise.
20816 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20818 PR tree-optimization/81635
20819 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
20820 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
20821 (test_round_for_mask): New functions.
20822 (wide_int_cc_tests): Call test_round_for_mask.
20823 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
20824 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
20825 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
20826 range returned by get_range_info.
20828 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
20831 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
20832 * symtab.c: Include builtins.h
20833 (symtab_node::output_to_lto_symbol_table_p): Move here
20834 from lto-streamer-out.c:output_symbol_p.
20835 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
20836 (output_symbol_p): Move all logic to symtab.c
20837 (produce_symtab): Update.
20839 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
20841 * config/s390/s390-opts.h (enum indirect_branch): Define.
20842 * config/s390/s390-protos.h (s390_return_addr_from_memory)
20843 (s390_indirect_branch_via_thunk)
20844 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
20845 (enum s390_indirect_branch_type): Define.
20846 * config/s390/s390.c (struct s390_frame_layout, struct
20847 machine_function): Remove.
20848 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
20849 (indirect_branch_table_label_no, indirect_branch_table_name):
20851 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
20852 (enum s390_indirect_branch_option): Define.
20853 (s390_return_addr_from_memory): New function.
20854 (s390_handle_string_attribute): New function.
20855 (s390_attribute_table): Add new attribute handler.
20856 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
20857 (s390_indirect_branch_via_thunk): New function.
20858 (s390_indirect_branch_via_inline_thunk): New function.
20859 (s390_function_ok_for_sibcall): When jumping via thunk disallow
20860 sibling call optimization for non z10 compiles.
20861 (s390_emit_call): Force indirect branch target to be a single
20862 register. Add r1 clobber for non-z10 compiles.
20863 (s390_emit_epilogue): Emit return jump via return_use expander.
20864 (s390_reorg): Handle JUMP_INSNs as execute targets.
20865 (s390_option_override_internal): Perform validity checks for the
20866 new command line options.
20867 (s390_indirect_branch_attrvalue): New function.
20868 (s390_indirect_branch_settings): New function.
20869 (s390_set_current_function): Invoke s390_indirect_branch_settings.
20870 (s390_output_indirect_thunk_function): New function.
20871 (s390_code_end): Implement target hook.
20872 (s390_case_values_threshold): Implement target hook.
20873 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
20875 * config/s390/s390.h (struct s390_frame_layout)
20876 (struct machine_function): Move here from s390.c.
20877 (TARGET_INDIRECT_BRANCH_NOBP_RET)
20878 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
20879 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
20880 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
20881 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
20882 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
20883 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
20884 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
20885 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
20886 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
20887 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
20888 (mnemonic attribute): Add values which aren't recognized
20890 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
20891 pattern for branch conversion. Fix mnemonic attribute.
20892 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
20893 indirect branch via thunk if requested.
20894 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
20895 ("*indirect_jump"): Disable for branch conversion using out of
20897 ("indirect_jump_via_thunk<mode>_z10")
20898 ("indirect_jump_via_thunk<mode>")
20899 ("indirect_jump_via_inlinethunk<mode>_z10")
20900 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
20901 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
20902 ("casesi_jump_via_inlinethunk<mode>_z10")
20903 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
20904 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
20905 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
20906 ("*indirect2_jump"): Disable for branch conversion.
20907 ("casesi_jump"): Turn into expander and expand patterns for branch
20909 ("return_use"): New expander.
20910 ("*return"): Emit return via thunk and rename it to ...
20911 ("*return<mode>"): ... this one.
20912 * config/s390/s390.opt: Add new options and and enum for the
20915 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
20917 * lra-constraints.c (match_reload): Unconditionally use
20918 gen_lowpart_SUBREG, rather than selecting between that
20919 and equivalent gen_rtx_SUBREG code.
20921 2018-02-08 Richard Biener <rguenther@suse.de>
20923 PR tree-optimization/84233
20924 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
20925 changed flag instead of boguously re-using phi_inserted.
20927 2018-02-08 Martin Jambor <mjambor@suse.cz>
20929 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
20930 static local variables.
20932 2018-02-08 Richard Biener <rguenther@suse.de>
20934 PR tree-optimization/84278
20935 * tree-vect-stmts.c (vectorizable_store): When looking for
20936 smaller vector types to perform grouped strided loads/stores
20937 make sure the mode is supported by the target.
20938 (vectorizable_load): Likewise.
20940 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
20942 * config/aarch64/aarch64.c (aarch64_components_for_bb):
20943 Increase LDP/STP opportunities by adding adjacent callee-saves.
20945 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
20947 PR rtl-optimization/84068
20948 PR rtl-optimization/83459
20949 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
20951 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
20953 PR tree-optimization/84224
20954 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
20955 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
20956 non-zero arguments.
20958 2018-02-07 Iain Sandoe <iain@codesourcery.com>
20961 * config/rs6000/altivec.md (*restore_world): Remove LR use.
20962 * config/rs6000/predicates.md (restore_world_operation): Adjust op
20963 count, remove one USE.
20965 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
20967 * doc/install.texi (Configuration): Document the
20968 --with-long-double-format={ibm,ieee} PowerPC configuration
20972 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
20973 Convert from define_expand to be define_insn_and_split. Rework
20974 float/double/_Float128 conversions to QI/HI/SImode to work with
20975 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
20976 conversions to QI/HImode types did a store and then a load to
20977 truncate the value. For conversions to VSX registers, don't split
20978 the insn, instead emit the code directly. Use the code iterator
20979 any_fix to combine signed and unsigned conversions.
20980 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
20981 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
20982 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20983 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
20984 (fix_<mode>di2_hw): Likewise.
20985 (fixuns_<mode>di2_hw): Likewise.
20986 (fix_<mode>si2_hw): Likewise.
20987 (fixuns_<mode>si2_hw): Likewise.
20988 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
20989 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20990 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
20991 fix<uns>_trunc<SFDF:mode>si2_p8.
20992 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
20993 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
20994 (fix<uns>_<mode>_mem): Likewise.
20995 (fctiw<u>z_<mode>_mem): Likewise.
20996 (fix<uns>_<mode>_mem): Likewise.
20997 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
20998 the register allocator from doing a direct move to the GPRs to do
20999 a store, and instead use the ISA 3.0 store byte/half-word from
21000 vector register instruction. For IEEE 128-bit floating point,
21001 also optimize stores of 32-bit ints.
21002 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
21004 2018-02-07 Alan Hayward <alan.hayward@arm.com>
21006 * genextract.c (push_pathstr_operand): New function to support
21008 (walk_rtx): Call push_pathstr_operand.
21009 (print_path): Support [a-zA-Z].
21011 2018-02-07 Richard Biener <rguenther@suse.de>
21013 PR tree-optimization/84037
21014 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
21015 (cse_and_gimplify_to_preheader): Declare.
21016 (vect_get_place_in_interleaving_chain): Likewise.
21017 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
21019 (_loop_vec_info::~_loop_vec_info): Delete it.
21020 (cse_and_gimplify_to_preheader): New function.
21021 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
21022 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
21023 (vectorizable_load): Likewise. For grouped stores always base
21024 the IV on the first element.
21025 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
21026 condition before gimplifying.
21028 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21030 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
21031 *DIV_EXPR and *MOD_EXPR.
21033 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
21036 * config/i386/i386.c (ix86_option_override_internal): Mask out
21037 the CF_SET bit when checking -fcf-protection.
21039 2018-02-07 Tom de Vries <tom@codesourcery.com>
21042 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
21045 2018-02-07 Richard Biener <rguenther@suse.de>
21047 PR tree-optimization/84204
21048 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
21051 PR tree-optimization/84205
21052 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
21053 special-case isl_ast_op_zdiv_r.
21055 PR tree-optimization/84223
21056 * graphite-scop-detection.c (gather_bbs::before_dom_children):
21057 Only add conditions from within the region.
21058 (gather_bbs::after_dom_children): Adjust.
21060 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
21063 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
21064 * config/avr/avr.md: Only post-reload split REG-REG moves if
21065 either register is GENERAL_REG_P.
21067 2018-02-07 Jakub Jelinek <jakub@redhat.com>
21069 PR tree-optimization/84235
21070 * tree-ssa-scopedtables.c
21071 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
21072 if the subtraction is performed in floating point type where NaNs are
21073 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
21074 build 1. Formatting fix.
21076 2018-02-06 Jakub Jelinek <jakub@redhat.com>
21079 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
21080 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
21081 and skip it regardless of bb boundaries. Use CALL_P macro,
21082 don't test INSN_P (insn) together with CALL_P or JUMP_P check
21083 unnecessarily, formatting fix.
21085 2018-02-06 Michael Collison <michael.collison@arm.com>
21087 * config/arm/thumb2.md:
21088 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
21089 (*thumb_mov_notscc): Ditto.
21091 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
21094 * config/rs6000/rs6000.md (su code attribute): Use "u" for
21095 unsigned_fix, not "s".
21097 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21099 * configure.ac (gcc_fn_eh_frame_ro): New function.
21100 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21101 correct .eh_frame permissions.
21102 * configure: Regenerate.
21104 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
21106 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21107 irrelevant options.
21109 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21111 * config/rs6000/rs6000.c (rs6000_option_override_internal):
21112 Display warning message for -mno-speculate-indirect-jumps.
21114 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
21116 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21118 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21120 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
21122 PR tree-optimization/84225
21123 * tree-eh.c (find_trapping_overflow): Only call
21124 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21126 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21129 * config/i386/i386.c: Reimplement the check of possible options
21130 -mibt/-mshstk conbination. Change error messages.
21131 * doc/invoke.texi: Fix a typo: remove extra '='.
21133 2018-02-06 Marek Polacek <polacek@redhat.com>
21135 PR tree-optimization/84228
21136 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21138 2018-02-06 Tamar Christina <tamar.christina@arm.com>
21141 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21142 emitted arch directives.
21143 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21144 __ARM_FEATURE_COPROC before changing architectures.
21146 2018-02-06 Richard Biener <rguenther@suse.de>
21148 * config/i386/i386.c (print_reg): Fix typo.
21149 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21151 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
21153 * configure: Regenerate.
21155 2018-02-05 Martin Sebor <msebor@redhat.com>
21157 PR tree-optimization/83369
21158 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21161 2018-02-05 Martin Liska <mliska@suse.cz>
21163 * doc/invoke.texi: Cherry-pick upstream r323995.
21165 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
21167 * ira.c (ira_init_register_move_cost): Adjust comment.
21169 2018-02-05 Martin Liska <mliska@suse.cz>
21171 PR gcov-profile/84137
21172 * doc/gcov.texi: Fix typo in documentation.
21174 2018-02-05 Martin Liska <mliska@suse.cz>
21176 PR gcov-profile/83879
21177 * doc/gcov.texi: Document necessity of --dynamic-list-data when
21178 using dlopen functionality.
21180 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
21182 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21183 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21184 _mm_maskz_range_ss, _mm_mask_range_round_ss,
21185 _mm_maskz_range_round_ss): New intrinsics.
21186 (__builtin_ia32_rangesd128_round)
21187 (__builtin_ia32_rangess128_round): Remove.
21188 (__builtin_ia32_rangesd128_mask_round,
21189 __builtin_ia32_rangess128_mask_round): New builtins.
21190 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21191 __builtin_ia32_rangess128_round): Remove.
21192 (__builtin_ia32_rangesd128_mask_round,
21193 __builtin_ia32_rangess128_mask_round): New builtins.
21194 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21195 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21196 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21197 "<round_saeonly_constraint>")): Changed to ...
21198 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21199 "<round_saeonly_scalar_constraint>")): ... this.
21200 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21201 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21202 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21203 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21204 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21206 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
21208 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21210 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21211 Remove all values except native, 8540 and 8548.
21213 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
21215 * config/i386/i386.c (ix86_output_function_return): Pass
21216 INVALID_REGNUM, instead of -1, as invalid register number to
21217 indirect_thunk_name and output_indirect_thunk.
21219 2018-02-02 Julia Koval <julia.koval@intel.com>
21221 * config.gcc: Add -march=icelake.
21222 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21223 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21224 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21225 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21226 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21227 (processor_target_table): Add icelake.
21228 (ix86_option_override_internal): Handle new PTAs.
21229 (get_builtin_code_for_version): Handle icelake.
21230 (M_INTEL_COREI7_ICELAKE): New.
21231 (fold_builtin_cpu): Handle icelake.
21232 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21233 * doc/invoke.texi: Add -march=icelake.
21235 2018-02-02 Julia Koval <julia.koval@intel.com>
21237 * config/i386/i386.c (ix86_option_override_internal): Change flags type
21238 to wide_int_bitmask.
21239 * wide-int-bitmask.h: New.
21241 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
21244 * config/i386/i386.md: Replace Pmode with word_mode in
21245 builtin_setjmp_setup and builtin_longjmp to support x32.
21247 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
21251 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
21253 (rs6000_supported_cpu_names): New static variable.
21254 (linux_cpu_translation_table): Likewise.
21255 (elf_platform) <cpu>: Define new static variable and use it.
21256 Translate kernel AT_PLATFORM name to canonical name if needed.
21257 Error if platform name is unknown.
21259 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
21262 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
21264 2018-02-01 Jeff Law <law@redhat.com>
21267 * config/i386/i386.c (release_scratch_register_on_entry): Add new
21268 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
21269 the scratch if RELEASE_VIA_POP is false.
21270 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
21271 If we have to save a temporary register, decrement SIZE appropriately.
21272 Pass new arguments to release_scratch_register_on_entry.
21273 (ix86_adjust_stack_and_probe): Likewise.
21274 (ix86_emit_probe_stack_range): Pass new arguments to
21275 release_scratch_register_on_entry.
21277 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
21279 PR rtl-optimization/84157
21280 * combine.c (change_zero_ext): Use REG_P predicate in
21281 front of HARD_REGISTER_P predicate.
21283 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
21285 * config/avr/avr.c (avr_option_override): Move disabling of
21286 -fdelete-null-pointer-checks to...
21287 * common/config/avr/avr-common.c (avr_option_optimization_table):
21290 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21292 PR tree-optimization/81635
21293 * tree-data-ref.c (split_constant_offset_1): For types that
21294 wrap on overflow, try to use range info to prove that wrapping
21297 2018-02-01 Renlin Li <renlin.li@arm.com>
21300 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
21301 TAILCALL_ADDR_REGS.
21302 (aarch64_register_move_cost): Likewise.
21303 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
21304 TAILCALL_ADDR_REGS.
21305 (REG_CLASS_NAMES): Likewise.
21306 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
21307 TAILCALL_ADDR_REGS. Remove IP registers.
21308 * config/aarch64/aarch64.md (Ucs): Update register constraint.
21310 2018-02-01 Richard Biener <rguenther@suse.de>
21312 * domwalk.h (dom_walker::dom_walker): Add additional constructor
21313 for specifying RPO order and allow NULL for that.
21314 * domwalk.c (dom_walker::dom_walker): Likewise.
21315 (dom_walker::walk): Handle NULL RPO order.
21316 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
21318 (rewrite_update_dom_walker): Likewise.
21319 (mark_def_dom_walker): Likewise.
21321 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21323 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
21324 (aarch64_maybe_expand_sve_subreg_move): Declare.
21325 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
21326 * config/aarch64/predicates.md (aarch64_any_register_operand): New
21328 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
21329 that are semantically a reverse operation.
21330 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
21331 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
21332 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
21334 (aarch64_can_change_mode_class): For big-endian, forbid changes
21335 between two SVE modes if they have different element sizes.
21337 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21339 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
21340 the TImode handling for big-endian targets.
21342 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21344 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
21345 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
21347 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
21348 Remove BSWAP handing for big-endian targets and use the form of
21349 LD1RQ appropariate for the mode.
21351 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21353 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
21354 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
21355 duplicated element.
21357 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
21360 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
21361 check for operands that need to go through aarch64_sve_reload_be.
21363 2018-02-01 Jakub Jelinek <jakub@redhat.com>
21365 PR tree-optimization/81661
21366 PR tree-optimization/84117
21367 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
21368 * tree-eh.c: Include gimplify.h.
21369 (find_trapping_overflow, replace_trapping_overflow,
21370 rewrite_to_non_trapping_overflow): New functions.
21371 * tree-vect-loop.c: Include tree-eh.h.
21372 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
21373 * tree-data-ref.c: Include tree-eh.h.
21374 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
21376 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
21378 PR rtl-optimization/84123
21379 * combine.c (change_zero_ext): Check if hard register satisfies
21380 can_change_dest_mode before calling gen_lowpart_SUBREG.
21382 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
21385 * ira.c (ira_init_register_move_cost): Remove assert.
21387 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21389 PR rtl-optimization/84071
21390 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
21391 * doc/tm.texi: Regenerate.
21393 2018-01-31 Richard Biener <rguenther@suse.de>
21395 PR tree-optimization/84132
21396 * tree-data-ref.c (analyze_miv_subscript): Properly
21397 check whether evolution_function_is_affine_multivariate_p
21398 before calling gcd_of_steps_may_divide_p.
21400 2018-01-31 Julia Koval <julia.koval@intel.com>
21403 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
21404 * config/i386/i386.md (rdpid_rex64) New.
21405 (rdpid): Make 32bit only.
21407 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
21410 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
21411 an IDENTIFIER_NODE for FUNCTION_TYPE's.
21413 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21416 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
21418 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21420 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
21422 PR rtl-optimization/84071
21423 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
21424 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
21426 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
21428 * config/arc/arc.c (arc_handle_aux_attribute): New function.
21429 (arc_attribute_table): Add 'aux' attribute.
21430 (arc_in_small_data_p): Consider aux like variables.
21431 (arc_is_aux_reg_p): New function.
21432 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
21433 (arc_get_aux_arg): New function.
21434 (prepare_move_operands): Handle aux-register access.
21435 (arc_handle_aux_attribute): New function.
21436 * doc/extend.texi (ARC Variable attributes): Add subsection.
21438 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
21440 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
21441 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
21442 (arc_attribute_table): Add 'uncached' attribute.
21443 (arc_print_operand): Print '.di' flag for uncached memory
21445 (arc_in_small_data_p): Do not consider for small data the uncached
21447 (arc_is_uncached_mem_p): New function.
21448 * config/arc/predicates.md (compact_store_memory_operand): Check
21449 for uncached memory accesses.
21450 (nonvol_nonimm_operand): Likewise.
21451 * doc/extend.texi (ARC Type Attribute): New subsection.
21453 2018-01-31 Jakub Jelinek <jakub@redhat.com>
21456 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
21457 falign-loops=): Add Optimization flag.
21459 2018-01-30 Jeff Law <law@redhat.com>
21462 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
21463 INT_REGISTERS_SAVED. Check it prior to calling
21464 get_scratch_register_on_entry.
21465 (ix86_adjust_stack_and_probe): Similarly.
21466 (ix86_emit_probe_stack_range): Similarly.
21467 (ix86_expand_prologue): Corresponding changes.
21469 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21472 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
21473 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
21475 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
21478 * lra-constraints.c (curr_insn_transform): Process AND in the
21481 2018-01-30 Jakub Jelinek <jakub@redhat.com>
21483 PR rtl-optimization/83986
21484 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
21485 dependence against last_pending_memory_flush in addition to
21486 pending_jump_insns.
21488 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
21490 PR tree-optimization/81611
21491 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
21494 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
21497 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
21500 2018-01-30 Richard Biener <rguenther@suse.de>
21501 Jakub Jelinek <jakub@redhat.com>
21503 PR tree-optimization/84111
21504 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
21505 inner loops added during recursion, as they don't have up-to-date
21508 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
21511 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
21512 (can_inline_edge_by_limits_p): ... here.
21513 (can_early_inline_edge_p, check_callers,
21514 update_caller_keys, update_callee_keys, recursive_inlining,
21515 add_new_edges_to_heap, speculation_useful_p,
21516 inline_small_functions,
21517 inline_small_functions, flatten_function,
21518 inline_to_all_callers_1): Update.
21520 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
21522 * profile-count.c (profile_count::combine_with_ipa_count): Handle
21525 2018-01-30 Richard Biener <rguenther@suse.de>
21527 PR tree-optimization/83008
21528 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
21529 invariant and constant vector uses in stmts when they need
21530 more than one stmt.
21532 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21535 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
21536 * configure: Regenerate.
21538 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21540 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
21542 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
21543 Use gen_rtx_REG rather than gen_lowpart.
21545 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21547 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
21548 rather than 0 when creating partial subregs.
21550 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
21552 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
21555 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
21558 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
21559 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
21560 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
21561 flags. This restores the settings used before the 2017-07-24.
21562 Turning off pre increment/decrement/modify allows IVOPTS to
21563 optimize DF/SF loops where the index is an int.
21565 2018-01-29 Richard Biener <rguenther@suse.de>
21566 Kelvin Nilsen <kelvin@gcc.gnu.org>
21569 * tree-vect-stmts.c (vectorizable_call): Don't call
21570 targetm.vectorize_builtin_md_vectorized_function if callee is
21573 2018-01-22 Carl Love <cel@us.ibm.com>
21575 * doc/extend.tex: Fix typo in second arg in
21576 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
21578 2018-01-29 Richard Biener <rguenther@suse.de>
21580 PR tree-optimization/84086
21581 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
21582 (flush_ssaname_freelist): When SSA names were released reset
21583 the SCEV hash table.
21585 2018-01-29 Richard Biener <rguenther@suse.de>
21587 PR tree-optimization/84057
21588 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
21589 removed paths when removing edges.
21591 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
21593 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
21594 -mfunction-return=@var{choice}.
21596 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
21598 PR diagnostic/84034
21599 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
21600 Handle CR like TAB.
21601 (layout::print_source_line): Likewise.
21602 (test_get_line_width_without_trailing_whitespace): Add test cases.
21604 2018-01-27 Jakub Jelinek <jakub@redhat.com>
21606 PR middle-end/84040
21607 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
21610 2018-01-26 Jim Wilson <jimw@sifive.com>
21612 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
21614 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
21617 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21619 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
21620 and CMP + SUB-immediate -> SUBS.
21622 2018-01-26 Martin Sebor <msebor@redhat.com>
21624 PR tree-optimization/83896
21625 * tree-ssa-strlen.c (get_string_len): Rename...
21626 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
21627 Avoid assuming length is constant.
21628 (handle_char_store): Use HOST_WIDE_INT for string length.
21630 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
21633 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
21634 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
21636 2018-01-26 Richard Biener <rguenther@suse.de>
21638 PR rtl-optimization/84003
21639 * dse.c (record_store): Only record redundant stores when
21640 the earlier store aliases at least all accesses the later one does.
21642 2018-01-26 Jakub Jelinek <jakub@redhat.com>
21644 PR rtl-optimization/83985
21645 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
21646 REG_CFA_RESTORE insns.
21647 (delete_unmarked_insns): Don't ignore separate shrink wrapping
21648 REG_CFA_RESTORE insns here.
21651 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
21652 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
21654 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21656 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
21657 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
21658 (arc_init): Likewise.
21659 (arc_override_options): Likewise.
21660 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
21662 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
21664 * config/arc/arc.h (TARGET_DBNZ): Define.
21665 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
21666 properly set the tune attribute.
21667 (dbnz): Use TARGET_DBNZ guard.
21668 * config/arc/arc.opt (mtune): Add core3 option.
21670 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21672 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
21673 recognize new pic like addresses.
21674 (arc_delegitimize_address): Clean up.
21676 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21678 * config/arc/arc-arches.def: Option mrf16 valid for all
21680 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
21681 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
21682 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
21683 * config/arc/arc-tables.opt: Regenerate.
21684 * config/arc/arc.c (arc_conditional_register_usage): Handle
21685 reduced register file case.
21686 (arc_file_start): Set must have build attributes.
21687 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
21688 mrf16 option value.
21689 * config/arc/arc.opt (mrf16): Add new option.
21690 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
21691 * config/arc/genmultilib.awk: Handle new mrf16 option.
21692 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
21693 * config/arc/t-multilib: Regenerate.
21694 * doc/invoke.texi (ARC Options): Document mrf16 option.
21696 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21698 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
21699 * config/arc/arc.c (arc_handle_secure_attribute): New function.
21700 (arc_attribute_table): Add 'secure_call' attribute.
21701 (arc_print_operand): Print secure call operand.
21702 (arc_function_ok_for_sibcall): Don't optimize tail calls when
21704 (arc_is_secure_call_p): New function. * config/arc/arc.md
21705 (call_i): Add support for sjli instruction.
21706 (call_value_i): Likewise.
21707 * config/arc/constraints.md (Csc): New constraint.
21709 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
21710 John Eric Martin <John.Martin@emmicro-us.com>
21712 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
21713 * config/arc/arc.c (_arc_jli_section): New struct.
21714 (arc_jli_section): New type.
21715 (rc_jli_sections): New static variable.
21716 (arc_handle_jli_attribute): New function.
21717 (arc_attribute_table): Add jli_always and jli_fixed attribute.
21718 (arc_file_end): New function.
21719 (TARGET_ASM_FILE_END): Define.
21720 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
21721 (arc_add_jli_section): New function.
21722 (jli_call_scan): Likewise.
21723 (arc_reorg): Call jli_call_scan.
21724 (arc_output_addsi): Remove 'S' from printing asm operand.
21725 (arc_is_jli_call_p): New function.
21726 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
21728 (movhi_insn): Likewise.
21729 (movsi_insn): Likewise.
21730 (movsi_set_cc_insn): Likewise.
21731 (loadqi_update): Likewise.
21732 (load_zeroextendqisi_update): Likewise.
21733 (load_signextendqisi_update): Likewise.
21734 (loadhi_update): Likewise.
21735 (load_zeroextendhisi_update): Likewise.
21736 (load_signextendhisi_update): Likewise.
21737 (loadsi_update): Likewise.
21738 (loadsf_update): Likewise.
21739 (movsicc_insn): Likewise.
21740 (bset_insn): Likewise.
21741 (bxor_insn): Likewise.
21742 (bclr_insn): Likewise.
21743 (bmsk_insn): Likewise.
21744 (bicsi3_insn): Likewise.
21745 (cmpsi_cc_c_insn): Likewise.
21746 (movsi_ne): Likewise.
21747 (movsi_cond_exec): Likewise.
21748 (clrsbsi2): Likewise.
21749 (norm_f): Likewise.
21759 (call_i): Remove 'S' asm letter, add jli instruction.
21760 (call_value_i): Likewise.
21761 * config/arc/arc.op (mjli-always): New option.
21762 * config/arc/constraints.md (Cji): New constraint.
21763 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
21765 (subsf3_fpx): Likewise.
21766 (mulsf3_fpx): Likewise.
21767 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
21769 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
21770 function attrbutes.
21771 * doc/invoke.texi (ARC): Document mjli-always option.
21773 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
21775 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
21776 avoid addition with 0 and use incw and decw where possible.
21778 2018-01-26 Richard Biener <rguenther@suse.de>
21780 PR tree-optimization/81082
21781 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
21782 association if it requires casting to unsigned.
21783 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
21784 from fold_plusminus_mult_expr to catch important cases late when
21785 range info is available.
21787 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
21789 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
21790 * configure.ac (hidden_linkonce): New test.
21791 * configure: Regenerate.
21792 * config.in: Regenerate.
21794 2018-01-26 Julia Koval <julia.koval@intel.com>
21796 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
21797 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
21798 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
21799 _mm_mask_bitshuffle_epi64_mask): Fix type.
21800 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
21801 USI_FTYPE_V4DI_V4DI_USI): Remove.
21802 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
21803 __builtin_ia32_vpshufbitqmb256_mask,
21804 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
21805 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
21806 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
21808 2018-01-26 Alan Modra <amodra@gmail.com>
21811 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
21812 UNSPEC_VBPERMQ. Sort other unspecs.
21814 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
21816 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
21818 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
21820 PR middle-end/83055
21821 * predict.c (drop_profile): Do not push/pop cfun; update also
21823 (handle_missing_profiles): Fix logic looking for zero profiles.
21825 2018-01-25 Jakub Jelinek <jakub@redhat.com>
21827 PR middle-end/83977
21828 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
21829 on functions with #pragma omp declare simd or functions with simd
21831 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
21832 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
21833 Remove trailing \n from warning_at calls.
21835 2018-01-25 Tom de Vries <tom@codesourcery.com>
21838 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21839 for neutered workers.
21841 2018-01-24 Joseph Myers <joseph@codesourcery.com>
21844 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
21845 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
21847 2018-01-24 Jeff Law <law@redhat.com>
21850 * i386.c (get_probe_interval): Move to earlier point.
21851 (ix86_compute_frame_layout): If -fstack-clash-protection and
21852 the frame is larger than the probe interval, then use pushes
21853 to save registers rather than reg->mem moves.
21854 (ix86_expand_prologue): Remove conditional for int_registers_saved
21857 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
21860 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
21861 min/max for never referenced object.
21863 2018-01-24 Jakub Jelinek <jakub@redhat.com>
21865 PR middle-end/83977
21866 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
21868 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
21869 attributes from DECL_ATTRIBUTES (decl) without affecting
21870 DECL_ATTRIBUTES (current_function_decl).
21871 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
21872 functions with non-NULL DECL_ABSTRACT_ORIGIN.
21874 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
21876 PR tree-optimization/83979
21877 * fold-const.c (fold_comparison): Use constant_boolean_node
21878 instead of boolean_{true,false}_node.
21880 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
21882 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
21885 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
21887 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21888 Simplify the clause that sets the length attribute.
21889 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21890 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
21891 clause that sets the length attribute.
21892 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21894 2018-01-24 Tom de Vries <tom@codesourcery.com>
21897 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
21898 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
21899 Add strict parameter.
21900 (prevent_branch_around_nothing): Insert dummy insn between branch to
21901 label and label with no ptx insn inbetween.
21902 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
21904 2018-01-24 Tom de Vries <tom@codesourcery.com>
21907 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21908 for neutered threads in warp.
21909 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
21911 2018-01-24 Richard Biener <rguenther@suse.de>
21913 PR tree-optimization/83176
21914 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
21917 2018-01-24 Richard Biener <rguenther@suse.de>
21919 PR tree-optimization/82819
21920 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
21921 code generating pluses that are no-ops in the target precision.
21923 2018-01-24 Richard Biener <rguenther@suse.de>
21925 PR middle-end/84000
21926 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
21928 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
21930 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
21931 to merge probabilities.
21932 * predict.c (probably_never_executed): Also mark as cold functions
21933 with global 0 profile and guessed local profile.
21934 * profile-count.c (profile_probability::combine_with_count): New
21936 * profile-count.h (profile_probability::operator*,
21937 profile_probability::operator*=, profile_probability::operator/,
21938 profile_probability::operator/=): Reduce precision to adjusted
21939 and set value to guessed on contradictory divisions.
21940 (profile_probability::combine_with_freq): Remove.
21941 (profile_probability::combine_wiht_count): Declare.
21942 (profile_count::force_nonzero):: Set to adjusted.
21943 (profile_count::probability_in):: Set quality to adjusted.
21944 * tree-ssa-tail-merge.c (replace_block_by): Use
21945 combine_with_count.
21947 2018-01-23 Andrew Waterman <andrew@sifive.com>
21948 Jim Wilson <jimw@sifive.com>
21950 * config/riscv/riscv.c (riscv_stack_boundary): New.
21951 (riscv_option_override): Set riscv_stack_boundary. Handle
21952 riscv_preferred_stack_boundary_arg.
21953 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
21954 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
21955 (STACK_BOUNDARY): Set to riscv_stack_boundary.
21956 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
21957 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
21958 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
21960 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
21963 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
21964 of struct ix86_frame.
21965 (ix86_expand_epilogue): Likewise. Add a local variable for
21966 the reg_save_offset field in struct ix86_frame.
21968 2018-01-23 Bin Cheng <bin.cheng@arm.com>
21970 PR tree-optimization/82604
21971 * tree-loop-distribution.c (enum partition_kind): New enum item
21972 PKIND_PARTIAL_MEMSET.
21973 (partition_builtin_p): Support above new enum item.
21974 (generate_code_for_partition): Ditto.
21975 (compute_access_range): Differentiate cases that equality can be
21976 proven at all loops, the innermost loops or no loops.
21977 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
21978 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
21979 (finalize_partitions, distribute_loop): Don't fuse partition of
21980 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
21981 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
21982 parloop is enabled.
21984 2018-01-23 Martin Liska <mliska@suse.cz>
21986 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
21987 order to ignore the predictor.
21988 (PRED_POLYMORPHIC_CALL): Likewise.
21989 (PRED_RECURSIVE_CALL): Likewise.
21991 2018-01-23 Martin Liska <mliska@suse.cz>
21993 * tree-profile.c (tree_profiling): Print function header to
21994 aware reader which function we are working on.
21995 * value-prof.c (gimple_find_values_to_profile): Do not print
21996 not interesting value histograms.
21998 2018-01-23 Martin Liska <mliska@suse.cz>
22000 * profile-count.h (enum profile_quality): Add
22001 profile_uninitialized as the first value. Do not number values
22002 as they are zero based.
22003 (profile_count::verify): Update sanity check.
22004 (profile_probability::verify): Likewise.
22006 2018-01-23 Nathan Sidwell <nathan@acm.org>
22008 * doc/invoke.texi (ffor-scope): Deprecate.
22010 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22012 PR tree-optimization/83510
22013 * domwalk.c (set_all_edges_as_executable): New function.
22014 (dom_walker::dom_walker): Convert bool param
22015 "skip_unreachable_blocks" to enum reachability. Move setup of
22016 edge flags to set_all_edges_as_executable and only do it when
22017 reachability is REACHABLE_BLOCKS.
22018 * domwalk.h (enum dom_walker::reachability): New enum.
22019 (dom_walker::dom_walker): Convert bool param
22020 "skip_unreachable_blocks" to enum reachability.
22021 (set_all_edges_as_executable): New decl.
22022 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
22023 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
22025 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
22026 but converting true to REACHABLE_BLOCKS.
22027 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
22029 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
22030 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
22031 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
22033 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
22034 if check_all_array_refs will be called.
22036 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22038 * tree.c (selftest::test_location_wrappers): Add more test
22041 2018-01-23 David Malcolm <dmalcolm@redhat.com>
22043 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
22044 (selftest::test_bit_in_range): Likewise.
22046 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22049 * doc/sourcebuild.texi (vect_float): Say that the selector
22050 only describes the situation when -funsafe-math-optimizations is on.
22051 (vect_float_strict): Document.
22053 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
22055 PR tree-optimization/83965
22056 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
22057 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
22058 instead of checking only for a reduction.
22059 (vect_recog_widen_sum_pattern): Likewise.
22061 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22063 * predict.c (probably_never_executed): Only use precise profile info.
22064 (compute_function_frequency): Skip after inlining hack since we now
22065 have quality checking.
22067 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
22069 * profile-count.h (profile_probability::very_unlikely,
22070 profile_probability::unlikely, profile_probability::even): Set
22071 precision to guessed.
22073 2018-01-23 Richard Biener <rguenther@suse.de>
22075 PR tree-optimization/83963
22076 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
22077 Properly terminate dominator walk when crossing the exit edge not
22078 when visiting its source block.
22080 2018-01-23 Jakub Jelinek <jakub@redhat.com>
22083 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
22084 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
22086 2018-01-22 Jakub Jelinek <jakub@redhat.com>
22088 PR tree-optimization/83957
22089 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
22090 semicolon after for body surrounded by braces.
22092 PR tree-optimization/83081
22093 * profile-count.h (profile_probability::split): New method.
22094 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
22095 Use profile_probability::split.
22096 (do_compare_rtx_and_jump): Fix adjustment of probabilities
22097 when splitting a single conditional jump into 2.
22099 2018-01-22 David Malcolm <dmalcolm@redhat.com>
22101 PR tree-optimization/69452
22102 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22105 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22107 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22108 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22109 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22111 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22113 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22115 * config/rl78/rl78.md (movdi): New define_expand.
22116 * config/rl78/rl78.c (rl78_split_movdi): New function.
22118 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
22121 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22123 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22124 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22125 128-bit to produce an UNSPEC move to get the double word with the
22126 signbit and then a shift directly to do signbit.
22127 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22128 implementation with a new version that just does either a direct
22129 move or a regular move. Move memory interface to separate insns.
22130 Move insns so they are next to the expander.
22131 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22132 with signbit move. Split big and little endian case.
22133 (signbit<mode>2_dm_mem_le): Likewise.
22134 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22135 (signbit<mode>2_dm2): Likewise.
22137 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22139 * config/rl78/rl78.md (anddi3): New define_expand.
22141 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22143 * config/rl78/rl78.md (umindi3): New define_expand.
22145 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22147 * config/rl78/rl78.md (smindi3): New define_expand.
22149 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22151 * config/rl78/rl78.md (smaxdi3): New define_expand.
22153 2018-01-22 Carl Love <cel@us.ibm.com>
22155 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22156 LVX_V1TI): Add macro expansion.
22157 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22158 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22159 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22160 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22161 Change check to determine if the instruction is a byte reversing
22162 entry. Fix typo in comment.
22163 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22164 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22165 Add def_builtin calls for new builtins.
22166 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22167 Add define_insn expansion.
22169 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22171 * config/rl78/rl78.md (umaxdi3): New define_expand.
22173 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
22175 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22176 for non-QImode registers.
22178 2018-01-22 Richard Biener <rguenther@suse.de>
22180 PR tree-optimization/83963
22181 * graphite-scop-detection.c (scop_detection::get_sese): Delay
22182 including the loop exit block.
22183 (scop_detection::merge_sese): Likewise.
22184 (scop_detection::add_scop): Do it here instead.
22186 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22188 * doc/sourcebuild.texi (arm_softfloat): Document.
22190 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
22193 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22194 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22195 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22197 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22198 David Edelsohn <dje.gcc@gmail.com>
22201 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22202 Change "crset eq" to "crset 2".
22203 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22204 (*call_indirect_aix<mode>_nospec): Likewise.
22205 (*call_value_indirect_aix<mode>_nospec): Likewise.
22206 (*call_indirect_elfv2<mode>_nospec): Likewise.
22207 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22208 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22209 change assembly output from . to $.
22210 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22211 (indirect_jump<mode>_nospec): Change assembly output from . to $.
22212 (*tablejump<mode>_internal1_nospec): Likewise.
22214 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
22217 * config/sh/sh_optimize_sett_clrt.cc:
22218 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22220 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
22222 PR tree-optimization/83940
22223 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22224 offset_dt to vect_constant_def rather than vect_unknown_def_type.
22225 (vect_check_load_store_mask): Add a mask_dt_out parameter and
22226 use it to pass back the definition type.
22227 (vect_check_store_rhs): Likewise rhs_dt_out.
22228 (vect_build_gather_load_calls): Add a mask_dt argument and use
22229 it instead of a call to vect_is_simple_use.
22230 (vectorizable_store): Update calls to vect_check_load_store_mask
22231 and vect_check_store_rhs. Use the dt returned by the latter instead
22232 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
22233 instead of calls to vect_is_simple_use. Pass the scalar rather
22234 than the vector operand to vect_is_simple_use when handling
22235 second and subsequent copies of an rhs value.
22236 (vectorizable_load): Update calls to vect_check_load_store_mask
22237 and vect_build_gather_load_calls. Use the cached mask_dt and
22238 gs_info.offset_dt instead of calls to vect_is_simple_use.
22240 2018-01-20 Jakub Jelinek <jakub@redhat.com>
22242 PR middle-end/83945
22243 * tree-emutls.c: Include gimplify.h.
22244 (lower_emutls_2): New function.
22245 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
22246 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
22247 it before further processing.
22250 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
22251 UINTVAL (trueop1) instead of INTVAL (op1).
22253 2018-01-19 Jakub Jelinek <jakub@redhat.com>
22257 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
22258 INCOMING_FRAME_SP_OFFSET if not defined.
22259 (scan_trace): Add ENTRY argument. If true and
22260 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
22261 emit a note to adjust the CFA offset.
22262 (create_cfi_notes): Adjust scan_trace callers.
22263 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
22264 INCOMING_FRAME_SP_OFFSET in the CIE.
22265 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
22266 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
22268 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
22269 * doc/tm.texi: Regenerated.
22271 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22273 PR rtl-optimization/83147
22274 * lra-constraints.c (remove_inheritance_pseudos): Use
22275 lra_substitute_pseudo_within_insn.
22277 2018-01-19 Tom de Vries <tom@codesourcery.com>
22278 Cesar Philippidis <cesar@codesourcery.com>
22281 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
22283 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
22286 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
22287 spaces for function labels.
22289 2018-01-19 Martin Liska <mliska@suse.cz>
22291 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
22292 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
22293 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
22294 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
22295 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
22296 (PRED_CONST_RETURN): Change from 69 to 65.
22297 (PRED_NULL_RETURN): Change from 91 to 71.
22298 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
22299 (PRED_LOOP_GUARD): Change from 66 to 73.
22301 2018-01-19 Martin Liska <mliska@suse.cz>
22303 * predict.c (predict_insn_def): Add new assert.
22304 (struct branch_predictor): Change type to signed integer.
22305 (test_prediction_value_range): Amend test to cover
22306 PROB_UNINITIALIZED.
22307 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
22308 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
22309 (PRED_LOOP_ITERATIONS_MAX): Likewise.
22310 (PRED_LOOP_IV_COMPARE): Likewise.
22311 * predict.h (PROB_UNINITIALIZED): Define new constant.
22313 2018-01-19 Martin Liska <mliska@suse.cz>
22315 * predict.c (dump_prediction): Add new format for
22316 analyze_brprob.py script which is enabled with -details
22318 * profile-count.h (precise_p): New function.
22320 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
22322 PR tree-optimization/83922
22323 * tree-vect-loop.c (vect_verify_full_masking): Return false if
22324 there are no statements that need masking.
22325 (vect_active_double_reduction_p): New function.
22326 (vect_analyze_loop_operations): Use it when handling phis that
22327 are not in the loop header.
22329 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
22331 PR tree-optimization/83914
22332 * tree-vect-loop.c (vectorizable_induction): Don't convert
22333 init_expr or apply the peeling adjustment for inductions
22334 that are nested within the vectorized loop.
22336 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22338 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
22341 2018-01-18 Jakub Jelinek <jakub@redhat.com>
22345 * function.h (gimplify_parameters): Add gimple_seq * argument.
22346 * function.c: Include gimple.h and options.h.
22347 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
22348 for the added local temporaries if needed.
22349 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
22350 if there are any parameter cleanups, wrap whole body into a
22351 try/finally with the cleanups.
22353 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
22356 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
22357 Use GET_MODE_CLASS for scalar floating point.
22359 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
22363 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
22364 Fix call of call_cgraph_insertion_hooks.
22366 2018-01-18 Martin Sebor <msebor@redhat.com>
22368 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
22370 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
22373 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
22376 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
22379 * common.opt: (-ffile-prefix-map): New option.
22380 * opts.c (common_handle_option): Defer it.
22381 * opts-global.c (handle_common_deferred_options): Handle it.
22382 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
22383 * file-prefix-map.h: New file.
22384 (remap_debug_filename, add_debug_prefix_map): ...here.
22385 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
22386 * final.c (debug_prefix_map, add_debug_prefix_map
22387 remap_debug_filename): Move to...
22388 * file-prefix-map.c: New file.
22389 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
22390 generalize, get rid of alloca(), use strrchr() instead of strchr().
22391 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
22392 Implement in terms of add_prefix_map().
22393 (remap_macro_filename, remap_debug_filename): Implement in term of
22395 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
22396 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
22397 * dbxout.c: Include file-prefix-map.h.
22398 * varasm.c: Likewise.
22399 * vmsdbgout.c: Likewise.
22400 * xcoffout.c: Likewise.
22401 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
22402 * doc/cppopts.texi (-fmacro-prefix-map): Document.
22403 * doc/invoke.texi (-ffile-prefix-map): Document.
22404 (-fdebug-prefix-map): Update description.
22406 2018-01-18 Martin Liska <mliska@suse.cz>
22408 * config/i386/i386.c (indirect_thunk_name): Document that also
22410 (output_indirect_thunk): Document why both instructions
22411 (pause and lfence) are generated.
22413 2018-01-18 Richard Biener <rguenther@suse.de>
22415 PR tree-optimization/83887
22416 * graphite-scop-detection.c
22417 (scop_detection::get_nearest_dom_with_single_entry): Remove.
22418 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
22419 (scop_detection::merge_sese): Re-implement with a flood-fill
22420 algorithm that properly finds a SESE region if it exists.
22422 2018-01-18 Jakub Jelinek <jakub@redhat.com>
22425 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
22426 pointer_diff optimizations use view_convert instead of convert.
22428 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22430 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22431 Generate different code for -mno-speculate-indirect-jumps.
22432 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22433 (*call_indirect_aix<mode>): Disable for
22434 -mno-speculate-indirect-jumps.
22435 (*call_indirect_aix<mode>_nospec): New define_insn.
22436 (*call_value_indirect_aix<mode>): Disable for
22437 -mno-speculate-indirect-jumps.
22438 (*call_value_indirect_aix<mode>_nospec): New define_insn.
22439 (*sibcall_nonlocal_sysv<mode>): Generate different code for
22440 -mno-speculate-indirect-jumps.
22441 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22443 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
22445 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
22446 long double type, set the flags for noting the default long double
22447 type, even if we don't pass or return a long double type.
22449 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
22452 * ipa-inline.c (flatten_function): Do not overwrite final inlining
22455 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
22457 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
22458 support for merge[hl].
22459 (fold_mergehl_helper): New helper function.
22460 (tree-vector-builder.h): New #include for tree_vector_builder usage.
22461 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
22462 (altivec_vmrglw_direct): Add xxmrglw insn.
22464 2018-01-17 Andrew Waterman <andrew@sifive.com>
22466 * config/riscv/riscv.c (riscv_conditional_register_usage): If
22467 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
22469 2018-01-17 David Malcolm <dmalcolm@redhat.com>
22472 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
22473 call the lto_location_cache before reading the
22474 DECL_SOURCE_LOCATION of the types.
22476 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
22477 Richard Sandiford <richard.sandiford@linaro.org>
22479 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
22480 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
22481 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
22482 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
22483 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
22485 * config/aarch64/constraints.md (aarch64_movti_operand):
22487 * config/aarch64/predicates.md (Uti): Add new constraint.
22489 2018-01-17 Carl Love <cel@us.ibm.com>
22491 * config/rs6000/vsx.md (define_expand xl_len_r,
22492 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
22493 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
22495 (define_expand, define_insn): Move the shift left from the
22496 define_insn to the define_expand for lxvl and stxvl instructions.
22497 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
22498 and XL_LEN_R definitions to PURE.
22500 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
22502 * config/i386/i386.c (indirect_thunk_name): Declare regno
22503 as unsigned int. Compare regno with INVALID_REGNUM.
22504 (output_indirect_thunk): Ditto.
22505 (output_indirect_thunk_function): Ditto.
22506 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
22507 in the call to output_indirect_thunk_function.
22509 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
22511 PR middle-end/83884
22512 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
22513 rather than the size of inner_type to determine the stack slot size
22514 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
22516 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
22519 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
22522 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
22524 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
22525 endian Linux systems to optionally enable multilibs for selecting
22526 the long double type if the user configured an explicit type.
22527 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
22528 have no long double multilibs if not defined.
22529 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
22530 warn if the user used -mabi={ieee,ibm}longdouble and we built
22531 multilibs for long double.
22532 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
22533 appropriate multilib option.
22534 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
22536 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
22537 for building long double multilibs.
22538 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
22540 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
22542 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
22545 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
22547 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
22550 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
22553 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
22556 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
22558 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
22559 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
22561 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
22563 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
22564 different rtl trees depending on TARGET_64BIT.
22565 (rs6000_gen_lvx): Likewise.
22567 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
22569 * config/visium/visium.md (nop): Tweak comment.
22570 (hazard_nop): Likewise.
22572 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
22574 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
22575 -mspeculate-indirect-jumps.
22576 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
22577 for -mno-speculate-indirect-jumps.
22578 (*call_indirect_elfv2<mode>_nospec): New define_insn.
22579 (*call_value_indirect_elfv2<mode>): Disable for
22580 -mno-speculate-indirect-jumps.
22581 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
22582 (indirect_jump): Emit different RTL for
22583 -mno-speculate-indirect-jumps.
22584 (*indirect_jump<mode>): Disable for
22585 -mno-speculate-indirect-jumps.
22586 (*indirect_jump<mode>_nospec): New define_insn.
22587 (tablejump): Emit different RTL for
22588 -mno-speculate-indirect-jumps.
22589 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
22590 (tablejumpsi_nospec): New define_expand.
22591 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
22592 (tablejumpdi_nospec): New define_expand.
22593 (*tablejump<mode>_internal1): Disable for
22594 -mno-speculate-indirect-jumps.
22595 (*tablejump<mode>_internal1_nospec): New define_insn.
22596 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
22599 2018-01-16 Artyom Skrobov tyomitch@gmail.com
22601 * caller-save.c (insert_save): Drop unnecessary parameter. All
22604 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22605 Richard Biener <rguenth@suse.de>
22608 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
22609 return early, inline manually is_gimple_sizepos. Make sure if we
22610 call gimplify_expr we don't end up with a gimple constant.
22611 * tree.c (variably_modified_type_p): Don't return true for
22612 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
22613 * gimplify.h (is_gimple_sizepos): Remove.
22615 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22617 PR tree-optimization/83857
22618 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
22619 vectorizable_live_operation for pure SLP statements.
22620 (vectorizable_live_operation): Handle PHIs.
22622 2018-01-16 Richard Biener <rguenther@suse.de>
22624 PR tree-optimization/83867
22625 * tree-vect-stmts.c (vect_transform_stmt): Precompute
22626 nested_in_vect_loop_p since the scalar stmt may get invalidated.
22628 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22631 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
22632 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
22633 If off is not INTEGER_CST, issue a may not be aligned warning
22634 rather than isn't aligned. Use isn%'t rather than isn't.
22635 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
22637 <case MULT_EXPR>: Improve the case when bottom and one of the
22638 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
22639 operand, in that case check if the other operand is multiple of
22640 bottom divided by the INTEGER_CST operand.
22642 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22645 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
22646 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
22647 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
22648 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
22649 * config/pa/pa.c (pa_function_arg_advance): Likewise.
22650 (pa_function_arg, pa_arg_partial_bytes): Likewise.
22651 (pa_function_arg_size): New function.
22653 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22655 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
22656 in a separate statement.
22658 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
22660 PR tree-optimization/83847
22661 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
22662 group gathers and scatters.
22664 2018-01-16 Jakub Jelinek <jakub@redhat.com>
22666 PR rtl-optimization/86620
22667 * params.def (max-sched-ready-insns): Bump minimum value to 1.
22669 PR rtl-optimization/83213
22670 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
22671 to last if both are JUMP_INSNs.
22673 PR tree-optimization/83843
22674 * gimple-ssa-store-merging.c
22675 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
22676 store_immediate_info for bswap/nop orig_stores.
22678 2018-01-15 Andrew Waterman <andrew@sifive.com>
22680 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
22682 <UDIV>: Increase cost if !TARGET_DIV.
22684 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
22686 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
22687 (define_attr "cr_logical_3op"): New.
22688 (cceq_ior_compare): Adjust.
22689 (cceq_ior_compare_complement): Adjust.
22690 (*cceq_rev_compare): Adjust.
22691 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
22692 (is_cracked_insn): Adjust.
22693 (insn_must_be_first_in_group): Adjust.
22694 * config/rs6000/40x.md: Adjust.
22695 * config/rs6000/440.md: Adjust.
22696 * config/rs6000/476.md: Adjust.
22697 * config/rs6000/601.md: Adjust.
22698 * config/rs6000/603.md: Adjust.
22699 * config/rs6000/6xx.md: Adjust.
22700 * config/rs6000/7450.md: Adjust.
22701 * config/rs6000/7xx.md: Adjust.
22702 * config/rs6000/8540.md: Adjust.
22703 * config/rs6000/cell.md: Adjust.
22704 * config/rs6000/e300c2c3.md: Adjust.
22705 * config/rs6000/e500mc.md: Adjust.
22706 * config/rs6000/e500mc64.md: Adjust.
22707 * config/rs6000/e5500.md: Adjust.
22708 * config/rs6000/e6500.md: Adjust.
22709 * config/rs6000/mpc.md: Adjust.
22710 * config/rs6000/power4.md: Adjust.
22711 * config/rs6000/power5.md: Adjust.
22712 * config/rs6000/power6.md: Adjust.
22713 * config/rs6000/power7.md: Adjust.
22714 * config/rs6000/power8.md: Adjust.
22715 * config/rs6000/power9.md: Adjust.
22716 * config/rs6000/rs64.md: Adjust.
22717 * config/rs6000/titan.md: Adjust.
22719 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22721 * config/i386/predicates.md (indirect_branch_operand): Rewrite
22722 ix86_indirect_branch_register logic.
22724 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22726 * config/i386/constraints.md (Bs): Update
22727 ix86_indirect_branch_register check. Don't check
22728 ix86_indirect_branch_register with GOT_memory_operand.
22730 * config/i386/predicates.md (GOT_memory_operand): Don't check
22731 ix86_indirect_branch_register here.
22732 (GOT32_symbol_operand): Likewise.
22734 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22736 * config/i386/predicates.md (constant_call_address_operand):
22737 Rewrite ix86_indirect_branch_register logic.
22738 (sibcall_insn_operand): Likewise.
22740 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22742 * config/i386/constraints.md (Bs): Replace
22743 ix86_indirect_branch_thunk_register with
22744 ix86_indirect_branch_register.
22746 * config/i386/i386.md (indirect_jump): Likewise.
22747 (tablejump): Likewise.
22748 (*sibcall_memory): Likewise.
22749 (*sibcall_value_memory): Likewise.
22750 Peepholes of indirect call and jump via memory: Likewise.
22751 * config/i386/i386.opt: Likewise.
22752 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22753 (GOT_memory_operand): Likewise.
22754 (call_insn_operand): Likewise.
22755 (sibcall_insn_operand): Likewise.
22756 (GOT32_symbol_operand): Likewise.
22758 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22760 PR middle-end/83837
22761 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
22762 type rather than type addr's type points to.
22763 (expand_omp_atomic_mutex): Likewise.
22764 (expand_omp_atomic): Likewise.
22766 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
22769 * config/i386/i386.c (output_indirect_thunk_function): Use
22770 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
22771 for __x86_return_thunk.
22773 2018-01-15 Richard Biener <rguenther@suse.de>
22775 PR middle-end/83850
22776 * expmed.c (extract_bit_field_1): Fix typo.
22778 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22781 * config/arm/iterators.md (VF): New mode iterator.
22782 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
22783 Remove integer-related logic from pattern.
22784 (neon_vabd<mode>_3): Likewise.
22786 2018-01-15 Jakub Jelinek <jakub@redhat.com>
22788 PR middle-end/82694
22789 * common.opt (fstrict-overflow): No longer an alias.
22790 (fwrapv-pointer): New option.
22791 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
22792 also for pointer types based on flag_wrapv_pointer.
22793 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
22794 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
22795 opts->x_flag_wrapv got set.
22796 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
22797 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
22798 POINTER_TYPE_OVERFLOW_UNDEFINED.
22799 * match.pd: Likewise in address comparison pattern.
22800 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
22802 2018-01-15 Richard Biener <rguenther@suse.de>
22805 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
22806 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
22807 Reset type names to their identifier if their TYPE_DECL doesn't
22808 have linkage (and thus is used for ODR and devirt).
22809 (save_debug_info_for_decl): Remove.
22810 (save_debug_info_for_type): Likewise.
22811 (add_tree_to_fld_list): Adjust.
22812 * tree-pretty-print.c (dump_generic_node): Make dumping of
22813 type names more robust.
22815 2018-01-15 Richard Biener <rguenther@suse.de>
22817 * BASE-VER: Bump to 8.0.1.
22819 2018-01-14 Martin Sebor <msebor@redhat.com>
22822 * builtins.c (check_access): Avoid warning when the no-warning bit
22825 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
22827 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
22828 * ira-color (allocno_hard_regs_compare): Likewise.
22830 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
22833 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
22834 Use .pushsection/.popsection.
22836 2018-01-14 Martin Sebor <msebor@redhat.com>
22839 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
22841 2018-01-14 Jakub Jelinek <jakub@redhat.com>
22843 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
22844 entry from extra_headers.
22845 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
22846 extra_headers, make the list bitwise identical to the i?86-*-* one.
22848 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22850 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
22851 -mcmodel=large with -mindirect-branch=thunk,
22852 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
22853 -mfunction-return=thunk-extern.
22854 * doc/invoke.texi: Document -mcmodel=large is incompatible with
22855 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
22856 -mfunction-return=thunk and -mfunction-return=thunk-extern.
22858 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22860 * config/i386/i386.c (print_reg): Print the name of the full
22861 integer register without '%'.
22862 (ix86_print_operand): Handle 'V'.
22863 * doc/extend.texi: Document 'V' modifier.
22865 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22867 * config/i386/constraints.md (Bs): Disallow memory operand for
22868 -mindirect-branch-register.
22870 * config/i386/predicates.md (indirect_branch_operand): Likewise.
22871 (GOT_memory_operand): Likewise.
22872 (call_insn_operand): Likewise.
22873 (sibcall_insn_operand): Likewise.
22874 (GOT32_symbol_operand): Likewise.
22875 * config/i386/i386.md (indirect_jump): Call convert_memory_address
22876 for -mindirect-branch-register.
22877 (tablejump): Likewise.
22878 (*sibcall_memory): Likewise.
22879 (*sibcall_value_memory): Likewise.
22880 Disallow peepholes of indirect call and jump via memory for
22881 -mindirect-branch-register.
22882 (*call_pop): Replace m with Bw.
22883 (*call_value_pop): Likewise.
22884 (*sibcall_pop_memory): Replace m with Bs.
22885 * config/i386/i386.opt (mindirect-branch-register): New option.
22886 * doc/invoke.texi: Document -mindirect-branch-register option.
22888 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22890 * config/i386/i386-protos.h (ix86_output_function_return): New.
22891 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
22892 set function_return_type.
22893 (indirect_thunk_name): Add ret_p to indicate thunk for function
22895 (output_indirect_thunk_function): Pass false to
22896 indirect_thunk_name.
22897 (ix86_output_indirect_branch_via_reg): Likewise.
22898 (ix86_output_indirect_branch_via_push): Likewise.
22899 (output_indirect_thunk_function): Create alias for function
22900 return thunk if regno < 0.
22901 (ix86_output_function_return): New function.
22902 (ix86_handle_fndecl_attribute): Handle function_return.
22903 (ix86_attribute_table): Add function_return.
22904 * config/i386/i386.h (machine_function): Add
22905 function_return_type.
22906 * config/i386/i386.md (simple_return_internal): Use
22907 ix86_output_function_return.
22908 (simple_return_internal_long): Likewise.
22909 * config/i386/i386.opt (mfunction-return=): New option.
22910 (indirect_branch): Mention -mfunction-return=.
22911 * doc/extend.texi: Document function_return function attribute.
22912 * doc/invoke.texi: Document -mfunction-return= option.
22914 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
22916 * config/i386/i386-opts.h (indirect_branch): New.
22917 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
22918 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
22919 with local indirect jump when converting indirect call and jump.
22920 (ix86_set_indirect_branch_type): New.
22921 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
22922 (indirectlabelno): New.
22923 (indirect_thunk_needed): Likewise.
22924 (indirect_thunk_bnd_needed): Likewise.
22925 (indirect_thunks_used): Likewise.
22926 (indirect_thunks_bnd_used): Likewise.
22927 (INDIRECT_LABEL): Likewise.
22928 (indirect_thunk_name): Likewise.
22929 (output_indirect_thunk): Likewise.
22930 (output_indirect_thunk_function): Likewise.
22931 (ix86_output_indirect_branch_via_reg): Likewise.
22932 (ix86_output_indirect_branch_via_push): Likewise.
22933 (ix86_output_indirect_branch): Likewise.
22934 (ix86_output_indirect_jmp): Likewise.
22935 (ix86_code_end): Call output_indirect_thunk_function if needed.
22936 (ix86_output_call_insn): Call ix86_output_indirect_branch if
22938 (ix86_handle_fndecl_attribute): Handle indirect_branch.
22939 (ix86_attribute_table): Add indirect_branch.
22940 * config/i386/i386.h (machine_function): Add indirect_branch_type
22941 and has_local_indirect_jump.
22942 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
22944 (tablejump): Likewise.
22945 (*indirect_jump): Use ix86_output_indirect_jmp.
22946 (*tablejump_1): Likewise.
22947 (simple_return_indirect_internal): Likewise.
22948 * config/i386/i386.opt (mindirect-branch=): New option.
22949 (indirect_branch): New.
22952 (thunk-inline): Likewise.
22953 (thunk-extern): Likewise.
22954 * doc/extend.texi: Document indirect_branch function attribute.
22955 * doc/invoke.texi: Document -mindirect-branch= option.
22957 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
22960 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
22962 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
22964 * ipa-inline.c (want_inline_small_function_p): Return false if
22965 inlining has already failed with CIF_FINAL_ERROR.
22966 (update_caller_keys): Call want_inline_small_function_p before
22968 (update_callee_keys): Likewise.
22970 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
22972 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
22974 (rs6000_quadword_masked_address_p): Likewise.
22975 (quad_aligned_load_p): Likewise.
22976 (quad_aligned_store_p): Likewise.
22977 (const_load_sequence_p): Add comment to describe the outer-most loop.
22978 (mimic_memory_attributes_and_flags): New function.
22979 (rs6000_gen_stvx): Likewise.
22980 (replace_swapped_aligned_store): Likewise.
22981 (rs6000_gen_lvx): Likewise.
22982 (replace_swapped_aligned_load): Likewise.
22983 (replace_swapped_load_constant): Capitalize argument name in
22984 comment describing this function.
22985 (rs6000_analyze_swaps): Add a third pass to search for vector loads
22986 and stores that access quad-word aligned addresses and replace
22987 with stvx or lvx instructions when appropriate.
22988 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
22989 New function prototype.
22990 (rs6000_quadword_masked_address_p): Likewise.
22991 (rs6000_gen_lvx): Likewise.
22992 (rs6000_gen_stvx): Likewise.
22993 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
22994 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
22995 when memory address is aligned.
22996 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
22997 this split to select lvx instruction when memory address is aligned.
22998 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
22999 instruction when memory address is aligned.
23000 (*vsx_le_perm_load_v16qi): Likewise.
23001 (four unnamed splitters): Modify to select the stvx instruction
23002 when memory is aligned.
23004 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
23006 * predict.c (determine_unlikely_bbs): Handle correctly BBs
23007 which appears in the queue multiple times.
23009 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23010 Alan Hayward <alan.hayward@arm.com>
23011 David Sherwood <david.sherwood@arm.com>
23013 * tree-vectorizer.h (vec_lower_bound): New structure.
23014 (_loop_vec_info): Add check_nonzero and lower_bounds.
23015 (LOOP_VINFO_CHECK_NONZERO): New macro.
23016 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
23017 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
23018 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
23019 fields. Make seg_len the distance travelled, not including the
23021 (dr_direction_indicator): Declare.
23022 (dr_zero_step_indicator): Likewise.
23023 (dr_known_forward_stride_p): Likewise.
23024 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
23026 (runtime_alias_check_p): Allow runtime alias checks with
23028 (operator ==): Compare access_size and align.
23029 (prune_runtime_alias_test_list): Rework for new distinction between
23030 the access_size and seg_len.
23031 (create_intersect_range_checks_index): Likewise. Cope with polynomial
23033 (get_segment_min_max): New function.
23034 (create_intersect_range_checks): Use it.
23035 (dr_step_indicator): New function.
23036 (dr_direction_indicator): Likewise.
23037 (dr_zero_step_indicator): Likewise.
23038 (dr_known_forward_stride_p): Likewise.
23039 * tree-loop-distribution.c (data_ref_segment_size): Return
23040 DR_STEP * (niters - 1).
23041 (compute_alias_check_pairs): Update call to the dr_with_seg_len
23043 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
23044 (vect_preserves_scalar_order_p): New function, split out from...
23045 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
23046 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
23047 (vect_vfa_access_size): New function.
23048 (vect_vfa_align): Likewise.
23049 (vect_compile_time_alias): Take access_size_a and access_b arguments.
23050 (dump_lower_bound): New function.
23051 (vect_check_lower_bound): Likewise.
23052 (vect_small_gap_p): Likewise.
23053 (vectorizable_with_step_bound_p): Likewise.
23054 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
23055 depencies if the vectorization factor is 1. Convert the checks
23056 for nonzero steps into checks on the bounds of DR_STEP. Try using
23057 a bunds check for variable steps if the minimum required step is
23058 relatively small. Update calls to the dr_with_seg_len
23059 constructor and to vect_compile_time_alias.
23060 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
23062 (vect_loop_versioning): Call it.
23063 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
23065 (vect_estimate_min_profitable_iters): Account for any bounds checks.
23067 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23068 Alan Hayward <alan.hayward@arm.com>
23069 David Sherwood <david.sherwood@arm.com>
23071 * doc/sourcebuild.texi (vect_scatter_store): Document.
23072 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
23074 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
23076 * genopinit.c (main): Add supports_vec_scatter_store and
23077 supports_vec_scatter_store_cached to target_optabs.
23078 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
23079 IFN_MASK_SCATTER_STORE.
23080 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
23082 * internal-fn.h (internal_store_fn_p): Declare.
23083 (internal_fn_stored_value_index): Likewise.
23084 * internal-fn.c (scatter_store_direct): New macro.
23085 (expand_scatter_store_optab_fn): New function.
23086 (direct_scatter_store_optab_supported_p): New macro.
23087 (internal_store_fn_p): New function.
23088 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
23089 IFN_MASK_SCATTER_STORE.
23090 (internal_fn_mask_index): Likewise.
23091 (internal_fn_stored_value_index): New function.
23092 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
23093 for scatter stores.
23094 * optabs-query.h (supports_vec_scatter_store_p): Declare.
23095 * optabs-query.c (supports_vec_scatter_store_p): New function.
23096 * tree-vectorizer.h (vect_get_store_rhs): Declare.
23097 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23098 true for scatter stores.
23099 (vect_gather_scatter_fn_p): Handle scatter stores too.
23100 (vect_check_gather_scatter): Consider using scatter stores if
23101 supports_vec_scatter_store_p.
23102 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23103 scatter stores too.
23104 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23105 internal_fn_stored_value_index.
23106 (check_load_store_masking): Handle scatter stores too.
23107 (vect_get_store_rhs): Make public.
23108 (vectorizable_call): Use internal_store_fn_p.
23109 (vectorizable_store): Handle scatter store internal functions.
23110 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23111 when deciding whether the end of the group has been reached.
23112 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23113 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23114 (mask_scatter_store<mode>): New insns.
23116 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23117 Alan Hayward <alan.hayward@arm.com>
23118 David Sherwood <david.sherwood@arm.com>
23120 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23121 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23122 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23124 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23125 Use vect_truncate_gather_scatter_offset if we can't treat the
23126 operation as a normal gather load or scatter store.
23127 (get_group_load_store_type): Take the gather_scatter_info
23128 as argument. Try using a gather load or scatter store for
23129 single-element groups.
23130 (get_load_store_type): Update calls to get_group_load_store_type
23131 and vect_use_strided_gather_scatters_p.
23133 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23134 Alan Hayward <alan.hayward@arm.com>
23135 David Sherwood <david.sherwood@arm.com>
23137 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23138 optional tree argument.
23139 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23141 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23142 but continue to use the current value as a fallback.
23143 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23144 to compare the updates.
23145 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23146 (get_load_store_type): Use it when handling a strided access.
23147 (vect_get_strided_load_store_ops): New function.
23148 (vect_get_data_ptr_increment): Likewise.
23149 (vectorizable_load): Handle strided gather loads. Always pass
23150 a step to vect_create_data_ref_ptr and bump_vector_ptr.
23152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23153 Alan Hayward <alan.hayward@arm.com>
23154 David Sherwood <david.sherwood@arm.com>
23156 * doc/md.texi (gather_load@var{m}): Document.
23157 (mask_gather_load@var{m}): Likewise.
23158 * genopinit.c (main): Add supports_vec_gather_load and
23159 supports_vec_gather_load_cached to target_optabs.
23160 * optabs-tree.c (init_tree_optimization_optabs): Use
23161 ggc_cleared_alloc to allocate target_optabs.
23162 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23163 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23165 * internal-fn.h (internal_load_fn_p): Declare.
23166 (internal_gather_scatter_fn_p): Likewise.
23167 (internal_fn_mask_index): Likewise.
23168 (internal_gather_scatter_fn_supported_p): Likewise.
23169 * internal-fn.c (gather_load_direct): New macro.
23170 (expand_gather_load_optab_fn): New function.
23171 (direct_gather_load_optab_supported_p): New macro.
23172 (direct_internal_fn_optab): New function.
23173 (internal_load_fn_p): Likewise.
23174 (internal_gather_scatter_fn_p): Likewise.
23175 (internal_fn_mask_index): Likewise.
23176 (internal_gather_scatter_fn_supported_p): Likewise.
23177 * optabs-query.c (supports_at_least_one_mode_p): New function.
23178 (supports_vec_gather_load_p): Likewise.
23179 * optabs-query.h (supports_vec_gather_load_p): Declare.
23180 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23181 and memory_type field.
23182 (NUM_PATTERNS): Bump to 15.
23183 * tree-vect-data-refs.c: Include internal-fn.h.
23184 (vect_gather_scatter_fn_p): New function.
23185 (vect_describe_gather_scatter_call): Likewise.
23186 (vect_check_gather_scatter): Try using internal functions for
23187 gather loads. Recognize existing calls to a gather load function.
23188 (vect_analyze_data_refs): Consider using gather loads if
23189 supports_vec_gather_load_p.
23190 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23191 (vect_get_gather_scatter_offset_type): Likewise.
23192 (vect_convert_mask_for_vectype): Likewise.
23193 (vect_add_conversion_to_patterm): Likewise.
23194 (vect_try_gather_scatter_pattern): Likewise.
23195 (vect_recog_gather_scatter_pattern): New pattern recognizer.
23196 (vect_vect_recog_func_ptrs): Add it.
23197 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23198 internal_fn_mask_index and internal_gather_scatter_fn_p.
23199 (check_load_store_masking): Take the gather_scatter_info as an
23200 argument and handle gather loads.
23201 (vect_get_gather_scatter_ops): New function.
23202 (vectorizable_call): Check internal_load_fn_p.
23203 (vectorizable_load): Likewise. Handle gather load internal
23205 (vectorizable_store): Update call to check_load_store_masking.
23206 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23207 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23208 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23209 (aarch64_gather_scale_operand_d): New predicates.
23210 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23211 (mask_gather_load<mode>): New insns.
23213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23214 Alan Hayward <alan.hayward@arm.com>
23215 David Sherwood <david.sherwood@arm.com>
23217 * optabs.def (fold_left_plus_optab): New optab.
23218 * doc/md.texi (fold_left_plus_@var{m}): Document.
23219 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23220 * internal-fn.c (fold_left_direct): Define.
23221 (expand_fold_left_optab_fn): Likewise.
23222 (direct_fold_left_optab_supported_p): Likewise.
23223 * fold-const-call.c (fold_const_fold_left): New function.
23224 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23225 * tree-parloops.c (valid_reduction_p): New function.
23226 (gather_scalar_reductions): Use it.
23227 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23228 (vect_finish_replace_stmt): Declare.
23229 * tree-vect-loop.c (fold_left_reduction_fn): New function.
23230 (needs_fold_left_reduction_p): New function, split out from...
23231 (vect_is_simple_reduction): ...here. Accept reductions that
23232 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23233 (vect_force_simple_reduction): Also store the reduction type in
23234 the assignment's STMT_VINFO_REDUC_TYPE.
23235 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
23236 (merge_with_identity): New function.
23237 (vect_expand_fold_left): Likewise.
23238 (vectorize_fold_left_reduction): Likewise.
23239 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
23240 scalar phi in place for it. Check for target support and reject
23241 cases that would reassociate the operation. Defer the transform
23242 phase to vectorize_fold_left_reduction.
23243 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
23244 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
23245 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
23247 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23249 * tree-if-conv.c (predicate_mem_writes): Remove redundant
23250 call to ifc_temp_var.
23252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23253 Alan Hayward <alan.hayward@arm.com>
23254 David Sherwood <david.sherwood@arm.com>
23256 * target.def (legitimize_address_displacement): Take the original
23257 offset as a poly_int.
23258 * targhooks.h (default_legitimize_address_displacement): Update
23260 * targhooks.c (default_legitimize_address_displacement): Likewise.
23261 * doc/tm.texi: Regenerate.
23262 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
23263 as an argument, moving assert of ad->disp == ad->disp_term to...
23264 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
23265 Try calling targetm.legitimize_address_displacement before expanding
23266 the address rather than afterwards, and adjust for the new interface.
23267 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
23268 Match the new hook interface. Handle SVE addresses.
23269 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
23270 new hook interface.
23272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23274 * Makefile.in (OBJS): Add early-remat.o.
23275 * target.def (select_early_remat_modes): New hook.
23276 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
23277 * doc/tm.texi: Regenerate.
23278 * targhooks.h (default_select_early_remat_modes): Declare.
23279 * targhooks.c (default_select_early_remat_modes): New function.
23280 * timevar.def (TV_EARLY_REMAT): New timevar.
23281 * passes.def (pass_early_remat): New pass.
23282 * tree-pass.h (make_pass_early_remat): Declare.
23283 * early-remat.c: New file.
23284 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
23286 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
23288 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23289 Alan Hayward <alan.hayward@arm.com>
23290 David Sherwood <david.sherwood@arm.com>
23292 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
23293 vfm1 with a bound_epilog parameter.
23294 (vect_do_peeling): Update calls accordingly, and move the prologue
23295 call earlier in the function. Treat the base bound_epilog as 0 for
23296 fully-masked loops and retain vf - 1 for other loops. Add 1 to
23297 this base when peeling for gaps.
23298 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
23299 with fully-masked loops.
23300 (vect_estimate_min_profitable_iters): Handle the single peeled
23301 iteration in that case.
23303 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23304 Alan Hayward <alan.hayward@arm.com>
23305 David Sherwood <david.sherwood@arm.com>
23307 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
23308 single-element interleaving even if the size is not a power of 2.
23309 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
23310 accesses for single-element interleaving if the group size is
23313 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23314 Alan Hayward <alan.hayward@arm.com>
23315 David Sherwood <david.sherwood@arm.com>
23317 * doc/md.texi (fold_extract_last_@var{m}): Document.
23318 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
23319 * optabs.def (fold_extract_last_optab): New optab.
23320 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
23321 * internal-fn.c (fold_extract_direct): New macro.
23322 (expand_fold_extract_optab_fn): Likewise.
23323 (direct_fold_extract_optab_supported_p): Likewise.
23324 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
23325 * tree-vect-loop.c (vect_model_reduction_cost): Handle
23326 EXTRACT_LAST_REDUCTION.
23327 (get_initial_def_for_reduction): Do not create an initial vector
23328 for EXTRACT_LAST_REDUCTION reductions.
23329 (vectorizable_reduction): Leave the scalar phi in place for
23330 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
23331 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
23332 epilogue code for EXTRACT_LAST_REDUCTION and defer the
23333 transform phase to vectorizable_condition.
23334 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
23336 (vect_finish_stmt_generation): ...here.
23337 (vect_finish_replace_stmt): New function.
23338 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
23339 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
23341 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
23343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23344 Alan Hayward <alan.hayward@arm.com>
23345 David Sherwood <david.sherwood@arm.com>
23347 * doc/md.texi (extract_last_@var{m}): Document.
23348 * optabs.def (extract_last_optab): New optab.
23349 * internal-fn.def (EXTRACT_LAST): New internal function.
23350 * internal-fn.c (cond_unary_direct): New macro.
23351 (expand_cond_unary_optab_fn): Likewise.
23352 (direct_cond_unary_optab_supported_p): Likewise.
23353 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
23354 loops using EXTRACT_LAST.
23355 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
23356 (extract_last_<mode>): ...this optab.
23357 (vec_extract<mode><Vel>): Update accordingly.
23359 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23360 Alan Hayward <alan.hayward@arm.com>
23361 David Sherwood <david.sherwood@arm.com>
23363 * target.def (empty_mask_is_expensive): New hook.
23364 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
23365 * doc/tm.texi: Regenerate.
23366 * targhooks.h (default_empty_mask_is_expensive): Declare.
23367 * targhooks.c (default_empty_mask_is_expensive): New function.
23368 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
23369 if the target says that empty masks are expensive.
23370 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
23372 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
23374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23375 Alan Hayward <alan.hayward@arm.com>
23376 David Sherwood <david.sherwood@arm.com>
23378 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
23379 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
23380 (vect_use_loop_mask_for_alignment_p): New function.
23381 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
23382 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
23383 niters_skip argument. Make sure that the first niters_skip elements
23384 of the first iteration are inactive.
23385 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
23386 Update call to vect_set_loop_masks_directly.
23387 (get_misalign_in_elems): New function, split out from...
23388 (vect_gen_prolog_loop_niters): ...here.
23389 (vect_update_init_of_dr): Take a code argument that specifies whether
23390 the adjustment should be added or subtracted.
23391 (vect_update_init_of_drs): Likewise.
23392 (vect_prepare_for_masked_peels): New function.
23393 (vect_do_peeling): Skip prologue peeling if we're using a mask
23394 instead. Update call to vect_update_inits_of_drs.
23395 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23397 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
23398 alignment. Do not include the number of peeled iterations in
23399 the minimum threshold in that case.
23400 (vectorizable_induction): Adjust the start value down by
23401 LOOP_VINFO_MASK_SKIP_NITERS iterations.
23402 (vect_transform_loop): Call vect_prepare_for_masked_peels.
23403 Take the number of skipped iterations into account when calculating
23405 * tree-vect-stmts.c (vect_gen_while_not): New function.
23407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23408 Alan Hayward <alan.hayward@arm.com>
23409 David Sherwood <david.sherwood@arm.com>
23411 * doc/sourcebuild.texi (vect_fully_masked): Document.
23412 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
23413 default value to 0.
23414 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
23416 (vect_analyze_loop_2): ...here. Don't check the vectorization
23417 factor against the number of loop iterations if the loop is
23420 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23421 Alan Hayward <alan.hayward@arm.com>
23422 David Sherwood <david.sherwood@arm.com>
23424 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
23425 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
23426 (dump_groups): Update accordingly.
23427 (iv_use::mem_type): New member variable.
23428 (address_p): New function.
23429 (record_use): Add a mem_type argument and initialize the new
23431 (record_group_use): Add a mem_type argument. Use address_p.
23432 Remove obsolete null checks of base_object. Update call to record_use.
23433 (find_interesting_uses_op): Update call to record_group_use.
23434 (find_interesting_uses_cond): Likewise.
23435 (find_interesting_uses_address): Likewise.
23436 (get_mem_type_for_internal_fn): New function.
23437 (find_address_like_use): Likewise.
23438 (find_interesting_uses_stmt): Try find_address_like_use before
23439 calling find_interesting_uses_op.
23440 (addr_offset_valid_p): Use the iv mem_type field as the type
23441 of the addressed memory.
23442 (add_autoinc_candidates): Likewise.
23443 (get_address_cost): Likewise.
23444 (split_small_address_groups_p): Use address_p.
23445 (split_address_groups): Likewise.
23446 (add_iv_candidate_for_use): Likewise.
23447 (autoinc_possible_for_pair): Likewise.
23448 (rewrite_groups): Likewise.
23449 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
23450 (determine_group_iv_cost): Update after split of USE_ADDRESS.
23451 (get_alias_ptr_type_for_ptr_address): New function.
23452 (rewrite_use_address): Rewrite address uses in calls that were
23453 identified by find_address_like_use.
23455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23456 Alan Hayward <alan.hayward@arm.com>
23457 David Sherwood <david.sherwood@arm.com>
23459 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
23461 * gimple-expr.h (is_gimple_addressable: Likewise.
23462 * gimple-expr.c (is_gimple_address): Likewise.
23463 * internal-fn.c (expand_call_mem_ref): New function.
23464 (expand_mask_load_optab_fn): Use it.
23465 (expand_mask_store_optab_fn): Likewise.
23467 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23468 Alan Hayward <alan.hayward@arm.com>
23469 David Sherwood <david.sherwood@arm.com>
23471 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
23472 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
23473 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
23474 (cond_umax@var{mode}): Document.
23475 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
23476 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
23477 (cond_umin_optab, cond_umax_optab): New optabs.
23478 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
23479 (COND_IOR, COND_XOR): New internal functions.
23480 * internal-fn.h (get_conditional_internal_fn): Declare.
23481 * internal-fn.c (cond_binary_direct): New macro.
23482 (expand_cond_binary_optab_fn): Likewise.
23483 (direct_cond_binary_optab_supported_p): Likewise.
23484 (get_conditional_internal_fn): New function.
23485 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
23486 Cope with reduction statements that are vectorized as calls rather
23488 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
23489 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
23490 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
23491 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
23492 (UNSPEC_COND_EOR): New unspecs.
23493 (optab): Add mappings for them.
23494 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
23495 (sve_int_op, sve_fp_op): New int attributes.
23497 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23498 Alan Hayward <alan.hayward@arm.com>
23499 David Sherwood <david.sherwood@arm.com>
23501 * optabs.def (while_ult_optab): New optab.
23502 * doc/md.texi (while_ult@var{m}@var{n}): Document.
23503 * internal-fn.def (WHILE_ULT): New internal function.
23504 * internal-fn.h (direct_internal_fn_supported_p): New override
23505 that takes two types as argument.
23506 * internal-fn.c (while_direct): New macro.
23507 (expand_while_optab_fn): New function.
23508 (convert_optab_supported_p): Likewise.
23509 (direct_while_optab_supported_p): New macro.
23510 * wide-int.h (wi::udiv_ceil): New function.
23511 * tree-vectorizer.h (rgroup_masks): New structure.
23512 (vec_loop_masks): New typedef.
23513 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
23514 and fully_masked_p.
23515 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
23516 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
23517 (vect_max_vf): New function.
23518 (slpeel_make_loop_iterate_ntimes): Delete.
23519 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
23520 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
23521 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
23522 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
23523 internal-fn.h, stor-layout.h and optabs-query.h.
23524 (vect_set_loop_mask): New function.
23525 (add_preheader_seq): Likewise.
23526 (add_header_seq): Likewise.
23527 (interleave_supported_p): Likewise.
23528 (vect_maybe_permute_loop_masks): Likewise.
23529 (vect_set_loop_masks_directly): Likewise.
23530 (vect_set_loop_condition_masked): Likewise.
23531 (vect_set_loop_condition_unmasked): New function, split out from
23532 slpeel_make_loop_iterate_ntimes.
23533 (slpeel_make_loop_iterate_ntimes): Rename to..
23534 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
23535 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
23536 (vect_do_peeling): Update call accordingly.
23537 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
23539 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23540 mask_compare_type, can_fully_mask_p and fully_masked_p.
23541 (release_vec_loop_masks): New function.
23542 (_loop_vec_info): Use it to free the loop masks.
23543 (can_produce_all_loop_masks_p): New function.
23544 (vect_get_max_nscalars_per_iter): Likewise.
23545 (vect_verify_full_masking): Likewise.
23546 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
23547 retries, and free the mask rgroups before retrying. Check loop-wide
23548 reasons for disallowing fully-masked loops. Make the final decision
23549 about whether use a fully-masked loop or not.
23550 (vect_estimate_min_profitable_iters): Do not assume that peeling
23551 for the number of iterations will be needed for fully-masked loops.
23552 (vectorizable_reduction): Disable fully-masked loops.
23553 (vectorizable_live_operation): Likewise.
23554 (vect_halve_mask_nunits): New function.
23555 (vect_double_mask_nunits): Likewise.
23556 (vect_record_loop_mask): Likewise.
23557 (vect_get_loop_mask): Likewise.
23558 (vect_transform_loop): Handle the case in which the final loop
23559 iteration might handle a partial vector. Call vect_set_loop_condition
23560 instead of slpeel_make_loop_iterate_ntimes.
23561 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
23562 (check_load_store_masking): New function.
23563 (prepare_load_store_mask): Likewise.
23564 (vectorizable_store): Handle fully-masked loops.
23565 (vectorizable_load): Likewise.
23566 (supportable_widening_operation): Use vect_halve_mask_nunits for
23568 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
23569 (vect_gen_while): New function.
23570 * config/aarch64/aarch64.md (umax<mode>3): New expander.
23571 (aarch64_uqdec<mode>): New insn.
23573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23574 Alan Hayward <alan.hayward@arm.com>
23575 David Sherwood <david.sherwood@arm.com>
23577 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
23578 (reduc_xor_scal_optab): New optabs.
23579 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
23580 (reduc_xor_scal_@var{m}): Document.
23581 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
23582 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
23583 internal functions.
23584 * fold-const-call.c (fold_const_call): Handle them.
23585 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
23586 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
23587 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
23588 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
23589 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
23590 (UNSPEC_XORV): New unspecs.
23591 (optab): Add entries for them.
23592 (BITWISEV): New int iterator.
23593 (bit_reduc_op): New int attributes.
23595 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23596 Alan Hayward <alan.hayward@arm.com>
23597 David Sherwood <david.sherwood@arm.com>
23599 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
23600 * internal-fn.def (VEC_SHL_INSERT): New internal function.
23601 * optabs.def (vec_shl_insert_optab): New optab.
23602 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
23603 (duplicate_and_interleave): Likewise.
23604 * tree-vect-loop.c: Include internal-fn.h.
23605 (neutral_op_for_slp_reduction): New function, split out from
23606 get_initial_defs_for_reduction.
23607 (get_initial_def_for_reduction): Handle option 2 for variable-length
23608 vectors by loading the neutral value into a vector and then shifting
23609 the initial value into element 0.
23610 (get_initial_defs_for_reduction): Replace the code argument with
23611 the neutral value calculated by neutral_op_for_slp_reduction.
23612 Use gimple_build_vector for constant-length vectors.
23613 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
23614 but the first group_size elements have a neutral value.
23615 Use duplicate_and_interleave otherwise.
23616 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
23617 Update call to get_initial_defs_for_reduction. Handle SLP
23618 reductions for variable-length vectors by creating one vector
23619 result for each scalar result, with the elements associated
23620 with other scalar results stubbed out with the neutral value.
23621 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
23622 Require IFN_VEC_SHL_INSERT for double reductions on
23623 variable-length vectors, or SLP reductions that have
23624 a neutral value. Require can_duplicate_and_interleave_p
23625 support for variable-length unchained SLP reductions if there
23626 is no neutral value, such as for MIN/MAX reductions. Also require
23627 the number of vector elements to be a multiple of the number of
23628 SLP statements when doing variable-length unchained SLP reductions.
23629 Update call to vect_create_epilog_for_reduction.
23630 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
23631 and remove initial values.
23632 (duplicate_and_interleave): Make public.
23633 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
23634 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
23636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23637 Alan Hayward <alan.hayward@arm.com>
23638 David Sherwood <david.sherwood@arm.com>
23640 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
23641 (can_duplicate_and_interleave_p): New function.
23642 (vect_get_and_check_slp_defs): Take the vector of statements
23643 rather than just the current one. Remove excess parentheses.
23644 Restriction rejectinon of vect_constant_def and vect_external_def
23645 for variable-length vectors to boolean types, or types for which
23646 can_duplicate_and_interleave_p is false.
23647 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
23648 (duplicate_and_interleave): New function.
23649 (vect_get_constant_vectors): Use gimple_build_vector for
23650 constant-length vectors and suitable variable-length constant
23651 vectors. Use duplicate_and_interleave for other variable-length
23652 vectors. Don't defer the update when inserting new statements.
23654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23655 Alan Hayward <alan.hayward@arm.com>
23656 David Sherwood <david.sherwood@arm.com>
23658 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
23659 min_profitable_iters doesn't go negative.
23661 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23662 Alan Hayward <alan.hayward@arm.com>
23663 David Sherwood <david.sherwood@arm.com>
23665 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
23666 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
23667 * optabs.def (vec_mask_load_lanes_optab): New optab.
23668 (vec_mask_store_lanes_optab): Likewise.
23669 * internal-fn.def (MASK_LOAD_LANES): New internal function.
23670 (MASK_STORE_LANES): Likewise.
23671 * internal-fn.c (mask_load_lanes_direct): New macro.
23672 (mask_store_lanes_direct): Likewise.
23673 (expand_mask_load_optab_fn): Handle masked operations.
23674 (expand_mask_load_lanes_optab_fn): New macro.
23675 (expand_mask_store_optab_fn): Handle masked operations.
23676 (expand_mask_store_lanes_optab_fn): New macro.
23677 (direct_mask_load_lanes_optab_supported_p): Likewise.
23678 (direct_mask_store_lanes_optab_supported_p): Likewise.
23679 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
23681 (vect_load_lanes_supported): Likewise.
23682 * tree-vect-data-refs.c (strip_conversion): New function.
23683 (can_group_stmts_p): Likewise.
23684 (vect_analyze_data_ref_accesses): Use it instead of checking
23685 for a pair of assignments.
23686 (vect_store_lanes_supported): Take a masked_p parameter.
23687 (vect_load_lanes_supported): Likewise.
23688 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
23689 vect_store_lanes_supported and vect_load_lanes_supported.
23690 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
23691 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
23692 parameter. Don't allow gaps for masked accesses.
23693 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
23694 and vect_load_lanes_supported.
23695 (get_load_store_type): Take a masked_p parameter and update
23696 call to get_group_load_store_type.
23697 (vectorizable_store): Update call to get_load_store_type.
23698 Handle IFN_MASK_STORE_LANES.
23699 (vectorizable_load): Update call to get_load_store_type.
23700 Handle IFN_MASK_LOAD_LANES.
23702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23703 Alan Hayward <alan.hayward@arm.com>
23704 David Sherwood <david.sherwood@arm.com>
23706 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
23708 * config/aarch64/aarch64-protos.h
23709 (aarch64_sve_struct_memory_operand_p): Declare.
23710 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
23711 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
23712 (VPRED, vpred): Handle SVE structure modes.
23713 * config/aarch64/constraints.md (Utx): New constraint.
23714 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
23715 (aarch64_sve_struct_nonimmediate_operand): New predicates.
23716 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
23717 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
23718 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
23719 structure modes. Split into pieces after RA.
23720 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
23721 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
23723 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
23724 SVE structure modes.
23725 (aarch64_classify_address): Likewise.
23726 (sizetochar): Move earlier in file.
23727 (aarch64_print_operand): Handle SVE register lists.
23728 (aarch64_array_mode): New function.
23729 (aarch64_sve_struct_memory_operand_p): Likewise.
23730 (TARGET_ARRAY_MODE): Redefine.
23732 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23733 Alan Hayward <alan.hayward@arm.com>
23734 David Sherwood <david.sherwood@arm.com>
23736 * target.def (array_mode): New target hook.
23737 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
23738 * doc/tm.texi: Regenerate.
23739 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
23740 * hooks.c (hook_optmode_mode_uhwi_none): New function.
23741 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
23742 targetm.array_mode.
23743 * stor-layout.c (mode_for_array): Likewise. Support polynomial
23746 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23747 Alan Hayward <alan.hayward@arm.com>
23748 David Sherwood <david.sherwood@arm.com>
23750 * fold-const.c (fold_binary_loc): Check the argument types
23751 rather than the result type when testing for a vector operation.
23753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23755 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
23756 * doc/tm.texi: Regenerate.
23758 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
23759 Alan Hayward <alan.hayward@arm.com>
23760 David Sherwood <david.sherwood@arm.com>
23762 * doc/invoke.texi (-msve-vector-bits=): Document new option.
23763 (sve): Document new AArch64 extension.
23764 * doc/md.texi (w): Extend the description of the AArch64
23765 constraint to include SVE vectors.
23766 (Upl, Upa): Document new AArch64 predicate constraints.
23767 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
23769 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
23770 (msve-vector-bits=): New option.
23771 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
23772 SVE when these are disabled.
23773 (sve): New extension.
23774 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
23775 modes. Adjust their number of units based on aarch64_sve_vg.
23776 (MAX_BITSIZE_MODE_ANY_MODE): Define.
23777 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
23778 aarch64_addr_query_type.
23779 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
23780 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
23781 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
23782 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
23783 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
23784 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
23785 (aarch64_simd_imm_zero_p): Delete.
23786 (aarch64_check_zero_based_sve_index_immediate): Declare.
23787 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23788 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23789 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23790 (aarch64_sve_float_mul_immediate_p): Likewise.
23791 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23792 rather than an rtx.
23793 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
23794 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
23795 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
23796 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
23797 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
23798 (aarch64_regmode_natural_size): Likewise.
23799 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
23800 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
23802 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
23803 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
23804 for VG and the SVE predicate registers.
23805 (V_ALIASES): Add a "z"-prefixed alias.
23806 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
23807 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
23808 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
23809 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
23810 (REG_CLASS_NAMES): Add entries for them.
23811 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
23812 and the predicate registers.
23813 (aarch64_sve_vg): Declare.
23814 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
23815 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
23816 (REGMODE_NATURAL_SIZE): Define.
23817 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
23819 * config/aarch64/aarch64.c: Include cfgrtl.h.
23820 (simd_immediate_info): Add a constructor for series vectors,
23821 and an associated step field.
23822 (aarch64_sve_vg): New variable.
23823 (aarch64_dbx_register_number): Handle VG and the predicate registers.
23824 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
23825 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
23826 (VEC_ANY_DATA, VEC_STRUCT): New constants.
23827 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
23828 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
23829 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
23830 (aarch64_get_mask_mode): New functions.
23831 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
23832 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
23833 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
23834 predicate modes and predicate registers. Explicitly restrict
23835 GPRs to modes of 16 bytes or smaller. Only allow FP registers
23836 to store a vector mode if it is recognized by
23837 aarch64_classify_vector_mode.
23838 (aarch64_regmode_natural_size): New function.
23839 (aarch64_hard_regno_caller_save_mode): Return the original mode
23841 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
23842 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
23843 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
23844 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
23846 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
23847 does not overlap dest if the function is frame-related. Handle
23849 (aarch64_split_add_offset): New function.
23850 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
23851 them aarch64_add_offset.
23852 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
23853 and update call to aarch64_sub_sp.
23854 (aarch64_add_cfa_expression): New function.
23855 (aarch64_expand_prologue): Pass extra temporary registers to the
23856 functions above. Handle the case in which we need to emit new
23857 DW_CFA_expressions for registers that were originally saved
23858 relative to the stack pointer, but now have to be expressed
23859 relative to the frame pointer.
23860 (aarch64_output_mi_thunk): Pass extra temporary registers to the
23862 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
23863 IP0 and IP1 values for SVE frames.
23864 (aarch64_expand_vec_series): New function.
23865 (aarch64_expand_sve_widened_duplicate): Likewise.
23866 (aarch64_expand_sve_const_vector): Likewise.
23867 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
23868 Handle SVE constants. Use emit_move_insn to move a force_const_mem
23869 into the register, rather than emitting a SET directly.
23870 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
23871 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
23872 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
23873 (offset_9bit_signed_scaled_p): New functions.
23874 (aarch64_replicate_bitmask_imm): New function.
23875 (aarch64_bitmask_imm): Use it.
23876 (aarch64_cannot_force_const_mem): Reject expressions involving
23877 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
23878 (aarch64_classify_index): Handle SVE indices, by requiring
23879 a plain register index with a scale that matches the element size.
23880 (aarch64_classify_address): Handle SVE addresses. Assert that
23881 the mode of the address is VOIDmode or an integer mode.
23882 Update call to aarch64_classify_symbol.
23883 (aarch64_classify_symbolic_expression): Update call to
23884 aarch64_classify_symbol.
23885 (aarch64_const_vec_all_in_range_p): New function.
23886 (aarch64_print_vector_float_operand): Likewise.
23887 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
23888 "vN" for FP registers with SVE modes. Handle (const ...) vectors
23889 and the FP immediates 1.0 and 0.5.
23890 (aarch64_print_address_internal): Handle SVE addresses.
23891 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
23892 (aarch64_regno_regclass): Handle predicate registers.
23893 (aarch64_secondary_reload): Handle big-endian reloads of SVE
23895 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
23896 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
23897 (aarch64_convert_sve_vector_bits): New function.
23898 (aarch64_override_options): Use it to handle -msve-vector-bits=.
23899 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23900 rather than an rtx.
23901 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
23902 Handle SVE vector and predicate modes. Accept VL-based constants
23903 that need only one temporary register, and VL offsets that require
23904 no temporary registers.
23905 (aarch64_conditional_register_usage): Mark the predicate registers
23906 as fixed if SVE isn't available.
23907 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
23908 Return true for SVE vector and predicate modes.
23909 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
23910 rather than an unsigned int. Handle SVE modes.
23911 (aarch64_preferred_simd_mode): Update call accordingly. Handle
23913 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
23915 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23916 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23917 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23918 (aarch64_sve_float_mul_immediate_p): New functions.
23919 (aarch64_sve_valid_immediate): New function.
23920 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
23921 Explicitly reject structure modes. Check for INDEX constants.
23922 Handle PTRUE and PFALSE constants.
23923 (aarch64_check_zero_based_sve_index_immediate): New function.
23924 (aarch64_simd_imm_zero_p): Delete.
23925 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
23926 vector modes. Accept constants in the range of CNT[BHWD].
23927 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
23928 ask for an Advanced SIMD mode.
23929 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
23930 (aarch64_simd_vector_alignment): Handle SVE predicates.
23931 (aarch64_vectorize_preferred_vector_alignment): New function.
23932 (aarch64_simd_vector_alignment_reachable): Use it instead of
23934 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
23935 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
23937 (MAX_VECT_LEN): Delete.
23938 (expand_vec_perm_d): Add a vec_flags field.
23939 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
23940 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
23941 (aarch64_evpc_ext): Don't apply a big-endian lane correction
23943 (aarch64_evpc_rev): Rename to...
23944 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
23945 (aarch64_evpc_rev_global): New function.
23946 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
23947 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
23949 (aarch64_evpc_sve_tbl): New function.
23950 (aarch64_expand_vec_perm_const_1): Update after rename of
23951 aarch64_evpc_rev. Handle SVE permutes too, trying
23952 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
23953 than aarch64_evpc_tbl.
23954 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
23955 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
23956 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
23957 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
23958 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
23959 (aarch64_expand_sve_vcond): New functions.
23960 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
23961 of aarch64_vector_mode_p.
23962 (aarch64_dwarf_poly_indeterminate_value): New function.
23963 (aarch64_compute_pressure_classes): Likewise.
23964 (aarch64_can_change_mode_class): Likewise.
23965 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
23966 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
23967 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
23968 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
23969 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
23970 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
23971 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
23972 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
23974 (Dn, Dl, Dr): Accept const as well as const_vector.
23975 (Dz): Likewise. Compare against CONST0_RTX.
23976 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
23977 of "vector" where appropriate.
23978 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
23979 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
23980 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
23981 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
23982 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
23983 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
23984 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
23985 (v_int_equiv): Extend to SVE modes.
23986 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
23988 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
23989 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
23990 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
23991 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
23992 (SVE_COND_FP_CMP): New int iterators.
23993 (perm_hilo): Handle the new unpack unspecs.
23994 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
23996 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
23997 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
23998 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
23999 (aarch64_equality_operator, aarch64_constant_vector_operand)
24000 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
24001 (aarch64_sve_nonimmediate_operand): Likewise.
24002 (aarch64_sve_general_operand): Likewise.
24003 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
24004 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
24005 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
24006 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
24007 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
24008 (aarch64_sve_float_arith_immediate): Likewise.
24009 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
24010 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
24011 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
24012 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
24013 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
24014 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
24015 (aarch64_sve_float_arith_operand): Likewise.
24016 (aarch64_sve_float_arith_with_sub_operand): Likewise.
24017 (aarch64_sve_float_mul_operand): Likewise.
24018 (aarch64_sve_vec_perm_operand): Likewise.
24019 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
24020 (aarch64_mov_operand): Accept const_poly_int and const_vector.
24021 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
24022 as well as const_vector.
24023 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
24024 in file. Use CONST0_RTX and CONSTM1_RTX.
24025 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
24026 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
24027 Use aarch64_simd_imm_zero.
24028 * config/aarch64/aarch64-sve.md: New file.
24029 * config/aarch64/aarch64.md: Include it.
24030 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
24031 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
24032 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
24033 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
24034 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
24035 (sve): New attribute.
24036 (enabled): Disable instructions with the sve attribute unless
24038 (movqi, movhi): Pass CONST_POLY_INT operaneds through
24039 aarch64_expand_mov_immediate.
24040 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
24041 CNT[BHSD] immediates.
24042 (movti): Split CONST_POLY_INT moves into two halves.
24043 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
24044 Split additions that need a temporary here if the destination
24045 is the stack pointer.
24046 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
24047 (*add<mode>3_poly_1): New instruction.
24048 (set_clobber_cc): New expander.
24050 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24052 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
24053 parameter and use it instead of GET_MODE_SIZE (innermode). Use
24054 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
24055 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
24056 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
24057 Change innermode from fixed_mode_size to machine_mode.
24058 (simplify_subreg): Update call accordingly. Handle a constant-sized
24059 subreg of a variable-length CONST_VECTOR.
24061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
24062 Alan Hayward <alan.hayward@arm.com>
24063 David Sherwood <david.sherwood@arm.com>
24065 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
24066 (add_offset_to_base): New function, split out from...
24067 (create_mem_ref): ...here. When handling a scale other than 1,
24068 check first whether the address is valid without the offset.
24069 Add it into the base if so, leaving the index and scale as-is.
24071 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24074 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
24075 fold_for_warn before checking if arg2 is INTEGER_CST.
24077 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
24079 * config/rs6000/predicates.md (load_multiple_operation): Delete.
24080 (store_multiple_operation): Delete.
24081 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
24082 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
24083 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
24084 guarded by TARGET_STRING.
24085 (rs6000_output_load_multiple): Delete.
24086 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
24087 OPTION_MASK_STRING / TARGET_STRING handling.
24088 (print_operand) <'N', 'O'>: Add comment that these are unused now.
24089 (const rs6000_opt_masks) <"string">: Change mask to 0.
24090 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
24091 (MASK_STRING): Delete.
24092 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
24094 (load_multiple): Delete.
24101 (store_multiple): Delete.
24108 (movmemsi_8reg): Delete.
24109 (corresponding unnamed define_insn): Delete.
24110 (movmemsi_6reg): Delete.
24111 (corresponding unnamed define_insn): Delete.
24112 (movmemsi_4reg): Delete.
24113 (corresponding unnamed define_insn): Delete.
24114 (movmemsi_2reg): Delete.
24115 (corresponding unnamed define_insn): Delete.
24116 (movmemsi_1reg): Delete.
24117 (corresponding unnamed define_insn): Delete.
24118 * config/rs6000/rs6000.opt (mno-string): New.
24119 (mstring): Replace by deprecation warning stub.
24120 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24122 2018-01-12 Jakub Jelinek <jakub@redhat.com>
24124 * regrename.c (regrename_do_replace): If replacing the same
24125 reg multiple times, try to reuse last created gen_raw_REG.
24128 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24129 main to workaround a bug in GDB.
24131 2018-01-12 Tom de Vries <tom@codesourcery.com>
24134 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24136 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
24138 PR rtl-optimization/80481
24139 * ira-color.c (get_cap_member): New function.
24140 (allocnos_conflict_by_live_ranges_p): Use it.
24141 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24142 (setup_slot_coalesced_allocno_live_ranges): Ditto.
24144 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
24147 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24148 (*saddl_se_1): Ditto.
24149 (*ssubsi_1): Ditto.
24150 (*ssubl_se_1): Ditto.
24152 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24154 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24155 rather than wi::to_widest for DR_INITs.
24156 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24157 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24158 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24160 (vect_analyze_group_access_1): Note that here.
24162 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24164 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24165 polynomial type sizes.
24167 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
24169 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24170 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24171 (gimple_add_tmp_var): Likewise.
24173 2018-01-12 Martin Liska <mliska@suse.cz>
24175 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24176 (gimple_alloc_sizes): Likewise.
24177 (dump_gimple_statistics): Use PRIu64 in printf format.
24178 * gimple.h: Change uint64_t to int.
24180 2018-01-12 Martin Liska <mliska@suse.cz>
24182 * tree-core.h: Use uint64_t instead of int.
24183 * tree.c (tree_node_counts): Likewise.
24184 (tree_node_sizes): Likewise.
24185 (dump_tree_statistics): Use PRIu64 in printf format.
24187 2018-01-12 Martin Liska <mliska@suse.cz>
24189 * Makefile.in: As qsort_chk is implemented in vec.c, add
24190 vec.o to linkage of gencfn-macros.
24191 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24192 passing the info to record_node_allocation_statistics.
24193 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24195 * ggc-common.c (struct ggc_usage): Add operator== and use
24196 it in operator< and compare function.
24197 * mem-stats.h (struct mem_usage): Likewise.
24198 * vec.c (struct vec_usage): Remove operator< and compare
24199 function. Can be simply inherited.
24201 2018-01-12 Martin Jambor <mjambor@suse.cz>
24204 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24205 * tree-ssa-math-opts.c: Include domwalk.h.
24206 (convert_mult_to_fma_1): New function.
24207 (fma_transformation_info): New type.
24208 (fma_deferring_state): Likewise.
24209 (cancel_fma_deferring): New function.
24210 (result_of_phi): Likewise.
24211 (last_fma_candidate_feeds_initial_phi): Likewise.
24212 (convert_mult_to_fma): Added deferring logic, split actual
24213 transformation to convert_mult_to_fma_1.
24214 (math_opts_dom_walker): New type.
24215 (math_opts_dom_walker::after_dom_children): New method, body moved
24216 here from pass_optimize_widening_mul::execute, added deferring logic
24218 (pass_optimize_widening_mul::execute): Moved most of code to
24219 math_opts_dom_walker::after_dom_children.
24220 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24221 * config/i386/i386.c (ix86_option_override_internal): Added
24222 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24224 2018-01-12 Richard Biener <rguenther@suse.de>
24227 * dwarf2out.c (gen_variable_die): Do not reset old_die for
24228 inline instance vars.
24230 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
24233 * config/rx/rx.c (rx_is_restricted_memory_address):
24234 Handle SUBREG case.
24236 2018-01-12 Richard Biener <rguenther@suse.de>
24238 PR tree-optimization/80846
24239 * target.def (split_reduction): New target hook.
24240 * targhooks.c (default_split_reduction): New function.
24241 * targhooks.h (default_split_reduction): Declare.
24242 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
24243 target requests first reduce vectors by combining low and high
24245 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
24246 (get_vectype_for_scalar_type_and_size): Export.
24247 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
24248 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
24249 * doc/tm.texi: Regenerate.
24250 * config/i386/i386.c (ix86_split_reduction): Implement
24251 TARGET_VECTORIZE_SPLIT_REDUCTION.
24253 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24256 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
24257 in PIC mode except for TARGET_VXWORKS_RTP.
24258 * config/sparc/sparc.c: Include cfgrtl.h.
24259 (TARGET_INIT_PIC_REG): Define.
24260 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
24261 (sparc_pic_register_p): New predicate.
24262 (sparc_legitimate_address_p): Use it.
24263 (sparc_legitimize_pic_address): Likewise.
24264 (sparc_delegitimize_address): Likewise.
24265 (sparc_mode_dependent_address_p): Likewise.
24266 (gen_load_pcrel_sym): Remove 4th parameter.
24267 (load_got_register): Adjust call to above. Remove obsolete stuff.
24268 (sparc_expand_prologue): Do not call load_got_register here.
24269 (sparc_flat_expand_prologue): Likewise.
24270 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
24271 (sparc_use_pseudo_pic_reg): New function.
24272 (sparc_init_pic_reg): Likewise.
24273 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24274 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
24276 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
24278 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
24279 Add item for branch_cost.
24281 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
24283 PR rtl-optimization/83565
24284 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
24285 not extend the result to a larger mode for rotate operations.
24286 (num_sign_bit_copies1): Likewise.
24288 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
24291 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
24293 Use values-Xc.o for -pedantic.
24294 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
24296 2018-01-12 Martin Liska <mliska@suse.cz>
24299 * ipa-devirt.c (final_warning_record::grow_type_warnings):
24301 (possible_polymorphic_call_targets): Use it.
24302 (ipa_devirt): Likewise.
24304 2018-01-12 Martin Liska <mliska@suse.cz>
24306 * profile-count.h (enum profile_quality): Use 0 as invalid
24307 enum value of profile_quality.
24309 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
24311 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
24312 -mext-string options.
24314 2018-01-12 Richard Biener <rguenther@suse.de>
24316 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
24317 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
24318 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
24320 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
24322 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
24324 * configure.ac (--with-long-double-format): Add support for the
24325 configuration option to change the default long double format on
24327 * config.gcc (powerpc*-linux*-*): Likewise.
24328 * configure: Regenerate.
24329 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
24330 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
24331 used without modification.
24333 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24335 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
24336 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
24337 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
24338 MISC_BUILTIN_SPEC_BARRIER.
24339 (rs6000_init_builtins): Likewise.
24340 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
24342 (speculation_barrier): New define_insn.
24343 * doc/extend.texi: Document __builtin_speculation_barrier.
24345 2018-01-11 Jakub Jelinek <jakub@redhat.com>
24348 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
24349 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
24350 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
24352 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
24353 integral modes instead of "ss" and "sd".
24354 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
24355 vectors with 32-bit and 64-bit elements.
24356 (vecdupssescalarmodesuffix): New mode attribute.
24357 (vec_dup<mode>): Use it.
24359 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
24362 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
24363 frame if argument is passed on stack.
24365 2018-01-11 Jakub Jelinek <jakub@redhat.com>
24368 * ree.c (combine_reaching_defs): Optimize also
24369 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
24370 reg2=any_extend(exp); reg1=reg2;, formatting fix.
24372 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
24374 PR middle-end/83189
24375 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
24377 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
24379 PR middle-end/83718
24380 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
24381 after they are computed.
24383 2018-01-11 Bin Cheng <bin.cheng@arm.com>
24385 PR tree-optimization/83695
24386 * gimple-loop-linterchange.cc
24387 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
24388 reset cached scev information after interchange.
24389 (pass_linterchange::execute): Remove call to scev_reset_htab.
24391 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24393 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
24394 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
24395 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
24396 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
24397 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
24398 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
24399 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
24400 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
24401 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
24402 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
24403 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
24404 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
24405 (V_lane_reg): Likewise.
24406 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
24408 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
24409 (vfmal_lane_low<mode>_intrinsic,
24410 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
24411 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
24412 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
24413 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
24414 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
24415 vfmsl_lane_high<mode>_intrinsic): New define_insns.
24417 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24419 * config/arm/arm-cpus.in (fp16fml): New feature.
24420 (ALL_SIMD): Add fp16fml.
24421 (armv8.2-a): Add fp16fml as an option.
24422 (armv8.3-a): Likewise.
24423 (armv8.4-a): Add fp16fml as part of fp16.
24424 * config/arm/arm.h (TARGET_FP16FML): Define.
24425 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
24427 * config/arm/arm-modes.def (V2HF): Define.
24428 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
24429 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
24430 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
24431 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
24432 vfmsl_low, vfmsl_high): New set of builtins.
24433 * config/arm/iterators.md (PLUSMINUS): New code iterator.
24434 (vfml_op): New code attribute.
24435 (VFMLHALVES): New int iterator.
24436 (VFML, VFMLSEL): New mode attributes.
24437 (V_reg): Define mapping for V2HF.
24438 (V_hi, V_lo): New mode attributes.
24439 (VF_constraint): Likewise.
24440 (vfml_half, vfml_half_selector): New int attributes.
24441 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
24443 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
24444 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
24446 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
24447 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
24448 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
24449 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
24451 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
24452 Document new effective target and option set.
24454 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24456 * config/arm/arm-cpus.in (armv8_4): New feature.
24457 (ARMv8_4a): New fgroup.
24458 (armv8.4-a): New arch.
24459 * config/arm/arm-tables.opt: Regenerate.
24460 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
24461 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
24462 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
24463 Add matching rules for -march=armv8.4-a and extensions.
24464 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
24466 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
24469 * config/rx/rx.md (BW): New mode attribute.
24470 (sync_lock_test_and_setsi): Add mode suffix to insn output.
24472 2018-01-11 Richard Biener <rguenther@suse.de>
24474 PR tree-optimization/83435
24475 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
24476 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
24477 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
24479 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24480 Alan Hayward <alan.hayward@arm.com>
24481 David Sherwood <david.sherwood@arm.com>
24483 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
24485 (aarch64_classify_address): Initialize it. Track polynomial offsets.
24486 (aarch64_print_address_internal): Use it to check for a zero offset.
24488 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24489 Alan Hayward <alan.hayward@arm.com>
24490 David Sherwood <david.sherwood@arm.com>
24492 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
24493 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
24494 Return a poly_int64 rather than a HOST_WIDE_INT.
24495 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
24496 rather than a HOST_WIDE_INT.
24497 * config/aarch64/aarch64.h (aarch64_frame): Protect with
24498 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
24499 hard_fp_offset, frame_size, initial_adjust, callee_offset and
24500 final_offset from HOST_WIDE_INT to poly_int64.
24501 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
24502 to_constant when getting the number of units in an Advanced SIMD
24504 (aarch64_builtin_vectorized_function): Check for a constant number
24506 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
24508 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
24509 attribute instead of GET_MODE_NUNITS.
24510 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
24511 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
24512 GET_MODE_SIZE for fixed-size registers.
24513 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
24514 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
24515 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
24516 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
24517 (aarch64_print_operand, aarch64_print_address_internal)
24518 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
24519 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
24520 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
24521 Handle polynomial GET_MODE_SIZE.
24522 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
24523 wider than SImode without modification.
24524 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
24525 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
24526 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
24527 passing and returning SVE modes.
24528 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
24529 rather than GEN_INT.
24530 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
24531 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
24532 (aarch64_allocate_and_probe_stack_space): Likewise.
24533 (aarch64_layout_frame): Cope with polynomial offsets.
24534 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
24535 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
24536 polynomial offsets.
24537 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
24538 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
24539 poly_int64 rather than a HOST_WIDE_INT.
24540 (aarch64_get_separate_components, aarch64_process_components)
24541 (aarch64_expand_prologue, aarch64_expand_epilogue)
24542 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
24543 (aarch64_anchor_offset): New function, split out from...
24544 (aarch64_legitimize_address): ...here.
24545 (aarch64_builtin_vectorization_cost): Handle polynomial
24546 TYPE_VECTOR_SUBPARTS.
24547 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
24549 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
24550 number of elements from the PARALLEL rather than the mode.
24551 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
24552 rather than GET_MODE_BITSIZE.
24553 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
24554 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
24555 (aarch64_expand_vec_perm_const_1): Handle polynomial
24556 d->perm.length () and d->perm elements.
24557 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
24558 Apply to_constant to d->perm elements.
24559 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
24560 polynomial CONST_VECTOR_NUNITS.
24561 (aarch64_move_pointer): Take amount as a poly_int64 rather
24563 (aarch64_progress_pointer): Avoid temporary variable.
24564 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
24565 the mode attribute instead of GET_MODE.
24567 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24568 Alan Hayward <alan.hayward@arm.com>
24569 David Sherwood <david.sherwood@arm.com>
24571 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
24572 x exists before using it.
24573 (aarch64_add_constant_internal): Rename to...
24574 (aarch64_add_offset_1): ...this. Replace regnum with separate
24575 src and dest rtxes. Handle the case in which they're different,
24576 including when the offset is zero. Replace scratchreg with an rtx.
24577 Use 2 additions if there is no spare register into which we can
24578 move a 16-bit constant.
24579 (aarch64_add_constant): Delete.
24580 (aarch64_add_offset): Replace reg with separate src and dest
24581 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
24582 Use aarch64_add_offset_1.
24583 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
24584 an rtx rather than an int. Take the delta as a poly_int64
24585 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
24586 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
24587 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
24588 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
24589 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
24590 and aarch64_add_sp.
24591 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
24592 aarch64_add_constant.
24594 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24596 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
24597 Use scalar_float_mode.
24599 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24601 * config/aarch64/aarch64-simd.md
24602 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
24603 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
24604 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
24605 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
24606 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
24607 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
24608 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
24609 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
24610 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
24611 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
24613 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24616 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
24617 targ_options->x_arm_arch_string is non NULL.
24619 2018-01-11 Tamar Christina <tamar.christina@arm.com>
24621 * config/aarch64/aarch64.h
24622 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
24624 2018-01-11 Sudakshina Das <sudi.das@arm.com>
24627 * expmed.c (emit_store_flag_force): Swap if const op0
24628 and change VOIDmode to mode of op0.
24630 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
24632 PR rtl-optimization/83761
24633 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
24634 than bytes to mode_for_size.
24636 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24638 PR middle-end/83189
24639 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
24640 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
24643 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
24645 PR middle-end/83575
24646 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
24647 when in layout mode.
24648 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
24649 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
24652 2018-01-10 Michael Collison <michael.collison@arm.com>
24654 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
24655 * config/aarch64/aarch64-option-extension.def: Add
24656 AARCH64_OPT_EXTENSION of 'fp16fml'.
24657 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24658 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
24659 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
24660 * config/aarch64/constraints.md (Ui7): New constraint.
24661 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
24662 (VFMLA_SEL_W): Ditto.
24665 (VFMLA16_LOW): New int iterator.
24666 (VFMLA16_HIGH): Ditto.
24667 (UNSPEC_FMLAL): New unspec.
24668 (UNSPEC_FMLSL): Ditto.
24669 (UNSPEC_FMLAL2): Ditto.
24670 (UNSPEC_FMLSL2): Ditto.
24671 (f16mac): New code attribute.
24672 * config/aarch64/aarch64-simd-builtins.def
24673 (aarch64_fmlal_lowv2sf): Ditto.
24674 (aarch64_fmlsl_lowv2sf): Ditto.
24675 (aarch64_fmlalq_lowv4sf): Ditto.
24676 (aarch64_fmlslq_lowv4sf): Ditto.
24677 (aarch64_fmlal_highv2sf): Ditto.
24678 (aarch64_fmlsl_highv2sf): Ditto.
24679 (aarch64_fmlalq_highv4sf): Ditto.
24680 (aarch64_fmlslq_highv4sf): Ditto.
24681 (aarch64_fmlal_lane_lowv2sf): Ditto.
24682 (aarch64_fmlsl_lane_lowv2sf): Ditto.
24683 (aarch64_fmlal_laneq_lowv2sf): Ditto.
24684 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
24685 (aarch64_fmlalq_lane_lowv4sf): Ditto.
24686 (aarch64_fmlsl_lane_lowv4sf): Ditto.
24687 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
24688 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
24689 (aarch64_fmlal_lane_highv2sf): Ditto.
24690 (aarch64_fmlsl_lane_highv2sf): Ditto.
24691 (aarch64_fmlal_laneq_highv2sf): Ditto.
24692 (aarch64_fmlsl_laneq_highv2sf): Ditto.
24693 (aarch64_fmlalq_lane_highv4sf): Ditto.
24694 (aarch64_fmlsl_lane_highv4sf): Ditto.
24695 (aarch64_fmlalq_laneq_highv4sf): Ditto.
24696 (aarch64_fmlsl_laneq_highv4sf): Ditto.
24697 * config/aarch64/aarch64-simd.md:
24698 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
24699 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24700 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
24701 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24702 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
24703 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
24704 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
24705 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
24706 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
24707 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
24708 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
24709 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
24710 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
24711 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
24712 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
24713 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
24714 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
24715 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
24716 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
24717 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
24718 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
24719 (vfmlsl_low_u32): Ditto.
24720 (vfmlalq_low_u32): Ditto.
24721 (vfmlslq_low_u32): Ditto.
24722 (vfmlal_high_u32): Ditto.
24723 (vfmlsl_high_u32): Ditto.
24724 (vfmlalq_high_u32): Ditto.
24725 (vfmlslq_high_u32): Ditto.
24726 (vfmlal_lane_low_u32): Ditto.
24727 (vfmlsl_lane_low_u32): Ditto.
24728 (vfmlal_laneq_low_u32): Ditto.
24729 (vfmlsl_laneq_low_u32): Ditto.
24730 (vfmlalq_lane_low_u32): Ditto.
24731 (vfmlslq_lane_low_u32): Ditto.
24732 (vfmlalq_laneq_low_u32): Ditto.
24733 (vfmlslq_laneq_low_u32): Ditto.
24734 (vfmlal_lane_high_u32): Ditto.
24735 (vfmlsl_lane_high_u32): Ditto.
24736 (vfmlal_laneq_high_u32): Ditto.
24737 (vfmlsl_laneq_high_u32): Ditto.
24738 (vfmlalq_lane_high_u32): Ditto.
24739 (vfmlslq_lane_high_u32): Ditto.
24740 (vfmlalq_laneq_high_u32): Ditto.
24741 (vfmlslq_laneq_high_u32): Ditto.
24742 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
24743 (AARCH64_FL_FOR_ARCH8_4): New.
24744 (AARCH64_ISA_F16FML): New ISA flag.
24745 (TARGET_F16FML): New feature flag for fp16fml.
24746 (doc/invoke.texi): Document new fp16fml option.
24748 2018-01-10 Michael Collison <michael.collison@arm.com>
24750 * config/aarch64/aarch64-builtins.c:
24751 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
24752 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24753 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
24754 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
24755 (AARCH64_ISA_SHA3): New ISA flag.
24756 (TARGET_SHA3): New feature flag for sha3.
24757 * config/aarch64/iterators.md (sha512_op): New int attribute.
24758 (CRYPTO_SHA512): New int iterator.
24759 (UNSPEC_SHA512H): New unspec.
24760 (UNSPEC_SHA512H2): Ditto.
24761 (UNSPEC_SHA512SU0): Ditto.
24762 (UNSPEC_SHA512SU1): Ditto.
24763 * config/aarch64/aarch64-simd-builtins.def
24764 (aarch64_crypto_sha512hqv2di): New builtin.
24765 (aarch64_crypto_sha512h2qv2di): Ditto.
24766 (aarch64_crypto_sha512su0qv2di): Ditto.
24767 (aarch64_crypto_sha512su1qv2di): Ditto.
24768 (aarch64_eor3qv8hi): Ditto.
24769 (aarch64_rax1qv2di): Ditto.
24770 (aarch64_xarqv2di): Ditto.
24771 (aarch64_bcaxqv8hi): Ditto.
24772 * config/aarch64/aarch64-simd.md:
24773 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
24774 (aarch64_crypto_sha512su0qv2di): Ditto.
24775 (aarch64_crypto_sha512su1qv2di): Ditto.
24776 (aarch64_eor3qv8hi): Ditto.
24777 (aarch64_rax1qv2di): Ditto.
24778 (aarch64_xarqv2di): Ditto.
24779 (aarch64_bcaxqv8hi): Ditto.
24780 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
24781 (vsha512h2q_u64): Ditto.
24782 (vsha512su0q_u64): Ditto.
24783 (vsha512su1q_u64): Ditto.
24784 (veor3q_u16): Ditto.
24785 (vrax1q_u64): Ditto.
24786 (vxarq_u64): Ditto.
24787 (vbcaxq_u16): Ditto.
24788 * config/arm/types.md (crypto_sha512): New type attribute.
24789 (crypto_sha3): Ditto.
24790 (doc/invoke.texi): Document new sha3 option.
24792 2018-01-10 Michael Collison <michael.collison@arm.com>
24794 * config/aarch64/aarch64-builtins.c:
24795 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
24796 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24797 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
24798 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
24799 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
24800 (AARCH64_ISA_SM4): New ISA flag.
24801 (TARGET_SM4): New feature flag for sm4.
24802 * config/aarch64/aarch64-simd-builtins.def
24803 (aarch64_sm3ss1qv4si): Ditto.
24804 (aarch64_sm3tt1aq4si): Ditto.
24805 (aarch64_sm3tt1bq4si): Ditto.
24806 (aarch64_sm3tt2aq4si): Ditto.
24807 (aarch64_sm3tt2bq4si): Ditto.
24808 (aarch64_sm3partw1qv4si): Ditto.
24809 (aarch64_sm3partw2qv4si): Ditto.
24810 (aarch64_sm4eqv4si): Ditto.
24811 (aarch64_sm4ekeyqv4si): Ditto.
24812 * config/aarch64/aarch64-simd.md:
24813 (aarch64_sm3ss1qv4si): Ditto.
24814 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
24815 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
24816 (aarch64_sm4eqv4si): Ditto.
24817 (aarch64_sm4ekeyqv4si): Ditto.
24818 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
24819 (sm3part_op): Ditto.
24820 (CRYPTO_SM3TT): Ditto.
24821 (CRYPTO_SM3PART): Ditto.
24822 (UNSPEC_SM3SS1): New unspec.
24823 (UNSPEC_SM3TT1A): Ditto.
24824 (UNSPEC_SM3TT1B): Ditto.
24825 (UNSPEC_SM3TT2A): Ditto.
24826 (UNSPEC_SM3TT2B): Ditto.
24827 (UNSPEC_SM3PARTW1): Ditto.
24828 (UNSPEC_SM3PARTW2): Ditto.
24829 (UNSPEC_SM4E): Ditto.
24830 (UNSPEC_SM4EKEY): Ditto.
24831 * config/aarch64/constraints.md (Ui2): New constraint.
24832 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
24833 * config/arm/types.md (crypto_sm3): New type attribute.
24834 (crypto_sm4): Ditto.
24835 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
24836 (vsm3tt1aq_u32): Ditto.
24837 (vsm3tt1bq_u32): Ditto.
24838 (vsm3tt2aq_u32): Ditto.
24839 (vsm3tt2bq_u32): Ditto.
24840 (vsm3partw1q_u32): Ditto.
24841 (vsm3partw2q_u32): Ditto.
24842 (vsm4eq_u32): Ditto.
24843 (vsm4ekeyq_u32): Ditto.
24844 (doc/invoke.texi): Document new sm4 option.
24846 2018-01-10 Michael Collison <michael.collison@arm.com>
24848 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
24849 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
24850 (AARCH64_FL_FOR_ARCH8_4): New.
24851 (AARCH64_FL_V8_4): New flag.
24852 (doc/invoke.texi): Document new armv8.4-a option.
24854 2018-01-10 Michael Collison <michael.collison@arm.com>
24856 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24857 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
24858 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
24859 * config/aarch64/aarch64-option-extension.def: Add
24860 AARCH64_OPT_EXTENSION of 'sha2'.
24861 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
24862 (crypto): Disable sha2 and aes if crypto disabled.
24863 (crypto): Enable aes and sha2 if enabled.
24864 (simd): Disable sha2 and aes if simd disabled.
24865 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
24867 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
24868 (TARGET_SHA2): New feature flag for sha2.
24869 (TARGET_AES): New feature flag for aes.
24870 * config/aarch64/aarch64-simd.md:
24871 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
24872 conditional on TARGET_AES.
24873 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
24874 (aarch64_crypto_sha1hsi): Make pattern conditional
24876 (aarch64_crypto_sha1hv4si): Ditto.
24877 (aarch64_be_crypto_sha1hv4si): Ditto.
24878 (aarch64_crypto_sha1su1v4si): Ditto.
24879 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
24880 (aarch64_crypto_sha1su0v4si): Ditto.
24881 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
24882 (aarch64_crypto_sha256su0v4si): Ditto.
24883 (aarch64_crypto_sha256su1v4si): Ditto.
24884 (doc/invoke.texi): Document new aes and sha2 options.
24886 2018-01-10 Martin Sebor <msebor@redhat.com>
24888 PR tree-optimization/83781
24889 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
24892 2018-01-11 Martin Sebor <msebor@gmail.com>
24893 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
24895 PR tree-optimization/83501
24896 PR tree-optimization/81703
24898 * tree-ssa-strlen.c (get_string_cst): Rename...
24899 (get_string_len): ...to this. Handle global constants.
24900 (handle_char_store): Adjust.
24902 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
24903 Jim Wilson <jimw@sifive.com>
24905 * config/riscv/riscv-protos.h (riscv_output_return): New.
24906 * config/riscv/riscv.c (struct machine_function): New naked_p field.
24907 (riscv_attribute_table, riscv_output_return),
24908 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
24909 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
24910 (riscv_compute_frame_info): Only compute frame->mask if not a naked
24912 (riscv_expand_prologue): Add early return for naked function.
24913 (riscv_expand_epilogue): Likewise.
24914 (riscv_function_ok_for_sibcall): Return false for naked function.
24915 (riscv_set_current_function): New.
24916 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
24917 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
24918 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
24919 * doc/extend.texi (RISC-V Function Attributes): New.
24921 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
24923 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
24924 check for 128-bit long double before checking TCmode.
24925 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
24926 128-bit long doubles before checking TFmode or TCmode.
24927 (FLOAT128_IBM_P): Likewise.
24929 2018-01-10 Martin Sebor <msebor@redhat.com>
24931 PR tree-optimization/83671
24932 * builtins.c (c_strlen): Unconditionally return zero for the empty
24934 Use -Warray-bounds for warnings.
24935 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
24936 for non-constant array indices with COMPONENT_REF, arrays of
24937 arrays, and pointers to arrays.
24938 (gimple_fold_builtin_strlen): Determine and set length range for
24939 non-constant character arrays.
24941 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
24943 PR middle-end/81897
24944 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
24947 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
24949 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
24951 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
24954 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
24955 VECTOR_MEM_ALTIVEC_OR_VSX_P.
24956 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
24957 indexed_or_indirect_operand predicate.
24958 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
24959 (*vsx_le_perm_load_v8hi): Likewise.
24960 (*vsx_le_perm_load_v16qi): Likewise.
24961 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
24962 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
24963 (*vsx_le_perm_store_v8hi): Likewise.
24964 (*vsx_le_perm_store_v16qi): Likewise.
24965 (eight unnamed splitters): Likewise.
24967 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
24969 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
24970 * config/rs6000/emmintrin.h: Likewise.
24971 * config/rs6000/mmintrin.h: Likewise.
24972 * config/rs6000/xmmintrin.h: Likewise.
24974 2018-01-10 David Malcolm <dmalcolm@redhat.com>
24977 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
24979 * tree.c (tree_nop_conversion): Return true for location wrapper
24981 (maybe_wrap_with_location): New function.
24982 (selftest::check_strip_nops): New function.
24983 (selftest::test_location_wrappers): New function.
24984 (selftest::tree_c_tests): Call it.
24985 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
24986 (maybe_wrap_with_location): New decl.
24987 (EXPR_LOCATION_WRAPPER_P): New macro.
24988 (location_wrapper_p): New inline function.
24989 (tree_strip_any_location_wrapper): New inline function.
24991 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
24994 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
24995 stack_realign_offset for the largest alignment of stack slot
24997 (ix86_find_max_used_stack_alignment): New function.
24998 (ix86_finalize_stack_frame_flags): Use it. Set
24999 max_used_stack_alignment if we don't realign stack.
25000 * config/i386/i386.h (machine_function): Add
25001 max_used_stack_alignment.
25003 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
25005 * config/arm/arm.opt (-mbranch-cost): New option.
25006 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
25009 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
25012 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
25013 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
25015 2018-01-10 Richard Biener <rguenther@suse.de>
25018 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
25019 early out so it also covers the case where we have a non-NULL
25022 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25024 PR tree-optimization/83753
25025 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
25026 for non-strided grouped accesses if the number of elements is 1.
25028 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
25031 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
25032 * i386.h (TARGET_USE_GATHER): Define.
25033 * x86-tune.def (X86_TUNE_USE_GATHER): New.
25035 2018-01-10 Martin Liska <mliska@suse.cz>
25038 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
25039 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
25041 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
25042 CLEANUP_NO_PARTITIONING is not set.
25044 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
25046 * doc/rtl.texi: Remove documentation of (const ...) wrappers
25047 for vectors, as a partial revert of r254296.
25048 * rtl.h (const_vec_p): Delete.
25049 (const_vec_duplicate_p): Don't test for vector CONSTs.
25050 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
25051 * expmed.c (make_tree): Likewise.
25054 * common.md (E, F): Use CONSTANT_P instead of checking for
25056 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
25057 checking for CONST_VECTOR.
25059 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25061 PR middle-end/83575
25062 * predict.c (force_edge_cold): Handle in more sane way edges
25063 with no prediction.
25065 2018-01-09 Carl Love <cel@us.ibm.com>
25067 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
25069 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
25070 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
25071 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
25072 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
25073 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
25074 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
25075 * config/rs6000/rs6000-protos.h: Add extern defition for
25076 rs6000_generate_float2_double_code.
25077 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
25079 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
25080 (float2_v2df): Add define_expand.
25082 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
25085 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
25086 op_mode in the force_to_mode call.
25088 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25090 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
25091 instead of checking each element individually.
25092 (aarch64_evpc_uzp): Likewise.
25093 (aarch64_evpc_zip): Likewise.
25094 (aarch64_evpc_ext): Likewise.
25095 (aarch64_evpc_rev): Likewise.
25096 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
25097 instead of checking each element individually. Return true without
25099 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25100 whether all selected elements come from the same input, instead of
25101 checking each element individually. Remove calls to gen_rtx_REG,
25102 start_sequence and end_sequence and instead assert that no rtl is
25105 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25107 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25108 order of HIGH and CONST checks.
25110 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
25112 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25113 if the destination isn't an SSA_NAME.
25115 2018-01-09 Richard Biener <rguenther@suse.de>
25117 PR tree-optimization/83668
25118 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25120 (canonicalize_loop_form): ... here, renamed from ...
25121 (canonicalize_loop_closed_ssa_form): ... this and amended to
25122 swap successor edges for loop exit blocks to make us use
25123 the RPO order we need for initial schedule generation.
25125 2018-01-09 Joseph Myers <joseph@codesourcery.com>
25127 PR tree-optimization/64811
25128 * match.pd: When optimizing comparisons with Inf, avoid
25129 introducing or losing exceptions from comparisons with NaN.
25131 2018-01-09 Martin Liska <mliska@suse.cz>
25134 * asan.c (shadow_mem_size): Add gcc_assert.
25136 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
25138 Don't save registers in main().
25141 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25142 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25143 * config/avr/avr.c (avr_set_current_function): Don't error if
25144 naked, OS_task or OS_main are specified at the same time.
25145 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25147 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25149 * common/config/avr/avr-common.c (avr_option_optimization_table):
25150 Switch on -mmain-is-OS_task for optimizing compilations.
25152 2018-01-09 Richard Biener <rguenther@suse.de>
25154 PR tree-optimization/83572
25155 * graphite.c: Include cfganal.h.
25156 (graphite_transform_loops): Connect infinite loops to exit
25157 and remove fake edges at the end.
25159 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25161 * ipa-inline.c (edge_badness): Revert accidental checkin.
25163 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
25166 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25167 symbols; not inline clones.
25169 2018-01-09 Jakub Jelinek <jakub@redhat.com>
25172 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25173 hard registers. Formatting fixes.
25175 PR preprocessor/83722
25176 * gcc.c (try_generate_repro): Pass
25177 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25178 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25181 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
25182 Kito Cheng <kito.cheng@gmail.com>
25184 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25185 (riscv_leaf_function_p): Delete.
25186 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25188 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25190 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25192 (do_ifelse): New function.
25193 (do_isel): New function.
25194 (do_sub3): New function.
25195 (do_add3): New function.
25196 (do_load_mask_compare): New function.
25197 (do_overlap_load_compare): New function.
25198 (expand_compare_loop): New function.
25199 (expand_block_compare): Call expand_compare_loop() when appropriate.
25200 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25201 option description.
25202 (-mblock-compare-inline-loop-limit): New option.
25204 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
25207 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25208 Reverse order of second and third operands in first alternative.
25209 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25210 of first and second elements in UNSPEC_VPERMR vector.
25211 (altivec_expand_vec_perm_le): Likewise.
25213 2018-01-08 Jeff Law <law@redhat.com>
25215 PR rtl-optimizatin/81308
25216 * tree-switch-conversion.c (cfg_altered): New file scoped static.
25217 (process_switch): If group_case_labels makes a change, then set
25219 (pass_convert_switch::execute): If a switch is converted, then
25220 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
25222 PR rtl-optimization/81308
25223 * recog.c (split_all_insns): Conditionally cleanup the CFG after
25226 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
25228 PR target/83663 - Revert r255946
25229 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25230 generation for cases where splatting a value is not useful.
25231 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25232 across a vec_duplicate and a paradoxical subreg forming a vector
25233 mode to a vec_concat.
25235 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
25237 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
25238 -march=armv8.3-a variants.
25239 * config/arm/t-multilib: Likewise.
25240 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
25242 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
25244 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
25246 (cceq_ior_compare_complement): Give it a name so I can use it, and
25247 change boolean_or_operator predicate to boolean_operator so it can
25248 be used to generate a crand.
25249 (eqne): New code iterator.
25250 (bd/bd_neg): New code_attrs.
25251 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
25252 a single define_insn.
25253 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
25254 decrement (bdnzt/bdnzf/bdzt/bdzf).
25255 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
25256 with the new names of the branch decrement patterns, and added the
25257 names of the branch decrement conditional patterns.
25259 2018-01-08 Richard Biener <rguenther@suse.de>
25261 PR tree-optimization/83563
25262 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
25265 2018-01-08 Richard Biener <rguenther@suse.de>
25267 PR middle-end/83713
25268 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
25270 2018-01-08 Richard Biener <rguenther@suse.de>
25272 PR tree-optimization/83685
25273 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
25274 references to abnormals.
25276 2018-01-08 Richard Biener <rguenther@suse.de>
25279 * dwarf2out.c (output_indirect_strings): Handle empty
25280 skeleton_debug_str_hash.
25281 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
25283 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
25285 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
25286 (emit_store_direct): Likewise.
25287 (arc_trampoline_adjust_address): Likewise.
25288 (arc_asm_trampoline_template): New function.
25289 (arc_initialize_trampoline): Use asm_trampoline_template.
25290 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
25291 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
25292 * config/arc/arc.md (flush_icache): Delete pattern.
25294 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
25296 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
25297 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
25300 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
25303 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
25304 by not USED_FOR_TARGET.
25305 (make_pass_resolve_sw_modes): Likewise.
25307 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
25309 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
25312 2018-01-08 Richard Biener <rguenther@suse.de>
25314 PR middle-end/83580
25315 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
25317 2018-01-08 Richard Biener <rguenther@suse.de>
25319 PR middle-end/83517
25320 * match.pd ((t * 2) / 2) -> t): Add missing :c.
25322 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
25324 PR middle-end/81897
25325 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
25326 basic blocks with a small number of successors.
25327 (convert_control_dep_chain_into_preds): Improve handling of
25329 (dump_predicates): Split apart into...
25330 (dump_pred_chain): ...here...
25331 (dump_pred_info): ...and here.
25332 (can_one_predicate_be_invalidated_p): Add debugging printfs.
25333 (can_chain_union_be_invalidated_p): Improve check for invalidation
25335 (uninit_uses_cannot_happen): Avoid unnecessary if
25336 convert_control_dep_chain_into_preds yielded nothing.
25338 2018-01-06 Martin Sebor <msebor@redhat.com>
25340 PR tree-optimization/83640
25341 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
25342 subtracting negative offset from size.
25343 (builtin_access::overlap): Adjust offset bounds of the access to fall
25344 within the size of the object if possible.
25346 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
25348 PR rtl-optimization/83699
25349 * expmed.c (extract_bit_field_1): Restrict the vector usage of
25350 extract_bit_field_as_subreg to cases in which the extracted
25351 value is also a vector.
25353 * lra-constraints.c (process_alt_operands): Test for the equivalence
25354 substitutions when detecting a possible reload cycle.
25356 2018-01-06 Jakub Jelinek <jakub@redhat.com>
25359 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
25360 by default if flag_selective_schedling{,2}. Formatting fixes.
25362 PR rtl-optimization/83682
25363 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
25364 if it has non-VECTOR_MODE element mode.
25365 (vec_duplicate_p): Likewise.
25367 PR middle-end/83694
25368 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
25369 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
25371 2018-01-05 Jakub Jelinek <jakub@redhat.com>
25374 * config/i386/i386-builtin.def
25375 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
25376 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
25377 Require also OPTION_MASK_ISA_AVX512F in addition to
25378 OPTION_MASK_ISA_GFNI.
25379 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
25380 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
25381 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
25382 to OPTION_MASK_ISA_GFNI.
25383 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
25384 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
25385 OPTION_MASK_ISA_AVX512BW.
25386 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
25387 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
25388 addition to OPTION_MASK_ISA_GFNI.
25389 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
25390 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
25391 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
25392 to OPTION_MASK_ISA_GFNI.
25393 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
25394 a requirement for all ISAs rather than any of them with a few
25396 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
25398 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
25399 bitmasks to be enabled with 3 exceptions, instead of requiring any
25400 enabled ISA with lots of exceptions.
25401 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
25402 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
25403 Change avx512bw in isa attribute to avx512f.
25404 * config/i386/sgxintrin.h: Add license boilerplate.
25405 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
25406 to __AVX512F__ and __AVX512VL to __AVX512VL__.
25407 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
25408 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
25410 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
25411 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
25412 temporarily sse2 rather than sse if not enabled already.
25415 * config/i386/sse.md (VI248_VLBW): Rename to ...
25416 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
25417 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
25418 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
25419 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
25420 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
25421 mode iterator instead of VI248_VLBW.
25423 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
25425 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
25426 (record_modified): Skip clobbers; add debug output.
25427 (param_change_prob): Use sreal frequencies.
25429 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
25431 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25432 punt for user-aligned variables.
25434 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
25436 * tree-chrec.c (chrec_contains_symbols): Return true for
25439 2018-01-05 Sudakshina Das <sudi.das@arm.com>
25442 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
25443 of (x|y) == x for BICS pattern.
25445 2018-01-05 Jakub Jelinek <jakub@redhat.com>
25447 PR tree-optimization/83605
25448 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
25449 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
25452 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
25454 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
25455 * config/epiphany/rtems.h: New file.
25457 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25458 Uros Bizjak <ubizjak@gmail.com>
25461 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
25462 QIreg_operand instead of register_operand predicate.
25463 * config/i386/i386.c (ix86_rop_should_change_byte_p,
25464 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
25465 comments instead of -fmitigate[-_]rop.
25467 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
25470 * cgraphunit.c (symbol_table::compile): Switch to text_section
25471 before calling assembly_start debug hook.
25472 * run-rtl-passes.c (run_rtl_passes): Likewise.
25475 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25477 * tree-vrp.c (extract_range_from_binary_expr_1): Check
25478 range_int_cst_p rather than !symbolic_range_p before calling
25479 extract_range_from_multiplicative_op_1.
25481 2018-01-04 Jeff Law <law@redhat.com>
25483 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
25484 redundant test in assertion.
25486 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25488 * doc/rtl.texi: Document machine_mode wrapper classes.
25490 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25492 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
25493 using tree_to_uhwi.
25495 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25497 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
25498 the VEC_PERM_EXPR fold to fail.
25500 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25503 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
25504 to switched_sections.
25506 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25509 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
25510 test for d.testing.
25512 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
25515 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
25516 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
25518 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25521 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
25522 is BLKmode and bitpos not zero or mode change is needed.
25524 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
25527 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
25530 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
25533 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
25534 instead of MULT rtx. Update all corresponding splitters.
25535 (*saddl_se): Ditto.
25536 (*ssub<modesuffix>): Ditto.
25537 (*ssubl_se): Ditto.
25538 (*cmp_sadd_di): Update split patterns.
25539 (*cmp_sadd_si): Ditto.
25540 (*cmp_sadd_sidi): Ditto.
25541 (*cmp_ssub_di): Ditto.
25542 (*cmp_ssub_si): Ditto.
25543 (*cmp_ssub_sidi): Ditto.
25544 * config/alpha/predicates.md (const23_operand): New predicate.
25545 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
25546 Look for ASHIFT, not MULT inner operand.
25547 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
25549 2018-01-04 Martin Liska <mliska@suse.cz>
25551 PR gcov-profile/83669
25552 * gcov.c (output_intermediate_file): Add version to intermediate
25554 * doc/gcov.texi: Document new field 'version' in intermediate
25555 file format. Fix location of '-k' option of gcov command.
25557 2018-01-04 Martin Liska <mliska@suse.cz>
25560 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
25562 2018-01-04 Jakub Jelinek <jakub@redhat.com>
25564 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
25566 2018-01-03 Martin Sebor <msebor@redhat.com>
25568 PR tree-optimization/83655
25569 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
25570 checking calls with invalid arguments.
25572 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25574 * tree-vect-stmts.c (vect_get_store_rhs): New function.
25575 (vectorizable_mask_load_store): Delete.
25576 (vectorizable_call): Return false for masked loads and stores.
25577 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
25578 instead of gimple_assign_rhs1.
25579 (vectorizable_load): Handle IFN_MASK_LOAD.
25580 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
25582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25584 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
25586 (vectorizable_mask_load_store): ...here.
25587 (vectorizable_load): ...and here.
25589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25591 * tree-vect-stmts.c (vect_build_all_ones_mask)
25592 (vect_build_zero_merge_argument): New functions, split out from...
25593 (vectorizable_load): ...here.
25595 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25597 * tree-vect-stmts.c (vect_check_store_rhs): New function,
25599 (vectorizable_mask_load_store): ...here.
25600 (vectorizable_store): ...and here.
25602 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25604 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
25606 (vectorizable_mask_load_store): ...here.
25608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25610 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
25611 (vect_model_store_cost): Take a vec_load_store_type instead of a
25613 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
25614 (vect_model_store_cost): Take a vec_load_store_type instead of a
25616 (vectorizable_mask_load_store): Update accordingly.
25617 (vectorizable_store): Likewise.
25618 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
25620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25622 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
25623 IFN_MASK_LOAD calls here rather than...
25624 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
25626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25627 Alan Hayward <alan.hayward@arm.com>
25628 David Sherwood <david.sherwood@arm.com>
25630 * expmed.c (extract_bit_field_1): For vector extracts,
25631 fall back to extract_bit_field_as_subreg if vec_extract
25634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25635 Alan Hayward <alan.hayward@arm.com>
25636 David Sherwood <david.sherwood@arm.com>
25638 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
25639 they are variable or constant sized.
25640 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
25641 slots for constant-sized data.
25643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25644 Alan Hayward <alan.hayward@arm.com>
25645 David Sherwood <david.sherwood@arm.com>
25647 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
25648 handling COND_EXPRs with boolean comparisons, try to find a better
25649 basis for the mask type than the boolean itself.
25651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25653 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
25654 is calculated and how it can be overridden.
25655 * genmodes.c (max_bitsize_mode_any_mode): New variable.
25656 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
25658 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
25661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25662 Alan Hayward <alan.hayward@arm.com>
25663 David Sherwood <david.sherwood@arm.com>
25665 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
25666 Remove the mode argument.
25667 (aarch64_simd_valid_immediate): Remove the mode and inverse
25669 * config/aarch64/iterators.md (bitsize): New iterator.
25670 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
25671 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
25672 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
25673 aarch64_simd_valid_immediate.
25674 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
25675 (aarch64_reg_or_bic_imm): Likewise.
25676 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
25677 with an insn_type enum and msl with a modifier_type enum.
25678 Replace element_width with a scalar_mode. Change the shift
25679 to unsigned int. Add constructors for scalar_float_mode and
25680 scalar_int_mode elements.
25681 (aarch64_vect_float_const_representable_p): Delete.
25682 (aarch64_can_const_movi_rtx_p)
25683 (aarch64_simd_scalar_immediate_valid_for_move)
25684 (aarch64_simd_make_constant): Update call to
25685 aarch64_simd_valid_immediate.
25686 (aarch64_advsimd_valid_immediate_hs): New function.
25687 (aarch64_advsimd_valid_immediate): Likewise.
25688 (aarch64_simd_valid_immediate): Remove mode and inverse
25689 arguments. Rewrite to use the above. Use const_vec_duplicate_p
25690 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
25691 and aarch64_float_const_representable_p on the result.
25692 (aarch64_output_simd_mov_immediate): Remove mode argument.
25693 Update call to aarch64_simd_valid_immediate and use of
25694 simd_immediate_info.
25695 (aarch64_output_scalar_simd_mov_immediate): Update call
25698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25699 Alan Hayward <alan.hayward@arm.com>
25700 David Sherwood <david.sherwood@arm.com>
25702 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
25703 (mode_nunits): Likewise CONST_MODE_NUNITS.
25704 * machmode.def (ADJUST_NUNITS): Document.
25705 * genmodes.c (mode_data::need_nunits_adj): New field.
25706 (blank_mode): Update accordingly.
25707 (adj_nunits): New variable.
25708 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
25710 (emit_mode_size_inline): Set need_bytesize_adj for all modes
25711 listed in adj_nunits.
25712 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
25713 listed in adj_nunits. Don't emit case statements for such modes.
25714 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
25715 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
25716 nothing if adj_nunits is nonnull.
25717 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
25718 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
25719 (emit_mode_fbit): Update use of print_maybe_const_decl.
25720 (emit_move_size): Likewise. Treat the array as non-const
25722 (emit_mode_adjustments): Handle adj_nunits.
25724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25726 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
25727 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
25728 (VECTOR_MODES): Use it.
25729 (make_vector_modes): Take the prefix as an argument.
25731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25732 Alan Hayward <alan.hayward@arm.com>
25733 David Sherwood <david.sherwood@arm.com>
25735 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
25736 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
25737 for MODE_VECTOR_BOOL.
25738 * machmode.def (VECTOR_BOOL_MODE): Document.
25739 * genmodes.c (VECTOR_BOOL_MODE): New macro.
25740 (make_vector_bool_mode): New function.
25741 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
25743 * lto-streamer-in.c (lto_input_mode_table): Likewise.
25744 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
25746 * stor-layout.c (int_mode_for_mode): Likewise.
25747 * tree.c (build_vector_type_for_mode): Likewise.
25748 * varasm.c (output_constant_pool_2): Likewise.
25749 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
25750 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
25751 for MODE_VECTOR_BOOL.
25752 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
25753 of mode class checks.
25754 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
25755 instead of a list of mode class checks.
25756 (expand_vector_scalar_condition): Likewise.
25757 (type_for_widest_vector_mode): Handle BImode as an inner mode.
25759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25760 Alan Hayward <alan.hayward@arm.com>
25761 David Sherwood <david.sherwood@arm.com>
25763 * machmode.h (mode_size): Change from unsigned short to
25765 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
25766 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25767 or if measurement_type is not polynomial.
25768 (fixed_size_mode::includes_p): Check for constant-sized modes.
25769 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
25770 return a poly_uint16 rather than an unsigned short.
25771 (emit_mode_size): Change the type of mode_size from unsigned short
25772 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
25773 (emit_mode_adjustments): Cope with polynomial vector sizes.
25774 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25776 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25778 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
25779 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
25780 * caller-save.c (setup_save_areas): Likewise.
25781 (replace_reg_with_saved_mem): Likewise.
25782 * calls.c (emit_library_call_value_1): Likewise.
25783 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
25784 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
25785 (gen_lowpart_for_combine): Likewise.
25786 * convert.c (convert_to_integer_1): Likewise.
25787 * cse.c (equiv_constant, cse_insn): Likewise.
25788 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
25789 (cselib_subst_to_values): Likewise.
25790 * dce.c (word_dce_process_block): Likewise.
25791 * df-problems.c (df_word_lr_mark_ref): Likewise.
25792 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
25793 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
25794 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
25795 (rtl_for_decl_location): Likewise.
25796 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
25797 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
25798 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
25799 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
25800 (expand_expr_real_1): Likewise.
25801 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
25802 (pad_below): Likewise.
25803 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25804 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
25805 * ira.c (get_subreg_tracking_sizes): Likewise.
25806 * ira-build.c (ira_create_allocno_objects): Likewise.
25807 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
25808 (ira_sort_regnos_for_alter_reg): Likewise.
25809 * ira-costs.c (record_operand_costs): Likewise.
25810 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
25811 (resolve_simple_move): Likewise.
25812 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
25813 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
25814 (lra_constraints): Likewise.
25815 (CONST_POOL_OK_P): Reject variable-sized modes.
25816 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
25817 (add_pseudo_to_slot, lra_spill): Likewise.
25818 * omp-low.c (omp_clause_aligned_alignment): Likewise.
25819 * optabs-query.c (get_best_extraction_insn): Likewise.
25820 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25821 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
25822 (expand_mult_highpart, valid_multiword_target_p): Likewise.
25823 * recog.c (offsettable_address_addr_space_p): Likewise.
25824 * regcprop.c (maybe_mode_change): Likewise.
25825 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
25826 * regrename.c (build_def_use): Likewise.
25827 * regstat.c (dump_reg_info): Likewise.
25828 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
25829 (find_reloads, find_reloads_subreg_address): Likewise.
25830 * reload1.c (eliminate_regs_1): Likewise.
25831 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
25832 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
25833 (simplify_binary_operation_1, simplify_subreg): Likewise.
25834 * targhooks.c (default_function_arg_padding): Likewise.
25835 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
25836 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
25837 (verify_gimple_assign_ternary): Likewise.
25838 * tree-inline.c (estimate_move_cost): Likewise.
25839 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25840 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
25841 (get_address_cost_ainc): Likewise.
25842 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
25843 (vect_supportable_dr_alignment): Likewise.
25844 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25845 (vectorizable_reduction): Likewise.
25846 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
25847 (vectorizable_operation, vectorizable_load): Likewise.
25848 * tree.c (build_same_sized_truth_vector_type): Likewise.
25849 * valtrack.c (cleanup_auto_inc_dec): Likewise.
25850 * var-tracking.c (emit_note_insn_var_location): Likewise.
25851 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
25852 (ADDR_VEC_ALIGN): Likewise.
25854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25855 Alan Hayward <alan.hayward@arm.com>
25856 David Sherwood <david.sherwood@arm.com>
25858 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
25860 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25861 or if measurement_type is polynomial.
25862 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
25863 * combine.c (make_extraction): Likewise.
25864 * dse.c (find_shift_sequence): Likewise.
25865 * dwarf2out.c (mem_loc_descriptor): Likewise.
25866 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
25867 (extract_bit_field, extract_low_bits): Likewise.
25868 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
25869 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
25870 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
25871 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
25872 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25873 * reload.c (find_reloads): Likewise.
25874 * reload1.c (alter_reg): Likewise.
25875 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
25876 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
25877 * tree-if-conv.c (predicate_mem_writes): Likewise.
25878 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
25879 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
25880 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
25881 * valtrack.c (dead_debug_insert_temp): Likewise.
25882 * varasm.c (mergeable_constant_section): Likewise.
25883 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
25885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25886 Alan Hayward <alan.hayward@arm.com>
25887 David Sherwood <david.sherwood@arm.com>
25889 * expr.c (expand_assignment): Cope with polynomial mode sizes
25890 when assigning to a CONCAT.
25892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25893 Alan Hayward <alan.hayward@arm.com>
25894 David Sherwood <david.sherwood@arm.com>
25896 * machmode.h (mode_precision): Change from unsigned short to
25898 (mode_to_precision): Return a poly_uint16 rather than an unsigned
25900 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
25901 or if measurement_type is not polynomial.
25902 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
25903 in which the mode is already known to be a scalar_int_mode.
25904 * genmodes.c (emit_mode_precision): Change the type of mode_precision
25905 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
25907 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25908 for GET_MODE_PRECISION.
25909 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25910 for GET_MODE_PRECISION.
25911 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
25913 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
25914 (expand_field_assignment, make_extraction): Likewise.
25915 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
25916 (get_last_value): Likewise.
25917 * convert.c (convert_to_integer_1): Likewise.
25918 * cse.c (cse_insn): Likewise.
25919 * expr.c (expand_expr_real_1): Likewise.
25920 * lra-constraints.c (simplify_operand_subreg): Likewise.
25921 * optabs-query.c (can_atomic_load_p): Likewise.
25922 * optabs.c (expand_atomic_load): Likewise.
25923 (expand_atomic_store): Likewise.
25924 * ree.c (combine_reaching_defs): Likewise.
25925 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
25926 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
25927 * tree.h (type_has_mode_precision_p): Likewise.
25928 * ubsan.c (instrument_si_overflow): Likewise.
25930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
25931 Alan Hayward <alan.hayward@arm.com>
25932 David Sherwood <david.sherwood@arm.com>
25934 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
25935 polynomial numbers of units.
25936 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
25937 (valid_vector_subparts_p): New function.
25938 (build_vector_type): Remove temporary shim and take the number
25939 of units as a poly_uint64 rather than an int.
25940 (build_opaque_vector_type): Take the number of units as a
25941 poly_uint64 rather than an int.
25942 * tree.c (build_vector_from_ctor): Handle polynomial
25943 TYPE_VECTOR_SUBPARTS.
25944 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
25945 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
25946 (build_vector_from_val): If the number of units is variable,
25947 use build_vec_duplicate_cst for constant operands and
25948 VEC_DUPLICATE_EXPR otherwise.
25949 (make_vector_type): Remove temporary is_constant ().
25950 (build_vector_type, build_opaque_vector_type): Take the number of
25951 units as a poly_uint64 rather than an int.
25952 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
25954 * cfgexpand.c (expand_debug_expr): Likewise.
25955 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
25956 (store_constructor, expand_expr_real_1): Likewise.
25957 (const_scalar_mask_from_tree): Likewise.
25958 * fold-const-call.c (fold_const_reduction): Likewise.
25959 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
25960 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
25961 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
25962 (fold_relational_const): Likewise.
25963 (native_interpret_vector): Likewise. Change the size from an
25964 int to an unsigned int.
25965 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
25966 TYPE_VECTOR_SUBPARTS.
25967 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
25968 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
25969 duplicating a non-constant operand into a variable-length vector.
25970 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
25971 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
25972 * ipa-icf.c (sem_variable::equals): Likewise.
25973 * match.pd: Likewise.
25974 * omp-simd-clone.c (simd_clone_subparts): Likewise.
25975 * print-tree.c (print_node): Likewise.
25976 * stor-layout.c (layout_type): Likewise.
25977 * targhooks.c (default_builtin_vectorization_cost): Likewise.
25978 * tree-cfg.c (verify_gimple_comparison): Likewise.
25979 (verify_gimple_assign_binary): Likewise.
25980 (verify_gimple_assign_ternary): Likewise.
25981 (verify_gimple_assign_single): Likewise.
25982 * tree-pretty-print.c (dump_generic_node): Likewise.
25983 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25984 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
25985 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
25986 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
25987 (vect_shift_permute_load_chain): Likewise.
25988 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
25989 (expand_vector_condition, optimize_vector_constructor): Likewise.
25990 (lower_vec_perm, get_compute_type): Likewise.
25991 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25992 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
25993 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
25994 (vect_recog_mask_conversion_pattern): Likewise.
25995 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
25996 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
25997 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25998 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
25999 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
26000 (vectorizable_shift, vectorizable_operation, vectorizable_store)
26001 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
26002 (supportable_widening_operation): Likewise.
26003 (supportable_narrowing_operation): Likewise.
26004 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
26006 * varasm.c (output_constant): Likewise.
26008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26009 Alan Hayward <alan.hayward@arm.com>
26010 David Sherwood <david.sherwood@arm.com>
26012 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
26013 so that both the length == 3 and length != 3 cases set up their
26014 own permute vectors. Add comments explaining why we know the
26015 number of elements is constant.
26016 (vect_permute_load_chain): Likewise.
26018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26019 Alan Hayward <alan.hayward@arm.com>
26020 David Sherwood <david.sherwood@arm.com>
26022 * machmode.h (mode_nunits): Change from unsigned char to
26024 (ONLY_FIXED_SIZE_MODES): New macro.
26025 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
26026 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
26027 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
26029 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
26030 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
26031 or if measurement_type is not polynomial.
26032 * genmodes.c (ZERO_COEFFS): New macro.
26033 (emit_mode_nunits_inline): Make mode_nunits_inline return a
26035 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
26036 Use ZERO_COEFFS when emitting initializers.
26037 * data-streamer.h (bp_pack_poly_value): New function.
26038 (bp_unpack_poly_value): Likewise.
26039 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
26040 for GET_MODE_NUNITS.
26041 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
26042 for GET_MODE_NUNITS.
26043 * tree.c (make_vector_type): Remove temporary shim and make
26044 the real function take the number of units as a poly_uint64
26045 rather than an int.
26046 (build_vector_type_for_mode): Handle polynomial nunits.
26047 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
26048 * emit-rtl.c (const_vec_series_p_1): Likewise.
26049 (gen_rtx_CONST_VECTOR): Likewise.
26050 * fold-const.c (test_vec_duplicate_folding): Likewise.
26051 * genrecog.c (validate_pattern): Likewise.
26052 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
26053 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
26054 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
26055 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
26056 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
26057 * rtlanal.c (subreg_get_info): Likewise.
26058 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26059 (vect_grouped_load_supported): Likewise.
26060 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
26061 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
26062 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
26063 (simplify_const_unary_operation, simplify_binary_operation_1)
26064 (simplify_const_binary_operation, simplify_ternary_operation)
26065 (test_vector_ops_duplicate, test_vector_ops): Likewise.
26066 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
26067 instead of CONST_VECTOR_NUNITS.
26068 * varasm.c (output_constant_pool_2): Likewise.
26069 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
26070 explicit-encoded elements in the XVEC for variable-length vectors.
26072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26074 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
26076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26077 Alan Hayward <alan.hayward@arm.com>
26078 David Sherwood <david.sherwood@arm.com>
26080 * coretypes.h (fixed_size_mode): Declare.
26081 (fixed_size_mode_pod): New typedef.
26082 * builtins.h (target_builtins::x_apply_args_mode)
26083 (target_builtins::x_apply_result_mode): Change type to
26084 fixed_size_mode_pod.
26085 * builtins.c (apply_args_size, apply_result_size, result_vector)
26086 (expand_builtin_apply_args_1, expand_builtin_apply)
26087 (expand_builtin_return): Update accordingly.
26089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26091 * cse.c (hash_rtx_cb): Hash only the encoded elements.
26092 * cselib.c (cselib_hash_rtx): Likewise.
26093 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
26094 CONST_VECTOR encoding.
26096 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26097 Jeff Law <law@redhat.com>
26100 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26101 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26102 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26103 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26106 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26107 explicitly probe *sp in a noreturn function if there were any callee
26108 register saves or frame pointer is needed.
26110 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26113 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26114 BLKmode for ternary, binary or unary expressions.
26117 * var-tracking.c (delete_vta_debug_insn): New inline function.
26118 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26119 insns from get_insns () to NULL instead of each bb separately.
26120 Use delete_vta_debug_insn. No longer static.
26121 (vt_debug_insns_local, variable_tracking_main_1): Adjust
26122 delete_vta_debug_insns callers.
26123 * rtl.h (delete_vta_debug_insns): Declare.
26124 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26125 instead of variable_tracking_main.
26127 2018-01-03 Martin Sebor <msebor@redhat.com>
26129 PR tree-optimization/83603
26130 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26131 arguments past the endof the argument list in functions declared
26132 without a prototype.
26133 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26134 Avoid checking when arguments are null.
26136 2018-01-03 Martin Sebor <msebor@redhat.com>
26139 * doc/extend.texi (attribute const): Fix a typo.
26140 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26141 issuing -Wsuggest-attribute for void functions.
26143 2018-01-03 Martin Sebor <msebor@redhat.com>
26145 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26146 offset_int::from instead of wide_int::to_shwi.
26147 (maybe_diag_overlap): Remove assertion.
26148 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26149 * gimple-ssa-sprintf.c (format_directive): Same.
26150 (parse_directive): Same.
26151 (sprintf_dom_walker::compute_format_length): Same.
26152 (try_substitute_return_value): Same.
26154 2018-01-03 Jeff Law <law@redhat.com>
26156 PR middle-end/83654
26157 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26158 non-constant residual for zero at runtime and avoid probing in
26159 that case. Reorganize code for trailing problem to mirror handling
26162 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
26164 PR tree-optimization/83501
26165 * tree-ssa-strlen.c (get_string_cst): New.
26166 (handle_char_store): Call get_string_cst.
26168 2018-01-03 Martin Liska <mliska@suse.cz>
26170 PR tree-optimization/83593
26171 * tree-ssa-strlen.c: Include tree-cfg.h.
26172 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26173 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26174 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26176 (strlen_dom_walker::before_dom_children): Call
26177 gimple_purge_dead_eh_edges. Dump tranformation with details
26179 (strlen_dom_walker::before_dom_children): Update call by adding
26180 new argument cleanup_eh.
26181 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26183 2018-01-03 Martin Liska <mliska@suse.cz>
26186 * cif-code.def (VARIADIC_THUNK): New enum value.
26187 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26190 2018-01-03 Jan Beulich <jbeulich@suse.com>
26192 * sse.md (mov<mode>_internal): Tighten condition for when to use
26193 vmovdqu<ssescalarsize> for TI and OI modes.
26195 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26197 Update copyright years.
26199 2018-01-03 Martin Liska <mliska@suse.cz>
26202 * ipa-visibility.c (function_and_variable_visibility): Skip
26203 functions with noipa attribure.
26205 2018-01-03 Jakub Jelinek <jakub@redhat.com>
26207 * gcc.c (process_command): Update copyright notice dates.
26208 * gcov-dump.c (print_version): Ditto.
26209 * gcov.c (print_version): Ditto.
26210 * gcov-tool.c (print_version): Ditto.
26211 * gengtype.c (create_file): Ditto.
26212 * doc/cpp.texi: Bump @copying's copyright year.
26213 * doc/cppinternals.texi: Ditto.
26214 * doc/gcc.texi: Ditto.
26215 * doc/gccint.texi: Ditto.
26216 * doc/gcov.texi: Ditto.
26217 * doc/install.texi: Ditto.
26218 * doc/invoke.texi: Ditto.
26220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26222 * vector-builder.h (vector_builder::m_full_nelts): Change from
26223 unsigned int to poly_uint64.
26224 (vector_builder::full_nelts): Update prototype accordingly.
26225 (vector_builder::new_vector): Likewise.
26226 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26227 (vector_builder::operator ==): Likewise.
26228 (vector_builder::finalize): Likewise.
26229 * int-vector-builder.h (int_vector_builder::int_vector_builder):
26230 Take the number of elements as a poly_uint64 rather than an
26232 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26233 from unsigned int to poly_uint64.
26234 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
26235 (vec_perm_indices::new_vector): Likewise.
26236 (vec_perm_indices::length): Likewise.
26237 (vec_perm_indices::nelts_per_input): Likewise.
26238 (vec_perm_indices::input_nelts): Likewise.
26239 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
26240 number of elements per input as a poly_uint64 rather than an
26241 unsigned int. Use the original encoding for variable-length
26242 vectors, rather than clamping each individual element.
26243 For the second and subsequent elements in each pattern,
26244 clamp the step and base before clamping their sum.
26245 (vec_perm_indices::series_p): Handle polynomial element counts.
26246 (vec_perm_indices::all_in_range_p): Likewise.
26247 (vec_perm_indices_to_tree): Likewise.
26248 (vec_perm_indices_to_rtx): Likewise.
26249 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
26250 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
26251 (tree_vector_builder::new_binary_operation): Handle polynomial
26252 element counts. Return false if we need to know the number
26253 of elements at compile time.
26254 * fold-const.c (fold_vec_perm): Punt if the number of elements
26255 isn't known at compile time.
26257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26259 * vec-perm-indices.h (vec_perm_builder): Change element type
26260 from HOST_WIDE_INT to poly_int64.
26261 (vec_perm_indices::element_type): Update accordingly.
26262 (vec_perm_indices::clamp): Handle polynomial element_types.
26263 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26264 (vec_perm_indices::all_in_range_p): Likewise.
26265 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
26267 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
26268 polynomial vec_perm_indices element types.
26269 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
26270 * fold-const.c (fold_vec_perm): Likewise.
26271 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
26272 * tree-vect-generic.c (lower_vec_perm): Likewise.
26273 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26274 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
26275 element type to HOST_WIDE_INT.
26277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26278 Alan Hayward <alan.hayward@arm.com>
26279 David Sherwood <david.sherwood@arm.com>
26281 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
26282 rather than an int. Use plus_constant.
26283 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
26284 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
26286 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26287 Alan Hayward <alan.hayward@arm.com>
26288 David Sherwood <david.sherwood@arm.com>
26290 * calls.c (emit_call_1, expand_call): Change struct_value_size from
26291 a HOST_WIDE_INT to a poly_int64.
26293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26294 Alan Hayward <alan.hayward@arm.com>
26295 David Sherwood <david.sherwood@arm.com>
26297 * calls.c (load_register_parameters): Cope with polynomial
26298 mode sizes. Require a constant size for BLKmode parameters
26299 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
26300 forces a parameter to be padded at the lsb end in order to
26301 fill a complete number of words, require the parameter size
26302 to be ordered wrt UNITS_PER_WORD.
26304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26305 Alan Hayward <alan.hayward@arm.com>
26306 David Sherwood <david.sherwood@arm.com>
26308 * reload1.c (spill_stack_slot_width): Change element type
26309 from unsigned int to poly_uint64_pod.
26310 (alter_reg): Treat mode sizes as polynomial.
26312 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26313 Alan Hayward <alan.hayward@arm.com>
26314 David Sherwood <david.sherwood@arm.com>
26316 * reload.c (complex_word_subreg_p): New function.
26317 (reload_inner_reg_of_subreg, push_reload): Use it.
26319 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26320 Alan Hayward <alan.hayward@arm.com>
26321 David Sherwood <david.sherwood@arm.com>
26323 * lra-constraints.c (process_alt_operands): Reject matched
26324 operands whose sizes aren't ordered.
26325 (match_reload): Refer to this check here.
26327 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26328 Alan Hayward <alan.hayward@arm.com>
26329 David Sherwood <david.sherwood@arm.com>
26331 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
26332 that the mode size is in the set {1, 2, 4, 8, 16}.
26334 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26335 Alan Hayward <alan.hayward@arm.com>
26336 David Sherwood <david.sherwood@arm.com>
26338 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
26339 Use plus_constant instead of gen_rtx_PLUS.
26341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26342 Alan Hayward <alan.hayward@arm.com>
26343 David Sherwood <david.sherwood@arm.com>
26345 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
26346 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
26347 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
26348 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
26349 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
26350 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
26351 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
26352 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
26353 * config/i386/i386.c (ix86_push_rounding): ...this new function.
26354 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
26356 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
26357 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
26358 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
26359 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
26360 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
26361 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
26362 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
26363 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
26364 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
26365 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
26367 * expr.c (emit_move_resolve_push): Treat the input and result
26368 of PUSH_ROUNDING as a poly_int64.
26369 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
26370 (emit_push_insn): Likewise.
26371 * lra-eliminations.c (mark_not_eliminable): Likewise.
26372 * recog.c (push_operand): Likewise.
26373 * reload1.c (elimination_effects): Likewise.
26374 * rtlanal.c (nonzero_bits1): Likewise.
26375 * calls.c (store_one_arg): Likewise. Require the padding to be
26376 known at compile time.
26378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26379 Alan Hayward <alan.hayward@arm.com>
26380 David Sherwood <david.sherwood@arm.com>
26382 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
26383 Use plus_constant instead of gen_rtx_PLUS.
26385 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26386 Alan Hayward <alan.hayward@arm.com>
26387 David Sherwood <david.sherwood@arm.com>
26389 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
26390 rather than an int.
26392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26393 Alan Hayward <alan.hayward@arm.com>
26394 David Sherwood <david.sherwood@arm.com>
26396 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
26397 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
26398 via stack temporaries. Treat the mode size as polynomial too.
26400 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26401 Alan Hayward <alan.hayward@arm.com>
26402 David Sherwood <david.sherwood@arm.com>
26404 * expr.c (expand_expr_real_2): When handling conversions involving
26405 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
26406 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
26407 as a poly_uint64 too.
26409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26410 Alan Hayward <alan.hayward@arm.com>
26411 David Sherwood <david.sherwood@arm.com>
26413 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
26415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26416 Alan Hayward <alan.hayward@arm.com>
26417 David Sherwood <david.sherwood@arm.com>
26419 * combine.c (can_change_dest_mode): Handle polynomial
26420 REGMODE_NATURAL_SIZE.
26421 * expmed.c (store_bit_field_1): Likewise.
26422 * expr.c (store_constructor): Likewise.
26423 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
26424 and polynomial REGMODE_NATURAL_SIZE.
26425 (gen_lowpart_common): Likewise.
26426 * reginfo.c (record_subregs_of_mode): Likewise.
26427 * rtlanal.c (read_modify_subreg_p): Likewise.
26429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26430 Alan Hayward <alan.hayward@arm.com>
26431 David Sherwood <david.sherwood@arm.com>
26433 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
26434 numbers of elements.
26436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26437 Alan Hayward <alan.hayward@arm.com>
26438 David Sherwood <david.sherwood@arm.com>
26440 * match.pd: Cope with polynomial numbers of vector elements.
26442 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26443 Alan Hayward <alan.hayward@arm.com>
26444 David Sherwood <david.sherwood@arm.com>
26446 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
26447 in a POINTER_PLUS_EXPR.
26449 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26450 Alan Hayward <alan.hayward@arm.com>
26451 David Sherwood <david.sherwood@arm.com>
26453 * omp-simd-clone.c (simd_clone_subparts): New function.
26454 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
26455 (ipa_simd_modify_function_body): Likewise.
26457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26458 Alan Hayward <alan.hayward@arm.com>
26459 David Sherwood <david.sherwood@arm.com>
26461 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
26462 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
26463 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
26464 (expand_vector_condition, vector_element): Likewise.
26465 (subparts_gt): New function.
26466 (get_compute_type): Use subparts_gt.
26467 (count_type_subparts): Delete.
26468 (expand_vector_operations_1): Use subparts_gt instead of
26469 count_type_subparts.
26471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26472 Alan Hayward <alan.hayward@arm.com>
26473 David Sherwood <david.sherwood@arm.com>
26475 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
26476 (vect_compile_time_alias): ...this new function. Do the calculation
26477 on poly_ints rather than trees.
26478 (vect_prune_runtime_alias_test_list): Update call accordingly.
26480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26481 Alan Hayward <alan.hayward@arm.com>
26482 David Sherwood <david.sherwood@arm.com>
26484 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
26486 (vect_schedule_slp_instance): Likewise.
26488 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26489 Alan Hayward <alan.hayward@arm.com>
26490 David Sherwood <david.sherwood@arm.com>
26492 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
26493 constant and extern definitions for variable-length vectors.
26494 (vect_get_constant_vectors): Note that the number of units
26495 is known to be constant.
26497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26498 Alan Hayward <alan.hayward@arm.com>
26499 David Sherwood <david.sherwood@arm.com>
26501 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
26502 of units as polynomial. Choose between WIDE and NARROW based
26505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26506 Alan Hayward <alan.hayward@arm.com>
26507 David Sherwood <david.sherwood@arm.com>
26509 * tree-vect-stmts.c (simd_clone_subparts): New function.
26510 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
26512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26513 Alan Hayward <alan.hayward@arm.com>
26514 David Sherwood <david.sherwood@arm.com>
26516 * tree-vect-stmts.c (vectorizable_call): Treat the number of
26517 vectors as polynomial. Use build_index_vector for
26518 IFN_GOMP_SIMD_LANE.
26520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26521 Alan Hayward <alan.hayward@arm.com>
26522 David Sherwood <david.sherwood@arm.com>
26524 * tree-vect-stmts.c (get_load_store_type): Treat the number of
26525 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
26526 for variable-length vectors.
26527 (vectorizable_mask_load_store): Treat the number of units as
26528 polynomial, asserting that it is constant if the condition has
26529 already been enforced.
26530 (vectorizable_store, vectorizable_load): Likewise.
26532 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26533 Alan Hayward <alan.hayward@arm.com>
26534 David Sherwood <david.sherwood@arm.com>
26536 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
26537 of units as polynomial. Punt if we can't tell at compile time
26538 which vector contains the final result.
26540 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26541 Alan Hayward <alan.hayward@arm.com>
26542 David Sherwood <david.sherwood@arm.com>
26544 * tree-vect-loop.c (vectorizable_induction): Treat the number
26545 of units as polynomial. Punt on SLP inductions. Use an integer
26546 VEC_SERIES_EXPR for variable-length integer reductions. Use a
26547 cast of such a series for variable-length floating-point
26550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26551 Alan Hayward <alan.hayward@arm.com>
26552 David Sherwood <david.sherwood@arm.com>
26554 * tree.h (build_index_vector): Declare.
26555 * tree.c (build_index_vector): New function.
26556 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
26557 of units as polynomial, forcibly converting it to a constant if
26558 vectorizable_reduction has already enforced the condition.
26559 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
26560 to create a {1,2,3,...} vector.
26561 (vectorizable_reduction): Treat the number of units as polynomial.
26562 Choose vectype_in based on the largest scalar element size rather
26563 than the smallest number of units. Enforce the restrictions
26566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26567 Alan Hayward <alan.hayward@arm.com>
26568 David Sherwood <david.sherwood@arm.com>
26570 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
26571 number of units as polynomial.
26573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26574 Alan Hayward <alan.hayward@arm.com>
26575 David Sherwood <david.sherwood@arm.com>
26577 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
26578 * target.def (autovectorize_vector_sizes): Return the vector sizes
26579 by pointer, using vector_sizes rather than a bitmask.
26580 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
26581 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
26582 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
26584 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
26585 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
26586 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
26587 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
26588 * omp-general.c (omp_max_vf): Likewise.
26589 * omp-low.c (omp_clause_aligned_alignment): Likewise.
26590 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
26591 * tree-vect-loop.c (vect_analyze_loop): Likewise.
26592 * tree-vect-slp.c (vect_slp_bb): Likewise.
26593 * doc/tm.texi: Regenerate.
26594 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
26596 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
26597 the vector size as a poly_uint64 rather than an unsigned int.
26598 (current_vector_size): Change from an unsigned int to a poly_uint64.
26599 (get_vectype_for_scalar_type): Update accordingly.
26600 * tree.h (build_truth_vector_type): Take the size and number of
26601 units as a poly_uint64 rather than an unsigned int.
26602 (build_vector_type): Add a temporary overload that takes
26603 the number of units as a poly_uint64 rather than an unsigned int.
26604 * tree.c (make_vector_type): Likewise.
26605 (build_truth_vector_type): Take the number of units as a poly_uint64
26606 rather than an unsigned int.
26608 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26609 Alan Hayward <alan.hayward@arm.com>
26610 David Sherwood <david.sherwood@arm.com>
26612 * target.def (get_mask_mode): Take the number of units and length
26613 as poly_uint64s rather than unsigned ints.
26614 * targhooks.h (default_get_mask_mode): Update accordingly.
26615 * targhooks.c (default_get_mask_mode): Likewise.
26616 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
26617 * doc/tm.texi: Regenerate.
26619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26620 Alan Hayward <alan.hayward@arm.com>
26621 David Sherwood <david.sherwood@arm.com>
26623 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
26624 * omp-general.c (omp_max_vf): Likewise.
26625 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
26626 (expand_omp_simd): Handle polynomial safelen.
26627 * omp-low.c (omplow_simd_context): Add a default constructor.
26628 (omplow_simd_context::max_vf): Change from int to poly_uint64.
26629 (lower_rec_simd_input_clauses): Update accordingly.
26630 (lower_rec_input_clauses): Likewise.
26632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26633 Alan Hayward <alan.hayward@arm.com>
26634 David Sherwood <david.sherwood@arm.com>
26636 * tree-vectorizer.h (vect_nunits_for_cost): New function.
26637 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
26638 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
26639 (vect_analyze_slp_cost): Likewise.
26640 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
26641 (vect_model_load_cost): Likewise.
26643 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26644 Alan Hayward <alan.hayward@arm.com>
26645 David Sherwood <david.sherwood@arm.com>
26647 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
26648 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
26649 from an unsigned int * to a poly_uint64_pod *.
26650 (calculate_unrolling_factor): New function.
26651 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
26653 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26654 Alan Hayward <alan.hayward@arm.com>
26655 David Sherwood <david.sherwood@arm.com>
26657 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
26658 from an unsigned int to a poly_uint64.
26659 (_loop_vec_info::slp_unrolling_factor): Likewise.
26660 (_loop_vec_info::vectorization_factor): Change from an int
26662 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
26663 (vect_get_num_vectors): New function.
26664 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
26665 (vect_get_num_copies): Use vect_get_num_vectors.
26666 (vect_analyze_data_ref_dependences): Change max_vf from an int *
26667 to an unsigned int *.
26668 (vect_analyze_data_refs): Change min_vf from an int * to a
26670 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26671 than an unsigned HOST_WIDE_INT.
26672 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
26673 (vect_analyze_data_ref_dependence): Change max_vf from an int *
26674 to an unsigned int *.
26675 (vect_analyze_data_ref_dependences): Likewise.
26676 (vect_compute_data_ref_alignment): Handle polynomial vf.
26677 (vect_enhance_data_refs_alignment): Likewise.
26678 (vect_prune_runtime_alias_test_list): Likewise.
26679 (vect_shift_permute_load_chain): Likewise.
26680 (vect_supportable_dr_alignment): Likewise.
26681 (dependence_distance_ge_vf): Take the vectorization factor as a
26682 poly_uint64 rather than an unsigned HOST_WIDE_INT.
26683 (vect_analyze_data_refs): Change min_vf from an int * to a
26685 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
26686 vfm1 as a poly_uint64 rather than an int. Make the same change
26687 for the returned bound_scalar.
26688 (vect_gen_vector_loop_niters): Handle polynomial vf.
26689 (vect_do_peeling): Likewise. Update call to
26690 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
26691 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
26693 * tree-vect-loop.c (vect_determine_vectorization_factor)
26694 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
26695 (vect_get_known_peeling_cost): Likewise.
26696 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
26697 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
26698 (vect_transform_loop): Likewise. Use the lowest possible VF when
26699 updating the upper bounds of the loop.
26700 (vect_min_worthwhile_factor): Make static. Return an unsigned int
26701 rather than an int.
26702 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
26703 polynomial unroll factors.
26704 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
26705 (vect_make_slp_decision): Likewise.
26706 (vect_supported_load_permutation_p): Likewise, and polynomial
26708 (vect_analyze_slp_cost): Handle polynomial vf.
26709 (vect_slp_analyze_node_operations): Likewise.
26710 (vect_slp_analyze_bb_1): Likewise.
26711 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26712 than an unsigned HOST_WIDE_INT.
26713 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
26714 (vectorizable_load): Handle polynomial vf.
26715 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
26717 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
26719 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26720 Alan Hayward <alan.hayward@arm.com>
26721 David Sherwood <david.sherwood@arm.com>
26723 * match.pd: Handle bit operations involving three constants
26724 and try to fold one pair.
26726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
26728 * tree-vect-loop-manip.c: Include gimple-fold.h.
26729 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
26730 niters_maybe_zero parameters. Handle other cases besides a step of 1.
26731 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
26732 Add a path that uses a step of VF instead of 1, but disable it
26734 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
26735 and niters_no_overflow parameters. Update calls to
26736 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
26737 Create a new SSA name if the latter choses to use a ste other
26738 than zero, and return it via niters_vector_mult_vf_var.
26739 * tree-vect-loop.c (vect_transform_loop): Update calls to
26740 vect_do_peeling, vect_gen_vector_loop_niters and
26741 slpeel_make_loop_iterate_ntimes.
26742 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
26743 (vect_gen_vector_loop_niters): Update declarations after above changes.
26745 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
26747 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
26748 128-bit round to integer instructions.
26749 (ceil<mode>2): Likewise.
26750 (btrunc<mode>2): Likewise.
26751 (round<mode>2): Likewise.
26753 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26755 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
26756 unaligned VSX load/store on P8/P9.
26757 (expand_block_clear): Allow the use of unaligned VSX
26758 load/store on P8/P9.
26760 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26762 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
26764 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
26765 swap associated with both a load and a store.
26767 2018-01-02 Andrew Waterman <andrew@sifive.com>
26769 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
26770 * config/riscv/riscv.md (clear_cache): Use it.
26772 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
26774 * web.c: Remove out-of-date comment.
26776 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26778 * expr.c (fixup_args_size_notes): Check that any existing
26779 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
26780 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
26781 (emit_single_push_insn): ...here.
26783 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26785 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
26786 (const_vector_encoded_nelts): New function.
26787 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
26788 (const_vector_int_elt, const_vector_elt): Declare.
26789 * emit-rtl.c (const_vector_int_elt_1): New function.
26790 (const_vector_elt): Likewise.
26791 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
26792 of CONST_VECTOR_ELT.
26794 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26796 * expr.c: Include rtx-vector-builder.h.
26797 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
26798 directly on the tree encoding.
26799 (const_vector_from_tree): Likewise.
26800 * optabs.c: Include rtx-vector-builder.h.
26801 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
26802 sequence of "u" values.
26803 * vec-perm-indices.c: Include rtx-vector-builder.h.
26804 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
26805 directly on the vec_perm_indices encoding.
26807 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26809 * doc/rtl.texi (const_vector): Describe new encoding scheme.
26810 * Makefile.in (OBJS): Add rtx-vector-builder.o.
26811 * rtx-vector-builder.h: New file.
26812 * rtx-vector-builder.c: Likewise.
26813 * rtl.h (rtx_def::u2): Add a const_vector field.
26814 (CONST_VECTOR_NPATTERNS): New macro.
26815 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
26816 (CONST_VECTOR_DUPLICATE_P): Likewise.
26817 (CONST_VECTOR_STEPPED_P): Likewise.
26818 (CONST_VECTOR_ENCODED_ELT): Likewise.
26819 (const_vec_duplicate_p): Check for a duplicated vector encoding.
26820 (unwrap_const_vec_duplicate): Likewise.
26821 (const_vec_series_p): Check for a non-duplicated vector encoding.
26822 Say that the function only returns true for integer vectors.
26823 * emit-rtl.c: Include rtx-vector-builder.h.
26824 (gen_const_vec_duplicate_1): Delete.
26825 (gen_const_vector): Call gen_const_vec_duplicate instead of
26826 gen_const_vec_duplicate_1.
26827 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
26828 (gen_const_vec_duplicate): Use rtx_vector_builder.
26829 (gen_const_vec_series): Likewise.
26830 (gen_rtx_CONST_VECTOR): Likewise.
26831 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
26832 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26833 Build a new vector rather than modifying a CONST_VECTOR in-place.
26834 (handle_special_swappables): Update call accordingly.
26835 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
26836 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26837 Build a new vector rather than modifying a CONST_VECTOR in-place.
26838 (handle_special_swappables): Update call accordingly.
26840 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26842 * simplify-rtx.c (simplify_const_binary_operation): Use
26843 CONST_VECTOR_ELT instead of XVECEXP.
26845 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26847 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
26848 the selector elements to be different from the data elements
26849 if the selector is a VECTOR_CST.
26850 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
26851 ssizetype for the selector.
26853 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26855 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
26856 before testing each element individually.
26857 * tree-vect-generic.c (lower_vec_perm): Likewise.
26859 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26861 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
26862 * selftest-run-tests.c (selftest::run_tests): Call it.
26863 * vector-builder.h (vector_builder::operator ==): New function.
26864 (vector_builder::operator !=): Likewise.
26865 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
26866 (vec_perm_indices::all_from_input_p): New function.
26867 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26868 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
26869 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
26870 instead of reading the VECTOR_CST directly. Detect whether both
26871 vector inputs are the same before constructing the vec_perm_indices,
26872 and update the number of inputs argument accordingly. Use the
26873 utility functions added above. Only construct sel2 if we need to.
26875 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26877 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
26878 the broadcast of the low byte.
26879 (expand_mult_highpart): Use an explicit encoding for the permutes.
26880 * optabs-query.c (can_mult_highpart_p): Likewise.
26881 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
26882 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26883 (vectorizable_bswap): Likewise.
26884 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
26885 explicit encoding for the power-of-2 permutes.
26886 (vect_permute_store_chain): Likewise.
26887 (vect_grouped_load_supported): Likewise.
26888 (vect_permute_load_chain): Likewise.
26890 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26892 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
26893 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
26894 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
26895 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26896 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
26897 (vect_gen_perm_mask_any): Likewise.
26899 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26901 * int-vector-builder.h: New file.
26902 * vec-perm-indices.h: Include int-vector-builder.h.
26903 (vec_perm_indices): Redefine as an int_vector_builder.
26904 (auto_vec_perm_indices): Delete.
26905 (vec_perm_builder): Redefine as a stand-alone class.
26906 (vec_perm_indices::vec_perm_indices): New function.
26907 (vec_perm_indices::clamp): Likewise.
26908 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
26909 (vec_perm_indices::new_vector): New function.
26910 (vec_perm_indices::new_expanded_vector): Update for new
26911 vec_perm_indices class.
26912 (vec_perm_indices::rotate_inputs): New function.
26913 (vec_perm_indices::all_in_range_p): Operate directly on the
26914 encoded form, without computing elided elements.
26915 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
26916 encoding. Update for new vec_perm_indices class.
26917 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
26918 the given vec_perm_builder.
26919 (expand_vec_perm_var): Update vec_perm_builder constructor.
26920 (expand_mult_highpart): Use vec_perm_builder instead of
26921 auto_vec_perm_indices.
26922 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
26923 vec_perm_indices instead of auto_vec_perm_indices. Use a single
26924 or double series encoding as appropriate.
26925 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
26926 vec_perm_indices instead of auto_vec_perm_indices.
26927 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26928 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26929 (vect_permute_store_chain): Likewise.
26930 (vect_grouped_load_supported): Likewise.
26931 (vect_permute_load_chain): Likewise.
26932 (vect_shift_permute_load_chain): Likewise.
26933 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26934 (vect_transform_slp_perm_load): Likewise.
26935 (vect_schedule_slp_instance): Likewise.
26936 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26937 (vectorizable_mask_load_store): Likewise.
26938 (vectorizable_bswap): Likewise.
26939 (vectorizable_store): Likewise.
26940 (vectorizable_load): Likewise.
26941 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
26942 vec_perm_indices instead of auto_vec_perm_indices. Use
26943 tree_to_vec_perm_builder to read the vector from a tree.
26944 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
26945 vec_perm_builder instead of a vec_perm_indices.
26946 (have_whole_vector_shift): Use vec_perm_builder and
26947 vec_perm_indices instead of auto_vec_perm_indices. Leave the
26948 truncation to calc_vec_perm_mask_for_shift.
26949 (vect_create_epilog_for_reduction): Likewise.
26950 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
26951 from auto_vec_perm_indices to vec_perm_indices.
26952 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26953 instead of changing individual elements.
26954 (aarch64_vectorize_vec_perm_const): Use new_vector to install
26955 the vector in d.perm.
26956 * config/arm/arm.c (expand_vec_perm_d::perm): Change
26957 from auto_vec_perm_indices to vec_perm_indices.
26958 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26959 instead of changing individual elements.
26960 (arm_vectorize_vec_perm_const): Use new_vector to install
26961 the vector in d.perm.
26962 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
26963 Update vec_perm_builder constructor.
26964 (rs6000_expand_interleave): Likewise.
26965 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
26966 (rs6000_expand_interleave): Likewise.
26968 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26970 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
26971 to qimode could truncate the indices.
26972 * optabs.c (expand_vec_perm_var): Likewise.
26974 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
26976 * Makefile.in (OBJS): Add vec-perm-indices.o.
26977 * vec-perm-indices.h: New file.
26978 * vec-perm-indices.c: Likewise.
26979 * target.h (vec_perm_indices): Replace with a forward class
26981 (auto_vec_perm_indices): Move to vec-perm-indices.h.
26982 * optabs.h: Include vec-perm-indices.h.
26983 (expand_vec_perm): Delete.
26984 (selector_fits_mode_p, expand_vec_perm_var): Declare.
26985 (expand_vec_perm_const): Declare.
26986 * target.def (vec_perm_const_ok): Replace with...
26987 (vec_perm_const): ...this new hook.
26988 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
26989 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
26990 * doc/tm.texi: Regenerate.
26991 * optabs.def (vec_perm_const): Delete.
26992 * doc/md.texi (vec_perm_const): Likewise.
26993 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
26994 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
26995 expand_vec_perm for constant permutation vectors. Assert that
26996 the mode of variable permutation vectors is the integer equivalent
26997 of the mode that is being permuted.
26998 * optabs-query.h (selector_fits_mode_p): Declare.
26999 * optabs-query.c: Include vec-perm-indices.h.
27000 (selector_fits_mode_p): New function.
27001 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
27002 is defined, instead of checking whether the vec_perm_const_optab
27003 exists. Use targetm.vectorize.vec_perm_const instead of
27004 targetm.vectorize.vec_perm_const_ok. Check whether the indices
27005 fit in the vector mode before using a variable permute.
27006 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
27007 vec_perm_indices instead of an rtx.
27008 (expand_vec_perm): Replace with...
27009 (expand_vec_perm_const): ...this new function. Take the selector
27010 as a vec_perm_indices rather than an rtx. Also take the mode of
27011 the selector. Update call to shift_amt_for_vec_perm_mask.
27012 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
27013 Use vec_perm_indices::new_expanded_vector to expand the original
27014 selector into bytes. Check whether the indices fit in the vector
27015 mode before using a variable permute.
27016 (expand_vec_perm_var): Make global.
27017 (expand_mult_highpart): Use expand_vec_perm_const.
27018 * fold-const.c: Includes vec-perm-indices.h.
27019 * tree-ssa-forwprop.c: Likewise.
27020 * tree-vect-data-refs.c: Likewise.
27021 * tree-vect-generic.c: Likewise.
27022 * tree-vect-loop.c: Likewise.
27023 * tree-vect-slp.c: Likewise.
27024 * tree-vect-stmts.c: Likewise.
27025 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
27027 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
27028 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
27029 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
27030 (aarch64_vectorize_vec_perm_const): ...this new function.
27031 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27032 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27033 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
27034 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
27035 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27036 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27037 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
27039 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
27040 check for NEON modes.
27041 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
27042 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
27043 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
27044 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
27046 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
27047 the old VEC_PERM_CONST conditions.
27048 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
27049 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
27050 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
27051 (ia64_vectorize_vec_perm_const_ok): Merge into...
27052 (ia64_vectorize_vec_perm_const): ...this new function.
27053 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
27054 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
27055 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
27056 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
27057 * config/mips/mips.c (mips_expand_vec_perm_const)
27058 (mips_vectorize_vec_perm_const_ok): Merge into...
27059 (mips_vectorize_vec_perm_const): ...this new function.
27060 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
27061 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
27062 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
27063 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
27064 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
27065 (rs6000_expand_vec_perm_const): Delete.
27066 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
27068 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27069 (altivec_expand_vec_perm_const_le): Take each operand individually.
27070 Operate on constant selectors rather than rtxes.
27071 (altivec_expand_vec_perm_const): Likewise. Update call to
27072 altivec_expand_vec_perm_const_le.
27073 (rs6000_expand_vec_perm_const): Delete.
27074 (rs6000_vectorize_vec_perm_const_ok): Delete.
27075 (rs6000_vectorize_vec_perm_const): New function.
27076 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27077 an element count and rtx array.
27078 (rs6000_expand_extract_even): Update call accordingly.
27079 (rs6000_expand_interleave): Likewise.
27080 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
27081 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
27082 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
27083 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
27084 (rs6000_expand_vec_perm_const): Delete.
27085 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
27086 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27087 (altivec_expand_vec_perm_const_le): Take each operand individually.
27088 Operate on constant selectors rather than rtxes.
27089 (altivec_expand_vec_perm_const): Likewise. Update call to
27090 altivec_expand_vec_perm_const_le.
27091 (rs6000_expand_vec_perm_const): Delete.
27092 (rs6000_vectorize_vec_perm_const_ok): Delete.
27093 (rs6000_vectorize_vec_perm_const): New function. Remove stray
27094 reference to the SPE evmerge intructions.
27095 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
27096 an element count and rtx array.
27097 (rs6000_expand_extract_even): Update call accordingly.
27098 (rs6000_expand_interleave): Likewise.
27099 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27100 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27102 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27104 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27106 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27107 vector mode and that that mode matches the mode of the data
27109 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27110 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
27111 directly using expand_vec_perm_1 when forcing selectors into
27113 (expand_vec_perm_var): New function, split out from expand_vec_perm.
27115 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27117 * optabs-query.h (can_vec_perm_p): Delete.
27118 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27119 * optabs-query.c (can_vec_perm_p): Split into...
27120 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27121 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27122 particular selector is valid.
27123 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27124 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27125 (vect_grouped_load_supported): Likewise.
27126 (vect_shift_permute_load_chain): Likewise.
27127 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27128 (vect_transform_slp_perm_load): Likewise.
27129 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27130 (vectorizable_bswap): Likewise.
27131 (vect_gen_perm_mask_checked): Likewise.
27132 * fold-const.c (fold_ternary_loc): Likewise. Don't take
27133 implementations of variable permutation vectors into account
27134 when deciding which selector to use.
27135 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27136 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27137 with a false third argument.
27138 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27139 to test whether the constant selector is valid and can_vec_perm_var_p
27140 to test whether a variable selector is valid.
27142 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27144 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27145 * optabs-query.c (can_vec_perm_p): Likewise.
27146 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27147 instead of vec_perm_indices.
27148 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27149 (vect_gen_perm_mask_checked): Likewise,
27150 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27151 (vect_gen_perm_mask_checked): Likewise,
27153 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
27155 * optabs-query.h (qimode_for_vec_perm): Declare.
27156 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27157 (qimode_for_vec_perm): ...this new function.
27158 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27160 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
27162 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27163 does not have a conditional at the top.
27165 2018-01-02 Richard Biener <rguenther@suse.de>
27167 * ipa-inline.c (big_speedup_p): Fix expression.
27169 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27172 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27175 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
27179 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27180 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27181 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27182 cond_taken_branch_cost 3->4.
27184 2018-01-01 Jakub Jelinek <jakub@redhat.com>
27186 PR tree-optimization/83581
27187 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27188 TODO_cleanup_cfg if any changes have been made.
27190 PR middle-end/83608
27191 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27192 convert_modes if target mode has the right side, but different mode
27195 PR middle-end/83609
27196 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27197 last argument when extracting from CONCAT. If either from_real or
27198 from_imag is NULL, use expansion through memory. If result is not
27199 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27200 the parts directly to inner mode, if even that fails, use expansion
27203 PR middle-end/83623
27204 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27205 check for bswap in mode rather than HImode and use that in expand_unop
27208 Copyright (C) 2018 Free Software Foundation, Inc.
27210 Copying and distribution of this file, with or without modification,
27211 are permitted in any medium without royalty provided the copyright
27212 notice and this notice are preserved.